Compare commits

..

No commits in common. "master" and "fix_decodeClassObj" have entirely different histories.

420 changed files with 12972 additions and 58119 deletions

12
.gitignore vendored
View File

@ -16,14 +16,4 @@ xcuserdata
/test.lyng
/sample_texts/1.txt.gz
/kotlin-js-store/wasm/yarn.lock
/distributables
.output*.txt
debug.log
/build.log
/test.md
/build_output.txt
/build_output_full.txt
/check_output.txt
/compile_jvm_output.txt
/compile_metadata_output.txt
test_output*.txt
/distributables

View File

@ -1,11 +0,0 @@
# Lyng Project Guidelines
This project uses the Lyng scripting language for multiplatform scripting.
## Coding in Lyng
When writing, refactoring, or analyzing Lyng code:
- **Reference**: Always use `LYNG_AI_SPEC.md` in the project root as the primary source of truth for syntax and idioms.
- **File Extensions**: Use `.lyng` for all script files.
- **Implicit Coroutines**: Remember that all Lyng functions are implicitly coroutines; do not look for `async/await`.
- **Everything is an Expression**: Leverage the fact that blocks, if-statements, and loops return values.
- **Maps vs Blocks**: Be careful: `{}` is a block/lambda, use `Map()` for an empty map.

View File

@ -1,20 +1,3 @@
<!--
~ Copyright 2025 Sergey S. Chernov real.sergeych@gmail.com
~
~ Licensed under the Apache License, Version 2.0 (the "License");
~ you may not use this file except in compliance with the License.
~ You may obtain a copy of the License at
~
~ http://www.apache.org/licenses/LICENSE-2.0
~
~ Unless required by applicable law or agreed to in writing, software
~ distributed under the License is distributed on an "AS IS" BASIS,
~ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
~ See the License for the specific language governing permissions and
~ limitations under the License.
~
-->
<component name="ProjectRunConfigurationManager">
<configuration default="false" name="lyng:site [jsBrowserDevelopmentRun]" type="GradleRunConfiguration" factoryName="Gradle">
<ExternalSystemSettings>
@ -34,11 +17,8 @@
</ExternalSystemSettings>
<ExternalSystemDebugServerProcess>true</ExternalSystemDebugServerProcess>
<ExternalSystemReattachDebugProcess>true</ExternalSystemReattachDebugProcess>
<ExternalSystemDebugDisabled>false</ExternalSystemDebugDisabled>
<DebugAllEnabled>false</DebugAllEnabled>
<RunAsTest>false</RunAsTest>
<GradleProfilingDisabled>true</GradleProfilingDisabled>
<GradleCoverageDisabled>true</GradleCoverageDisabled>
<method v="2" />
</configuration>
</component>

View File

@ -1,23 +0,0 @@
# AI Agent Notes
## Kotlin/Wasm generation guardrails
- Avoid creating suspend lambdas for compiler runtime statements. Prefer explicit `object : Statement()` with `override suspend fun execute(...)`.
- Do not use `statement { ... }` or other inline suspend lambdas in compiler hot paths (e.g., parsing/var declarations, initializer thunks).
- If you need a wrapper for delegated properties, check for `getValue` explicitly and return a concrete `Statement` object when missing; avoid `onNotFoundResult` lambdas.
- If wasmJs browser tests hang, first run `:lynglib:wasmJsNodeTest` and look for wasm compilation errors; hangs usually mean module instantiation failed.
- Do not increase test timeouts to mask wasm generation errors; fix the invalid IR instead.
## Type inference notes (notes/new_lyng_type_system_spec.md)
- Nullability is Kotlin-style: `T` non-null, `T?` nullable, `!!` asserts non-null.
- `void` is a singleton of class `Void` (syntax sugar for return type).
- Object members are always allowed even on unknown types; non-Object members require explicit casts. Remove `inspect` from Object and use `toInspectString()` instead.
- Type expression checks: `x is T` is value instance check; `T1 is T2` is type-subset; `A in T` means `A` is subset of `T`; `==` is structural type equality.
- Type aliases: `type Name = TypeExpr` (generic allowed) expand to their underlying type expressions; no nominal distinctness.
- Bounds and variance: `T: A & B` / `T: A | B` for bounds; declaration-site variance with `out` / `in`.
- Do not reintroduce bytecode fallback opcodes (e.g., `GET_NAME`, `EVAL_*`, `CALL_FALLBACK`) or runtime name-resolution fallbacks; all symbol resolution must stay compile-time only.
## Bytecode frame-first migration plan
- Treat frame slots as the only storage for locals/temps by default; avoid pre-creating scope slot mappings for compiled functions.
- Create closure references only when a capture is detected; use a direct frame+slot reference (foreign slot ref) instead of scope slots.
- Keep Scope as a lazy reflection facade: resolve name -> slot only on demand for Kotlin interop (no eager name mapping on every call).
- Avoid PUSH_SCOPE/POP_SCOPE in bytecode for loops/functions unless dynamic name access or Kotlin reflection is requested.

View File

@ -1,88 +1,8 @@
## 1.5.0-SNAPSHOT
## Changelog
### Language Features
- Added `return` statement with local and non-local exit support (`return@label`).
- Support for `abstract` classes, methods, and variables.
- Introduced `interface` as a synonym for `abstract class`.
- Multiple Inheritance (MI) completed and enabled by default (C3 MRO).
- Class properties with custom accessors (`get`, `set`).
- Restricted setter visibility (`private set`, `protected set`).
- Late-initialized `val` fields in classes with `Unset` protection.
- Named arguments (`name: value`) and named splats (`...Map`).
- Assign-if-null operator `?=`.
- Refined `protected` visibility rules and `closed` modifier.
- Transient attribute `@Transient` for serialization and equality.
- Unified Delegation model for `val`, `var`, and `fun`.
- Singleton objects (`object`) and object expressions.
### Unreleased
### Standard Library
- Added `with(self, block)` for scoped execution.
- Added `clamp()` function and extension.
- Improved `Exception` and `StackTraceEntry` reporting.
### Tooling and IDE
- **CLI**: Added `fmt` as a first-class subcommand for code formatting.
- **IDEA Plugin**: Lightweight autocompletion (experimental), improved docs, and Grazie integration.
- **Highlighters**: Updated TextMate bundle and website highlighters for new syntax.
### Detailed Changes:
- Language: Refined `protected` visibility rules
- Ancestor classes can now access `protected` members of their descendants, provided the ancestor also defines or inherits a member with the same name (indicating an override of a member known to the ancestor).
- This allows patterns where a base class calls a `protected` method that is implemented in a subclass.
- Fixed a regression where self-calls to unmangled members sometimes bypassed visibility checks incorrectly; these are now handled by the refined logic.
- Language: Added `return` statement
- `return [expression]` exits the innermost enclosing callable (function or lambda).
- Supports non-local returns using `@label` syntax (e.g., `return@outer 42`).
- Named functions automatically provide their name as a label for non-local returns.
- Labeled lambdas: lambdas can be explicitly labeled using `@label { ... }`.
- Restriction: `return` is forbidden in shorthand function definitions (e.g., `fun f(x) = return x` is a syntax error).
- Control Flow: `return` and `break` are now protected from being caught by user-defined `try-catch` blocks in Lyng.
- Documentation: New `docs/return_statement.md` and updated `tutorial.md`.
- Language: stdlib improvements
- Added `with(self, block)` function to `root.lyng` which executes a block with `this` set to the provided object.
- Language: Abstract Classes and Interfaces
- Support for `abstract` modifier on classes, methods, and variables.
- Introduced `interface` as a synonym for `abstract class`, supporting full state (constructors, fields, `init` blocks) and implementation by parts via MI.
- New `closed` modifier (antonym to `open`) to prevent overriding class members.
- Refined `override` logic: mandatory keyword when re-declaring members that exist in the ancestor chain (MRO).
- MI Satisfaction: Abstract requirements are automatically satisfied by matching concrete members found later in the C3 MRO chain without requiring explicit proxy methods.
- Integration: Updated highlighters (lynglib, lyngweb, IDEA plugin), IDEA completion, and Grazie grammar checking.
- Documentation: Updated `docs/OOP.md` with sections on "Abstract Classes and Members", "Interfaces", and "Overriding and Virtual Dispatch".
- IDEA plugin: Improved natural language support and spellchecking
- Disabled the limited built-in English and Technical dictionaries.
- Enforced usage of the platform's standard Natural Languages (Grazie) and Spellchecker components.
- Integrated `SpellCheckerManager` for word suggestions and validation, respecting users' personal and project dictionaries.
- Added project-specific "learned words" support via `Lyng Formatter` settings and quick-fixes.
- Enhanced fallback spellchecker for technical terms and Lyng-specific vocabulary.
- Language: Class properties with accessors
- Support for `val` (read-only) and `var` (read-write) properties in classes.
- Syntax: `val name [ : Type ] get() { body }` or `var name [ : Type ] get() { body } set(value) { body }`.
- Laconic Expression Shorthand: `val prop get() = expression` and `var prop get() = read set(v) = write`.
- Properties are pure accessors and do **not** have automatic backing fields.
- Validation: `var` properties must have both accessors; `val` must have only a getter.
- Integration: Updated TextMate grammar and IntelliJ plugin (highlighting + keywords).
- Documentation: New "Properties" section in `docs/OOP.md`.
- Language: Restricted Setter Visibility
- Support for `private set` and `protected set` modifiers on `var` fields and properties.
- Allows members to be publicly readable but only writable from within the declaring class or its subclasses.
- Enforcement at runtime: throws `AccessException` on unauthorized writes.
- Supported only for declarations in class bodies (fields and properties).
- Documentation: New "Restricted Setter Visibility" section in `docs/OOP.md`.
- Language: Late-initialized `val` fields in classes
- Support for declaring `val` without an immediate initializer in class bodies.
- Compulsory initialization: every late-init `val` must be assigned at least once within the class body or an `init` block.
- Write-once enforcement: assigning to a `val` is allowed only if its current value is `Unset`.
- Access protection: reading a late-init `val` before it is assigned returns the `Unset` singleton; using `Unset` for most operations throws an `UnsetException`.
- Extension properties do not support late-init.
- Documentation: New "Late-initialized `val` fields" and "The `Unset` singleton" sections in `docs/OOP.md`.
- Docs: OOP improvements
- Docs: Scopes and Closures guidance
- New page: `docs/scopes_and_closures.md` detailing `ClosureScope` resolution order, recursion‑safe helpers (`chainLookupIgnoreClosure`, `chainLookupWithMembers`, `baseGetIgnoreClosure`), cycle prevention, and capturing lexical environments for callbacks (`snapshotForClosure`).
- Updated: `docs/advanced_topics.md` (link to the new page), `docs/parallelism.md` (closures in `launch`/`flow`), `docs/OOP.md` (visibility from closures with preserved `currentClassCtx`), `docs/exceptions_handling.md` (compatibility alias `SymbolNotFound`).
- Tutorial: added quick link to Scopes and Closures.
@ -126,7 +46,7 @@
- Header-specified constructor arguments are passed to direct bases.
- Visibility enforcement under MI:
- `private` visible only inside the declaring class body.
- `protected` visible inside the declaring class and its transitive subclasses. Additionally, ancestor classes can access protected members of their descendants if it's an override of a member known to the ancestor. Unrelated contexts cannot access it (qualification/casts do not bypass).
- `protected` visible inside the declaring class and any of its transitive subclasses; unrelated contexts cannot access it (qualification/casts do not bypass).
- Diagnostics improvements:
- Missing member/field messages include receiver class and linearization order; hints for `this@Type` or casts when helpful.
- Invalid `this@Type` reports that the qualifier is not an ancestor and shows the receiver lineage.
@ -134,6 +54,16 @@
- Documentation updated (docs/OOP.md and tutorial quick-start) to reflect MI with active C3 MRO.
Notes:
- Existing single-inheritance code continues to work; resolution reduces to the single base.
- If code previously relied on non-deterministic parent set iteration, C3 MRO provides a predictable order; disambiguate explicitly if needed using `this@Type`/casts.
# Changelog
All notable changes to this project will be documented in this file.
## Unreleased
- CLI: Added `fmt` as a first-class Clikt subcommand.
- Default behavior: formats files to stdout (no in-place edits by default).
- Options:
@ -144,9 +74,6 @@
- Mutually exclusive: `--check` and `--in-place` together now produce an error and exit with code 1.
- Multi-file stdout prints headers `--- <path> ---` per file.
- `lyng --help` shows `fmt`; `lyng fmt --help` displays dedicated help.
- Fix: Property accessors (`get`, `set`, `private set`, `protected set`) are now correctly indented relative to the property declaration.
- Fix: Indentation now correctly carries over into blocks that start on extra‑indented lines (e.g., nested `if` statements or property accessor bodies).
- Fix: Formatting Markdown files no longer deletes content in `.lyng` code fences and works correctly with injected files (resolves clobbering, `StringIndexOutOfBoundsException`, and `nonempty text is not covered by block` errors).
- CLI: Preserved legacy script invocation fast-paths:
- `lyng script.lyng [args...]` executes the script directly.

View File

@ -1,130 +0,0 @@
# Lyng Language AI Specification (V1.5.0-SNAPSHOT)
High-density specification for LLMs. Reference this for all Lyng code generation.
## 1. Core Philosophy & Syntax
- **Everything is an Expression**: Blocks, `if`, `when`, `for`, `while`, `do-while` return their last expression (or `void`).
- **Static Types + Inference**: Every declaration has a compile-time type (explicit or inferred). Types are Kotlin‑style: non‑null by default, nullable with `?`.
- **Loops with `else`**: `for`, `while`, and `do-while` support an optional `else` block.
- `else` executes **only if** the loop finishes normally (without a `break`).
- `break <value>` exits the loop and sets its return value.
- Loop Return Value:
1. Value from `break <value>`.
2. Result of `else` block (if loop finished normally and `else` exists).
3. Result of the last iteration (if loop finished normally and no `else`).
4. `void` (if loop body never executed and no `else`).
- **Implicit Coroutines**: All functions are coroutines. No `async/await`. Use `launch { ... }` (returns `Deferred`) or `flow { ... }`.
- **Functions**: Use `fun` or the short form `fn`. Function declarations are expressions returning a callable.
- **Variables**: `val` (read-only), `var` (mutable). Supports late-init `val` in classes (must be assigned in `init` or body).
- **Serialization**: Use `@Transient` attribute before `val`/`var` or constructor parameters to exclude them from Lynon/JSON serialization. Transient fields are also ignored during `==` structural equality checks.
- **Null Safety**: `?` (nullable type), `?.` (safe access), `?( )` (safe invoke), `?{ }` (safe block invoke), `?[ ]` (safe index), `?:` or `??` (elvis), `?=` (assign-if-null).
- **Equality**: `==` (equals), `!=` (not equals), `===` (ref identity), `!==` (ref not identity).
- **Comparison**: `<`, `>`, `<=`, `>=`, `<=>` (shuttle/spaceship, returns -1, 0, 1).
- **Destructuring**: `val [a, b, rest...] = list`. Supports nested `[a, [b, c]]` and splats.
- **Compile-Time Resolution Only**: All names/members must resolve at compile time. No runtime name lookup or fallback opcodes.
## 2. Object-Oriented Programming (OOP)
- **Multiple Inheritance**: Supported with **C3 MRO** (Python-style). Diamond-safe.
- **Header Arguments**: `class Foo(a, b) : Base(a)` defines fields `a`, `b` and passes `a` to `Base`.
- **Members**: `fun name(args) { ... }`, `val`, `var`, `static val`, `static fun`.
- **Properties (Get/Set)**: Pure accessors, no auto-backing fields.
```lyng
var age
get() = _age
private set(v) { if(v >= 0) _age = v }
// Laconic syntax:
val area get = π * r * r
```
- **Mandatory `override`**: Required for all members existing in the ancestor chain.
- **Visibility**: `public` (default), `protected` (subclasses and ancestors for overrides), `private` (this class instance only). `private set` / `protected set` allowed on properties.
- **Disambiguation**: `this@Base.member()` or `(obj as Base).member()`. `as` returns a qualified view.
- **Abstract/Interface**: `interface` is a synonym for `abstract class`. Both support state and constructors.
- **Extensions**: `fun Class.ext()` or `val Class.ext get = ...`. Scope-isolated.
- **Member Access**: Object members (`toString`, `toInspectString`, `let`, `also`, `apply`, `run`) are allowed on unknown types; all other members require a statically known receiver type or explicit cast.
## 2.1 Type System (2026)
- **Root Type**: Everything is an `Object` (root of the hierarchy).
- **Nullability**: Non-null by default (`T`), nullable with `T?`, `!!` asserts non-null.
- **Untyped params**: `fun foo(x)` -> `x: Object`, `fun foo(x?)` -> `x: Object?`.
- **Untyped vars**: `var x` is `Unset` until first assignment locks the type (including nullability).
- `val x = null` -> type `Null`; `var x = null` -> type `Object?`.
- **Inference**:
- List literals infer union element types; empty list defaults to `List<Object>` unless constrained.
- Map literals infer key/value types; empty map defaults to `Map<Object, Object>` unless constrained.
- Mixed numeric ops promote `Int` + `Real` to `Real`.
- **Type aliases**: `type Name = TypeExpr` (generic allowed). Aliases expand to their underlying type expressions (no nominal distinctness).
- **Generics**: Bounds with `T: A & B` or `T: A | B`; variance uses `out`/`in` (declaration‑site only).
- **Casts**: `as` is a runtime-checked cast; `as?` is safe-cast returning `null`. If the value is nullable, `as T` implies `!!`.
## 2.2 Type Expressions and Checks
- **Value checks**: `x is T` (runtime instance check).
- **Type checks**: `T1 is T2` and `A in T` are subset checks between type expressions (compile-time where possible).
- **Type equality**: `T1 == T2` is structural (unions/intersections are order‑insensitive).
- **Compile-time enforcement**: Bounds are checked at call sites; runtime checks only appear when the compile‑time type is too general.
## 3. Delegation (`by`)
Unified model for `val`, `var`, and `fun`.
```lyng
val x by MyDelegate()
var y by Map() // Uses "y" as key in map
fn f(a, b) by RemoteProxy() // Calls Proxy.invoke(thisRef, "f", a, b)
```
Delegate Methods:
- `getValue(thisRef, name)`: for `val`/`var`.
- `setValue(thisRef, name, val)`: for `var`.
- `invoke(thisRef, name, args...)`: for `fn` (called if `getValue` is absent).
- `bind(name, access, thisRef)`: optional hook called at declaration/binding time. `access` is `DelegateAccess.Val`, `Var`, or `Callable`.
## 4. Standard Library & Functional Built-ins
- **Scope Functions**:
- `obj.let { it... }`: result of block. `it` is `obj`.
- `obj.apply { this... }`: returns `obj`. `this` is `obj`.
- `obj.also { it... }`: returns `obj`. `it` is `obj`.
- `obj.run { this... }`: result of block. `this` is `obj`.
- `with(obj, { ... })`: result of block. `this` is `obj`.
- **Functional**: `forEach`, `map`, `filter`, `any`, `all`, `sum`, `count`, `sortedBy`, `flatten`, `flatMap`, `associateBy`.
- **Lazy**: `val x = cached { expensive() }` (call as `x()`) or `val x by lazy { ... }`.
- **Collections**: `List` ( `[a, b]` ), `Map` ( `Map(k => v)` ), `Set` ( `Set(a, b)` ). `MapEntry` ( `k => v` ).
## 5. Patterns & Shorthands
- **Map Literals**: `{ key: value, identifier: }` (identifier shorthand `x:` is `x: x`). Empty map is `{:}`.
- **Named Arguments**: `fun(y: 10, x: 5)`. Shorthand: `Point(x:, y:)`.
- **Varargs & Splats**: `fun f(args...)`, `f(...otherList)`.
- **Labels**: `loop@ for(x in list) { if(x == 0) break@loop }`.
- **Dynamic**: `val d = dynamic { get { name -> ... } }` allows `d.anyName` via explicit dynamic handler (not implicit fallback).
## 6. Operators & Methods to Overload
| Op | Method | Op | Method |
| :--- | :--- | :--- | :--- |
| `+` | `plus` | `==` | `equals` |
| `-` | `minus` | `<=>` | `compareTo` |
| `*` | `mul` | `[]` | `getAt` / `putAt` |
| `/` | `div` | `!` | `logicalNot` |
| `%` | `mod` | `-` | `negate` (unary) |
| `=~` | `operatorMatch` | `+=` | `plusAssign` |
## 7. Common Snippets
```lyng
// Multiple Inheritance and Properties
class Warrior(id, hp) : Character(id), HealthPool(hp) {
override fun toString() = "Warrior #%s (%s HP)"(id, hp)
}
// Map entry and merging
val m = Map("a" => 1) + ("b" => 2)
m += "c" => 3
// Destructuring with splat
val [first, middle..., last] = [1, 2, 3, 4, 5]
// Safe Navigation and Elvis
val companyName = person?.job?.company?.name ?: "Freelancer"
```
## 8. Standard Library Discovery
To collect data on the standard library and available APIs, AI should inspect:
- **Global Symbols**: `lynglib/src/commonMain/kotlin/net/sergeych/lyng/Script.kt` (root functions like `println`, `sqrt`, `assert`).
- **Core Type Members**: `lynglib/src/commonMain/kotlin/net/sergeych/lyng/obj/*.kt` (e.g., `ObjList.kt`, `ObjString.kt`, `ObjMap.kt`) for methods on built-in types.
- **Lyng-side Extensions**: `lynglib/stdlib/lyng/root.lyng` for high-level functional APIs (e.g., `map`, `filter`, `any`, `lazy`).
- **I/O & Processes**: `lyngio/src/commonMain/kotlin/net/sergeych/lyng/io/` for `fs` and `process` modules.
- **Documentation**: `docs/*.md` (e.g., `tutorial.md`, `lyngio.md`) for high-level usage and module overviews.

View File

@ -1,62 +1,47 @@
# Lyng: ideal scripting for kotlin multiplatform
__Please visit the project homepage: [https://lynglang.com](https://lynglang.com) and a [telegram channel](https://t.me/lynglang).__
__Main development site:__ [https://gitea.sergeych.net/SergeychWorks/lyng](https://gitea.sergeych.net/SergeychWorks/lyng)
__github mirror__: [https://github.com/sergeych/lyng](https://github.com/sergeych/lyng)
We keep github as a mirror and backup for the project, while the main development site is hosted on gitea.sergeych.net. We use gitea for issues and pull requests, and as a main point of trust, as github access now is a thing that can momentarily be revoked for no apparent reason.
We encourage using the github if the main site is not accessible from your country and vice versa. We recommend to `publishToMavenLocal` and not depend on politics.
# Lyng: modern scripting for kotlin multiplatform
Please visit the project homepage: [https://lynglang.com](https://lynglang.com) and a [telegram channel](https://t.me/lynglang) for updates.
- simple, compact, intuitive and elegant modern code:
```lyng
class Point(x, y) {
```
class Point(x,y) {
fun dist() { sqrt(x*x + y*y) }
}
// Auto-named arguments shorthand (x: is x: x):
val x = 3
val y = 4
Point(x:, y:).dist() //< 5
Point(3,4).dist() //< 5
fun swapEnds(first, args..., last, f) {
f( last, ...args, first)
}
class A {
class B(x?)
object Inner { val foo = "bar" }
enum E* { One, Two }
}
val ab = A.B()
assertEquals(null, ab.x)
assertEquals("bar", A.Inner.foo)
assertEquals(A.E.One, A.One)
```
- extremely simple Kotlin integration on any platform (JVM, JS, WasmJS, Lunux, MacOS, iOS, Windows)
- 100% secure: no access to any API you didn't explicitly provide
- 100% coroutines! Every function/script is a coroutine, it does not block the thread, no async/await/suspend keyword garbage, see [parallelism]. it is multithreaded on platforms supporting it (automatically, no code changes required, just `launch` more coroutines and they will be executed concurrently if possible). See [parallelism]
- functional style and OOP together: multiple inheritance (so you got it all - mixins, interfaces, etc.), delegation, sigletons, anonymous classes,extensions.
- nice literals for maps and arrays, destructuring assignment, ranges.
- 100% coroutines! Every function/script is a coroutine, it does not block the thread, no async/await/suspend keyword garbage, see [parallelism]
```
val deferred = launch {
delay(1.5) // coroutine is delayed for 1.5s, thread is not blocked!
"done"
}
// ...
// suspend current coroutine, no thread is blocked again,
// and wait for deferred to return something:
assertEquals("donw", deferred.await())
```
and it is multithreaded on platforms supporting it (automatically, no code changes required, just
`launch` more coroutines and they will be executed concurrently if possible). See [parallelism]
- functional style and OOP together, multiple inheritance, implementing interfaces for existing classes, writing extensions.
- Any Unicode letters can be used as identifiers: `assert( sin(π/2) == 1 )`.
## Resources:
- [Language home](https://lynglang.com)
- [introduction and tutorial](docs/tutorial.md) - start here please
- [What's New in 1.5](docs/whats_new_1_5.md)
- [Testing and Assertions](docs/Testing.md)
- [Filesystem and Processes (lyngio)](docs/lyngio.md)
- [Return Statement](docs/return_statement.md)
- [Efficient Iterables in Kotlin Interop](docs/EfficientIterables.md)
- [Samples directory](docs/samples)
- [Formatter (core + CLI + IDE)](docs/formatter.md)
- [Books directory](docs)
- [AI agent guidance](AGENTS.md)
## Integration in Kotlin multiplatform
@ -64,7 +49,7 @@ assertEquals(A.E.One, A.One)
```kotlin
// update to current please:
val lyngVersion = "1.5.0-SNAPSHOT"
val lyngVersion = "0.6.1-SNAPSHOT"
repositories {
// ...
@ -88,7 +73,7 @@ Now you can import lyng and use it:
### Execute script:
```kotlin
import net.sergeych.lyng.*
import net.sergeyh.lyng.*
// we need a coroutine to start, as Lyng
// is a coroutine based language, async topdown
@ -104,7 +89,9 @@ Script is executed over some `Scope`. Create instance,
add your specific vars and functions to it, and call:
```kotlin
import net.sergeych.lyng.*
import com.sun.source.tree.Scope
import new.sergeych.lyng.*
// simple function
val scope = Script.newScope().apply {
@ -152,12 +139,6 @@ Tips:
- After a dot, globals are intentionally suppressed (e.g., `lines().Path` is not valid), only the receiver’s members are suggested.
- If completion seems sparse, make sure related modules are imported (e.g., `import lyng.io.fs` so that `Path` and its methods are known).
## AI Assistant Support
To help AI assistants (like Cursor, Windsurf, or GitHub Copilot) understand Lyng with minimal effort, we provide a high-density language specification:
- **[LYNG_AI_SPEC.md](LYNG_AI_SPEC.md)**: A concise guide for AI models to learn Lyng syntax, idioms, and core philosophy. We recommend pointing your AI tool to this file or including it in your project's custom instructions.
## Why?
Designed to add scripting to kotlin multiplatform application in easy and efficient way. This is attempt to achieve what Lua is for C/++.
@ -177,7 +158,7 @@ Designed to add scripting to kotlin multiplatform application in easy and effici
# Language Roadmap
We are now at **v1.5.0-SNAPSHOT** (stable development cycle): basic optimization performed, battery included: standard library is 90% here, initial
We are now at **v1.0**: basic optimization performed, battery included: standard library is 90% here, initial
support in HTML, popular editors, and IDEA; tools to syntax highlight and format code are ready. It was released closed to schedule.
Ready features:
@ -189,7 +170,6 @@ Ready features:
- [x] ranges, lists, strings, interfaces: Iterable, Iterator, Collection, Array
- [x] when(value), if-then-else
- [x] exception handling: throw, try-catch-finally, exception classes.
- [x] user-defined exception classes
- [x] multiplatform maven publication
- [x] documentation for the current state
- [x] maps, sets and sequences (flows?)
@ -204,20 +184,8 @@ Ready features:
- [x] better stack reporting
- [x] regular exceptions + extended `when`
- [x] multiple inheritance for user classes
- [x] class properties (accessors)
- [x] `return` statement for local and non-local exit
- [x] Unified Delegation model: val, var and fun
- [x] `lazy val` using delegation
- [x] singletons `object TheOnly { ... }`
- [x] object expressions `object: List { ... }`
- [x] late-init vals in classes
- [x] properties with getters and setters
- [x] assign-if-null operator `?=`
- [x] user-defined exception classes
All of this is documented in the [language site](https://lynglang.com) and locally [docs/language.md](docs/tutorial.md). the current nightly builds published on the site and in the private maven repository.
## plan: towards v2.0 Next Generation
## plan: towards v1.5 Enhancing
- [x] site with integrated interpreter to give a try
- [x] kotlin part public API good docs, integration focused
@ -243,4 +211,4 @@ __Sergey Chernov__ @sergeych: Initial idea and architecture, language concept, d
__Yulia Nezhinskaya__ @AlterEgoJuliaN: System analysis, math and features design.
[parallelism]: docs/parallelism.md
[parallelism]: docs/parallelism.md

View File

@ -1,4 +0,0 @@
# Obsolete files
__Do not rely on contents of the files in this directory. They are kept for historical reference only and may not be up-to-date or relevant.__

View File

@ -1,117 +0,0 @@
/*
This is a tech proposal under construction, please do not use it yet
for any purpose
*/
/*
Abstract delegate can be used to proxy read/wrtie field access
or method call. Default implementation reports error.
*/
interface Delegate {
fun getValue() = Unset
fun setValue(newValue) { throw NotImplementedException("delegate setter is not implemented") }
fun invoke(args...) { throw NotImplementedException("delegate setter is not implemented") }
}
/*
Delegate cam be used to implement a val, var or fun, so there are
access type enum to distinguish:
*/
enum DelegateAccess {
Val,
Var,
Callable
}
// Delegate can be associated by a val/var/fun in a declaraion site using `by` keyword
val proxiedVal by proxy(1)
var proxiedVar by proxy(2, 3)
fun proxiedFun by proxy()
// each proxy is a Lyng expression returning instance of the Proxy interface:
/*
Proxy interface is connecting some named property of a given kind with the `Delegate`.
It removes the burden of dealing with property name and this ref on each get/set value
or invoke allowing having one delegate per instance, execution buff.
*/
interface Proxy {
fun getDelegate(propertyName: String,access: DelegateAccess,thisRef: Obj?): Delegate
}
// val, var and fun can be delegated, local or class instance:
class TestProxy: Proxy {
override getDelegate(name,access,thisRef) {
Delegate()
}
}
val proxy = TestProxy()
class Allowed {
val v1 by proxy
var v2 by proxy
fun f1 by proxy
}
val v3 by proxy
var v4 by proxy
fun f2 by proxy
/*
It means that for example
Allowed().f1("foo")
would call a delegate.invoke("foo") on the `Delegate` instance supplied by `proxy`, etc.
*/
// The practic sample: lazy value
/*
The delegate that caches single time evaluated value
*/
class LazyDelegate(creator): Delegate {
private var currentValue=Unset
override fun getValue() {
if( currentValue == Unset )
currentValue = creator()
currentValue
}
}
/*
The proxy to assign it
*/
class LazyProxy(creator) {
fun getDelegate(name,access,thisRef) {
if( access != DelegateAccess.Val )
throw IllegalArgumentException("only lazy val are allowed")
LazyDelegate(creator)
}
}
/*
A helper function to simplify creation:
*/
fun lazy(creator) {
LazyProxy(creator)
}
// Usage sample and the test:
var callCounter = 0
assertEquals(0, clallCounter)
val lazyText by lazy { "evaluated text" }
// the lazy property is not yet evaluated:
assertEquals(0, clallCounter)
// now evaluate it by using it:
assertEquals("evaluated text", lazyText)
assertEquals(1, callCounter)
// lazy delegate should fail on vars or funs:
assertThrows { var bad by lazy { "should not happen" } }
assertThrows { fun bad by lazy { 42 } }

View File

@ -1,40 +0,0 @@
#!/bin/bash
#
# Copyright 2026 Sergey S. Chernov real.sergeych@gmail.com
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
#
#
set -e
echo "publishing all artifacts"
echo
./gradlew publishToMavenLocal
./gradlew publish
echo
echo "Creating plugin"
echo
./gradlew buildInstallablePlugin
echo
echo "building CLI tools"
echo
bin/local_jrelease
bin/local_release
echo
echo "Deploying site"
echo
./bin/deploy_site

View File

@ -1,26 +0,0 @@
#!/bin/sh
#
# Copyright 2026 Sergey S. Chernov real.sergeych@gmail.com
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
#
#
set -e
echo
echo "Creating plugin"
echo
./gradlew buildInstallablePlugin
deploy_site -u

View File

@ -1,7 +1,7 @@
#!/bin/bash
#
# Copyright 2026 Sergey S. Chernov real.sergeych@gmail.com
# Copyright 2025 Sergey S. Chernov real.sergeych@gmail.com
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
@ -17,14 +17,6 @@
#
#
upload_only=false
for arg in "$@"; do
if [[ "$arg" == "-u" || "$arg" == "--upload-only" ]]; then
upload_only=true
break
fi
done
function checkState() {
if [[ $? != 0 ]]; then
echo
@ -32,10 +24,9 @@ function checkState() {
echo
exit 100
fi
}
# Update docs/idea_plugin.md to point to the latest built IDEA plugin zip
# from ./distributables before building the site. The change is temporary and
# the original file is restored right after the build.
@ -91,7 +82,7 @@ function updateIdeaPluginDownloadLink() {
# default target settings
case "com" in
com)
SSH_HOST=sergeych@d.lynglang.com # host to deploy to
SSH_HOST=sergeych@lynglang.com # host to deploy to
SSH_PORT=22 # ssh port on it
ROOT=/bigstore/sergeych_pub/lyng # directory to rsync to
;;
@ -107,37 +98,27 @@ esac
die() { echo "ERROR: $*" 1>&2 ; exit 1; }
function refreshTextmateZip() {
echo "Refreshing distributables/lyng-textmate.zip from editors/..."
mkdir -p distributables
# We use -r for recursive and -q for quiet (optional)
# -j can be used if we want to junk paths, but the request says "contents of editors/"
# usually we want to preserve the structure inside editors/
(cd editors && zip -rq ../distributables/lyng-textmate.zip .)
}
# Update the IDEA plugin download link in docs (temporarily), then build, then restore the doc
updateIdeaPluginDownloadLink || echo "WARN: proceeding without updating IDEA plugin download link"
./gradlew site:clean site:jsBrowserDistribution
BUILD_RC=$?
if [[ "$upload_only" == false ]]; then
# Update the IDEA plugin download link in docs (temporarily), then build, then restore the doc
refreshTextmateZip
updateIdeaPluginDownloadLink || echo "WARN: proceeding without updating IDEA plugin download link"
./gradlew site:jsBrowserDistribution
BUILD_RC=$?
# Always restore original doc if backup exists
if [[ -f "$DOC_IDEA_PLUGIN_BACKUP" ]]; then
mv -f "$DOC_IDEA_PLUGIN_BACKUP" "$DOC_IDEA_PLUGIN"
fi
if [[ $BUILD_RC -ne 0 ]]; then
echo
echo -- build failed. deploy aborted.
echo
exit 100
fi
# Always restore original doc if backup exists
if [[ -f "$DOC_IDEA_PLUGIN_BACKUP" ]]; then
mv -f "$DOC_IDEA_PLUGIN_BACKUP" "$DOC_IDEA_PLUGIN"
fi
if [[ $BUILD_RC -ne 0 ]]; then
echo
echo -- build failed. deploy aborted.
echo
exit 100
fi
#exit 0
# Prepare working dir
ssh -p ${SSH_PORT} ${SSH_HOST} "
cd ${ROOT}
@ -152,15 +133,12 @@ ssh -p ${SSH_PORT} ${SSH_HOST} "
fi
";
if [[ "$upload_only" == false ]]; then
# sync files
SRC=./site/build/dist/js/productionExecutable
rsync -e "ssh -p ${SSH_PORT}" -avz -r -d --delete ${SRC}/* ${SSH_HOST}:${ROOT}/build/dist
checkState
#rsync -e "ssh -p ${SSH_PORT}" -avz ./static/* ${SSH_HOST}:${ROOT}/build/dist
#checkState
fi
# sync files
SRC=./site/build/dist/js/productionExecutable
rsync -e "ssh -p ${SSH_PORT}" -avz -r -d --delete ${SRC}/* ${SSH_HOST}:${ROOT}/build/dist
checkState
#rsync -e "ssh -p ${SSH_PORT}" -avz ./static/* ${SSH_HOST}:${ROOT}/build/dist
#checkState
rsync -e "ssh -p ${SSH_PORT}" -avz -r -d --delete distributables/* ${SSH_HOST}:${ROOT}/build/dist/distributables
checkState

View File

@ -1,256 +0,0 @@
#!/bin/bash
#
# Copyright 2026 Sergey S. Chernov real.sergeych@gmail.com
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
#
#
set -e
# Configuration
DOCS_DIR="docs"
OUTPUT_DIR="distributables"
TEMP_DIR="build/temp_docs"
MERGED_MD="$TEMP_DIR/merged.md"
OUTPUT_HTML="$OUTPUT_DIR/lyng_documentation.html"
mkdir -p "$OUTPUT_DIR"
mkdir -p "$TEMP_DIR"
# Files that should come first in specific order
PRIORITY_FILES=(
"tutorial.md"
"OOP.md"
"advanced_topics.md"
"declaring_arguments.md"
"scopes_and_closures.md"
"exceptions_handling.md"
"when.md"
"parallelism.md"
"Testing.md"
)
# Files that should come next (reference)
REFERENCE_FILES=(
"Collection.md"
"Iterable.md"
"Iterator.md"
"List.md"
"Set.md"
"Map.md"
"Array.md"
"Buffer.md"
"RingBuffer.md"
"Range.md"
"Real.md"
"Regex.md"
"math.md"
"time.md"
)
# Files that are about integration/tools
INTEGRATION_FILES=(
"serialization.md"
"json_and_kotlin_serialization.md"
"embedding.md"
"lyng_cli.md"
"lyng.io.fs.md"
"formatter.md"
"EfficientIterables.md"
)
# Tracking processed files to avoid duplicates
PROCESSED_PATHS=()
is_excluded() {
local full_path="$1"
if grep -q "excludeFromIndex" "$full_path"; then
return 0 # true in bash
fi
return 1 # false
}
process_file() {
local rel_path="$1"
local full_path="$DOCS_DIR/$rel_path"
if [[ ! -f "$full_path" ]]; then
return
fi
if is_excluded "$full_path"; then
echo "Skipping excluded: $rel_path"
return
fi
# Check for duplicates
for p in "${PROCESSED_PATHS[@]}"; do
if [[ "$p" == "$rel_path" ]]; then
return
fi
done
PROCESSED_PATHS+=("$rel_path")
echo "Processing: $rel_path"
# 1. Add an anchor for the file based on its path
local anchor_name=$(echo "$rel_path" | sed 's/\//_/g')
echo "<div id=\"$anchor_name\"></div>" >> "$MERGED_MD"
echo "" >> "$MERGED_MD"
# 2. Append content with fixed links
# - [text](file.md) -> [text](#file.md)
# - [text](dir/file.md) -> [text](#dir_file.md)
# - [text](file.md#anchor) -> [text](#anchor)
# - Fix image links: [alt](../images/...) -> [alt](images/...) if needed, but none found yet.
cat "$full_path" | \
perl -pe 's/\[([^\]]+)\]\(([^)]+)\.md\)/"[$1](#" . ($2 =~ s|\/|_|gr) . ".md)"/ge' | \
perl -pe 's/\[([^\]]+)\]\(([^)]+)\.md#([^)]+)\)/[$1](#$3)/g' >> "$MERGED_MD"
echo -e "\n\n---\n\n" >> "$MERGED_MD"
}
# Start with an empty merged file
echo "% Lyng Language Documentation" > "$MERGED_MD"
echo "" >> "$MERGED_MD"
# 1. Process priority files
for f in "${PRIORITY_FILES[@]}"; do
process_file "$f"
done
# 2. Process reference files
for f in "${REFERENCE_FILES[@]}"; do
process_file "$f"
done
# 3. Process integration files
for f in "${INTEGRATION_FILES[@]}"; do
process_file "$f"
done
# 4. Process remaining files in docs root
for f in "$DOCS_DIR"/*.md; do
rel_f=${f#"$DOCS_DIR/"}
process_file "$rel_f"
done
# 5. Process remaining files in subdirs (like samples)
find "$DOCS_DIR" -name "*.md" | sort | while read -r f; do
rel_f=${f#"$DOCS_DIR/"}
process_file "$rel_f"
done
echo "Running pandoc to generate $OUTPUT_HTML..."
# Use a basic but clean CSS
pandoc "$MERGED_MD" -o "$OUTPUT_HTML" \
--toc --toc-depth=2 \
--standalone \
--embed-resources \
--metadata title="Lyng Language Documentation" \
--css <(echo "
body {
font-family: -apple-system, BlinkMacSystemFont, \"Segoe UI\", Helvetica, Arial, sans-serif;
line-height: 1.6;
max-width: 1000px;
margin: 0 auto;
padding: 2em;
color: #24292e;
background-color: #fff;
}
code {
background-color: rgba(27,31,35,0.05);
padding: 0.2em 0.4em;
border-radius: 3px;
font-family: SFMono-Regular, Consolas, \"Liberation Mono\", Menlo, monospace;
font-size: 85%;
}
pre {
background-color: #f6f8fa;
padding: 16px;
overflow: auto;
border-radius: 3px;
line-height: 1.45;
}
pre code {
background-color: transparent;
padding: 0;
font-size: 100%;
}
h1 {
border-bottom: 1px solid #eaecef;
padding-bottom: 0.3em;
margin-top: 24px;
margin-bottom: 16px;
font-weight: 600;
}
h2 {
border-bottom: 1px solid #eaecef;
padding-bottom: 0.3em;
margin-top: 24px;
margin-bottom: 16px;
font-weight: 600;
}
hr {
height: 0.25em;
padding: 0;
margin: 24px 0;
background-color: #e1e4e8;
border: 0;
}
blockquote {
padding: 0 1em;
color: #6a737d;
border-left: 0.25em solid #dfe2e5;
margin: 0 0 16px 0;
}
nav#TOC {
background: #f9f9f9;
padding: 1em;
border: 1px solid #eee;
margin-bottom: 2.5em;
border-radius: 6px;
}
nav#TOC ul {
list-style: none;
padding-left: 1.5em;
}
nav#TOC > ul {
padding-left: 0;
}
table {
border-spacing: 0;
border-collapse: collapse;
margin-top: 0;
margin-bottom: 16px;
}
table th, table td {
padding: 6px 13px;
border: 1px solid #dfe2e5;
}
table tr {
background-color: #fff;
border-top: 1px solid #c6cbd1;
}
table tr:nth-child(2n) {
background-color: #f6f8fa;
}
")
echo "-------------------------------------------------------"
echo "Done! Documentation generated successfully."
echo "Location: $OUTPUT_HTML"
echo "-------------------------------------------------------"

View File

@ -1,5 +1,5 @@
/*
* Copyright 2026 Sergey S. Chernov real.sergeych@gmail.com
* Copyright 2025 Sergey S. Chernov real.sergeych@gmail.com
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
@ -29,33 +29,3 @@ tasks.register<org.gradle.api.DefaultTask>("runIde") {
description = "Run IntelliJ IDEA with the Lyng plugin (:lyng-idea)"
dependsOn(":lyng-idea:runIde")
}
tasks.register<Exec>("generateDocs") {
group = "documentation"
description = "Generates a single-file documentation HTML using bin/generate_docs.sh"
commandLine("./bin/generate_docs.sh")
}
// Sample generator task for .lyng.d definition files (not wired into build).
// Usage: ./gradlew generateLyngDefsSample
tasks.register("generateLyngDefsSample") {
group = "lyng"
description = "Generate a sample .lyng.d file under build/generated/lyng/defs"
outputs.dir(layout.buildDirectory.dir("generated/lyng/defs"))
doLast {
val outDir = layout.buildDirectory.dir("generated/lyng/defs").get().asFile
outDir.mkdirs()
val outFile = outDir.resolve("sample.lyng.d")
outFile.writeText(
"""
/** Generated API */
extern fun ping(): Int
/** Generated class */
class Generated(val name: String) {
fun greet(): String = "hi " + name
}
""".trimIndent()
)
}
}

View File

@ -1,7 +0,0 @@
# Bytecode Migration Plan (Archived)
Status: completed.
Historical reference:
- `notes/archive/bytecode_migration_plan.md` (full plan)
- `notes/archive/bytecode_migration_plan_completed.md` (summary)

View File

@ -23,11 +23,11 @@ There are a lo of ways to construct a buffer:
assertEquals( 5, Buffer("hello").size )
// from bytes, e.g. integers in range 0..255
assertEquals( 255, Buffer(1,2,3,255).last )
assertEquals( 255, Buffer(1,2,3,255).last() )
// from whatever iterable that produces bytes, e.g.
// integers in 0..255 range:
assertEquals( 129, Buffer([1,2,129]).last )
assertEquals( 129, Buffer([1,2,129]).last() )
// Empty buffer of fixed size:
assertEquals(100, Buffer(100).size)

View File

@ -1,280 +0,0 @@
# Lyng Bytecode VM Spec v0 (Draft)
This document describes a register-like (3-address) bytecode for Lyng with
dynamic slot width (8/16/32-bit slot IDs), a slot-tail argument model, and
typed lanes for Obj/Int/Real/Bool. The VM is intended to run as a suspendable
interpreter and fall back to the existing AST execution when needed.
## 1) Frame & Slot Model
### Frame metadata
- localCount: number of local slots for this function (fixed at compile time).
- argCount: number of arguments passed at call time.
- scopeSlotNames: optional debug names for scope slots (locals/params), aligned to slot mapping.
- argBase = localCount.
### Slot layout
slots[0 .. localCount-1] locals
slots[localCount .. localCount+argCount-1] arguments
### Typed lanes
- slotType[]: UNKNOWN/OBJ/INT/REAL/BOOL
- objSlots[], intSlots[], realSlots[], boolSlots[]
- A slot is a logical index; active lane is selected by slotType.
### Parameter access
- param i => slot localCount + i
- variadic extra => slot localCount + declaredParamCount + k
### Debug metadata (optional)
- scopeSlotNames: array sized scopeSlotCount, each entry nullable.
- Intended for disassembly/debug tooling; VM semantics do not depend on it.
### Constant pool extras
- SlotPlan: map of name -> slot index, used by PUSH_SCOPE to pre-allocate and map loop locals.
- CallArgsPlan: ordered argument specs (name/splat) + tailBlock flag, used when argCount has the plan flag set.
## 2) Slot ID Width
Per frame, select:
- 8-bit if localCount + argCount < 256
- 16-bit if < 65536
- 32-bit otherwise
The decoder uses a dedicated loop per width. All slot operands are expanded to
Int internally.
## 3) CALL Semantics (Model A)
Instruction:
CALL_DIRECT fnId, argBase, argCount, dst
Behavior:
- Allocate a callee frame sized localCount + argCount.
- Copy caller slots [argBase .. argBase+argCount-1] into callee slots
[localCount .. localCount+argCount-1].
- Callee returns via RET slot or RET_VOID.
- Caller stores return value to dst.
Other calls:
- CALL_VIRTUAL recvSlot, methodId, argBase, argCount, dst
- CALL_FALLBACK stmtId, argBase, argCount, dst
- CALL_SLOT calleeSlot, argBase, argCount, dst
## 4) Binary Encoding Layout
All instructions are:
[opcode:U8] [operands...]
Operand widths:
- slotId: S = 1/2/4 bytes (per frame slot width)
- constId: K = 2 bytes (U16), extend to 4 if needed
- ip: I = 2 bytes (U16) or 4 bytes (U32) per function size
- fnId/methodId/stmtId: F/M/T = 2 bytes (U16) unless extended
- argCount: C = 2 bytes (U16), extend to 4 if needed
Endianness: little-endian for multi-byte operands.
Common operand patterns:
- S: one slot
- SS: two slots
- SSS: three slots
- K S: constId + dst slot
- S I: slot + jump target
- I: jump target
- F S C S: fnId, argBase slot, argCount, dst slot
Arg count flag:
- If high bit of C is set (0x8000), the low 15 bits encode a CallArgsPlan constId.
- When not set, C is the raw positional count and tailBlockMode=false.
## 5) Opcode Table
Note: Any opcode can be compiled to FALLBACK if not implemented in a VM pass.
### Data movement
- NOP
- MOVE_OBJ S -> S
- MOVE_INT S -> S
- MOVE_REAL S -> S
- MOVE_BOOL S -> S
- BOX_OBJ S -> S
- CONST_OBJ K -> S
- CONST_INT K -> S
- CONST_REAL K -> S
- CONST_BOOL K -> S
- CONST_NULL -> S
### Numeric conversions
- INT_TO_REAL S -> S
- REAL_TO_INT S -> S
- BOOL_TO_INT S -> S
- INT_TO_BOOL S -> S
### Arithmetic: INT
- ADD_INT S, S -> S
- SUB_INT S, S -> S
- MUL_INT S, S -> S
- DIV_INT S, S -> S
- MOD_INT S, S -> S
- NEG_INT S -> S
- INC_INT S
- DEC_INT S
### Arithmetic: REAL
- ADD_REAL S, S -> S
- SUB_REAL S, S -> S
- MUL_REAL S, S -> S
- DIV_REAL S, S -> S
- NEG_REAL S -> S
### Arithmetic: OBJ
- ADD_OBJ S, S -> S
- SUB_OBJ S, S -> S
- MUL_OBJ S, S -> S
- DIV_OBJ S, S -> S
- MOD_OBJ S, S -> S
### Bitwise: INT
- AND_INT S, S -> S
- OR_INT S, S -> S
- XOR_INT S, S -> S
- SHL_INT S, S -> S
- SHR_INT S, S -> S
- USHR_INT S, S -> S
- INV_INT S -> S
### Comparisons (typed)
- CMP_EQ_INT S, S -> S
- CMP_NEQ_INT S, S -> S
- CMP_LT_INT S, S -> S
- CMP_LTE_INT S, S -> S
- CMP_GT_INT S, S -> S
- CMP_GTE_INT S, S -> S
- CMP_EQ_REAL S, S -> S
- CMP_NEQ_REAL S, S -> S
- CMP_LT_REAL S, S -> S
- CMP_LTE_REAL S, S -> S
- CMP_GT_REAL S, S -> S
- CMP_GTE_REAL S, S -> S
- CMP_EQ_BOOL S, S -> S
- CMP_NEQ_BOOL S, S -> S
### Mixed numeric comparisons
- CMP_EQ_INT_REAL S, S -> S
- CMP_EQ_REAL_INT S, S -> S
- CMP_LT_INT_REAL S, S -> S
- CMP_LT_REAL_INT S, S -> S
- CMP_LTE_INT_REAL S, S -> S
- CMP_LTE_REAL_INT S, S -> S
- CMP_GT_INT_REAL S, S -> S
- CMP_GT_REAL_INT S, S -> S
- CMP_GTE_INT_REAL S, S -> S
- CMP_GTE_REAL_INT S, S -> S
- CMP_NEQ_INT_REAL S, S -> S
- CMP_NEQ_REAL_INT S, S -> S
- CMP_EQ_OBJ S, S -> S
- CMP_NEQ_OBJ S, S -> S
- CMP_REF_EQ_OBJ S, S -> S
- CMP_REF_NEQ_OBJ S, S -> S
- CMP_LT_OBJ S, S -> S
- CMP_LTE_OBJ S, S -> S
- CMP_GT_OBJ S, S -> S
- CMP_GTE_OBJ S, S -> S
### Boolean ops
- NOT_BOOL S -> S
- AND_BOOL S, S -> S
- OR_BOOL S, S -> S
### Control flow
- JMP I
- JMP_IF_TRUE S, I
- JMP_IF_FALSE S, I
- RET S
- RET_VOID
- PUSH_SCOPE K
- POP_SCOPE
### Scope setup
- PUSH_SCOPE uses const `SlotPlan` (name -> slot index) to create a child scope and apply slot mapping.
- POP_SCOPE restores the parent scope.
### Calls
- CALL_DIRECT F, S, C, S
- CALL_VIRTUAL S, M, S, C, S
- CALL_FALLBACK T, S, C, S
- CALL_SLOT S, S, C, S
### Object access (optional, later)
- GET_FIELD S, M -> S
- SET_FIELD S, M, S
- GET_INDEX S, S -> S
- SET_INDEX S, S, S
### Fallback
- EVAL_FALLBACK T -> S
## 6) Const Pool Encoding (v0)
Each const entry is encoded as:
[tag:U8] [payload...]
Tags:
- 0x00: NULL
- 0x01: BOOL (payload: U8 0/1)
- 0x02: INT (payload: S64, little-endian)
- 0x03: REAL (payload: F64, IEEE-754, little-endian)
- 0x04: STRING (payload: U32 length + UTF-8 bytes)
- 0x05: OBJ_REF (payload: U32 index into external Obj table)
Notes:
- OBJ_REF is reserved for embedding prebuilt Obj handles if needed.
- Strings use UTF-8; length is bytes, not chars.
## 7) Function Header (binary container)
Suggested layout for a bytecode function blob:
- magic: U32 ("LYBC")
- version: U16 (0x0001)
- slotWidth: U8 (1,2,4)
- ipWidth: U8 (2,4)
- constIdWidth: U8 (2,4)
- localCount: U32
- codeSize: U32 (bytes)
- constCount: U32
- constPool: [const entries...]
- code: [bytecode...]
Const pool entries use the encoding described in section 6.
## 8) Sample Bytecode (illustrative)
Example Lyng:
val x = 2
val y = 3
val z = x + y
Assume:
- localCount = 3 (x,y,z)
- argCount = 0
- slot width = 1 byte
- const pool: [INT 2, INT 3]
Bytecode:
CONST_INT k0 -> s0
CONST_INT k1 -> s1
ADD_INT s0, s1 -> s2
RET_VOID
Encoded (opcode values symbolic):
[OP_CONST_INT][k0][s0]
[OP_CONST_INT][k1][s1]
[OP_ADD_INT][s0][s1][s2]
[OP_RET_VOID]
## 9) Notes
- Mixed-mode is allowed: compiler can emit FALLBACK ops for unsupported nodes.
- The VM must be suspendable; on suspension, store ip + minimal operand state.
- Source mapping uses a separate ip->Pos table, not part of core bytecode.

View File

@ -12,7 +12,7 @@ Is a [Iterable] with known `size`, a finite [Iterable]:
(1)
: `comparator(a,b)` should return -1 if `a < b`, +1 if `a > b` or zero.
See [List], [Set], [Iterable] and [Efficient Iterables in Kotlin Interop](EfficientIterables.md)
See [List], [Set] and [Iterable]
[Iterable]: Iterable.md
[List]: List.md

View File

@ -1,92 +0,0 @@
# Efficient Iterables in Kotlin Interop
Lyng provides high-performance iteration mechanisms that allow Kotlin-side code to interact with Lyng iterables efficiently and vice versa.
## 1. Enumerating Lyng Objects from Kotlin
To iterate over a Lyng object (like a `List`, `Set`, or `Range`) from Kotlin code, use the virtual `enumerate` method:
```kotlin
val lyngList: Obj = ...
lyngList.enumerate(scope) { item ->
println("Processing $item")
true // return true to continue, false to break
}
```
### Why it's efficient:
- **Zero allocation**: Unlike traditional iterators, it doesn't create a `LyngIterator` object or any intermediate wrappers.
- **Direct access**: Subclasses like `ObjList` override `enumerate` to iterate directly over their internal Kotlin collections.
- **Reduced overhead**: It avoids multiple `invokeInstanceMethod` calls for `hasNext()` and `next()` on every step, which would normally involve dynamic dispatch and scope overhead.
## 2. Reactive Enumeration with Flow
If you prefer a reactive approach or need to integrate with Kotlin Coroutines flows, use `toFlow()`:
```kotlin
lyngList.toFlow(scope).collect { item ->
// ...
}
```
*Note: `toFlow()` internally uses the Lyng iterator protocol (`iterator()`, `hasNext()`, `next()`), so it's slightly less efficient than `enumerate()` for performance-critical loops, but more idiomatic for flow-based processing.*
## 3. Creating Efficient Iterables for Lyng in Kotlin
When implementing a custom object in Kotlin that should be iterable in Lyng (e.g., usable in `for (x in myObj) { ... }`), follow these steps to ensure maximum performance.
### A. Inherit from `Obj` and use `ObjIterable`
Ensure your object's class has `ObjIterable` as a parent so the Lyng compiler recognizes it as an iterable.
```kotlin
class MyCollection(val items: List<Obj>) : Obj() {
override val objClass = MyCollection.type
companion object {
val type = ObjClass("MyCollection", ObjIterable).apply {
// Provide a Lyng-side iterator for compatibility with
// manual iterator usage in Lyng scripts.
// Using ObjKotlinObjIterator if items are already Obj instances:
addFn("iterator") {
ObjKotlinObjIterator(thisAs<MyCollection>().items.iterator())
}
}
}
}
```
### B. Override `enumerate` for Maximum Performance
The Lyng compiler's `for` loops use the `enumerate` method. By overriding it in your Kotlin class, you provide a "fast path" for iteration.
```kotlin
class MyCollection(val items: List<Obj>) : Obj() {
// ...
override suspend fun enumerate(scope: Scope, callback: suspend (Obj) -> Boolean) {
for (item in items) {
// If callback returns false, it means 'break' was called in Lyng
if (!callback(item)) break
}
}
}
```
### C. Use `ObjInt.of()` for Numeric Data
If your iterable contains integers, always use `ObjInt.of(Long)` instead of the `ObjInt(Long)` constructor. Lyng maintains a cache for small integers (-128 to 127), which significantly reduces object allocations and GC pressure during tight loops.
```kotlin
// Efficiently creating an integer object
val obj = ObjInt.of(42L)
// Or using extension methods which also use the cache:
val obj2 = 42.toObj()
val obj3 = 42L.toObj()
```
#### Note on `toObj()` extensions:
While `<reified T> T.toObj()` is convenient, using specific extensions like `Int.toObj()` or `Long.toObj()` is slightly more efficient as they use the `ObjInt` cache.
## 4. Summary of Best Practices
- **To Consume**: Use `enumerate(scope) { item -> ... true }`.
- **To Implement**: Override `enumerate` in your `Obj` subclass.
- **To Register**: Use `ObjIterable` (or `ObjCollection`) as a parent class in your `ObjClass` definition.
- **To Optimize**: Use `ObjInt.of()` (or `.toObj()`) for all integer object allocations.

View File

@ -40,13 +40,13 @@ available, for example
## joinToString
This methods convert any iterable to a string joining string representation of each element, optionally transforming it
and joining using specified separator.
and joining using specified suffix.
Iterable.joinToString(separator=' ', transformer=null)
Iterable.joinToString(suffux=' ', transform=null)
- if `Iterable` `isEmpty`, the empty string `""` is returned.
- `separator` is inserted between items when there are more than one.
- `transformer` of specified is applied to each element, otherwise its `toString()` method is used.
- `suffix` is inserted between items when there are more than one.
- `transform` of specified is applied to each element, otherwise its `toString()` method is used.
Here is the sample:
@ -55,7 +55,7 @@ Here is the sample:
assertEquals( (1..3).joinToString { it * 10 }, "10 20 30")
>>> void
## `sum` and `sumOf`
## `sum` and `sumBy`
These, again, does the thing:
@ -68,79 +68,19 @@ These, again, does the thing:
>>> void
## map, filter and their variations
## map and mapNotNull
Used to transform or filter the whole iterable stream:
Used to transform either the whole iterable stream or also skipping som elements from it:
val source = [1,2,3,4]
// transform every element to string or null:
assertEquals(["n1", "n2", null, "n4"], source.map { if( it == 3 ) null else "n"+it } )
// map: transform every element to something else
assertEquals(["n1", "n2", "n3", "n4"], source.map { "n"+it } )
// filter: keep only elements matching the predicate
assertEquals([2, 4], source.filter { it % 2 == 0 } )
// count: count elements matching the predicate
assertEquals(2, source.count { it % 2 == 0 } )
// mapNotNull: transform every element, skipping null results:
// transform every element to stirng, skipping 3:
assertEquals(["n1", "n2", "n4"], source.mapNotNull { if( it == 3 ) null else "n"+it } )
// filterNotNull: skip all null elements:
assertEquals([1, 2, 4], [1, 2, null, 4].filterNotNull())
>>> void
You can also use flow variations that return a cold `Flow` instead of a `List`, which is useful for large or infinite sequences:
val source = [1, 2, 3, 4]
// filterFlow: returns a Flow of filtered elements
assert( source.filterFlow { it % 2 == 0 } is Flow )
// filterFlowNotNull: returns a Flow of non-null elements
assert( [1, null, 2].filterFlowNotNull() is Flow )
>>> void
## minOf and maxOf
Find the minimum or maximum value of a function applied to each element:
val source = ["abc", "de", "fghi"]
assertEquals(2, source.minOf { (it as String).length })
assertEquals(4, source.maxOf { (it as String).length })
>>> void
## flatten and flatMap
Work with nested collections:
val nested = [[1, 2], [3, 4]]
// flatten: combine nested collections into one list
assertEquals([1, 2, 3, 4], nested.flatten())
// flatMap: map each element to a collection and flatten the result
assertEquals([1, 10, 2, 20], [1, 2].flatMap { [it, it*10] })
>>> void
## findFirst and findFirstOrNull
Search for the first element that satisfies the given predicate:
val source = [1, 2, 3, 4]
assertEquals( 2, source.findFirst { it % 2 == 0 } )
assertEquals( 2, source.findFirstOrNull { it % 2 == 0 } )
// findFirst throws if not found:
assertThrows( NoSuchElementException ) { source.findFirst { it > 10 } }
// findFirstOrNull returns null if not found:
assertEquals( null, source.findFirstOrNull { it > 10 } )
>>> void
## Instance methods:
@ -149,56 +89,52 @@ Search for the first element that satisfies the given predicate:
| toList() | create a list from iterable |
| toSet() | create a set from iterable |
| contains(i) | check that iterable contains `i` |
| `i in iterable` | same as `contains(i)` |
| `i in iterator` | same as `contains(i)` |
| isEmpty() | check iterable is empty |
| forEach(f) | call f for each element |
| toMap() | create a map from list of key-value pairs (arrays of 2 items or like) |
| any(p) | true if any element matches predicate `p` |
| all(p) | true if all elements match predicate `p` |
| map(f) | create a list of values returned by `f` called for each element of the iterable |
| indexOf(i) | return index if the first encounter of i or a negative value if not found |
| associateBy(kf) | create a map where keys are returned by kf that will be called for each element |
| filter(p) | create a list of elements matching predicate `p` |
| count(p) | count elements matching predicate `p` |
| filterFlow(p) | create a [Flow] of elements matching predicate `p` |
| filterNotNull() | create a list of non-null elements |
| filterFlowNotNull() | create a [Flow] of non-null elements |
| minOf(f) | return minimum value of `f` applied to elements |
| maxOf(f) | return maximum value of `f` applied to elements |
| flatten() | flatten nested collections into a single [List] |
| flatMap(f) | map each element with `f` and flatten results into a [List] |
| findFirst(p) | return first element matching predicate `p` or throw (1) |
| findFirstOrNull(p) | return first element matching predicate `p` or `null` |
| first | first element (1) |
| last | last element (1) |
| take(n) | return [Iterable] of up to n first elements |
| takeLast(n) | return [Iterable] of up to n last elements |
| taleLast(n) | return [Iterable] of up to n last elements |
| drop(n) | return new [Iterable] without first n elements |
| dropLast(n) | return new [Iterable] without last n elements |
| sum() | return sum of the collection applying `+` to its elements (3) |
| sumOf(f) | sum of the modified collection items (3) |
| sumOf(predicate) | sum of the modified collection items (3) |
| sorted() | return [List] with collection items sorted naturally |
| sortedWith(comparator) | sort using a comparator that compares elements (1) |
| sortedBy(predicate) | sort by comparing results of the predicate function |
| joinToString(s,t) | convert iterable to string, see (2) |
| reversed() | create a list containing items from this in reverse order |
| shuffled() | create a list of shuffled elements |
| shuffled() | create a listof shiffled elements |
(1)
:: throws `NoSuchElementException` if there is no such element
: throws `NoSuchElementException` if there is no such element
(2)
:: `joinToString(separator=" ", transformer=null)`: separator is inserted between items if there are more than one, transformer is
: `joinToString(suffix=" ",transform=null)`: suffix is inserted between items if there are more than one, trasnfom is
optional function applied to each item that must return result string for an item, otherwise `item.toString()` is used.
(3)
:: sum of empty collection is `null`
: sum of empty collection is `null`
fun Iterable.toList(): List
fun Iterable.toSet(): Set
fun Iterable.indexOf(element): Int
fun Iterable.contains(element): Bool
fun Iterable.isEmpty(element): Bool
fun Iterable.forEach(block: (Any?)->Void ): Void
fun Iterable.map(block: (Any?)->Void ): List
fun Iterable.associateBy( keyMaker: (Any?)->Any): Map
## Abstract methods:
fun iterator(): Iterator
For high-performance Kotlin-side interop and custom iterable implementation details, see [Efficient Iterables in Kotlin Interop](EfficientIterables.md).
Creates a list by iterating to the end. So, the Iterator should be finite to be used with it.
## Included in interfaces:
@ -212,10 +148,8 @@ For high-performance Kotlin-side interop and custom iterable implementation deta
[List]: List.md
[Flow]: parallelism.md#flow
[Range]: Range.md
[Set]: Set.md
[RingBuffer]: RingBuffer.md
[RingBuffer]: RingBuffer.md

View File

@ -23,8 +23,6 @@ must throw `ObjIterationFinishedError`.
Iterators are returned when implementing [Iterable] interface.
For high-performance Kotlin-side interop and custom iterable implementation details, see [Efficient Iterables in Kotlin Interop](EfficientIterables.md).
## Implemented for classes:
- [List], [Range]

View File

@ -92,34 +92,6 @@ Open end ranges remove head and tail elements:
assert( [1, 2, 3] !== [1, 2, 3])
>>> void
## Destructuring
Lists can be used as L-values for destructuring assignments. This allows you to unpack list elements into multiple variables.
### Basic Destructuring
```lyng
val [a, b, c] = [1, 2, 3]
```
### With Splats (Variadic)
A single ellipsis `...` can be used to capture remaining elements into a list. It can be placed at the beginning, middle, or end of the pattern.
```lyng
val [head, rest...] = [1, 2, 3] // head=1, rest=[2, 3]
val [first, middle..., last] = [1, 2, 3, 4, 5] // first=1, middle=[2, 3, 4], last=5
```
### Nested Patterns
Destructuring patterns can be nested to unpack multi-dimensional lists.
```lyng
val [a, [b, c...], d] = [1, [2, 3, 4], 5]
```
### Reassignment
Destructuring can also be used to reassign existing variables:
```lyng
[x, y] = [y, x] // Swap values
```
## In-place sort
List could be sorted in place, just like [Collection] provide sorted copies, in a very like way:
@ -158,8 +130,7 @@ List could be sorted in place, just like [Collection] provide sorted copies, in
| `sort()` | in-place sort, natural order | void |
| `sortBy(predicate)` | in-place sort bu `predicate` call result (3) | void |
| `sortWith(comparator)` | in-place sort using `comarator` function (4) | void |
| `shuffle()` | in-place shuffle contents | |
| `toString()` | string representation like `[a,b,c]` | |
| `shiffle()` | in-place shiffle contents | |
(1)
: optimized implementation that override `Array` one

View File

@ -94,8 +94,7 @@ Or iterate its key-value pairs that are instances of [MapEntry] class:
val map = Map( ["foo", 1], ["bar", "buzz"], [42, "answer"] )
for( entry in map ) {
val e: MapEntry = entry as MapEntry
println("map[%s] = %s"(e.key, e.value))
println("map[%s] = %s"(entry.key, entry.value))
}
void
>>> map[foo] = 1
@ -173,7 +172,7 @@ Maps and entries can also be merged with `+` and `+=`:
Notes:
- Map literals always use string keys (identifier keys are converted to strings).
- Spreads inside map literals and `+`/`+=` merges allow any objects as keys.
- When you need computed or non-string keys, use the constructor form `Map(...)`, map literals with computed keys (if supported), or build entries with `=>` and then merge.
- Spreads inside map literals and `+`/`+=` merges require string keys on the right-hand side; this aligns with named-argument splats.
- When you need computed or non-string keys, use the constructor form `Map(...)` or build entries with `=>` and then merge.
[Collection](Collection.md)
[Collection](Collection.md)

View File

@ -9,7 +9,7 @@ Lyng supports first class OOP constructs, based on classes with multiple inherit
The class clause looks like
class Point(x,y)
assertEquals("Point", Point.className)
assert( Point is Class )
>>> void
It creates new `Class` with two fields. Here is the more practical sample:
@ -42,271 +42,6 @@ a _constructor_ that requires two parameters for fields. So when creating it wit
Form now on `Point` is a class, it's type is `Class`, and we can create instances with it as in the
example above.
## Singleton Objects
Singleton objects are declared using the `object` keyword. An `object` declaration defines both a class and a single instance of that class at the same time. This is perfect for stateless utilities, global configuration, or shared delegates.
```lyng
object Config {
val version = "1.0.0"
val debug = true
fun printInfo() {
println("App version: " + version)
}
}
// Usage:
println(Config.version)
Config.printInfo()
```
Objects can also inherit from classes or interfaces:
```lyng
object DefaultLogger : Logger("Default") {
override fun log(msg) {
println("[DEFAULT] " + msg)
}
}
```
## Object Expressions
Object expressions allow you to create an instance of an anonymous class. This is useful when you need to provide a one-off implementation of an interface or inherit from a class without declaring a new named subclass.
```lyng
val worker = object : Runnable {
override fun run() {
println("Working...")
}
}
```
Object expressions can implement multiple interfaces and inherit from one class:
```lyng
val x = object : Base(arg1), Interface1, Interface2 {
val property = 42
override fun method() = property * 2
}
```
### Scoping and `this@object`
Object expressions capture their lexical scope, meaning they can access local variables and members of the outer class. When `this` is rebound (for example, inside an `apply` block), you can use the reserved alias `this@object` to refer to the innermost anonymous object instance.
```lyng
val handler = object {
fun process() {
this@object.apply {
// here 'this' is rebound to the map/context
// but we can still access the anonymous object via this@object
println("Processing in " + this@object)
}
}
}
```
### Serialization and Identity
- **Serialization**: Anonymous objects are **not serializable**. Attempting to encode an anonymous object via `Lynon` will throw a `SerializationException`. This is because their class definition is transient and cannot be safely restored in a different session or process.
- **Type Identity**: Every object expression creates a unique anonymous class. Two identical object expressions will result in two different classes with distinct type identities.
## Nested Declarations
Lyng allows classes, objects, enums, and type aliases to be declared inside another class. These declarations live in the **class namespace** (not the instance), so they do not capture an outer instance and are accessed with a qualifier.
```lyng
class A {
class B(x?)
object Inner { val foo = "bar" }
type Alias = B
enum E { One, Two }
}
val ab = A.B()
assertEquals(ab.x, null)
assertEquals(A.Inner.foo, "bar")
```
Rules:
- **Qualified access**: use `Outer.Inner` for nested classes/objects/enums/aliases. Inside `Outer` you can refer to them by unqualified name unless shadowed.
- **No inner semantics**: nested declarations do not capture an instance of the outer class. They are resolved at compile time.
- **Visibility**: `private` restricts a nested declaration to the declaring class body (not visible from outside or subclasses).
- **Reflection name**: a nested class reports `Outer.Inner` (e.g., `A.B::class.name` is `"A.B"`).
- **Type aliases**: behave as aliases of the qualified nested type and are expanded by the type system.
### Lifted Enum Entries
Enums can optionally lift their entries into the surrounding class namespace using `*`:
```lyng
class A {
enum E* { One, Two }
}
assertEquals(A.One, A.E.One)
assertEquals(A.Two, A.E.Two)
```
Notes:
- `E*` exposes entries in `A` as if they were direct members (`A.One`).
- If a name would conflict with an existing class member, compilation fails (no implicit fallback).
- Without `*`, use the normal `A.E.One` form.
## Properties
Properties allow you to define member accessors that look like fields but execute code when read or written. Unlike regular fields, properties in Lyng do **not** have automatic backing fields; they are pure accessors.
### Basic Syntax
Properties are declared using `val` (read-only) or `var` (read-write) followed by a name and a `get` (and optionally `set`) accessor. Unlike fields, properties do not have automatic storage and must compute their values or delegate to other members.
The standard syntax uses parentheses:
```lyng
class Person(private var _age: Int) {
// Read-only property
val ageCategory
get() {
if (_age < 18) "Minor" else "Adult"
}
// Read-write property
var age: Int
get() { _age }
set(value) {
if (value >= 0) _age = value
}
}
```
### Laconic Syntax (Optional Parentheses)
For even cleaner code, you can omit the parentheses for `get` and `set`. This is especially useful for simple expression shorthand:
```lyng
class Circle(val radius: Real) {
// Laconic expression shorthand
val area get = π * radius * radius
val circumference get = 2 * π * radius
fun diameter() = radius * 2
}
fun median(a, b) = (a + b) / 2
class Counter {
private var _count = 0
var count get = _count set(v) = _count = v
}
```
### Key Rules
- **`val` properties** must have a `get` accessor (with or without parentheses) and cannot have a `set`.
- **`var` properties** must have both `get` and `set` accessors.
- **Functions and methods** can use the `=` shorthand to return the result of a single expression.
- **`override` is mandatory**: If you are overriding a member from a base class, you MUST use the `override` keyword.
- **No Backing Fields**: There is no magic `field` identifier. If you need to store state, you must declare a separate (usually `private`) field.
- **Type Inference**: You can omit the type declaration if it can be inferred or if you don't need strict typing.
### Lazy Properties with `cached`
When you want to define a property that is computed only once (on demand) and then remembered, use the built-in `cached` function. This is more efficient than a regular property with `get()` if the computation is expensive, as it avoids re-calculating the value on every access.
```lyng
class DataService(val id: Int) {
// The lambda passed to cached is only executed once, the first time data() is called.
val data = cached {
println("Fetching data for " + id)
// Perform expensive operation
"Record " + id
}
}
val service = DataService(42)
// No printing yet
println(service.data()) // Prints "Fetching data for 42", then returns "Record 42"
println(service.data()) // Returns "Record 42" immediately (no second fetch)
```
Note that `cached` returns a lambda, so you access the value by calling it like a method: `service.data()`. This is a powerful pattern for lazy-loading resources, caching results of database queries, or delaying expensive computations until they are truly needed.
## Delegation
Delegation allows you to hand over the logic of a property or function to another object. This is done using the `by` keyword.
### Property Delegation
Instead of providing `get()` and `set()` accessors, you can delegate them to an object that implements the `getValue` and `setValue` methods.
```lyng
class User {
var name by MyDelegate()
}
```
### Function Delegation
You can also delegate a whole function to an object. When the function is called, it will invoke the delegate's `invoke` method.
```lyng
fun remoteAction by RemoteProxy("actionName")
```
### The Unified Delegate Interface
A delegate is any object that provides the following methods (all optional depending on usage):
- `getValue(thisRef, name)`: Called when a delegated `val` or `var` is read.
- `setValue(thisRef, name, newValue)`: Called when a delegated `var` is written.
- `invoke(thisRef, name, args...)`: Called when a delegated `fun` is invoked.
- `bind(name, access, thisRef)`: Called once during initialization to configure or validate the delegate.
### Map as a Delegate
Maps can also be used as delegates. When delegated to a property, the map uses the property name as the key:
```lyng
val settings = { "theme": "dark", "fontSize": 14 }
val theme by settings
var fontSize by settings
println(theme) // "dark"
fontSize = 16 // Updates settings["fontSize"]
```
For more details and advanced patterns (like `lazy`, `observable`, and shared stateless delegates), see the [Delegation Guide](delegation.md).
## Instance initialization: init block
In addition to the primary constructor arguments, you can provide an `init` block that runs on each instance creation. This is useful for more complex initializations, side effects, or setting up fields that depend on multiple constructor parameters.
class Point(val x, val y) {
var magnitude
init {
magnitude = Math.sqrt(x*x + y*y)
}
}
Key features of `init` blocks:
- **Scope**: They have full access to `this` members and all primary constructor parameters.
- **Order**: In a single-inheritance scenario, `init` blocks run immediately after the instance fields are prepared but before the primary constructor body logic.
- **Multiple blocks**: You can have multiple `init` blocks; they will be executed in the order they appear in the class body.
### Initialization in Multiple Inheritance
In cases of multiple inheritance, `init` blocks are executed following the constructor chaining rule:
1. All ancestors are initialized first, following the inheritance hierarchy (diamond-safe: each ancestor is initialized exactly once).
2. The `init` blocks of each class are executed after its parents have been fully initialized.
3. For a hierarchy `class D : B, C`, the initialization order is: `B`'s chain, then `C`'s chain (skipping common ancestors with `B`), and finally `D`'s own `init` blocks.
### Initialization during Deserialization
When an object is restored from a serialized form (e.g., using `Lynon`), `init` blocks are **re-executed**. This ensures that transient state or derived fields are correctly recalculated upon restoration. However, primary constructors are **not** re-called during deserialization; only the `init` blocks and field initializers are executed to restore the instance state.
Class point has a _method_, or a _member function_ `length()` that uses its _fields_ `x` and `y` to
calculate the magnitude. Length is called
@ -318,68 +53,19 @@ statements discussed later, there could be default values, ellipsis, etc.
class Point(x=0,y=0)
val p = Point()
assert( p.x == 0 && p.y == 0 )
// Named arguments in constructor calls use colon syntax:
val p2 = Point(y: 10, x: 5)
assert( p2.x == 5 && p2.y == 10 )
// Auto-substitution shorthand for named arguments:
val x = 1
val y = 2
val p3 = Point(x:, y:)
assert( p3.x == 1 && p3.y == 2 )
>>> void
Note that unlike **Kotlin**, which uses `=` for named arguments, Lyng uses `:` to avoid ambiguity with assignment expressions.
### Late-initialized `val` fields
You can declare a `val` field without an immediate initializer if you provide an assignment for it within an `init` block or the class body. This is useful when the initial value depends on logic that cannot be expressed in a single expression.
```lyng
class DataProcessor(data: Object) {
val result: Object
init {
// Complex initialization logic
result = transform(data)
}
}
```
Key rules for late-init `val`:
- **Compile-time Check**: The compiler ensures that every `val` declared without an initializer in a class body has at least one assignment within that class body (including `init` blocks). Failing to do so results in a syntax error.
- **Write-Once**: A `val` can only be assigned once. Even if it was declared without an initializer, once it is assigned a value (e.g., in `init`), any subsequent assignment will throw an `IllegalAssignmentException`.
- **Access before Initialization**: If you attempt to read a late-init `val` before it has been assigned (for example, by calling a method in `init` that reads the field before its assignment), it will hold a special `Unset` value. Using `Unset` for most operations (like arithmetic or method calls) will throw an `UnsetException`.
- **No Extensions**: Extension properties do not support late initialization as they do not have per-instance storage. Extension `val`s must always have an initializer or a `get()` accessor.
### The `Unset` singleton
The `Unset` singleton represents a field that has been declared but not yet initialized. While it can be compared and converted to a string, most other operations on it are forbidden to prevent accidental use of uninitialized data.
```lyng
class T {
val x
fun check() {
if (x == Unset) println("Not ready")
}
init {
check() // Prints "Not ready"
x = 42
}
}
```
## Methods
Functions defined inside a class body are methods, and unless declared
`private` are available to be called from outside the class:
class Point(x,y) {
// private method:
private fun d2() { x*x + y*y }
// public method declaration:
fun length() { sqrt(d2()) }
// private method:
private fun d2() {x*x + y*y}
}
val p = Point(3,4)
// private called from inside public: OK
@ -393,7 +79,7 @@ Functions defined inside a class body are methods, and unless declared
Lyng supports declaring a class with multiple direct base classes. The syntax is:
```lyng
```
class Foo(val a) {
var tag = "F"
fun runA() { "ResultA:" + a }
@ -442,16 +128,16 @@ Key rules and features:
- Syntax
- `class Derived(args) : Base1(b1Args), Base2(b2Args)`
- Each direct base may receive constructor arguments specified in the header. Only direct bases receive header args; indirect bases must either be default‑constructible or receive their args through their direct child.
- Each direct base may receive constructor arguments specified in the header. Only direct bases receive header args; indirect bases must either be default‑constructible or receive their args through their direct child (future extensions may add more control).
- Resolution order (C3 MRO)
- Resolution order (C3 MRO — active)
- Member lookup is deterministic and follows C3 linearization (Python‑like), which provides a monotonic, predictable order for complex hierarchies and diamonds.
- Intuition: for `class D() : B(), C()` where `B()` and `C()` both derive from `A()`, the C3 order is `D → B → C → A`.
- The first visible match along this order wins.
- Qualified dispatch
- Inside a class body, use `this@Type.member(...)` to start lookup at the specified ancestor.
- For arbitrary receivers, use casts: `(expr as Type).member(...)` or `(expr as? Type)?.member(...)`.
- For arbitrary receivers, use casts: `(expr as Type).member(...)` or `(expr as? Type)?.member(...)` (safe‑call `?.` is already available in Lyng).
- Qualified access does not relax visibility.
- Field inheritance (`val`/`var`) and collisions
@ -466,215 +152,12 @@ Key rules and features:
- Visibility
- `private`: accessible only inside the declaring class body; not visible in subclasses and cannot be accessed via `this@Type` or casts.
- `protected`: accessible in the declaring class and in any of its transitive subclasses (including MI). Additionally, ancestor classes can access protected members of their descendants if it's an override of a member known to the ancestor. Protected members are not visible from unrelated contexts; qualification/casts do not bypass it.
- `protected`: accessible in the declaring class and in any of its transitive subclasses (including MI), but not from unrelated contexts; qualification/casts do not bypass it.
## Abstract Classes and Members
An `abstract` class is a class that cannot be instantiated and is intended to be inherited by other classes. It can contain `abstract` members that have no implementation and must be implemented by concrete subclasses.
### Abstract Classes
To declare an abstract class, use the `abstract` modifier:
```lyng
abstract class Shape {
abstract fun area(): Real
}
```
Abstract classes can have constructors, fields, and concrete methods, just like regular classes.
### Abstract Members
Methods and variables (`val`/`var`) can be marked as `abstract`. Abstract members must not have a body or initializer.
```lyng
abstract class Base {
abstract fun foo(): Int
abstract var bar: String
}
```
- **Safety**: `abstract` members cannot be `private`, as they must be visible to subclasses for implementation.
- **Contract of Capability**: An `abstract val/var` represents a requirement for a capability. It can be implemented by either a **field** (storage) or a **property** (logic) in a subclass.
## Interfaces
An `interface` in Lyng is a synonym for an `abstract class`. Following the principle that Lyng's Multiple Inheritance system is powerful enough to handle stateful contracts, interfaces support everything classes do, including constructors, fields, and `init` blocks.
```lyng
interface Named(val name: String) {
fun greet() { "Hello, " + name }
}
class Person(name) : Named(name)
```
Using `interface` instead of `abstract class` is a matter of semantic intent, signaling that the class is primarily intended to be used as a contract in MI.
### Implementation by Parts
One of the most powerful benefits of Lyng's Multiple Inheritance and C3 MRO is the ability to satisfy an interface's requirements "by parts" from different parent classes. Since an `interface` can have state and requirements, a subclass can inherit these requirements and satisfy them using members inherited from other parents in the MRO chain.
Example:
```lyng
// Interface with state (id) and abstract requirements
interface Character(val id) {
var health
var mana
fun isAlive() = health > 0
fun status() = name + " (#" + id + "): " + health + " HP, " + mana + " MP"
}
// Parent class 1: provides health
class HealthPool(var health)
// Parent class 2: provides mana and name
class ManaPool(var mana) {
val name = "Hero"
}
// Composite class: implements Character by combining HealthPool and ManaPool
class Warrior(id, h, m) : HealthPool(h), ManaPool(m), Character(id)
val w = Warrior(1, 100, 50)
assertEquals("Hero (#1): 100 HP, 50 MP", w.status())
```
In this example, `Warrior` inherits from `HealthPool`, `ManaPool`, and `Character`. The abstract requirements `health` and `mana` from `Character` are automatically satisfied by the matching members inherited from `HealthPool` and `ManaPool`. The `status()` method also successfully finds the `name` field from `ManaPool`. This pattern allows for highly modular and reusable "trait-like" classes that can be combined to fulfill complex contracts without boilerplate proxy methods.
## Overriding and Virtual Dispatch
When a class defines a member that already exists in one of its parents, it is called **overriding**.
### The `override` Keyword
In Lyng, the `override` keyword is **mandatory when declaring a member** that exists in the ancestor chain (MRO).
```lyng
class Parent {
fun foo() = 1
}
class Child : Parent() {
override fun foo() = 2 // Mandatory override keyword
}
```
- **Implicit Satisfaction**: If a class inherits an abstract requirement and a matching implementation from different parents, the requirement is satisfied automatically without needing an explicit `override` proxy.
- **No Accidental Overrides**: If you define a member that happens to match a parent's member but you didn't use `override`, the compiler will throw an error. This prevents the "Fragile Base Class" problem.
- **Private Members**: Private members in parent classes are NOT part of the virtual interface and cannot be overridden. Defining a member with the same name in a subclass is allowed without `override` and is treated as a new, independent member.
### Visibility Widening
A subclass can increase the visibility of an overridden member (e.g., `protected``public`), but it is strictly forbidden from narrowing it (e.g., `public``protected`).
### The `closed` Modifier
To prevent a member from being overridden in subclasses, use the `closed` modifier (equivalent to `final` in other languages).
```lyng
class Critical {
closed fun secureStep() { ... }
}
```
Attempting to override a `closed` member results in a compile-time error.
## Operator Overloading
Lyng allows you to overload standard operators by defining specific named methods in your classes. When an operator expression is evaluated, Lyng delegates the operation to these methods if they are available.
### Binary Operators
To overload a binary operator, define the corresponding method that takes one argument:
| Operator | Method Name |
| :--- | :--- |
| `a + b` | `plus(other)` |
| `a - b` | `minus(other)` |
| `a * b` | `mul(other)` |
| `a / b` | `div(other)` |
| `a % b` | `mod(other)` |
| `a && b` | `logicalAnd(other)` |
| `a \|\| b` | `logicalOr(other)` |
| `a =~ b` | `operatorMatch(other)` |
| `a & b` | `bitAnd(other)` |
| `a \| b` | `bitOr(other)` |
| `a ^ b` | `bitXor(other)` |
| `a << b` | `shl(other)` |
| `a >> b` | `shr(other)` |
Example:
```lyng
class Vector(val x, val y) {
fun plus(other) = Vector(x + other.x, y + other.y)
override fun toString() = "Vector(${x}, ${y})"
}
val v1 = Vector(1, 2)
val v2 = Vector(3, 4)
assertEquals(Vector(4, 6), v1 + v2)
```
### Unary Operators
Unary operators are overloaded by defining methods with no arguments:
| Operator | Method Name |
| :--- | :--- |
| `-a` | `negate()` |
| `!a` | `logicalNot()` |
| `~a` | `bitNot()` |
### Assignment Operators
Assignment operators like `+=` first attempt to call a specific assignment method. If that method is not defined, they fall back to a combination of the binary operator and a regular assignment (e.g., `a = a + b`).
| Operator | Method Name | Fallback |
| :--- | :--- | :--- |
| `a += b` | `plusAssign(other)` | `a = a + b` |
| `a -= b` | `minusAssign(other)` | `a = a - b` |
| `a *= b` | `mulAssign(other)` | `a = a * b` |
| `a /= b` | `divAssign(other)` | `a = a / b` |
| `a %= b` | `modAssign(other)` | `a = a % b` |
Example of in-place mutation:
```lyng
class Counter(var value) {
fun plusAssign(n) {
value = value + n
}
}
val c = Counter(10)
c += 5
assertEquals(15, c.value)
```
### Comparison Operators
Comparison operators use `compareTo` and `equals`.
| Operator | Method Name |
| :--- | :--- |
| `a == b`, `a != b` | `equals(other)` |
| `<`, `>`, `<=`, `>=`, `<=>` | `compareTo(other)` |
- `compareTo` should return:
- `0` if `a == b`
- A negative integer if `a < b`
- A positive integer if `a > b`
- The `<=>` (shuttle) operator returns the result of `compareTo` directly.
- `equals` returns a `Bool`. If `equals` is not explicitly defined, Lyng falls back to `compareTo(other) == 0`.
> **Note**: Methods that are already defined in the base `Obj` class (like `equals`, `toString`, or `contains`) require the `override` keyword when redefined in your class or as an extension. Other operator methods (like `plus` or `negate`) do not require `override` unless they are already present in your class's hierarchy.
### Increment and Decrement
`++` and `--` operators are implemented using `plus(1)` or `minus(1)` combined with an assignment back to the variable. If the variable is a field or local variable, it will be updated with the result of the operation.
- Diagnostics
- When a member/field is not found, error messages include the receiver class name and the considered linearization order, with suggestions to disambiguate using `this@Type` or casts if appropriate.
- Qualifying with a non‑ancestor in `this@Type` reports a clear error mentioning the receiver lineage.
- `as`/`as?` cast errors mention the actual and target types.
Compatibility notes:
@ -759,23 +242,6 @@ Notes and limitations (current version):
- `name` and `ordinal` are read‑only properties of an entry.
- `entries` is a read‑only list owned by the enum type.
## Exception Classes
You can define your own exception classes by inheriting from the built-in `Exception` class. User-defined exceptions are regular classes and can have their own properties and methods.
```lyng
class MyError(val code, m) : Exception(m)
try {
throw MyError(500, "Internal Server Error")
}
catch(e: MyError) {
println("Error " + e.code + ": " + e.message)
}
```
For more details on error handling, see the [Exceptions Handling Guide](exceptions_handling.md).
## fields and visibility
It is possible to add non-constructor fields:
@ -808,69 +274,6 @@ Are declared with var
assert( p.isSpecial == true )
>>> void
### Restricted Setter Visibility
You can restrict the visibility of a `var` field's or property's setter by using `private set` or `protected set` modifiers. This allows the member to be publicly readable but only writable from within the class or its subclasses.
#### On Fields
```lyng
class SecretCounter {
var count = 0
private set // Can be read anywhere, but written only in SecretCounter
fun increment() { count++ }
}
val c = SecretCounter()
println(c.count) // OK
c.count = 10 // Throws IllegalAccessException
c.increment() // OK
```
#### On Properties
You can also apply restricted visibility to custom property setters:
```lyng
class Person(private var _age: Int) {
var age
get() = _age
private set(v) { if (v >= 0) _age = v }
}
```
#### Protected Setters and Inheritance
A `protected set` allows subclasses to modify a field that is otherwise read-only to the public:
```lyng
class Base {
var state = "initial"
protected set
}
class Derived : Base() {
fun changeState(newVal) {
state = newVal // OK: protected access from subclass
}
}
val d = Derived()
println(d.state) // OK: "initial"
d.changeState("updated")
println(d.state) // OK: "updated"
d.state = "bad" // Throws IllegalAccessException: public write not allowed
```
### Key Rules and Limitations
- **Only for `var`**: Restricted setter visibility cannot be used with `val` declarations, as they are inherently read-only. Attempting to use it with `val` results in a syntax error.
- **Class Body Only**: These modifiers can only be used on members declared within the class body. They are not supported for primary constructor parameters.
- **`private set`**: The setter is only accessible within the same class context (specifically, when `this` is an instance of that class).
- **`protected set`**: The setter is accessible within the declaring class and all its transitive subclasses.
- **Multiple Inheritance**: In MI scenarios, visibility is checked against the class that actually declared the member. Qualified access (e.g., `this@Base.field = value`) also respects restricted setter visibility.
### Private fields
Private fields are visible only _inside the class instance_:
@ -898,44 +301,23 @@ Private fields are visible only _inside the class instance_:
void
>>> void
### Transient fields
You can mark a field or a constructor parameter as transient using the `@Transient` attribute. Transient members are ignored during serialization (Lynon and JSON) and are also excluded from structural equality (`==`) checks.
```lyng
class Session(@Transient val token, val userId) {
@Transient var lastAccess = time.now()
var data = Map()
}
```
For more details on how transient fields behave during restoration, see the [Serialization Guide](serialization.md).
### Protected members
Protected members are available to the declaring class and all of its transitive subclasses (including via MI). Additionally, an ancestor class can access a `protected` member of its descendant if the ancestor also defines or inherits a member with the same name (i.e., it is an override of something the ancestor knows about).
Protected members are available to the declaring class and all of its transitive subclasses (including via MI), but not from unrelated contexts:
Protected members are not available from unrelated contexts:
```lyng
class Base {
abstract protected fun foo()
fun bar() {
// Ancestor can see foo() because it's an override
// of a member it defines (even as abstract):
foo()
}
```
class A() {
protected fun ping() { "pong" }
}
class B() : A() {
fun call() { this@A.ping() }
}
class Derived : Base {
override protected fun foo() { "ok" }
}
assertEquals("ok", Derived().bar())
val b = B()
assertEquals("pong", b.call())
// Unrelated access is forbidden, even via cast
assertThrows { (Derived() as Base).foo() }
assertThrows { (b as A).ping() }
```
It is possible to provide private constructor parameters so they can be
@ -978,7 +360,7 @@ You can mark a field or a method as static. This is borrowed from Java as more p
static fun exclamation() {
// here foo is a regular var:
Value.foo.x + "!"
foo.x + "!"
}
}
assertEquals( Value.foo.x, "foo" )
@ -989,98 +371,59 @@ You can mark a field or a method as static. This is borrowed from Java as more p
assertEquals( "bar!", Value.exclamation() )
>>> void
Static fields can be accessed from static methods via the class qualifier:
As usual, private statics are not accessible from the outside:
class Test {
static var data = "foo"
static fun getData() { Test.data }
// private, inacessible from outside protected data:
private static var data = null
// the interface to access and change it:
static fun getData() { data }
static fun setData(value) { data = value }
}
assertEquals( "foo", Test.getData() )
Test.data = "bar"
assertEquals("bar", Test.getData() )
// no direct access:
assertThrows { Test.data }
// accessible with the interface:
assertEquals( null, Test.getData() )
Test.setData("fubar")
assertEquals("fubar", Test.getData() )
>>> void
# Extending classes
It sometimes happen that the class is missing some particular functionality that can be _added to it_ without rewriting its inner logic and using its private state. In this case _extension members_ could be used.
It sometimes happen that the class is missing some particular functionality that can be _added to it_ without rewriting its inner logic and using its private state. In this case _extension methods_ could be used, for example. we want to create an extension method
that would test if some object of unknown type contains something that can be interpreted
as an integer. In this case we _extend_ class `Object`, as it is the parent class for any instance of any type:
## Extension methods
fun Object.isInteger() {
when(this) {
// already Int?
is Int -> true
For example, we want to create an extension method that would test if a value can be interpreted as an integer:
// real, but with no declimal part?
is Real -> toInt() == this
fun Int.isInteger() { true }
fun Real.isInteger() { this.toInt() == this }
fun String.isInteger() { (this.toReal() as Real).isInteger() }
// string with int or real reuusig code above
is String -> toReal().isInteger()
// otherwise, no:
else -> false
}
}
// Let's test:
// Let's test:
assert( 12.isInteger() == true )
assert( 12.1.isInteger() == false )
assert( "5".isInteger() )
assert( ! "5.2".isInteger() )
>>> void
## Extension properties
__Important note__ as for version 0.6.9, extensions are in __global scope__. It means, that once applied to a global type (Int in our sample), they will be available for _all_ contexts, even new created,
as they are modifying the type, not the context.
Just like methods, you can extend existing classes with properties. These can be defined using simple initialization (for `val` only) or with custom accessors.
### Simple val extension
A read-only extension can be defined by assigning an expression:
```lyng
val String.isLong = length > 10
val s = "Hello, world!"
assert(s.isLong)
```
### Properties with accessors
For more complex logic, use `get()` and `set()` blocks:
```lyng
class Box(var value: Int)
var Box.doubledValue
get() = value * 2
set(v) = value = v / 2
val b = Box(10)
assertEquals(20, b.doubledValue)
b.doubledValue = 30
assertEquals(15, b.value)
```
Extension members are strictly barred from accessing private members of the class they extend, maintaining encapsulation.
### Extension Scoping and Isolation
Extensions in Lyng are **scope-isolated**. This means an extension is only visible within the scope where it is defined and its child scopes. This reduces the "attack surface" and prevents extensions from polluting the global space or other modules.
#### Scope Isolation Example
You can define different extensions with the same name in different scopes:
```lyng
fun scopeA() {
val Int.description = "Number: " + toString()
assertEquals("Number: 42", 42.description)
}
fun scopeB() {
val Int.description = "Value: " + toString()
assertEquals("Value: 42", 42.description)
}
scopeA()
scopeB()
// Outside those scopes, Int.description is not defined
assertThrows { 42.description }
```
This isolation ensures that libraries can use extensions internally without worrying about name collisions with other libraries or the user's code. When a module is imported using `use`, its top-level extensions become available in the importing scope.
Beware of it. We might need to reconsider it later.
## dynamic symbols
@ -1115,7 +458,7 @@ The same we can provide writable dynamic fields (var-type), adding set method:
// mutable field
"bar" -> storedValueForBar
else -> throw SymbolNotFound()
else -> throw SymbolNotFoundException()
}
}
set { name, value ->
@ -1199,12 +542,12 @@ request](https://gitea.sergeych.net/SergeychWorks/lyng/issues).
- ObjClass sole parent is Obj
- ObjClass contains code for instance methods, class fields, hierarchy information.
- Class information is also scoped.
- We avoid imported classes duplication using packages and import caching, so the same imported module is the same object in all its classes.
- We acoid imported classes duplication using packages and import caching, so the same imported module is the same object in all its classes.
## Instances
Result of executing of any expression or statement in the Lyng is the object that
inherits `Obj`, but is not `Obj`. For example, it could be Int, void, null, real, string, bool, etc.
inherits `Obj`, but is not `Obj`. For example it could be Int, void, null, real, string, bool, etc.
This means whatever expression returns or the variable holds, is the first-class
object, no differenes. For example:

View File

@ -45,11 +45,10 @@ are equal or within another, taking into account the end-inclusiveness:
assert( (1..<3) in (1..3) )
>>> void
## Ranges are iterable
## Finite Ranges are iterable
Finite ranges are [Iterable] and can be used in loops and list conversions.
Open-ended ranges are iterable only with an explicit `step`, and open-start
ranges are never iterable.
So given a range with both ends, you can assume it is [Iterable]. This automatically let
use finite ranges in loops and convert it to lists:
assert( [-2, -1, 0, 1] == (-2..1).toList() )
>>> void
@ -63,8 +62,6 @@ In spite of this you can use ranges in for loops:
>>> 3
>>> void
The loop variable is read-only inside the loop body (behaves like a `val`).
but
for( i in 1..<3 )
@ -73,26 +70,6 @@ but
>>> 2
>>> void
### Stepped ranges
Use `step` to change the iteration increment. The range bounds still define membership,
so iteration ends when the next value is no longer in the range.
assert( [1,3,5] == (1..5 step 2).toList() )
assert( [1,3] == (1..<5 step 2).toList() )
assert( ['a','c','e'] == ('a'..'e' step 2).toList() )
>>> void
Real ranges require an explicit step:
assert( [0,0.25,0.5,0.75,1.0] == (0.0..1.0 step 0.25).toList() )
>>> void
Open-ended ranges require an explicit step to iterate:
(0.. step 1).take(3).toList()
>>> [0,1,2]
## Character ranges
You can use Char as both ends of the closed range:
@ -121,12 +98,10 @@ Exclusive end char ranges are supported too:
| isEndInclusive | true for '..' | Bool |
| isOpen | at any end | Bool |
| isIntRange | both start and end are Int | Bool |
| step | explicit iteration step | Any? |
| start | | Any? |
| end | | Any? |
| start | | Bool |
| end | | Bool |
| size | for finite ranges, see above | Long |
| [] | see above | |
| | | |
Ranges are also used with the `clamp(value, range)` function and the `value.clamp(range)` extension method to limit values within boundaries.
[Iterable]: Iterable.md
[Iterable]: Iterable.md

View File

@ -19,7 +19,6 @@ you can use it's class to ensure type:
|-----------------|-------------------------------------------------------------|------|
| `.roundToInt()` | round to nearest int like round(x) | Int |
| `.toInt()` | convert integer part of real to `Int` dropping decimal part | Int |
| `.clamp(range)` | clamp value within range boundaries | Real |
| | | |
| | | |
| | | |

View File

@ -24,14 +24,13 @@ counterpart, _not match_ operator `!~`:
When you need to find groups, and more detailed match information, use `Regex.find`:
val result: RegexMatch? = Regex("abc(\d)(\d)(\d)").find( "bad456 good abc123")
val result = Regex("abc(\d)(\d)(\d)").find( "bad456 good abc123")
assert( result != null )
val match: RegexMatch = result as RegexMatch
assertEquals( 12 ..< 17, match.range )
assertEquals( "abc123", match[0] )
assertEquals( "1", match[1] )
assertEquals( "2", match[2] )
assertEquals( "3", match[3] )
assertEquals( 12 .. 17, result.range )
assertEquals( "abc123", result[0] )
assertEquals( "1", result[1] )
assertEquals( "2", result[2] )
assertEquals( "3", result[3] )
>>> void
Note that the object `RegexMatch`, returned by [Regex.find], behaves much like in many other languages: it provides the
@ -40,12 +39,11 @@ index range and groups matches as indexes.
Match operator actually also provides `RegexMatch` in `$~` reserved variable (borrowed from Ruby too):
assert( "bad456 good abc123" =~ "abc(\d)(\d)(\d)".re )
val match2: RegexMatch = $~ as RegexMatch
assertEquals( 12 ..< 17, match2.range )
assertEquals( "abc123", match2[0] )
assertEquals( "1", match2[1] )
assertEquals( "2", match2[2] )
assertEquals( "3", match2[3] )
assertEquals( 12 .. 17, $~.range )
assertEquals( "abc123", $~[0] )
assertEquals( "1", $~[1] )
assertEquals( "2", $~[2] )
assertEquals( "3", $~[3] )
>>> void
This is often more readable than calling `find`.
@ -61,7 +59,7 @@ string can be either left or right operator, but not both:
Also, string indexing is Regex-aware, and works like `Regex.find` (_not findall!_):
assert( "cd" == ("abcdef"[ "c.".re ] as RegexMatch).value )
assert( "cd" == "abcdef"[ "c.".re ].value )
>>> void
@ -90,3 +88,4 @@ Also, string indexing is Regex-aware, and works like `Regex.find` (_not findall!
[List]: List.md
[Range]: Range.md

View File

@ -26,8 +26,8 @@ no indexing. Use [set.toList] as needed.
// intersection
assertEquals( Set(1,4), Set(3, 1, 4).intersect(Set(2, 4, 1)) )
// or simple (intersection)
assertEquals( Set(1,4), Set(3, 1, 4).intersect(Set(2, 4, 1)) )
// or simple
assertEquals( Set(1,4), Set(3, 1, 4) * Set(2, 4, 1) )
// To find collection elements not present in another collection, use the
// subtract() or `-`:
@ -91,4 +91,4 @@ Sets are only equal when contains exactly same elements, order, as was said, is
Also, it inherits methods from [Iterable].
[Range]: Range.md
[Range]: Range.md

View File

@ -1,121 +0,0 @@
# Testing and Assertions
Lyng provides several built-in functions for testing and verifying code behavior. These are available in all scripts.
## Basic Assertions
### `assert`
Assert that a condition is true.
assert(condition, message=null)
- `condition`: A boolean expression.
- `message` (optional): A string message to include in the exception if the assertion fails.
If the condition is false, it throws an `AssertionFailedException`.
```lyng
assert(1 + 1 == 2)
assert(true, "This should be true")
```
### `assertEquals` and `assertEqual`
Assert that two values are equal. `assertEqual` is an alias for `assertEquals`.
assertEquals(expected, actual)
assertEqual(expected, actual)
If `expected != actual`, it throws an `AssertionFailedException` with a message showing both values.
```lyng
assertEquals(4, 2 * 2)
assertEqual("hello", "hel" + "lo")
```
### `assertNotEquals`
Assert that two values are not equal.
assertNotEquals(unexpected, actual)
If `unexpected == actual`, it throws an `AssertionFailedException`.
```lyng
assertNotEquals(5, 2 * 2)
```
## Exception Testing
### `assertThrows`
Assert that a block of code throws an exception.
assertThrows(code)
assertThrows(expectedExceptionClass, code)
- `expectedExceptionClass` (optional): The class of the exception that is expected to be thrown.
- `code`: A lambda block or statement to execute.
If the code does not throw an exception, an `AssertionFailedException` is raised.
If an `expectedExceptionClass` is provided, the thrown exception must be of that class (or its subclass), otherwise an error is raised.
`assertThrows` returns the caught exception object if successful.
```lyng
// Just assert that something is thrown
assertThrows { 1 / 0 }
// Assert that a specific exception class is thrown
assertThrows(NoSuchElementException) {
[1, 2, 3].findFirst { it > 10 }
}
// You can use the returned exception
val ex = assertThrows { throw Exception("custom error") }
assertEquals("custom error", ex.message)
```
## Other Validation Functions
While not strictly for testing, these functions help in defensive programming:
### `require`
require(condition, message="requirement not met")
Throws an `IllegalArgumentException` if the condition is false. Use this for validating function arguments.
If we want to evaluate the message lazily:
require(condition) { "requirement not met: %s"(someData) }
In this case, formatting will only occur if the condition is not met.
### `check`
check(condition, message="check failed")
Throws an `IllegalStateException` if the condition is false. Use this for validating internal state.
With lazy message evaluation:
check(condition) { "check failed: %s"(someData) }
In this case, formatting will only occur if the condition is not met.
### TODO
It is easy to mark some code and make it throw a special exception at cone with:
TODO()
or
TODO("some message")
It raises an `NotImplementedException` with the given message. You can catch it
as any other exception when necessary.
Many IDE and editors have built-in support for marking code with TODOs.

View File

@ -105,7 +105,6 @@ arguments list in almost arbitrary ways. For example:
var result = ""
for( a in args ) result += a
}
// loop variables are read-only inside the loop body
assertEquals(
"4231",
@ -154,10 +153,9 @@ Function annotation can have more args specified at call time. There arguments m
@Registered("bar")
fun foo2() { "called foo2" }
val fooFn: Callable = registered["foo"] as Callable
val barFn: Callable = registered["bar"] as Callable
assertEquals(fooFn(), "called foo")
assertEquals(barFn(), "called foo2")
assertEquals(registered["foo"](), "called foo")
assertEquals(registered["bar"](), "called foo2")
>>> void
[parallelism]: parallelism.md

View File

@ -1,15 +0,0 @@
# AI notes: avoid Kotlin/Wasm invalid IR with suspend lambdas
## Do
- Prefer explicit `object : Statement()` with `override suspend fun execute(...)` when building compiler statements.
- Keep `Statement` objects non-lambda, especially in compiler hot paths like parsing/var declarations.
- If you need conditional behavior, return early in `execute` instead of wrapping `parseExpression()` with `statement(...) { ... }`.
- When wasmJs tests hang in the browser, first check `wasmJsNodeTest` for a compile error; hangs often mean module instantiation failed.
## Don't
- Do not create suspend lambdas inside `Statement` factories (`statement { ... }`) for wasm targets.
- Do not "fix" hangs by increasing browser timeouts; it masks invalid wasm generation.
## Debugging tips
- Look for `$invokeCOROUTINE$` in wasm function names when mapping failures.
- If node test logs a wasm compile error, the browser hang is likely the same root cause.

View File

@ -34,18 +34,13 @@ Valid examples:
Ellipsis are used to declare variadic arguments. It basically means "all the arguments available here". It means, ellipsis argument could be in any part of the list, being, end or middle, but there could be only one ellipsis argument and it must not have default value, its default value is always `[]`, en empty list.
Ellipsis can also appear in **function types** to denote a variadic position:
var f: (Int, Object..., String)->Real
var anyArgs: (...)->Int // shorthand for (Object...)->Int
Ellipsis argument receives what is left from arguments after processing regular one that could be before or after.
Ellipsis could be a first argument:
fun testCountArgs(data...,size) {
assert(size is Int)
assertEquals(size, (data as List).size)
assertEquals(size, data.size)
}
testCountArgs( 1, 2, "three", 3)
>>> void
@ -54,7 +49,7 @@ Ellipsis could also be a last one:
fun testCountArgs(size, data...) {
assert(size is Int)
assertEquals(size, (data as List).size)
assertEquals(size, data.size)
}
testCountArgs( 3, 10, 2, "three")
>>> void
@ -63,7 +58,7 @@ Or in the middle:
fun testCountArgs(size, data..., textToReturn) {
assert(size is Int)
assertEquals(size, (data as List).size)
assertEquals(size, data.size)
textToReturn
}
testCountArgs( 3, 10, 2, "three", "All OK")
@ -80,13 +75,6 @@ destructuring arrays when calling functions and lambdas:
getFirstAndLast( ...(1..10) ) // see "splats" section below
>>> [1,10]
Note that array destructuring can also be used in assignments:
val [first, middle..., last] = [1, 2, 3, 4, 5]
[x, y] = [y, x] // Swap
See [tutorial] and [List] documentation for more details on destructuring assignments.
# Splats
Ellipsis allows to convert argument lists to lists. The inversa algorithm that converts [List],
@ -112,54 +100,42 @@ There could be any number of splats at any positions. You can splat any other [I
## Named arguments in calls
Lyng supports named arguments at call sites using colon syntax `name: value`.
### Shorthand for Named Arguments
If you want to pass a variable as a named argument and the variable has the same name as the parameter, you can omit the value and use the shorthand `name:`. This is highly readable and matches the shorthand for map literals.
Lyng supports named arguments at call sites using colon syntax `name: value`:
```lyng
fun test(a, b, c) { [a, b, c] }
val a = 1
val b = 2
val c = 3
// Explicit:
assertEquals([1, 2, 3], test(a: a, b: b, c: c))
// Shorthand (preferred):
assertEquals([1, 2, 3], test(a:, b:, c:))
fun test(a="foo", b="bar", c="bazz") { [a, b, c] }
assertEquals(["foo", "b", "bazz"], test(b: "b"))
assertEquals(["a", "bar", "c"], test("a", c: "c"))
```
This shorthand is elegant, reduces boilerplate, and is consistent with Lyng's map literal syntax. It works for both function calls and class constructors.
Rules for named arguments:
Rules:
- Named arguments must follow positional arguments. After the first named argument, no positional arguments may appear inside the parentheses.
- The only exception is the syntactic trailing block after the call: `f(args) { ... }`. This block is outside the parentheses and is handled specially (see below).
- A named argument cannot reassign a parameter already set positionally.
- If the last parameter has already been assigned by a named argument (or named splat), a trailing block is not allowed and results in an error.
Why `:` and not `=` at call sites? In Lyng, `=` is an expression (assignment), so we use `:` to avoid ambiguity. This is a key difference from **Kotlin**, which uses `=` for named arguments. Declarations in Lyng continue to use `:` for types, while call sites use `as` / `as?` for type operations.
Why `:` and not `=` at call sites? In Lyng, `=` is an expression (assignment), so we use `:` to avoid ambiguity. Declarations continue to use `:` for types, while call sites use `as` / `as?` for type operations.
## Named splats (map splats)
Splat (`...`) of a Map provides named arguments to the call. Only string keys are allowed. You can use the same auto-substitution shorthand inside map literals used for splats:
Splat (`...`) of a Map provides named arguments to the call. Only string keys are allowed:
```lyng
fun test(a="a", b="b", c="c", d="d") { [a, b, c, d] }
val b = "B!"
val d = "D!"
// Auto-substitution in map literal:
val patch = { d:, b: }
val r = test("A?", ...patch)
val r = test("A?", ...Map("d" => "D!", "b" => "B!"))
assertEquals(["A?","B!","c","D!"], r)
```
The same with a map literal is often more concise. Define the literal, then splat the variable:
fun test(a="a", b="b", c="c", d="d") { [a, b, c, d] }
val patch = { d: "D!", b: "B!" }
val r = test("A?", ...patch)
assertEquals(["A?","B!","c","D!"], r)
>>> void
Constraints:
- Map splat keys must be strings; otherwise, a clean error is thrown.
@ -179,4 +155,3 @@ If a call is immediately followed by a block `{ ... }`, it is treated as an extr
[tutorial]: tutorial.md
[List]: List.md

View File

@ -1,194 +0,0 @@
# Delegation in Lyng
Delegation is a powerful pattern that allows you to outsource the logic of properties (`val`, `var`) and functions (`fun`) to another object. This enables code reuse, separation of concerns, and the implementation of common patterns like lazy initialization, observable properties, and remote procedure calls (RPC) with minimal boilerplate.
## The `by` Keyword
Delegation is triggered using the `by` keyword in a declaration. The expression following `by` is evaluated once when the member is initialized, and the resulting object becomes the **delegate**.
```lyng
val x by MyDelegate()
var y by MyDelegate()
fun f by MyDelegate()
```
## The Unified Delegate Model
A delegate object can implement any of the following methods to intercept member access. All methods receive the `thisRef` (the instance containing the member) and the `name` of the member.
```lyng
interface Delegate {
// Called when a 'val' or 'var' is read
fun getValue(thisRef, name)
// Called when a 'var' is assigned
fun setValue(thisRef, name, newValue)
// Called when a 'fun' is invoked
fun invoke(thisRef, name, args...)
// Optional: Called once during initialization to "bind" the delegate
// Can be used for validation or to return a different delegate instance
fun bind(name, access, thisRef) = this
}
```
### Delegate Access Types
The `bind` method receives an `access` parameter of type `DelegateAccess`, which can be one of:
- `DelegateAccess.Val`
- `DelegateAccess.Var`
- `DelegateAccess.Callable` (for `fun`)
## Usage Cases and Examples
### 1. Lazy Initialization
The classic `lazy` pattern ensures a value is computed only when first accessed and then cached. In Lyng, `lazy` is implemented as a class that follows this pattern. While classes typically start with an uppercase letter, `lazy` is an exception to make its usage feel like a native language feature.
```lyng
class lazy(val creator) : Delegate {
private var value = Unset
override fun bind(name, access, thisRef) {
if (access != DelegateAccess.Val) throw "lazy delegate can only be used with 'val'"
this
}
override fun getValue(thisRef, name) {
if (value == Unset) {
// calculate value using thisRef as this:
value = with(thisRef) creator()
}
value
}
}
// Usage:
val expensiveData by lazy {
println("Performing expensive computation...")
42
}
println(expensiveData) // Computes and prints 42
println(expensiveData) // Returns 42 immediately
```
### 2. Observable Properties
Delegates can be used to react to property changes.
```lyng
class Observable(initialValue, val onChange) {
private var value = initialValue
fun getValue(thisRef, name) = value
fun setValue(thisRef, name, newValue) {
val oldValue = value
value = newValue
onChange(name, oldValue, newValue)
}
}
class User {
var name by Observable("Guest") { name, old, new ->
println("Property %s changed from %s to %s"(name, old, new))
}
}
val u = User()
u.name = "Alice" // Prints: Property name changed from Guest to Alice
```
### 3. Function Delegation (Proxies)
You can delegate an entire function to an object. This is particularly useful for implementing decorators or RPC clients.
```lyng
object LoggerDelegate {
fun invoke(thisRef, name, args...) {
println("Calling function: " + name + " with args: " + args)
// Logic here...
"Result of " + name
}
}
fun remoteAction by LoggerDelegate
println(remoteAction(1, 2, 3))
// Prints: Calling function: remoteAction with args: [1, 2, 3]
// Prints: Result of remoteAction
```
### 4. Stateless Delegates (Shared Singletons)
Because `getValue`, `setValue`, and `invoke` receive `thisRef`, a single object can act as a delegate for multiple properties across many instances without any per-property memory overhead.
```lyng
object Constant42 {
fun getValue(thisRef, name) = 42
}
class Foo {
val a by Constant42
val b by Constant42
}
val f = Foo()
assertEquals(42, f.a)
assertEquals(42, f.b)
```
### 5. Local Delegation
Delegation is not limited to class members; you can also use it for local variables inside functions.
```lyng
fun test() {
val x by LocalProxy(123)
println(x)
}
```
### 6. Map as a Delegate
Maps can be used as delegates for `val` and `var` properties. When a map is used as a delegate, it uses the property name as a key to read from or write to the map.
```lyng
val m = { "a": 1, "b": 2 }
val a by m
var b by m
println(a) // 1
println(b) // 2
b = 42
println(m["b"]) // 42
```
Because `Map` implements `getValue` and `setValue`, it works seamlessly with any object that needs to store its properties in a map (e.g., when implementing dynamic schemas or JSON-backed objects).
## The `bind` Hook
The `bind(name, access, thisRef)` method is called exactly once when the member is being initialized. It allows the delegate to:
1. **Validate usage**: Throw an error if the delegate is used with the wrong member type (e.g., `lazy` on a `var`).
2. **Initialize state**: Set up internal state based on the property name or the containing instance.
3. **Substitute itself**: Return a different object that will act as the actual delegate.
```lyng
class ValidatedDelegate() {
fun bind(name, access, thisRef) {
if (access == DelegateAccess.Var) {
throw "This delegate cannot be used with 'var'"
}
this
}
fun getValue(thisRef, name) = "Validated"
}
```
## Summary
Delegation in Lyng combines the elegance of Kotlin-style properties with the flexibility of dynamic function interception. By unifying `val`, `var`, and `fun` delegation into a single model, Lyng provides a consistent and powerful tool for meta-programming and code reuse.

View File

@ -103,207 +103,13 @@ scope.addVoidFn("log") {
println(items.joinToString(" ") { it.toString(this).value })
}
// When adding a member function to a class, you can use isOverride = true
// myClass.addFn("toString", isOverride = true) {
// ObjString("Custom string representation")
// }
// Call them from Lyng
scope.eval("val y = inc(41); log('Answer:', y)")
```
You can register multiple names (aliases) at once: `addFn<ObjInt>("inc", "increment") { ... }`.
Scope-backed Kotlin lambdas receive a `ScopeFacade` (not a full `Scope`). For migration and convenience, these utilities are available on the facade:
- Access: `args`, `pos`, `thisObj`, `get(name)`
- Invocation: `call(...)`, `resolve(...)`, `assign(...)`, `toStringOf(...)`, `inspect(...)`, `trace(...)`
- Args helpers: `requiredArg<T>()`, `requireOnlyArg<T>()`, `requireExactCount(...)`, `requireNoArgs()`, `thisAs<T>()`
- Errors: `raiseError(...)`, `raiseClassCastError(...)`, `raiseIllegalArgument(...)`, `raiseIllegalState(...)`, `raiseNoSuchElement(...)`,
`raiseSymbolNotFound(...)`, `raiseNotImplemented(...)`, `raiseNPE()`, `raiseIndexOutOfBounds(...)`, `raiseIllegalAssignment(...)`,
`raiseUnset(...)`, `raiseNotFound(...)`, `raiseAssertionFailed(...)`, `raiseIllegalOperation(...)`, `raiseIterationFinished()`
If you truly need the full `Scope` (e.g., for low-level interop), use `requireScope()` explicitly.
### 5) Add Kotlin‑backed fields
If you need a simple field (with a value) instead of a computed property, use `createField`. This adds a field to the class that will be present in all its instances.
```kotlin
val myClass = ObjClass("MyClass")
// Add a read-only field (constant)
myClass.createField("version", ObjString("1.0.0"), isMutable = false)
// Add a mutable field with an initial value
myClass.createField("count", ObjInt(0), isMutable = true)
// If you are overriding a field from a base class, use isOverride = true
// myClass.createField("someBaseField", ObjInt(42), isOverride = true)
scope.addConst("MyClass", myClass)
```
In Lyng:
```lyng
val instance = MyClass()
println(instance.version) // -> "1.0.0"
instance.count = 5
println(instance.count) // -> 5
```
### 6) Add Kotlin‑backed properties
Properties in Lyng are pure accessors (getters and setters) and do not have automatic backing fields. You can add them to a class using `addProperty`.
```kotlin
val myClass = ObjClass("MyClass")
var internalValue: Long = 10
myClass.addProperty(
name = "value",
getter = {
// Return current value as a Lyng object
ObjInt(internalValue)
},
setter = { newValue ->
// newValue is passed as a Lyng object (the first and only argument)
internalValue = (newValue as ObjInt).value
}
)
// You can also create an ObjProperty explicitly
val explicitProp = ObjProperty(
name = "hexValue",
getter = statement { ObjString(internalValue.toString(16)) }
)
myClass.addProperty("hexValue", prop = explicitProp)
// Use isOverride = true when overriding a property from a base class
// myClass.addProperty("baseProp", getter = { ... }, isOverride = true)
scope.addConst("MyClass", myClass)
```
Usage in Lyng:
```lyng
val instance = MyClass()
println(instance.value) // -> 10
instance.value = 42
println(instance.value) // -> 42
```
### 6.5) Preferred: bind Kotlin implementations to declared Lyng classes
For extensions and libraries, the **preferred** workflow is Lyng‑first: declare the class and its members in Lyng, then bind the Kotlin implementations using the bridge.
This keeps Lyng semantics (visibility, overrides, type checks) in Lyng, while Kotlin supplies the behavior.
```lyng
// Lyng side (in a module)
class Counter {
extern var value: Int
extern fun inc(by: Int): Int
}
```
Note: members must be marked `extern` so the compiler emits the ABI slots that Kotlin bindings attach to. This applies to functions and properties bound via `addFun` / `addVal` / `addVar`.
```kotlin
// Kotlin side (binding)
val moduleScope = Script.newScope() // or an existing module scope
moduleScope.eval("class Counter { extern var value: Int; extern fun inc(by: Int): Int }")
moduleScope.bind("Counter") {
addVar(
name = "value",
get = { _, self -> self.readField(this, "value").value },
set = { _, self, v -> self.writeField(this, "value", v) }
)
addFun("inc") { _, self, args ->
val by = args.requiredArg<ObjInt>(0).value
val current = self.readField(this, "value").value as ObjInt
val next = ObjInt(current.value + by)
self.writeField(this, "value", next)
next
}
}
```
Notes:
- Binding must happen **before** the first instance is created.
- Use [LyngClassBridge] to bind by name/module, or by an already resolved `ObjClass`.
- Use `ObjInstance.data` / `ObjClass.classData` to attach Kotlin‑side state when needed.
### 6.5a) Bind Kotlin implementations to declared Lyng objects
For `extern object` declarations, bind implementations to the singleton instance using `ModuleScope.bindObject`.
This mirrors class binding but targets an already created object instance.
```lyng
// Lyng side (in a module)
extern object HostObject {
extern fun add(a: Int, b: Int): Int
extern val status: String
extern var count: Int
}
```
```kotlin
// Kotlin side (binding)
val moduleScope = importManager.createModuleScope(Pos.builtIn, "bridge.obj")
moduleScope.bindObject("HostObject") {
classData = "OK"
init { _ -> data = 0L }
addFun("add") { _, _, args ->
val a = args.requiredArg<ObjInt>(0).value
val b = args.requiredArg<ObjInt>(1).value
ObjInt.of(a + b)
}
addVal("status") { _, _ -> ObjString(classData as String) }
addVar(
"count",
get = { _, inst -> ObjInt.of((inst as ObjInstance).data as Long) },
set = { _, inst, value -> (inst as ObjInstance).data = (value as ObjInt).value }
)
}
```
Notes:
- Members must be marked `extern` so the compiler emits ABI slots for Kotlin bindings.
- You can also bind by name/module via `LyngObjectBridge.bind(...)`.
### 6.6) Preferred: Kotlin reflection bridge for call‑by‑name
For Kotlin code that needs dynamic access to Lyng variables, functions, or members, use the bridge resolver.
It provides explicit, cached handles and predictable lookup rules.
```kotlin
val scope = Script.newScope()
scope.eval("""
val x = 40
fun add(a, b) = a + b
class Box { var value = 1 }
""")
val resolver = scope.resolver()
// Read a top‑level value
val x = resolver.resolveVal("x").get(scope)
// Call a function by name (cached inside the resolver)
val sum = (resolver as BridgeCallByName).callByName(scope, "add", Arguments(ObjInt(1), ObjInt(2)))
// Member access
val box = scope.eval("Box()")
val valueHandle = resolver.resolveMemberVar(box, "value")
valueHandle.set(scope, ObjInt(10))
val value = valueHandle.get(scope)
```
### 7) Read variable values back in Kotlin
### 5) Read variable values back in Kotlin
The simplest approach: evaluate an expression that yields the value and convert it.
@ -318,7 +124,7 @@ val kotlinName = scope.eval("name").toKotlin(scope) // -> "Lyng rocks!"
Advanced: you can also grab a variable record directly via `scope.get(name)` and work with its `Obj` value, but evaluating `"name"` is often clearer and enforces Lyng semantics consistently.
### 8) Execute scripts with parameters; call Lyng functions from Kotlin
### 6) Execute scripts with parameters; call Lyng functions from Kotlin
There are two convenient patterns.
@ -351,7 +157,7 @@ val result = resultObj.toKotlin(scope) // -> 42
If you need to pass complex data (lists, maps), construct the corresponding Lyng `Obj` types (`ObjList`, `ObjMap`, etc.) and pass them in `Arguments`.
### 9) Create your own packages and import them in Lyng
### 7) Create your own packages and import them in Lyng
Lyng supports packages that are imported from scripts. You can register packages programmatically via `ImportManager` or by providing source texts that declare `package ...`.
@ -406,7 +212,7 @@ val s = scope.eval("s").toKotlin(scope) // -> 144
You can also register from parsed `Source` instances via `addSourcePackages(source)`.
### 10) Executing from files, security, and isolation
### 8) Executing from files, security, and isolation
- To run code from a file, read it and pass to `scope.eval(text)` or compile with `Compiler.compile(Source(fileName, text))`.
- `ImportManager` takes an optional `SecurityManager` if you need to restrict what packages or operations are available. By default, `Script.defaultImportManager` allows everything suitable for embedded use; clamp it down in sandboxed environments.
@ -417,7 +223,7 @@ You can also register from parsed `Source` instances via `addSourcePackages(sour
val isolated = net.sergeych.lyng.Scope.new()
```
### 11) Tips and troubleshooting
### 9) Tips and troubleshooting
- All values that cross the boundary must be Lyng `Obj` instances. Convert Kotlin values explicitly (e.g., `ObjInt`, `ObjReal`, `ObjString`).
- Use `toKotlin(scope)` to get Kotlin values back. Collections convert to Kotlin collections recursively.
@ -425,46 +231,6 @@ val isolated = net.sergeych.lyng.Scope.new()
- When registering packages, names must be unique. Register before you compile/evaluate scripts that import them.
- To debug scope content, `scope.toString()` and `scope.trace()` can help during development.
### 12) Handling and serializing exceptions
When Lyng code throws an exception, it is caught in Kotlin as an `ExecutionError`. This error wraps the actual Lyng `Obj` that was thrown (which could be a built-in `ObjException` or a user-defined `ObjInstance`).
To simplify handling these objects from Kotlin, several extension methods are provided on the `Obj` class. These methods work uniformly regardless of whether the exception is built-in or user-defined.
#### Uniform Exception API
| Method | Description |
| :--- | :--- |
| `obj.isLyngException()` | Returns `true` if the object is an instance of `Exception`. |
| `obj.isInstanceOf("ClassName")` | Returns `true` if the object is an instance of the named Lyng class or its ancestors. |
| `obj.getLyngExceptionMessage(scope?=null)` | Returns the exception message as a Kotlin `String`. |
| `obj.getLyngExceptionMessageWithStackTrace(scope?=null)` | Returns a detailed message with a formatted stack trace. |
| `obj.getLyngExceptionString(scope)` | Returns a formatted string including the class name, message, and primary throw site. |
| `obj.getLyngExceptionStackTrace(scope)` | Returns the stack trace as an `ObjList` of `StackTraceEntry`. |
| `obj.getLyngExceptionExtraData(scope)` | Returns the extra data associated with the exception. |
| `obj.raiseAsExecutionError(scope?=null)` | Rethrows the object as a Kotlin `ExecutionError`. |
#### Example: Serialization and Rethrowing
You can serialize Lyng exception objects using `Lynon` to transmit them across boundaries and then rethrow them.
```kotlin
try {
scope.eval("throw MyUserException(404, \"Not Found\")")
} catch (e: ExecutionError) {
// 1. Serialize the Lyng exception object
val encoded: UByteArray = lynonEncodeAny(scope, e.errorObject)
// ... (transmit 'encoded' byte array) ...
// 2. Deserialize it back to an Obj in a different context
val decoded: Obj = lynonDecodeAny(scope, encoded)
// 3. Properly rethrow it on the Kotlin side using the uniform API
decoded.raiseAsExecutionError(scope)
}
```
---
That’s it. You now have Lyng embedded in your Kotlin app: you can expose your app’s API, evaluate user scripts, and organize your own packages to import from Lyng code.

View File

@ -128,17 +128,15 @@ Serializable class that conveys information about the exception. Important membe
| name | description |
|-------------------|--------------------------------------------------------|
| message | String message |
| stackTrace() | lyng stack trace, list of `StackTraceEntry`, see below |
| printStackTrace() | format and print stack trace using println() |
> **Note for Kotlin users**: When working with Lyng exceptions from Kotlin, you can use extension methods like `getLyngExceptionMessageWithStackTrace()`. See [Embedding Lyng](embedding.md#12-handling-and-serializing-exceptions) for the full API.
| message | String message |
| stackTrace | lyng stack trace, list of `StackTraceEntry`, see below |
| printStackTrace() | format and print stack trace using println() |
## StackTraceEntry
A simple structire that stores single entry in Lyng stack, it is created automatically on exception creation:
```lyng
```kotlin
class StackTraceEntry(
val sourceName: String,
val line: Int,
@ -152,103 +150,24 @@ class StackTraceEntry(
# Custom error classes
You can define your own exception classes by inheriting from the built-in `Exception` class. This allows you to create specific error types for your application logic and catch them specifically.
## Defining a custom exception
To define a custom exception, create a class that inherits from `Exception`:
```lyng
class MyUserException : Exception("something went wrong")
```
You can also pass the message dynamically:
```lyng
class MyUserException(m) : Exception(m)
throw MyUserException("custom error message")
```
If you don't provide a message to the `Exception` constructor, the class name will be used as the default message:
```lyng
class SimpleException : Exception
val e = SimpleException()
assertEquals("SimpleException", e.message)
```
## Throwing and catching custom exceptions
Custom exceptions are thrown using the `throw` keyword and can be caught using `catch` blocks, just like standard exceptions:
```lyng
class ValidationException(m) : Exception(m)
try {
throw ValidationException("Invalid input")
}
catch(e: ValidationException) {
println("Caught validation error: " + e.message)
}
catch(e: Exception) {
println("Caught other exception: " + e.message)
}
```
Since user exceptions are real classes, inheritance works as expected:
```lyng
class BaseError : Exception
class DerivedError : BaseError
try {
throw DerivedError()
}
catch(e: BaseError) {
// This will catch DerivedError as well
assert(e is DerivedError)
}
```
## Accessing extra data
You can add your own fields to custom exception classes to carry additional information:
```lyng
class NetworkException(m, val statusCode) : Exception(m)
try {
throw NetworkException("Not Found", 404)
}
catch(e: NetworkException) {
println("Error " + e.statusCode + ": " + e.message)
}
```
_this functionality is not yet released_
# Standard exception classes
| class | notes |
|----------------------------|-------------------------------------------------------|
| Exception | root of all throwable objects |
| Exception | root of al throwable objects |
| NullReferenceException | |
| AssertionFailedException | |
| ClassCastException | |
| IndexOutOfBoundsException | |
| IllegalArgumentException | |
| IllegalStateException | |
| NoSuchElementException | |
| IllegalAssignmentException | assigning to val, etc. |
| SymbolNotDefinedException | |
| IterationEndException | attempt to read iterator past end, `hasNext == false` |
| IllegalAccessException | attempt to access private members or like |
| UnknownException | unexpected internal exception caught |
| NotFoundException | |
| IllegalOperationException | |
| UnsetException | access to uninitialized late-init val |
| NotImplementedException | used by `TODO()` |
| SyntaxError | |
| AccessException | attempt to access private members or like |
| UnknownException | unexpected kotlin exception caught |
| | |
### Symbol resolution errors

View File

@ -1,128 +0,0 @@
# Generics and type expressions
This document covers generics, bounds, unions/intersections, and the rules for type expressions in Lyng.
# Generic parameters
Declare type parameters with `<...>` on functions and classes:
fun id<T>(x: T): T = x
class Box<T>(val value: T)
Type arguments are usually inferred at call sites:
val b = Box(10) // Box<Int>
val s = id("ok") // T is String
# Bounds
Use `:` to set bounds. Bounds may be unions (`|`) or intersections (`&`):
fun sum<T: Int | Real>(x: T, y: T) = x + y
class Named<T: Iterable & Comparable>(val data: T)
Bounds are checked at compile time. For union bounds, the argument must fit at least one option. For intersection bounds, it must fit all options.
# Variance
Generic types are invariant by default. You can specify declaration-site variance:
class Source<out T>(val value: T)
class Sink<in T> { fun accept(x: T) { ... } }
`out` makes the type covariant (produced), `in` makes it contravariant (consumed).
# Type aliases
Type aliases name type expressions (including unions/intersections):
type Num = Int | Real
type AB = A & B
Aliases can be generic and can use bounds and defaults:
type Maybe<T> = T?
type IntList<T: Int> = List<T>
Aliases expand to their underlying type expressions. They can be used anywhere a type expression is expected.
# Inference rules
- Literals set obvious types (`1` is `Int`, `1.0` is `Real`, etc.).
- Empty list literals default to `List<Object>` unless constrained by context.
- Non-empty list literals infer element type as a union of element types.
- Map literals infer key and value types; named keys are `String`.
Examples:
val a = [1, 2, 3] // List<Int>
val b = [1, "two", true] // List<Int | String | Bool>
val c: List<Int> = [] // List<Int>
val m1 = { "a": 1, "b": 2 } // Map<String, Int>
val m2 = { "a": 1, "b": "x" } // Map<String, Int | String>
val m3 = { ...m1, "c": true } // Map<String, Int | Bool>
Map spreads carry key/value types when possible.
Spreads propagate element type when possible:
val base = [1, 2]
val mix = [...base, 3] // List<Int>
# Type expressions
Type expressions include simple types, generics, unions, and intersections:
Int
List<String>
Int | String
Iterable & Comparable
These type expressions can appear in casts and `is` checks.
# `is`, `in`, and `==` with type expressions
There are two categories of `is` checks:
1) Value checks: `x is T`
- `x` is a value, `T` is a type expression.
- This is a runtime instance check.
2) Type checks: `T1 is T2`
- both sides are type expressions (class objects or unions/intersections).
- This is a *type-subset* check: every value of `T1` must fit in `T2`.
Exact type expression equality uses `==` and is structural (union/intersection order does not matter).
Includes checks use `in` with type expressions:
A in T
This means `A` is a subset of `T` (the same relation as `A is T`).
Examples (T = A | B):
T == A // false
T is A // false
A in T // true
B in T // true
T is A | B // true
# Practical examples
fun acceptInts<T: Int>(xs: List<T>) { }
acceptInts([1, 2, 3])
// acceptInts([1, "a"]) -> compile-time error
fun f<T>(list: List<T>) {
assert( T is Int | String | Bool )
assert( !(T is Int) )
assert( Int in T )
}
f([1, "two", true])
# Notes
- `T` is reified as a type expression when needed (e.g., union/intersection). When it is a single class, `T` is that class object.
- Type expression checks are compile-time where possible; runtime checks only happen for `is` on values and explicit casts.

View File

@ -9,12 +9,9 @@ should be compatible with other IDEA flavors, notably [OpenIDE](https://openide.
- reformat code (indents, spaces)
- reformat on paste
- smart enter key
- `.lyng.d` definition files (merged into analysis for completion, navigation, Quick Docs, and error checking)
Features are configurable via the plugin settings page, in system settings.
See `docs/lyng_d_files.md` for `.lyng.d` syntax and examples.
> Recommended for IntelliJ-based IDEs: While IntelliJ can import TextMate bundles
> (Settings/Preferences → Editor → TextMate Bundles), the native Lyng plugin provides
> better support (formatting, smart enter, background analysis, etc.). Prefer installing
@ -29,4 +26,4 @@ See `docs/lyng_d_files.md` for `.lyng.d` syntax and examples.
### [Download plugin v0.0.2-SNAPSHOT](https://lynglang.com/distributables/lyng-idea-0.0.2-SNAPSHOT.zip)
Your ideas and bugreports are welcome on the [project gitea page](https://gitea.sergeych.net/SergeychWorks/lyng/issues)
Your ideas and bugreports are welcome on the [project gitea page](https://gitea.sergeych.net/SergeychWorks/lyng/issues)

View File

@ -20,18 +20,7 @@ Simple classes serialization is supported:
assertEquals( "{\"foo\":1,\"bar\":2}", Point(1,2).toJsonString() )
>>> void
Note that mutable members are serialized by default. You can exclude any member (including constructor parameters) from JSON serialization using the `@Transient` attribute:
import lyng.serialization
class Point2(@Transient val foo, val bar) {
@Transient var reason = 42
var visible = 100
}
assertEquals( "{\"bar\":2,\"visible\":100}", Point2(1,2).toJsonString() )
>>> void
Note that if you override json serialization:
Note that mutable members are serialized:
import lyng.serialization

View File

@ -8,7 +8,7 @@ This module provides a uniform, suspend-first filesystem API to Lyng scripts, ba
It exposes a Lyng class `Path` with methods for file and directory operations, including streaming readers for large files.
It is a separate library because access to the filesystem is a security risk we compensate with a separate API that user must explicitly include to the dependency and allow. Together with `FsAccessPolicy` that is required to `createFs()` which actually adds the filesystem to the scope, the security risk is isolated.
It is a separate library because access to teh filesystem is a security risk we compensate with a separate API that user must explicitly include to the dependency and allow. Together with `FsAceessPolicy` that is required to `createFs()` which actually adds the filesystem to the scope, the security risk is isolated.
Also, it helps keep Lyng core small and focused.
@ -23,7 +23,7 @@ dependencies {
implementation("net.sergeych:lyngio:0.0.1-SNAPSHOT")
}
```
Note on maven repository. Lyngio uses the same maven as Lyng code (`lynglib`) so it is most likely already in your project. If not, add it to the proper section of your `build.gradle.kts` or settings.gradle.kts:
Note on maven repository. Lyngio uses ths same maven as Lyng code (`lynglib`) so it is most likely already in your project. If ont, add it to the proper section of your `build.gradle.kts` or settings.gradle.kts:
```kotlin
repositories {
@ -43,13 +43,9 @@ This brings in:
The filesystem module is not installed automatically. You must explicitly register it in the scope’s `ImportManager` using the installer. You can customize access control via `FsAccessPolicy`.
Kotlin (host) bootstrap example:
Kotlin (host) bootstrap example (imports omitted for brevity):
```kotlin
import net.sergeych.lyng.Scope
import net.sergeych.lyng.io.fs.createFs
import net.sergeych.lyngio.fs.security.PermitAllAccessPolicy
val scope: Scope = Scope.new()
val installed: Boolean = createFs(PermitAllAccessPolicy, scope)
// installed == true on first registration in this ImportManager, false on repeats

View File

@ -1,136 +0,0 @@
### lyng.io.process — Process execution and control for Lyng scripts
This module provides a way to run external processes and shell commands from Lyng scripts. It is designed to be multiplatform and uses coroutines for non-blocking execution.
> **Note:** `lyngio` is a separate library module. It must be explicitly added as a dependency to your host application and initialized in your Lyng scopes.
---
#### Add the library to your project (Gradle)
If you use this repository as a multi-module project, add a dependency on `:lyngio`:
```kotlin
dependencies {
implementation("net.sergeych:lyngio:0.0.1-SNAPSHOT")
}
```
For external projects, ensure you have the appropriate Maven repository configured (see `lyng.io.fs` documentation).
---
#### Install the module into a Lyng Scope
The process module is not installed automatically. You must explicitly register it in the scope’s `ImportManager` using `createProcessModule`. You can customize access control via `ProcessAccessPolicy`.
Kotlin (host) bootstrap example:
```kotlin
import net.sergeych.lyng.Scope
import net.sergeych.lyng.Script
import net.sergeych.lyng.io.process.createProcessModule
import net.sergeych.lyngio.process.security.PermitAllProcessAccessPolicy
// ... inside a suspend function or runBlocking
val scope: Scope = Script.newScope()
createProcessModule(PermitAllProcessAccessPolicy, scope)
// In scripts (or via scope.eval), import the module:
scope.eval("import lyng.io.process")
```
---
#### Using from Lyng scripts
```lyng
import lyng.io.process
// Execute a process with arguments
val p = Process.execute("ls", ["-l", "/tmp"])
for (line in p.stdout) {
println("OUT: " + line)
}
val exitCode = p.waitFor()
println("Process exited with: " + exitCode)
// Run a shell command
val sh = Process.shell("echo 'Hello from shell' | wc -w")
for (line in sh.stdout) {
println("Word count: " + line.trim())
}
// Platform information
val details = Platform.details()
println("OS: " + details.name + " " + details.version + " (" + details.arch + ")")
if (details.kernelVersion != null) {
println("Kernel: " + details.kernelVersion)
}
if (Platform.isSupported()) {
println("Processes are supported!")
}
```
---
#### API Reference
##### `Process` (static methods)
- `execute(executable: String, args: List<String>): RunningProcess` — Start an external process.
- `shell(command: String): RunningProcess` — Run a command through the system shell (e.g., `/bin/sh` or `cmd.exe`).
##### `RunningProcess` (instance methods)
- `stdout: Flow` — Standard output stream as a Lyng Flow of lines.
- `stderr: Flow` — Standard error stream as a Lyng Flow of lines.
- `waitFor(): Int` — Wait for the process to exit and return the exit code.
- `signal(name: String)` — Send a signal to the process (e.g., `"SIGINT"`, `"SIGTERM"`, `"SIGKILL"`).
- `destroy()` — Forcefully terminate the process.
##### `Platform` (static methods)
- `details(): Map` — Get platform details. Returned map keys: `name`, `version`, `arch`, `kernelVersion`.
- `isSupported(): Bool` — True if process execution is supported on the current platform.
---
#### Security Policy
Process execution is a sensitive operation. `lyngio` uses `ProcessAccessPolicy` to control access to `execute` and `shell` operations.
- `ProcessAccessPolicy` — Interface for custom policies.
- `PermitAllProcessAccessPolicy` — Allows all operations.
- `ProcessAccessOp` (sealed) — Operations to check:
- `Execute(executable, args)`
- `Shell(command)`
Example of a restricted policy in Kotlin:
```kotlin
import net.sergeych.lyngio.fs.security.AccessDecision
import net.sergeych.lyngio.fs.security.Decision
import net.sergeych.lyngio.process.security.ProcessAccessOp
import net.sergeych.lyngio.process.security.ProcessAccessPolicy
val restrictedPolicy = object : ProcessAccessPolicy {
override suspend fun check(op: ProcessAccessOp, ctx: AccessContext): AccessDecision {
return when (op) {
is ProcessAccessOp.Execute -> {
if (op.executable == "ls") AccessDecision(Decision.Allow)
else AccessDecision(Decision.Deny, "Only 'ls' is allowed")
}
is ProcessAccessOp.Shell -> AccessDecision(Decision.Deny, "Shell is forbidden")
}
}
}
createProcessModule(restrictedPolicy, scope)
```
---
#### Platform Support
- **JVM:** Full support using `ProcessBuilder`.
- **Native (Linux/macOS):** Support via POSIX.
- **Windows:** Support planned.
- **Android/JS/iOS/Wasm:** Currently not supported; `isSupported()` returns `false` and attempts to run processes will throw `UnsupportedOperationException`.

View File

@ -1,116 +0,0 @@
# `.lyng.d` Definition Files
`.lyng.d` files declare Lyng symbols for tooling without shipping runtime implementations. The IntelliJ IDEA plugin merges
all `*.lyng.d` files from the current directory and its parent directories into the active file’s analysis, enabling:
- completion
- navigation
- error checking for declared symbols
- Quick Docs for declarations defined in `.lyng.d`
Place `*.lyng.d` files next to the code they describe (or in a parent folder). The plugin will pick them up automatically.
## Writing `.lyng.d` Files
You can declare any language-level symbol in a `.lyng.d` file. Use doc comments before declarations to make Quick Docs
work in the IDE. The doc parser accepts standard comments (`/** ... */` or `// ...`) and supports tags like `@param`.
### Full Example
```lyng
/** Library entry point */
extern fun connect(url: String, timeoutMs: Int = 5000): Client
/** Type alias with generics */
type NameMap = Map<String, String>
/** Multiple inheritance via interfaces */
interface A { abstract fun a(): Int }
interface B { abstract fun b(): Int }
/** A concrete class implementing both */
class Multi(name: String) : A, B {
/** Public field */
val id: Int = 0
/** Mutable property with accessors */
var size: Int
get() = 0
set(v) { }
/** Instance method */
fun a(): Int = 1
fun b(): Int = 2
}
/** Nullable and dynamic types */
extern val dynValue: dynamic
extern var dynVar: dynamic?
/** Delegated property */
class LazyBox(val create) {
fun getValue(thisRef, name) = create()
}
val cached by LazyBox { 42 }
/** Delegated function */
object RpcDelegate {
fun invoke(thisRef, name, args...) = Unset
}
fun remoteCall by RpcDelegate
/** Singleton object */
object Settings {
val version: String = "1.0"
}
/** Class with documented members */
class Client {
/** Returns a greeting. */
fun greet(name: String): String = "hi " + name
}
```
See a runnable sample file in `docs/samples/definitions.lyng.d`.
Notes:
- Use real bodies if the declaration is not `extern` or `abstract`.
- If you need purely declarative stubs, prefer `extern` members (see `embedding.md`).
## Doc Comment Format
Doc comments are picked up when they immediately precede a declaration.
```lyng
/**
* A sample function.
* @param name user name
* @return greeting string
*/
fun greet(name: String): String = "hi " + name
```
## Generating `.lyng.d` Files
You can generate `.lyng.d` as part of your build. A common approach is to write a Gradle task that emits a file from a
template or a Kotlin data model.
Example (pseudo-code):
```kotlin
tasks.register("generateLyngDefs") {
doLast {
val out = file("src/main/lyng/api.lyng.d")
out.writeText(
"""
/** Generated API */
fun ping(): Int
""".trimIndent()
)
}
}
```
Place the generated file in your source tree, and the IDE will load it automatically.

View File

@ -1,87 +0,0 @@
### lyngio — Extended I/O and System Library for Lyng
`lyngio` is a separate library that extends the Lyng core (`lynglib`) with powerful, multiplatform, and secure I/O capabilities.
#### Why a separate module?
1. **Security:** I/O and process execution are sensitive operations. By keeping them in a separate module, we ensure that the Lyng core remains 100% safe by default. You only enable what you explicitly need.
2. **Footprint:** Not every script needs filesystem or process access. Keeping these as a separate module helps minimize the dependency footprint for small embedded projects.
3. **Control:** `lyngio` provides fine-grained security policies (`FsAccessPolicy`, `ProcessAccessPolicy`) that allow you to control exactly what a script can do.
#### Included Modules
- **[lyng.io.fs](lyng.io.fs.md):** Async filesystem access. Provides the `Path` class for file/directory operations, streaming, and globbing.
- **[lyng.io.process](lyng.io.process.md):** External process execution and shell commands. Provides `Process`, `RunningProcess`, and `Platform` information.
---
#### Quick Start: Embedding lyngio
##### 1. Add Dependencies (Gradle)
```kotlin
repositories {
maven("https://gitea.sergeych.net/api/packages/SergeychWorks/maven")
}
dependencies {
// Both are required for full I/O support
implementation("net.sergeych:lynglib:0.0.1-SNAPSHOT")
implementation("net.sergeych:lyngio:0.0.1-SNAPSHOT")
}
```
##### 2. Initialize in Kotlin (JVM or Native)
To use `lyngio` modules in your scripts, you must install them into your Lyng scope and provide a security policy.
```kotlin
import net.sergeych.lyng.Script
import net.sergeych.lyng.io.fs.createFs
import net.sergeych.lyng.io.process.createProcessModule
import net.sergeych.lyngio.fs.security.PermitAllAccessPolicy
import net.sergeych.lyngio.process.security.PermitAllProcessAccessPolicy
suspend fun runMyScript() {
val scope = Script.newScope()
// Install modules with policies
createFs(PermitAllAccessPolicy, scope)
createProcessModule(PermitAllProcessAccessPolicy, scope)
// Now scripts can import them
scope.eval("""
import lyng.io.fs
import lyng.io.process
println("Working dir: " + Path(".").readUtf8())
println("OS: " + Platform.details().name)
""")
}
```
---
#### Security Tools
`lyngio` is built with a "Secure by Default" philosophy. Every I/O or process operation is checked against a policy.
- **Filesystem Security:** Implement `FsAccessPolicy` to restrict access to specific paths or operations (e.g., read-only access to a sandbox directory).
- **Process Security:** Implement `ProcessAccessPolicy` to restrict which executables can be run or to disable shell execution entirely.
For more details, see the specific module documentation:
- [Filesystem Security Details](lyng.io.fs.md#access-policy-security)
- [Process Security Details](lyng.io.process.md#security-policy)
---
#### Platform Support Overview
| Platform | lyng.io.fs | lyng.io.process |
| :--- | :---: | :---: |
| **JVM** | ✅ | ✅ |
| **Native (Linux/macOS)** | ✅ | ✅ |
| **Native (Windows)** | ✅ | 🚧 (Planned) |
| **Android** | ✅ | ❌ |
| **NodeJS** | ✅ | ❌ |
| **Browser / Wasm** | ✅ (In-memory) | ❌ |

View File

@ -92,7 +92,6 @@ or transformed `Real` otherwise.
| pow(x, y) | ${x^y}$ |
| sqrt(x) | $ \sqrt {x}$ |
| abs(x) | absolute value of x. Int if x is Int, Real otherwise |
| clamp(x, range) | limit x to be inside range boundaries |
For example:
@ -103,11 +102,6 @@ For example:
// abs() keeps the argument type:
assert( abs(-1) is Int)
assert( abs(-2.21) == 2.21 )
// clamp() limits value to the range:
assert( clamp(15, 0..10) == 10 )
assert( clamp(-5, 0..10) == 0 )
assert( 5.clamp(0..10) == 5 )
>>> void
## Scientific constant

View File

@ -1,25 +0,0 @@
# Migration of Instant and Clock
## History
Before kotlin 2.0, there was an excellent library, kotlinx.datetime, which was widely used everywhere, also in Lyng and its dependencies.
When Kotlin 2.0 was released, or soon after, JetBrains made a perplexing decision to remove `Instant` and `Clock` from kotlinx.datetime and replace it with _yet experimental_ analogs in `kotlin.time`.
The problem is, these were not quite the same (these weren't `@Serializable`!), so people didn't migrate with ease. Okay, then JetBrains decided to not only deprecate it but also make them unusable on Apple targets. It sort of split auditories of many published libraries to those who hate JetBrains and Apple and continue to use 1.9-2.0 compatible versions that no longer work with Kotlin 2.2 on Apple targets (but work pretty well with earlier Kotlin or on other platforms).
Later JetBrains added serializers for their new `Instant` and `Clock` types, but strangely not in the stdlib, but in newer versions of `kotlinx.serialization`. This means that plain upgrade of dependencies to 2.2 is not enough to make them work.
## Solution
We hereby publish a new version of Lyng, 1.0.8-SNAPSHOT, which uses `kotlin.time.Instant` and `kotlin.time.Clock` instead of `kotlinx.datetime.Instant` and `kotlinx.datetime.Clock`. It is in other aspects compatible also with Lynon encoded binaries. You might need to migrate your code to use `kotlin.time` types. (LocalDateTime/TimeZone still come from `kotlinx.datetime`.)
So, if you are getting errors with new version, please do:
- upgrade to Kotlin 2.2
- upgrade to Lyng 1.0.8-SNAPSHOT
- replace in your code imports (or other uses) of `kotlinx.datetime.Clock` to `kotlin.time.Clock` and `kotlinx.datetime.Instant` to `kotlin.time.Instant`.
This should solve the problem and hopefully we'll see no more such "brilliant" ideas from IDEA ideologspersons.
Sorry for inconvenience and send a ray of hate to JetBrains ;)

View File

@ -49,7 +49,7 @@ Suppose we have a resource, that could be used concurrently, a counter in our ca
delay(100)
counter = c + 1
}
}.forEach { (it as Deferred).await() }
}.forEach { it.await() }
assert(counter < 50) { "counter is "+counter }
>>> void
@ -64,12 +64,13 @@ Using [Mutex] makes it all working:
launch {
// slow increment:
mutex.withLock {
val c = counter ?: 0
val c = counter
delay(10)
counter = c + 1
}
}
}.forEach { (it as Deferred).await() }
assert(counter in 1..4)
}.forEach { it.await() }
assertEquals(4, counter)
>>> void
now everything works as expected: `mutex.withLock` makes them all be executed in sequence, not in parallel.
@ -223,14 +224,19 @@ Future work: introduce thread‑safe pooling (e.g., per‑thread pools or confin
### Closures inside coroutine helpers (launch/flow)
Closures executed by `launch { ... }` and `flow { ... }` use **compile‑time resolution** just like any other Lyng code:
Closures executed by `launch { ... }` and `flow { ... }` resolve names using the `ClosureScope` rules:
- **Captured locals are slots**: outer locals are resolved at compile time and captured as frame‑slot references, so they remain visible across suspension points.
- **Members are statically resolved**: member access requires a statically known receiver type or an explicit cast (except `Object` members).
- **No runtime fallbacks**: there is no dynamic name lookup or “search parent scopes” at runtime for missing symbols.
1. Closure frame locals/arguments
2. Captured receiver instance/class members
3. Closure ancestry locals + each frame’s `this` members (cycle‑safe)
4. Caller `this` members
5. Caller ancestry locals + each frame’s `this` members (cycle‑safe)
6. Module pseudo‑symbols (e.g., `__PACKAGE__`)
7. Direct module/global fallback (nearest `ModuleScope` and its parent/root)
Implications:
- Global helpers like `delay(ms)` and `yield()` must be imported/known at compile time.
- If you need dynamic access, use explicit helpers (e.g., `dynamic { ... }`) rather than relying on scope resolution.
- Outer locals (e.g., `counter`) stay visible across suspension points.
- Global helpers like `delay(ms)` and `yield()` are available from inside closures.
- If you write your own async helpers, execute user lambdas under `ClosureScope(callScope, capturedCreatorScope)` and avoid manual ancestry walking.
See also: [Scopes and Closures: compile-time resolution](scopes_and_closures.md)
See also: [Scopes and Closures: resolution and safety](scopes_and_closures.md)

View File

@ -33,7 +33,7 @@ PerfProfiles.restore(snap) // restore previous flags
- `ARG_BUILDER` — Efficient argument building: small‑arity no‑alloc and pooled builder on JVM (ON JVM default).
- `ARG_SMALL_ARITY_12` — Extends small‑arity no‑alloc call paths from 0–8 to 0–12 arguments (JVM‑first exploration; OFF by default). Use for codebases with many 9–12 arg calls; A/B before enabling.
- `SKIP_ARGS_ON_NULL_RECEIVER` — Early return on optional‑null receivers before building args (semantics‑compatible). A/B only.
- `SCOPE_POOL` — Scope frame pooling for calls (per‑thread ThreadLocal pool on JVM/Android/Native; global deque on JS/Wasm). ON by default on all platforms; togglable at runtime.
- `SCOPE_POOL` — Scope frame pooling for calls (JVM, per‑thread ThreadLocal pool). ON by default on JVM; togglable at runtime.
- `FIELD_PIC` — 2‑entry polymorphic inline cache for field reads/writes keyed by `(classId, layoutVersion)` (ON JVM default).
- `METHOD_PIC` — 2‑entry PIC for instance method calls keyed by `(classId, layoutVersion)` (ON JVM default).
- `FIELD_PIC_SIZE_4` — Increases Field PIC size from 2 to 4 entries (JVM-first tuning; OFF by default). Use for sites with >2 receiver shapes.

View File

@ -1,86 +0,0 @@
# The `return` statement
The `return` statement is used to terminate the execution of the innermost enclosing callable (a function or a lambda) and optionally return a value to the caller.
## Basic Usage
By default, Lyng functions and blocks return the value of their last expression. However, `return` allows you to exit early, which is particularly useful for guard clauses.
```lyng
fun divide(a, b) {
if (b == 0) return null // Guard clause: early exit
a / b
}
```
If no expression is provided, `return` returns `void`:
```lyng
fun logIfDebug(msg) {
if (!DEBUG) return
println("[DEBUG] " + msg)
}
```
## Scoping Rules
In Lyng, `return` always exits the **innermost enclosing callable**. Callables include:
* Named functions (`fun` or `fn`)
* Anonymous functions/lambdas (`{ ... }`)
Standard control flow blocks like `if`, `while`, `do`, and `for` are **not** callables; `return` inside these blocks will return from the function or lambda that contains them.
```lyng
fun findFirstPositive(list) {
list.forEach {
if (it > 0) return it // ERROR: This returns from the lambda, not findFirstPositive!
}
null
}
```
*Note: To return from an outer scope, use [Non-local Returns](#non-local-returns).*
## Non-local Returns
Lyng supports returning from outer scopes using labels. This is a powerful feature for a closure-intensive language.
### Named Functions as Labels
Every named function automatically provides its name as a label.
```lyng
fun findFirstPositive(list) {
list.forEach {
if (it > 0) return@findFirstPositive it // Returns from findFirstPositive
}
null
}
```
### Labeled Lambdas
You can explicitly label a lambda using the `@label` syntax to return from it specifically when nested.
```lyng
val process = @outer { x ->
val result = {
if (x < 0) return@outer "negative" // Returns from the outer lambda
x * 2
}()
"Result: " + result
}
```
## Restriction on Shorthand Functions
To maintain Lyng's clean, expression-oriented style, the `return` keyword is **forbidden** in shorthand function definitions (those using `=`).
```lyng
fun square(x) = x * x // Correct
fun square(x) = return x * x // Syntax Error: 'return' not allowed here
```
## Summary
* `return [expression]` exits the innermost `fun` or `{}`.
* Use `return@label` for non-local returns.
* Named functions provide automatic labels.
* Cannot be used in `=` shorthand functions.
* Consistency: Mirrors the syntax and behavior of `break@label expression`.

View File

@ -1,65 +0,0 @@
/**
* Sample .lyng.d file for IDE support.
* Demonstrates declarations and doc comments.
*/
/** Simple function with default and named parameters. */
extern fun connect(url: String, timeoutMs: Int = 5000): Client
/** Type alias with generics. */
type NameMap = Map<String, String>
/** Multiple inheritance via interfaces. */
interface A { abstract fun a(): Int }
interface B { abstract fun b(): Int }
/** A concrete class implementing both. */
class Multi(name: String) : A, B {
/** Public field. */
val id: Int = 0
/** Mutable property with accessors. */
var size: Int
get() = 0
set(v) { }
/** Instance method. */
fun a(): Int = 1
fun b(): Int = 2
}
/** Nullable and dynamic types. */
extern val dynValue: dynamic
extern var dynVar: dynamic?
/** Delegated property provider. */
class LazyBox(val create) {
fun getValue(thisRef, name) = create()
}
/** Delegated property using provider. */
val cached by LazyBox { 42 }
/** Delegated function. */
object RpcDelegate {
fun invoke(thisRef, name, args...) = Unset
}
/** Remote function proxy. */
fun remoteCall by RpcDelegate
/** Singleton object. */
object Settings {
/** Version string. */
val version: String = "1.0"
}
/**
* Client API entry.
* @param name user name
* @return greeting string
*/
class Client {
/** Returns a greeting. */
fun greet(name: String): String = "hi " + name
}

View File

@ -1,6 +1,7 @@
#!/bin/env lyng
import lyng.io.fs
import lyng.stdlib
val files = Path("../..").list().toList()
// most long is longest?

View File

@ -4,21 +4,15 @@
test the Lyng way. It is not meant to be effective.
*/
fun naiveCountHappyNumbers(): Int {
fun naiveCountHappyNumbers() {
var count = 0
for( n1 in 0..9 ) {
for( n2 in 0..9 ) {
for( n3 in 0..9 ) {
for( n4 in 0..9 ) {
for( n5 in 0..9 ) {
for( n6 in 0..9 ) {
for( n1 in 0..9 )
for( n2 in 0..9 )
for( n3 in 0..9 )
for( n4 in 0..9 )
for( n5 in 0..9 )
for( n6 in 0..9 )
if( n1 + n2 + n3 == n4 + n5 + n6 ) count++
}
}
}
}
}
}
count
}
@ -34,3 +28,4 @@ for( r in 1..900 ) {
assert( found == 55252 )
delay(0.05)
}

View File

@ -1,63 +0,0 @@
// Sample: Operator Overloading in Lyng
class Vector<T>(val x: T, val y: T) {
// Overload +
fun plus(other: Vector<U>) = Vector(x + other.x, y + other.y)
// Overload -
fun minus(other: Vector<U>) = Vector(x - other.x, y - other.y)
// Overload unary -
fun negate() = Vector(-x, -y)
// Overload ==
fun equals(other) {
if (other is Vector<U>) x == other.x && y == other.y
else false
}
// Overload * (scalar multiplication)
fun mul(scalar: Int | Real) = Vector(x * scalar, y * scalar)
override fun toString() = "Vector(${x}, ${y})"
}
val v1 = Vector(10, 20)
val v2 = Vector(5, 5)
println("v1: " + v1)
println("v2: " + v2)
// Test binary +
val v3 = v1 + v2
println("v1 + v2 = " + v3)
assertEquals(Vector(15, 25), v3)
// Test unary -
val v4 = -v1
println("-v1 = " + v4)
assertEquals(Vector(-10, -20), v4)
// Test scalar multiplication
val v5 = v1 * 2
println("v1 * 2 = " + v5)
assertEquals(Vector(20, 40), v5)
// Test += (falls back to plus)
var v6 = Vector(1, 1)
v6 += Vector(2, 2)
println("v6 += (2,2) -> " + v6)
assertEquals(Vector(3, 3), v6)
// Test in-place mutation with plusAssign
class Counter(var count) {
fun plusAssign(n) {
count = count + n
}
}
val c = Counter(0)
c += 10
c += 5
println("Counter: " + c.count)
assertEquals(15, c.count)

View File

@ -13,6 +13,7 @@ fun findSumLimit(f) {
println("limit reached after "+n+" rounds")
break sum
}
n++
}
else {
println("limit not reached")

View File

@ -1,18 +1,75 @@
# Scopes and Closures: compile-time resolution
# Scopes and Closures: resolution and safety
Attention to AI: name lookup in runtime `Scope` is legacy. The bytecode compiler uses **compile-time name/member resolution only**.
This page documents how name resolution works with `ClosureScope`, how to avoid recursion pitfalls, and how to safely capture and execute callbacks that need access to outer locals.
This page documents the **current** rules: static name resolution, closure captures, and the limited role of runtime `Scope` in Kotlin interop and explicit dynamic helpers.
## Why this matters
Name lookup across nested scopes and closures can accidentally form recursive resolution paths or hide expected symbols (outer locals, module/global functions). The rules below ensure predictable resolution and prevent infinite recursion.
## Current rules (bytecode compiler)
- **All names resolve at compile time**: locals, parameters, captures, members, imports, and module globals must be known when compiling. Missing symbols are compile-time errors.
- **No runtime fallbacks**: there is no dynamic name lookup, no fallback opcodes, and no “search parent scopes” at runtime for missing names.
- **Object members on unknown types only**: `toString`, `toInspectString`, `let`, `also`, `apply`, `run` are allowed on unknown types; all other members require a statically known receiver type or an explicit cast.
- **Closures capture slots**: lambdas and nested functions capture **frame slots** directly. Captures are resolved at compile time and compiled to slot references.
- **Scope is a reflection facade**: `Scope` is used only for Kotlin interop or explicit dynamic helpers. It must **not** be used for general symbol resolution in compiled Lyng code.
## Resolution order in ClosureScope
When evaluating an identifier `name` inside a closure, `ClosureScope.get(name)` resolves in this order:
## Explicit dynamic access (opt-in only)
Dynamic name access is available only via explicit helpers (e.g., `dynamic { get { name -> ... } }`). It is **not** a fallback for normal member or variable access.
1. Closure frame locals and arguments
2. Captured receiver (`closureScope.thisObj`) instance/class members
3. Closure ancestry locals + each frame’s `thisObj` members (cycle‑safe)
4. Caller `this` members
5. Caller ancestry locals + each frame’s `thisObj` members (cycle‑safe)
6. Module pseudo‑symbols (e.g., `__PACKAGE__`) from the nearest `ModuleScope`
7. Direct module/global fallback (nearest `ModuleScope` and its parent/root scope)
8. Final fallback: base local/parent lookup for the current frame
## Legacy interpreter behavior (reference only)
The old runtime `Scope`-based resolution order (locals → captured → `this` → caller → globals) is obsolete for bytecode compilation. Keep it only for legacy interpreter paths and tooling that explicitly opts into it.
This preserves intuitive visibility (locals → captured receiver → closure chain → caller members → caller chain → module/root) while preventing infinite recursion between scope types.
## Use raw‑chain helpers for ancestry walks
When authoring new scope types or advanced lookups, avoid calling virtual `get` while walking parents. Instead, use the non‑dispatch helpers on `Scope`:
- `chainLookupIgnoreClosure(name)`
- Walk raw `parent` chain and check only per‑frame locals/bindings/slots.
- Ignores overridden `get` (e.g., in `ClosureScope`). Cycle‑safe.
- `chainLookupWithMembers(name)`
- Like above, but after locals/bindings it also checks each frame’s `thisObj` members.
- Ignores overridden `get`. Cycle‑safe.
- `baseGetIgnoreClosure(name)`
- For the current frame only: check locals/bindings, then walk raw parents (locals/bindings), then fallback to this frame’s `thisObj` members.
These helpers avoid ping‑pong recursion and make structural cycles harmless (lookups terminate).
## Preventing structural cycles
- Don’t construct parent chains that can point back to a descendant.
- A debug‑time guard throws if assigning a parent would create a cycle; keep it enabled for development builds.
- Even with a cycle, chain helpers break out via a small `visited` set keyed by `frameId`.
## Capturing lexical environments for callbacks
For dynamic objects or custom builders, capture the creator’s lexical scope so callbacks can see outer locals/parameters:
1. Use `snapshotForClosure()` on the caller scope to capture locals/bindings/slots and parent.
2. Store this snapshot and run callbacks under `ClosureScope(callScope, captured)`.
Kotlin sketch:
```kotlin
val captured = scope.snapshotForClosure()
val execScope = ClosureScope(currentCallScope, captured)
callback.execute(execScope)
```
This ensures expressions like `contractName` used inside dynamic `get { name -> ... }` resolve to outer variables defined at the creation site.
## Closures in coroutines (launch/flow)
- The closure frame still prioritizes its own locals/args.
- Outer locals declared before suspension points remain visible through slot‑aware ancestry lookups.
- Global functions like `delay(ms)` and `yield()` are resolved via module/root fallbacks from within closures.
Tip: If a closure unexpectedly cannot see an outer local, check whether an intermediate runtime helper introduced an extra call frame; the built‑in lookup already traverses caller ancestry, so prefer the standard helpers rather than custom dispatch.
## Local variable references and missing symbols
- Unqualified identifier resolution first prefers locals/bindings/slots before falling back to `this` members.
- If neither locals nor members contain the symbol, missing field lookups map to `SymbolNotFound` (compatibility alias for `SymbolNotDefinedException`).
## Performance notes
- The `visited` sets used for cycle detection are tiny and short‑lived; in typical scripts the overhead is negligible.
- If profiling shows hotspots, consider limiting ancestry depth in your custom helpers or using small fixed arrays instead of hash sets—only for extremely hot code paths.
## Dos and Don’ts
- Do use `chainLookupIgnoreClosure` / `chainLookupWithMembers` for ancestry traversals.
- Do maintain the resolution order above for predictable behavior.
- Don’t call virtual `get` while walking parents; it risks recursion across scope types.
- Don’t attach instance scopes to transient/pool frames; bind to a stable parent scope instead.

View File

@ -17,40 +17,23 @@ It is as simple as:
assertEquals( text, Lynon.decode(encodedBits) )
// compression was used automatically
assert( text.length > (encodedBits.toBuffer() as Buffer).size )
assert( text.length > encodedBits.toBuffer().size )
>>> void
Any class you create is serializable by default; lynon serializes first constructor fields, then any `var` member fields.
Any class you create is serializable by default; lynon serializes first constructor fields, then any `var` member fields:
## Transient Fields
import lyng.serialization
Sometimes you have fields that should not be serialized, for example, temporary caches, secret data, or derived values that are recomputed in `init` blocks. You can mark such fields with the `@Transient` attribute:
class Point(x,y)
```lyng
class MyData(@Transient val tempSecret, val publicData) {
@Transient var cachedValue = 0
var persistentValue = 42
val p = Lynon.decode( Lynon.encode( Point(5,6) ) )
init {
// cachedValue can be recomputed here upon deserialization
cachedValue = computeCache(publicData)
}
}
```
assertEquals( 5, p.x )
assertEquals( 6, p.y )
>>> void
Transient fields:
- Are **omitted** from Lynon binary streams.
- Are **omitted** from JSON output (via `toJson`).
- Are **ignored** during structural equality checks (`==`).
- If a transient constructor parameter has a **default value**, it will be restored to that default value during deserialization. Otherwise, it will be `null`.
- Class body fields marked as `@Transient` will keep their initial values (or values assigned in `init`) after deserialization.
## Serialization of Objects and Classes
- **Singleton Objects**: `object` declarations are serializable by name. Their state (mutable fields) is also serialized and restored, respecting `@Transient`.
- **Classes**: Class objects themselves can be serialized. They are serialized by their full qualified name. When converted to JSON, a class object includes its public static fields (excluding those marked `@Transient`).
## Custom Serialization
just as expected.
Important is to understand that normally `Lynon.decode` wants [BitBuffer], as `Lynon.encode` produces. If you have the regular [Buffer], be sure to convert it:

View File

@ -2,105 +2,166 @@
Lyng date and time support requires importing `lyng.time` packages. Lyng uses simple yet modern time object models:
- `Instant` class for absolute time stamps with platform-dependent resolution.
- `DateTime` class for calendar-aware points in time within a specific time zone.
- `Duration` to represent amount of time not depending on the calendar (e.g., milliseconds, seconds).
- `Instant` class for time stamps with platform-dependent resolution
- `Duration` to represent amount of time not depending on the calendar, e.g. in absolute units (milliseconds, seconds,
hours, days)
## Time instant: `Instant`
Represent some moment of time not depending on the calendar. It is similar to `TIMESTAMP` in SQL or `Instant` in Kotlin.
Represent some moment of time not depending on the calendar (calendar for example may b e changed, daylight saving time
can be for example introduced or dropped). It is similar to `TIMESTAMP` in SQL or `Instant` in Kotlin. Some moment of
time; not the calendar date.
### Constructing and converting
Instant is comparable to other Instant. Subtracting instants produce `Duration`, period in time that is not dependent on
the calendar, e.g. absolute time period.
It is possible to add or subtract `Duration` to and from `Instant`, that gives another `Instant`.
Instants are converted to and from `Real` number of seconds before or after Unix Epoch, 01.01.1970. Constructor with
single number parameter constructs from such number of seconds,
and any instance provide `.epochSeconds` member:
import lyng.time
// default constructor returns time now:
val t1 = Instant()
// constructing from a number is treated as seconds since unix epoch:
val t2 = Instant(1704110400) // 2024-01-01T12:00:00Z
// from RFC3339 string:
val t3 = Instant("2024-01-01T12:00:00.123456Z")
// truncation:
val t4 = t3.truncateToMinute
assertEquals(t4.toRFC3339(), "2024-01-01T12:00:00Z")
// to localized DateTime (uses system default TZ if not specified):
val dt = t3.toDateTime("+02:00")
assertEquals(dt.hour, 14)
val t2 = Instant()
assert( t2 - t1 < 1.millisecond )
assert( t2.epochSeconds - t1.epochSeconds < 0.001 )
>>> void
### Instant members
## Constructing
import lyng.time
// empty constructor gives current time instant using system clock:
val now = Instant()
// constructor with Instant instance makes a copy:
assertEquals( now, Instant(now) )
// constructing from a number is trated as seconds since unix epoch:
val copyOfNow = Instant( now.epochSeconds )
// note that instant resolution is higher that Real can hold
// so reconstructed from real slightly differs:
assert( abs( (copyOfNow - now).milliseconds ) < 0.01 )
>>> void
The resolution of system clock could be more precise and double precision real number of `Real`, keep it in mind.
## Comparing and calculating periods
import lyng.time
val now = Instant()
// you cam add or subtract periods, and compare
assert( now - 5.minutes < now )
val oneHourAgo = now - 1.hour
assertEquals( now, oneHourAgo + 1.hour)
>>> void
## Getting the max precision
Normally, subtracting instants gives precision to microseconds, which is well inside the jitter
the language VM adds. Still `Instant()` or `Instant.now()` capture most precise system timer at hand and provide inner
value of 12 bytes, up to nanoseconds (hopefully). To access it use:
import lyng.time
// capture time
val now = Instant.now()
// this is Int value, number of whole epoch
// milliseconds to the moment, it fits 8 bytes Int well
val seconds = now.epochWholeSeconds
assert(seconds is Int)
// and this is Int value of nanoseconds _since_ the epochMillis,
// it effectively add 4 more mytes int:
val nanos = now.nanosecondsOfSecond
assert(nanos is Int)
assert( nanos in 0..999_999_999 )
// we can construct epochSeconds from these parts:
assertEquals( now.epochSeconds, nanos * 1e-9 + seconds )
>>> void
## Truncating to more realistic precision
Full precision Instant is way too long and impractical to store, especially when serializing,
so it is possible to truncate it to milliseconds, microseconds or seconds:
import lyng.time
import lyng.serialization
// max supported size (now microseconds for serialized value):
// note that encoding return _bit array_ and this is a _bit size_:
val s0 = Lynon.encode(Instant.now()).size
// shorter: milliseconds only
val s1 = Lynon.encode(Instant.now().truncateToMillisecond()).size
// truncated to seconds, good for file mtime, etc:
val s2 = Lynon.encode(Instant.now().truncateToSecond()).size
assert( s1 < s0 )
assert( s2 < s1 )
>>> void
## Formatting instants
You can freely use `Instant` in string formatting. It supports usual sprintf-style formats:
import lyng.time
val now = Instant()
// will be something like "now: 12:10:05"
val currentTimeOnly24 = "now: %tT"(now)
// we can extract epoch second with formatting too,
// this was since early C time
// get epoch while seconds from formatting
val unixEpoch = "Now is %ts since unix epoch"(now)
// and it is the same as now.epochSeconds, int part:
assertEquals( unixEpoch, "Now is %d since unix epoch"(now.epochSeconds.toInt()) )
>>> void
See
the [complete list of available formats](https://github.com/sergeych/mp_stools?tab=readme-ov-file#datetime-formatting)
and the [formatting reference](https://github.com/sergeych/mp_stools?tab=readme-ov-file#printf--sprintf): it all works
in Lyng as `"format"(args...)`!
## Instant members
| member | description |
|--------------------------------|---------------------------------------------------------|
| epochSeconds: Real | positive or negative offset in seconds since Unix epoch |
| epochWholeSeconds: Int | same, but in _whole seconds_. Slightly faster |
| nanosecondsOfSecond: Int | offset from epochWholeSeconds in nanos |
| nanosecondsOfSecond: Int | offset from epochWholeSeconds in nanos (1) |
| isDistantFuture: Bool | true if it `Instant.distantFuture` |
| isDistantPast: Bool | true if it `Instant.distantPast` |
| truncateToMinute: Instant | create new instance truncated to minute |
| truncateToSecond: Instant | create new instance truncated to second |
| truncateToMillisecond: Instant | truncate new instance to millisecond |
| truncateToSecond: Intant | create new instnce truncated to second |
| truncateToMillisecond: Instant | truncate new instance with to millisecond |
| truncateToMicrosecond: Instant | truncate new instance to microsecond |
| toRFC3339(): String | format as RFC3339 string (UTC) |
| toDateTime(tz?): DateTime | localize to a TimeZone (ID string or offset seconds) |
## Calendar time: `DateTime`
(1)
: The value of nanoseconds is to be added to `epochWholeSeconds` to get exact time point. It is in 0..999_999_999 range.
The precise time instant value therefore needs as for now 12 bytes integer; we might use bigint later (it is planned to
be added)
`DateTime` represents a point in time in a specific timezone. It provides access to calendar components like year,
month, and day.
## Class members
### Constructing
| member | description |
|--------------------------------|----------------------------------------------|
| Instant.now() | create new instance with current system time |
| Instant.distantPast: Instant | most distant instant in past |
| Instant.distantFuture: Instant | most distant instant in future |
import lyng.time
// Current time in system default timezone
val now = DateTime.now()
// Specific timezone
val offsetTime = DateTime.now("+02:00")
// From Instant
val dt = Instant().toDateTime("Z")
// By components (year, month, day, hour=0, minute=0, second=0, timeZone="UTC")
val dt2 = DateTime(2024, 1, 1, 12, 0, 0, "Z")
// From RFC3339 string
val dt3 = DateTime.parseRFC3339("2024-01-01T12:00:00+02:00")
### DateTime members
| member | description |
|----------------------------------|-----------------------------------------------|
| year: Int | year component |
| month: Int | month component (1..12) |
| day: Int | day of month (alias `dayOfMonth`) |
| hour: Int | hour component (0..23) |
| minute: Int | minute component (0..59) |
| second: Int | second component (0..59) |
| dayOfWeek: Int | day of week (1=Monday, 7=Sunday) |
| timeZone: String | timezone ID string |
| toInstant(): Instant | convert back to absolute Instant |
| toUTC(): DateTime | shortcut to convert to UTC |
| toTimeZone(tz): DateTime | convert to another timezone |
| addMonths(n): DateTime | add/subtract months (normalizes end of month) |
| addYears(n): DateTime | add/subtract years |
| toRFC3339(): String | format with timezone offset |
| static now(tz?): DateTime | create DateTime with current time |
| static parseRFC3339(s): DateTime | parse RFC3339 string |
### Arithmetic and normalization
`DateTime` handles calendar arithmetic correctly:
val leapDay = Instant("2024-02-29T12:00:00Z").toDateTime("Z")
val nextYear = leapDay.addYears(1)
assertEquals(nextYear.day, 28) // Feb 29, 2024 -> Feb 28, 2025
# `Duration` class
# `Duraion` class
Represent absolute time distance between two `Instant`.

View File

@ -8,9 +8,7 @@ __Other documents to read__ maybe after this one:
- [Advanced topics](advanced_topics.md), [declaring arguments](declaring_arguments.md), [Scopes and Closures](scopes_and_closures.md)
- [OOP notes](OOP.md), [exception handling](exceptions_handling.md)
- [math in Lyng](math.md), [the `when` statement](when.md), [return statement](return_statement.md)
- [Testing and Assertions](Testing.md)
- [Generics and type expressions](generics.md)
- [math in Lyng](math.md), [the `when` statement](when.md)
- [time](time.md) and [parallelism](parallelism.md)
- [parallelism] - multithreaded code, coroutines, etc.
- Some class
@ -33,15 +31,6 @@ any block also returns it's last expression:
}
>>> 6
If you want to exit a function or lambda earlier, use the `return` statement:
fn divide(a, b) {
if( b == 0 ) return null
a / b
}
See [return statement](return_statement.md) for more details on scoping and non-local returns.
If you don't want block to return anything, use `void`:
fn voidFunction() {
@ -97,44 +86,6 @@ Lyng supports simple enums for a fixed set of named constants. Declare with `enu
For more details (usage patterns, `when` switching, serialization), see OOP notes: [Enums in detail](OOP.md#enums).
## Singleton Objects
Singleton objects are declared using the `object` keyword. They define a class and create its single instance immediately.
object Logger {
fun log(msg) { println("[LOG] " + msg) }
}
Logger.log("Hello singleton!")
## Nested Declarations (short)
Classes, objects, and enums can be declared inside another class. They live in the class namespace (no outer instance capture), so you access them with a qualifier:
class A {
class B(x?)
object Inner { val foo = "bar" }
enum E* { One, Two }
}
val ab = A.B()
assertEquals(ab.x, null)
assertEquals(A.Inner.foo, "bar")
assertEquals(A.One, A.E.One)
See [OOP notes](OOP.md#nested-declarations) for rules, visibility, and enum lifting details.
## Delegation (briefly)
You can delegate properties and functions to other objects using the `by` keyword. This is perfect for patterns like `lazy` initialization.
val expensiveData by lazy {
// computed only once on demand
"computed"
}
For more details on these features, see [Delegation in Lyng](delegation.md) and [OOP notes](OOP.md).
When putting multiple statments in the same line it is convenient and recommended to use `;`:
var from; var to
@ -155,41 +106,6 @@ Assignemnt is an expression that changes its lvalue and return assigned value:
>>> 11
>>> 6
### Destructuring assignments
Lyng supports destructuring assignments for lists. This allows you to unpack list elements into multiple variables at once:
val [a, b, c] = [1, 2, 3]
assertEquals(1, a)
assertEquals(2, b)
assertEquals(3, c)
It also supports *splats* (ellipsis) to capture multiple elements into a list:
val [head, rest...] = [1, 2, 3]
assertEquals(1, head)
assertEquals([2, 3], rest)
val [first, middle..., last] = [1, 2, 3, 4, 5]
assertEquals(1, first)
assertEquals([2, 3, 4], middle)
assertEquals(5, last)
Destructuring can be nested:
val [x, [y, z...]] = [1, [2, 3, 4]]
assertEquals(1, x)
assertEquals(2, y)
assertEquals([3, 4], z)
And it can be used for reassigning existing variables, for example, to swap values:
var x = 5
var y = 10
[x, y] = [y, x]
assertEquals(10, x)
assertEquals(5, y)
As the assignment itself is an expression, you can use it in strange ways. Just remember
to use parentheses as assignment operation insofar is left-associated and will not
allow chained assignments (we might fix it later). Use parentheses insofar:
@ -206,13 +122,14 @@ Note that assignment operator returns rvalue, it can't be assigned.
## Modifying arithmetics
There is a set of assigning operations: `+=`, `-=`, `*=`, `/=` and even `%=`.
There is also a special null-aware assignment operator `?=`: it performs the assignment only if the lvalue is `null`.
var x = null
x ?= 10
assertEquals(10, x)
x ?= 20
assertEquals(10, x)
var x = 5
assert( 25 == (x*=5) )
assert( 25 == x)
assert( 24 == (x-=1) )
assert( 12 == (x/=2) )
x
>>> 12
Notice the parentheses here: the assignment has low priority!
@ -229,8 +146,9 @@ Naturally, assignment returns its value:
rvalue means you cant assign the result if the assignment
var x
// compile-time error: can't assign to rvalue
(x = 11) = 5
assertThrows { (x = 11) = 5 }
void
>>> void
This also prevents chain assignments so use parentheses:
@ -241,48 +159,29 @@ This also prevents chain assignments so use parentheses:
## Nullability
Nullability is part of the type. `String` is non-null, `String?` is nullable. Use `!!` to assert non-null and throw
`NullReferenceException` if the value is `null`.
When the value is `null`, it might throws `NullReferenceException`, the name is somewhat a tradition. To avoid it
one can check it against null or use _null coalescing_. The null coalescing means, if the operand (left) is null,
the operation won't be performed and the result will be null. Here is the difference:
class Sample {
var field = 1
fun method() { 2 }
var list = [1, 2, 3]
}
val ref: Sample? = null
val list: List<Int>? = null
// direct access throws NullReferenceException:
// ref.field
// ref.method()
// ref.list[1]
// list[1]
val ref = null
assertThrows { ref.field }
assertThrows { ref.method() }
assertThrows { ref.array[1] }
assertThrows { ref[1] }
assertThrows { ref() }
assert( ref?.field == null )
assert( ref?.method() == null )
assert( ref?.list?[1] == null )
assert( list?[1] == null )
assert( ref?.array?[1] == null )
assert( ref?[1] == null )
assert( ref?() == null )
>>> void
Note: `?.` is still a typed operation. The receiver must have a compile-time type that declares the member; if the
receiver is `Object`, cast it first or declare a more specific type.
There is also "elvis operator", null-coalesce infix operator '?:' that returns rvalue if lvalue is `null`:
null ?: "nothing"
>>> "nothing"
There is also a null-aware assignment operator `?=`, which assigns a value only if the target is `null`:
var config = null
config ?= { port: 8080 }
config ?= { port: 9000 } // no-op, config is already not null
assertEquals(8080, config.port)
## Utility functions
The following functions simplify nullable values processing and
@ -327,8 +226,8 @@ Much like let, but it does not alter returned value:
While it is not altering return value, the source object could be changed:
also
class Point(var x: Int, var y: Int)
val p: Point = Point(1,2).also { it.x++ }
class Point(x,y)
val p = Point(1,2).also { it.x++ }
assertEquals(p.x, 2)
>>> void
@ -336,22 +235,12 @@ also
It works much like `also`, but is executed in the context of the source object:
class Point(var x: Int, var y: Int)
class Point(x,y)
// see the difference: apply changes this to newly created Point:
val p = Point(1,2).apply { this@Point.x++; this@Point.y++ }
val p = Point(1,2).apply { x++; y++ }
assertEquals(p, Point(2,3))
>>> void
## with
Sets `this` to the first argument and executes the block. Returns the value returned by the block:
class Point(var x: Int, var y: Int)
val p = Point(1,2)
val sum = with(p) { x + y }
assertEquals(3, sum)
>>> void
## run
Executes a block after it returning the value passed by the block. for example, can be used with elvis operator:
@ -422,6 +311,8 @@ Reference quality and object equality example:
assert( null == null) // singletons
assert( null === null)
// but, for non-singletons:
assert( 5 == 5)
assert( 5 !== 5)
assert( "foo" !== "foo" )
>>> void
@ -438,7 +329,7 @@ will be thrown:
// WRONG! Exception will be thrown at next line:
foo + "bar"
The correct pattern is:
Correct pattern is:
foo = "foo"
// now is OK:
@ -454,6 +345,8 @@ Almost the same, using `val`:
val foo = 1
foo += 1 // this will throw exception
# Constants
Same as in kotlin:
val HalfPi = π / 2
@ -461,151 +354,6 @@ Same as in kotlin:
Note using greek characters in identifiers! All letters allowed, but remember who might try to read your script, most
likely will know some English, the rest is the pure uncertainty.
# Types and inference
Lyng uses Kotlin-style static types with inference. You can always write explicit types, but in most places the compiler
can infer them from literals, defaults, and flow analysis.
## Type annotations
Use `:` to specify a type:
var x: Int = 10
val label: String = "count"
fun clamp(x: Int, min: Int, max: Int): Int { ... }
`Object` is the top type. If you omit a type and there is no default value, the parameter is `Object` by default:
fun show(x) { println(x) } // x is Object
For nullable types, add `?`:
fun showMaybe(x: Object?) { ... }
fun parseInt(s: String?): Int? { ... }
There is also a nullable shorthand for untyped parameters and constructor args: `x?` means `x: Object?`.
It cannot be combined with an explicit type annotation.
class A(x?) { ... } // x: Object?
fun f(x?) { x == null } // x: Object?
Type aliases name type expressions and can be generic:
type Num = Int | Real
type Maybe<T> = T?
Aliases expand to their underlying type expressions. See `docs/generics.md` for details.
`void` is a singleton value of the class `Void`. `Void` can be used as an explicit return type:
fun log(msg): Void { println(msg); void }
`Null` is the class of `null`. It is a singleton type and mostly useful for type inference results.
## Type inference
The compiler infers types from:
- literals: `1` is `Int`, `1.0` is `Real`, `"s"` is `String`, `'c'` is `Char`
- defaults: `fun f(x=1, name="n")` infers `x: Int`, `name: String`
- assignments: `val x = call()` uses the return type of `call`
- returns and branches: the result type of a block is the last expression, and if any branch is nullable,
the inferred type becomes nullable
- numeric ops: `Int` and `Real` stay `Int` when both sides are `Int`, and promote to `Real` on mixed arithmetic
Examples:
fun inc(x=0) = x + 1 // (Int)->Int
fun maybe(flag) { if(flag) 1 else null } // ()->Int?
Function types are written as `(T1, T2, ...)->R`. You can include ellipsis in function *types* to
express a variadic position:
var fmt: (String, Object...)->String
var f: (Int, Object..., String)->Real
var anyArgs: (...)->Int // shorthand for (Object...)->Int
Untyped locals are allowed, but their type is fixed on the first assignment:
var x
x = 1 // x becomes Int
x = "one" // compile-time error
var y = null // y is Object?
val z = null // z is Null
Empty list/map literals default to `List<Object>` and `Map<Object,Object>` until a more specific type is known:
val xs = [] // List<Object>
val ys: List<Int> = [] // List<Int>
Map literals infer key/value types from entries; named keys are `String`. See `docs/generics.md` for details.
## Flow analysis
Lyng uses flow analysis to narrow types inside branches:
fun len(x: String?): Int {
if( x == null ) return 0
// x is String (non-null) in this branch
return x.length
}
`is` checks and `when` branches also narrow types:
fun kind(x: Object) {
if( x is Int ) return "int"
if( x is String ) return "string"
return "other"
}
Narrowing is local to the branch; after the branch, the original type is restored.
## Casts and unknown types
Use `as` for explicit casts. The compiler inserts casts only when it can be valid and necessary. If a cast fails at
runtime, it throws `ClassCastException`. If the value is nullable, `as T` implies a non-null assertion.
Member access is resolved at compile time. Only members of `Object` are available on unknown types; non-Object members
require an explicit cast:
fun f(x) { // x is Object
x.toString() // ok (Object member)
x.size() // compile-time error
(x as List).size() // ok
}
This avoids runtime name-resolution fallbacks; all symbols must be known at compile time.
## Generics and bounds
Generic parameters are declared with `<...>`:
fun id<T>(x: T): T = x
class Box<T>(val value: T)
Bounds use `:` and can combine with `&` (intersection) and `|` (union):
fun sum<T: Int | Real>(x: T, y: T) = x + y
class Named<T: Iterable & Comparable>(val data: T)
Type arguments are usually inferred from call sites:
val b = Box(10) // Box<Int>
val s = id("ok") // T is String
See [Generics and type expressions](generics.md) for bounds, unions/intersections, and type-checking rules.
## Variance
Generic types are invariant by default, so `List<Int>` is not a `List<Object>`.
Use declaration-site variance when you need it:
class Source<out T>(val value: T)
class Sink<in T> { fun accept(x: T) { ... } }
`out` makes the type covariant (only produced), `in` makes it contravariant (only consumed).
# Defining functions
fun check(amount) {
@ -647,28 +395,14 @@ There are default parameters in Lyng:
It is possible to define also vararg using ellipsis:
fun sum(args...) {
val list = args as List
var result = list[0]
for( i in 1 ..< list.size ) result += list[i]
var result = args[0]
for( i in 1 ..< args.size ) result += args[i]
}
sum(10,20,30)
>>> 60
See the [arguments reference](declaring_arguments.md) for more details.
## Named arguments
When calling functions, you can use named arguments with the colon syntax `name: value`. This is particularly useful when you have many parameters with default values.
```lyng
fun test(a="foo", b="bar", c="bazz") { [a, b, c] }
assertEquals(["foo", "b", "bazz"], test(b: "b"))
assertEquals(["a", "bar", "c"], test("a", c: "c"))
```
**Note for Kotlin users:** Lyng uses `:` instead of `=` for named arguments at call sites. This is because in Lyng, `=` is an expression that returns the assigned value, and using it in an argument list would create ambiguity.
## Closures
Each __block has an isolated context that can be accessed from closures__. For example:
@ -742,22 +476,14 @@ one could be with ellipsis that means "the rest pf arguments as List":
assert( { a, b...-> [a,...b] }(100, 1, 2, 3) == [100, 1, 2, 3])
void
Type-annotated lambdas can use variadic *function types* as well:
val f: (Int, Object..., String)->Real = { a, rest..., b -> 0.0 }
val anyArgs: (...)->Int = { -> 0 }
### Using lambda as the parameter
See also: [Testing and Assertions](Testing.md)
// note that fun returns its last calculated value,
// in our case, result after in-place addition:
fun mapValues(iterable, transform) {
var result = []
for( x in iterable ) result += transform(x)
}
// loop variables are read-only inside the loop body
assert( [11, 21, 31] == mapValues( [1,2,3], { it*10+1 }))
>>> void
@ -793,7 +519,7 @@ Lists can contain any type of objects, lists too:
assert( list is Array ) // general interface
assert(list.size == 3)
// second element is a list too:
assert((list[1] as List).size == 2)
assert(list[1].size == 2)
>>> void
Notice usage of indexing. You can use negative indexes to offset from the end of the list; see more in [Lists](List.md).
@ -1241,8 +967,8 @@ ends normally, without breaks. It allows override loop result value, for example
to not calculate it in every iteration. For example, consider this naive prime number
test function (remember function return it's last expression result):
fun naive_is_prime(candidate: Int) {
val x = candidate
fun naive_is_prime(candidate) {
val x = if( candidate !is Int) candidate.toInt() else candidate
var divisor = 1
while( ++divisor < x/2 || divisor == 2 ) {
if( x % divisor == 0 ) break false
@ -1317,9 +1043,8 @@ For loop are intended to traverse collections, and all other objects that suppor
size and index access, like lists:
var letters = 0
val words: List<String> = ["hello", "world"]
for( w in words) {
letters += (w as String).length
for( w in ["hello", "wolrd"]) {
letters += w.length
}
"total letters: "+letters
>>> "total letters: 10"
@ -1432,7 +1157,7 @@ The same with `--`:
sum
>>> 5050
There is a self-assigning version for operators too:
There are self-assigning version for operators too:
var count = 100
var sum = 0
@ -1510,7 +1235,7 @@ than enum arrays, until `Lynon.encodeTyped` will be implemented.
var result = null // here we will store the result
>>> null
# Built-in types
# Integral data types
| type | description | literal samples |
|--------|---------------------------------|---------------------|
@ -1520,7 +1245,6 @@ than enum arrays, until `Lynon.encodeTyped` will be implemented.
| Char | single unicode character | `'S'`, `'\n'` |
| String | unicode string, no limits | "hello" (see below) |
| List | mutable list | [1, "two", 3] |
| Object | top type for all values | |
| Void | no value could exist, singleton | void |
| Null | missing value, singleton | null |
| Fn | callable type | |
@ -1643,51 +1367,42 @@ Concatenation is a `+`: `"hello " + name` works as expected. No confusion. There
Extraction:
("abcd42def"[ "\d+".re ] as RegexMatch).value
"abcd42def"[ "\d+".re ].value
>>> "42"
Part match:
assert( "abc foo def" =~ "f[oO]+".re )
assert( "foo" == ($~ as RegexMatch).value )
assert( "foo" == $~.value )
>>> void
Repeating the fragment:
Typical set of String functions includes:
assertEquals("hellohello", "hello"*2)
assertEquals("", "hello"*0)
>>> void
A typical set of String functions includes:
| fun/prop | description / notes |
|----------------------|------------------------------------------------------------|
| lower(), lowercase() | change case to unicode upper |
| upper(), uppercase() | change case to unicode lower |
| trim() | trim space chars from both ends |
| isEmpty() | true if string is empty |
| isNotEmpty() | true if string is not empty |
| isBlank() | true if empty or contains only whitespace |
| startsWith(prefix) | true if starts with a prefix |
| endsWith(prefix) | true if ends with a prefix |
| last() | get last character of a string or throw |
| take(n) | get a new string from up to n first characters |
| takeLast(n) | get a new string from up to n last characters |
| drop(n) | get a new string dropping n first chars, or empty string |
| dropLast(n) | get a new string dropping n last chars, or empty string |
| size | size in characters like `length` because String is [Array] |
| (args...) | sprintf-like formatting, see [string formatting] |
| [index] | character at index |
| [Range] | substring at range (2) |
| [Regex] | find first match of regex, like [Regex.find] (2) |
| s1 + s2 | concatenation |
| s1 += s2 | self-modifying concatenation |
| toReal() | attempts to parse string as a Real value |
| toInt() | parse string to Int value |
| characters | create [List] of characters (1) |
| encodeUtf8() | returns [Buffer] with characters encoded to utf8 |
| matches(re) | matches the regular expression (2) |
| | |
| fun/prop | description / notes |
|--------------------|------------------------------------------------------------|
| lower() | change case to unicode upper |
| upper() | change case to unicode lower |
| trim() | trim space chars from both ends |
| startsWith(prefix) | true if starts with a prefix |
| endsWith(prefix) | true if ends with a prefix |
| last() | get last character of a string or throw |
| take(n) | get a new string from up to n first characters |
| takeLast(n) | get a new string from up to n last characters |
| drop(n) | get a new string dropping n first chars, or empty string |
| dropLast(n) | get a new string dropping n last chars, or empty string |
| size | size in characters like `length` because String is [Array] |
| (args...) | sprintf-like formatting, see [string formatting] |
| [index] | character at index |
| [Range] | substring at range (2) |
| [Regex] | find first match of regex, like [Regex.find] (2) |
| s1 + s2 | concatenation |
| s1 += s2 | self-modifying concatenation |
| toReal() | attempts to parse string as a Real value |
| toInt() | parse string to Int value |
| characters() | create [List] of characters (1) |
| encodeUtf8() | returns [Buffer] with characters encoded to utf8 |
| matches(re) | matches the regular expression (2) |
| | |
(1)
: List is mutable therefore a new copy is created on each call.
@ -1733,8 +1448,8 @@ See [math functions](math.md). Other general purpose functions are:
| print(args...) | Open for overriding, it prints to stdout without newline. |
| flow {} | create flow sequence, see [parallelism] |
| delay, launch, yield | see [parallelism] |
| cached(builder) | [Lazy evaluation with `cached`](#lazy-evaluation-with-cached) |
| let, also, apply, run, with | see above, flow controls |
| cached(builder) | remembers builder() on first invocation and return it then |
| let, also, apply, run | see above, flow controls |
(1)
: `fn` is optional lambda returning string message to add to exception string.
@ -1751,8 +1466,6 @@ Lambda avoid unnecessary execution if assertion is not failed. for example:
[List]: List.md
[Testing]: Testing.md
[Iterable]: Iterable.md
[Iterator]: Iterator.md
@ -1761,7 +1474,7 @@ Lambda avoid unnecessary execution if assertion is not failed. for example:
[Range]: Range.md
[String]: ../archived/development/String.md
[String]: development/String.md
[string formatting]: https://github.com/sergeych/mp_stools?tab=readme-ov-file#sprintf-syntax-summary
@ -1781,50 +1494,6 @@ Lambda avoid unnecessary execution if assertion is not failed. for example:
[Regex]: Regex.md
## Lazy evaluation with `cached`
Sometimes you have an expensive computation that you only want to perform if and when it is actually needed, and then remember (cache) the result for all future calls. Lyng provides the `cached(builder)` function for this purpose.
It is extremely simple to use: you pass it a block (lambda) that performs the computation, and it returns a zero-argument function that manages the caching for you.
### Basic Example
```lyng
val expensive = cached {
println("Performing expensive calculation...")
2 + 2
}
println(expensive()) // Prints "Performing expensive calculation...") then "4"
println(expensive()) // Prints only "4" (result is cached)
```
### Benefits and Simplicity
1. **Lazy Execution:** The code inside the `cached` block doesn't run until you actually call the resulting function.
2. **Automatic State Management:** You don't need to manually check if a value has been computed or store it in a separate variable.
3. **Closures and Class Support:** `cached` works perfectly with closures. If you use it inside a class, it will correctly capture the instance variables, and each instance will have its own independent cache.
### Use Case: Lazy Properties in Classes
This is the most common use case for `cached`. It allows you to define expensive "fields" that are only computed if someone actually uses them:
```lyng
class User(val id: Int) {
// The details will be fetched only once, on demand
val details = cached {
println("Fetching details for user " + id)
// Db.query is a hypothetical example
Db.query("SELECT * FROM users WHERE id = " + id)
}
}
val u = User(101)
// ... nothing happens yet ...
val d = u.details() // Computation happens here
val sameD = u.details() // Returns the same result immediately
```
## Multiple Inheritance (quick start)
Lyng supports multiple inheritance (MI) with simple, predictable rules. For a full reference see OOP notes, this is a quick, copy‑paste friendly overview.
@ -1885,31 +1554,7 @@ assertEquals(null, (buzz as? Foo)?.runA())
Notes:
- Resolution order uses C3 MRO (active): deterministic, monotonic order suitable for diamonds and complex hierarchies. Example: for `class D() : B(), C()` where both `B()` and `C()` derive from `A()`, the C3 order is `D → B → C → A`. The first visible match wins.
- `private` is visible only inside the declaring class; `protected` is visible from the declaring class and its subclasses. Additionally, ancestors can access protected members of descendants if they override a member known to the ancestor. Qualification (`this@Type`) or casts do not bypass visibility.
- `private` is visible only inside the declaring class; `protected` is visible from the declaring class and any of its transitive subclasses. Qualialsofication (`this@Type`) or casts do not bypass visibility.
- Safe‑call `?.` works with `as?` for optional dispatch.
## Extension members
You can add new methods and properties to existing classes without modifying them.
### Extension functions
fun String.shout() = this.upper() + "!!!"
"hello".shout()
>>> "HELLO!!!"
### Extension properties
val Int.isEven get() = this % 2 == 0
4.isEven
>>> true
Example with custom accessors:
val String.firstChar get() = this[0]
"abc".firstChar
>>> 'a'
Extension members are **scope-isolated**: they are visible only in the scope where they are defined and its children. This prevents name collisions and improves security.
To get details on OOP in Lyng, see [OOP notes](OOP.md).
To get details on OOP in Lyng, see [OOP notes](oop.md).

View File

@ -1,27 +0,0 @@
# Wasm generation hang in wasmJs browser tests
## Summary
The wasmJs browser test runner hung after commit 5f819dc. The root cause was invalid WebAssembly generated by the Kotlin/Wasm backend when certain compiler paths emitted suspend lambdas for `Statement` execution. The invalid module failed to instantiate in the browser, and Karma kept the browser connected but never ran tests.
## Symptoms
- `:lynglib:wasmJsBrowserTest` hangs indefinitely in ChromeHeadless.
- `:lynglib:wasmJsNodeTest` fails with a WebAssembly compile error similar to:
- `struct.set expected type (ref null XXXX), found global.get of type (ref null YYYY)`
- The failing function name in the wasm name section looks like:
- `net.sergeych.lyng.$invokeCOROUTINE$.doResume`
## Root cause
The delegation/var-declaration changes introduced compiler-generated suspend lambdas inside `Statement` construction (e.g., `statement { ... }` wrappers). Kotlin/Wasm generates extra coroutine state for those suspend lambdas, which in this case produced invalid wasm IR (mismatched GC reference types). The browser loader then waits forever because the module fails to instantiate.
## Fix
Avoid suspend-lambda `Statement` construction in compiler code paths. Replace `statement { ... }` and other anonymous suspend lambdas with explicit `object : Statement()` implementations and move logic into `override suspend fun execute(...)`. This keeps the resulting wasm IR valid while preserving behavior.
## Where it was fixed
- `lynglib/src/commonMain/kotlin/net/sergeych/lyng/Compiler.kt`
- `lynglib/src/commonMain/kotlin/net/sergeych/lyng/Scope.kt`
## Verification
- `./gradlew :lynglib:wasmJsNodeTest --info`
- `./gradlew :lynglib:wasmJsBrowserTest --info`
Both tests finish quickly after the change.

View File

@ -1,253 +0,0 @@
# What's New in Lyng
This document highlights the latest additions and improvements to the Lyng language and its ecosystem.
For a programmer-focused migration summary, see `docs/whats_new_1_5.md`.
## Language Features
### Class Properties with Accessors
Classes now support properties with custom `get()` and `set()` accessors. Properties in Lyng do **not** have automatic backing fields; they are pure accessors.
```lyng
class Person(private var _age: Int) {
// Read-only property
val ageCategory get() = if (_age < 18) "Minor" else "Adult"
// Read-write property
var age: Int
get() = _age
set(v) {
if (v >= 0) _age = v
}
}
```
### Private and Protected Setters
You can now restrict the visibility of a property's or field's setter using `private set` or `protected set`. This allows members to be publicly readable but only writable from within the declaring class or its subclasses.
### Refined Protected Visibility
Ancestor classes can now access `protected` members of their descendants if it is an override of a member known to the ancestor. This enables base classes to call protected methods that are implemented or overridden in subclasses.
```lyng
class Counter {
var count = 0
private set // Field with private setter
fun increment() { count++ }
}
class AdvancedCounter : Counter {
var totalOperations = 0
protected set // Settable here and in further subclasses
}
let c = Counter()
c.increment() // OK
// c.count = 10 // Error: setter is private
```
### Late-initialized `val` Fields
`val` fields in classes can be declared without an immediate initializer, provided they are assigned exactly once. If accessed before initialization, they hold the special `Unset` singleton.
```lyng
class Service {
val logger
fun check() {
if (logger == Unset) println("Not initialized yet")
}
init {
logger = Logger("Service")
}
}
```
### Named Arguments and Named Splats
Function calls now support named arguments using the `name: value` syntax. If the variable name matches the parameter name, you can use the `name:` shorthand.
```lyng
fun greet(name, greeting = "Hello") {
println("$greeting, $name!")
}
val name = "Alice"
greet(name:) // Shorthand for greet(name: name)
greet(greeting: "Hi", name: "Bob")
let params = { name: "Charlie", greeting: "Hey")
greet(...params) // Named splat expansion
```
### Multiple Inheritance (MI)
Lyng now supports multiple inheritance using the C3 Method Resolution Order (MRO). Use `this@Type` or casts for disambiguation.
```lyng
class A { fun foo() = "A" }
class B { fun foo() = "B" }
class Derived : A, B {
fun test() {
println(foo()) // Resolves to A.foo (leftmost)
println(this@B.foo()) // Qualified dispatch to B.foo
}
}
let d = Derived()
println((d as B).foo()) // Disambiguation via cast
```
### Singleton Objects
Singleton objects are declared using the `object` keyword. They provide a convenient way to define a class and its single instance in one go.
```lyng
object Config {
val version = "1.5.0-SNAPSHOT"
fun show() = println("Config version: " + version)
}
Config.show()
```
### Nested Declarations and Lifted Enums
You can now declare classes, objects, enums, and type aliases inside another class. These nested declarations live in the class namespace (no outer instance capture) and are accessed with a qualifier.
```lyng
class A {
class B(x?)
object Inner { val foo = "bar" }
enum E* { One, Two }
}
val ab = A.B()
assertEquals(ab.x, null)
assertEquals(A.Inner.foo, "bar")
assertEquals(A.One, A.E.One)
```
The `*` on `enum E*` lifts entries into the enclosing class namespace (compile-time error on ambiguity).
### Object Expressions
You can now create anonymous objects that inherit from classes or interfaces using the `object : Base { ... }` syntax. These expressions capture their lexical scope and support multiple inheritance.
```lyng
val worker = object : Runnable {
override fun run() = println("Working...")
}
val x = object : Base(arg1), Interface1 {
val property = 42
override fun method() = this@object.property * 2
}
```
Use `this@object` to refer to the innermost anonymous object instance when `this` is rebound.
### Unified Delegation Model
A powerful new delegation system allows `val`, `var`, and `fun` members to delegate their logic to other objects using the `by` keyword.
```lyng
// Property delegation
val lazyValue by lazy { "expensive" }
// Function delegation
fun remoteAction by myProxy
// Observable properties
var name by Observable("initial") { n, old, new ->
println("Changed!")
}
```
The system features a unified interface (`getValue`, `setValue`, `invoke`) and a `bind` hook for initialization-time validation and configuration. See the [Delegation Guide](delegation.md) for more.
### User-Defined Exception Classes
You can now create custom exception types by inheriting from the built-in `Exception` class. Custom exceptions are real classes that can have their own fields and methods, and they work seamlessly with `throw` and `try-catch` blocks.
```lyng
class ValidationException(val field, m) : Exception(m)
try {
throw ValidationException("email", "Invalid format")
}
catch(e: ValidationException) {
println("Error in " + e.field + ": " + e.message)
}
```
### Assign-if-null Operator (`?=`)
The new `?=` operator provides a concise way to assign a value only if the target is `null`. It is especially useful for setting default values or lazy initialization.
```lyng
var x = null
x ?= 42 // x is now 42
x ?= 100 // x remains 42 (not null)
// Works with properties and index access
config.port ?= 8080
settings["theme"] ?= "dark"
```
The operator returns the final value of the receiver (the original value if it was not `null`, or the new value if the assignment occurred).
### Transient Attribute (`@Transient`)
The `@Transient` attribute can now be applied to class fields, constructor parameters, and static fields to exclude them from serialization.
```lyng
class MyData(@Transient val tempSecret, val publicData) {
@Transient var cachedValue = 0
var persistentValue = 42
}
```
Key features:
- **Serialization**: Transient members are omitted from both Lynon binary streams and JSON output.
- **Structural Equality**: Transient fields are automatically ignored during `==` equality checks.
- **Deserialization**: Transient constructor parameters with default values are correctly restored to those defaults upon restoration.
### Value Clamping (`clamp`)
A new `clamp()` function has been added to the standard library to limit a value within a specified range. It is available as both a global function and an extension method on all objects.
```lyng
// Global function
clamp(15, 0..10) // returns 10
clamp(-5, 0..10) // returns 0
// Extension method
val x = 15
x.clamp(0..10) // returns 10
// Exclusive and open-ended ranges
15.clamp(0..<10) // returns 9
15.clamp(..10) // returns 10
-5.clamp(0..) // returns 0
```
`clamp()` correctly handles inclusive (`..`) and exclusive (`..<`) ranges. For discrete types like `Int` and `Char`, clamping to an exclusive upper bound returns the previous value.
## Tooling and Infrastructure
### CLI: Formatting Command
A new `fmt` subcommand has been added to the Lyng CLI.
```bash
lyng fmt MyFile.lyng # Print formatted code to stdout
lyng fmt --in-place MyFile.lyng # Format file in-place
lyng fmt --check MyFile.lyng # Check if file needs formatting
```
### IDEA Plugin: Autocompletion
Experimental lightweight autocompletion is now available in the IntelliJ plugin. It features type-aware member suggestions and inheritance-aware completion.
You can enable it in **Settings | Lyng Formatter | Enable Lyng autocompletion**.
### Kotlin API: Exception Handling
The `Obj.getLyngExceptionMessageWithStackTrace()` extension method has been added to simplify retrieving detailed error information from Lyng exception objects in Kotlin. Additionally, `getLyngExceptionMessage()` and `raiseAsExecutionError()` now accept an optional `Scope`, making it easier to use them when a scope is not immediately available.
### Kotlin API: Bridge Reflection and Class Binding (Preferred Extensions)
Lyng now provides a public Kotlin reflection bridge and a Lyng‑first class binding workflow. This is the **preferred** way to write Kotlin extensions and library integrations:
- **Bridge resolver**: explicit handles for values, vars, and callables with predictable lookup rules.
- **Class bridge binding**: declare classes/members in Lyng (marked `extern`) and bind the implementations in Kotlin before the first instance is created.
See **Embedding Lyng** for full samples and usage details.

View File

@ -1,133 +0,0 @@
# What's New in Lyng 1.3 (vs 1.2.* / master)
This is a programmer-focused summary of what changed since the 1.2.* line on `master`. It highlights new language and type-system features, runtime/IDE improvements, and how to migrate code safely.
## Highlights
- Generics are now a first-class part of the type system, with bounds, variance, unions, and intersections.
- Type aliases and type-expression checks (`T1 is T2`, `A in T`) enable richer static modeling.
- Nested declarations inside classes, plus lifted enum entries via `enum E*`.
- Stepped ranges (`step`) including iterable open-ended and real ranges.
- Runtime and compiler speedups: more bytecode coverage, direct slot access, call-site caching.
## Language and type system
### Generics, bounds, and variance
You can declare generic functions/classes with `<...>`, restrict them with bounds, and control variance.
```lyng
fun id<T>(x: T): T = x
class Box<out T>(val value: T)
fun sum<T: Int | Real>(x: T, y: T) = x + y
class Named<T: Iterable & Comparable>(val data: T)
```
### Type aliases and type expressions
Type aliases can name any type expression, including unions and intersections.
```lyng
type Num = Int | Real
type Maybe<T> = T?
```
Type expressions can be checked directly:
```lyng
fun f<T>(xs: List<T>) {
assert( T is Int | String | Bool ) // type-subset check
assert( Int in T ) // same relation as `Int is T`
}
```
Value checks remain `x is T`. Type expression equality uses `==` and is structural.
### Nullable shorthand for parameters
Untyped parameters and constructor args can use `x?` as shorthand for `x: Object?`:
```lyng
class A(x?) { ... }
fun f(x?) { x == null }
```
### List/map literal inference
The compiler now infers element and key/value types from literals and spreads. Mixed element types produce unions.
```lyng
val a = [1, 2, 3] // List<Int>
val b = [1, "two", true] // List<Int | String | Bool>
val m = { "a": 1, "b": "x" } // Map<String, Int | String>
```
### Compile-time member access only
Member access is resolved at compile time. On unknown types, only `Object` members are visible; other members require an explicit cast.
```lyng
fun f(x) { // x: Object
x.toString() // ok
x.size() // compile-time error
(x as List).size()
}
```
This removes runtime name-resolution fallbacks and makes errors deterministic.
### Nested declarations and lifted enums
Classes, objects, enums, and type aliases can be declared inside another class and accessed by qualifier. Enums can lift entries into the outer namespace with `*`.
```lyng
class A {
class B(x?)
object Inner { val foo = "bar" }
type Alias = B
enum E* { One, Two }
}
val b = A.B()
assertEquals(A.One, A.E.One)
```
### Stepped ranges
Ranges now support `step`, and open-ended/real ranges are iterable only with an explicit step.
```lyng
(1..5 step 2).toList() // [1,3,5]
(0.0..1.0 step 0.25).toList() // [0,0.25,0.5,0.75,1.0]
(0.. step 1).take(3).toList() // [0,1,2]
```
## Tooling and performance
- Bytecode compiler/VM coverage expanded (loops, expressions, calls), improving execution speed and consistency.
- Direct frame-slot access and scoped slot addressing reduce lookup overhead, including in closures.
- Call-site caching and numeric fast paths reduce hot-loop overhead.
- IDE tooling updated for the new type system and nested declarations; MiniAst-based completion work continues.
## Migration guide (from 1.2.*)
1) Replace dynamic member access on unknown types
- If you relied on runtime name resolution, add explicit casts or annotate types so the compiler can resolve members.
2) Adopt new type-system constructs where helpful
- Consider `type` aliases for complex unions/intersections.
- Prefer generic signatures over `Object` when the API is parametric.
3) Update range iteration where needed
- Use `step` for open-ended or real ranges you want to iterate.
4) Nullable shorthand is optional
- If you used untyped nullable params, you can keep `x` (Object) or switch to `x?` (Object?) for clarity.
## References
- `docs/generics.md`
- `docs/Range.md`
- `docs/OOP.md`
- `docs/BytecodeSpec.md`

View File

@ -1,140 +0,0 @@
# What's New in Lyng 1.5 (vs 1.3.* / master)
This document summarizes the significant changes and new features introduced in the 1.5 development cycle.
## Principal changes
### JIT compiler and compile-time types and symbols.
This major improvement gives the following big advantages:
- **Blazing Fast execution**: several times faster! (three to six times speedup in different scenarios).
- **Better IDE support**: autocompletion, early error detection, types check.
- **Error safety**: all symbols and types are checked at bound at compile-time. Many errors are detected earlier. Also, no risk that external or caller code would shadow some internally used symbols (especially in closures and inheritance).
In particular, it means no slow and flaky runtime lookups. Once compiled, code guarantees that it will always call the symbol known at compile-time; runtime name lookup though does not guarantee it and can be source of hard to trace bugs.
### New stable API to create Kotlin extensions
The API is fixed and will be kept with further Lyng core changes. It is now the recommended way to write Lyng extensions in Kotlin. It is much simpler and more elegant than the internal one. See [Kotlin Bridge Binding](../notes/kotlin_bridge_binding.md).
### Smart types system
- **Deep inference**: The compiler analyzes types of symbols along the execution path and in many cases eliminates unnecessary casts or type specifications.
- **Union and intersection types**: `A & B`, `A | B`.
- **Generics**: Generic types are first-class citizens with support for [bounds and variance](generics.md). Type params are erased by default and are reified only when needed (e.g., `T::class`, `T is ...`, `as T`, or in extern-facing APIs), which enables checks like `A in T` when `T` is reified.
- **Inner classes and enums**: Full support for nested declarations, including [Enums with lifting](OOP.md#lifted-enum-entries).
## Other highlights
- **The `return` Statement**: Added support for local and non-local returns using labels.
- **Abstract Classes and Interfaces**: Full support for `abstract` members and the `interface` keyword.
- **Singleton Objects**: Define singletons using the `object` keyword or use anonymous object expressions.
- **Multiple Inheritance**: Enhanced multi-inheritance with predictable [C3 MRO resolution](OOP.md#multiple-inheritance-and-mro).
- **Unified Delegation**: Powerful delegation model for `val`, `var`, and `fun` members. See [Delegation](delegation.md).
- **Class Properties with Accessors**: Define `val` and `var` properties with custom `get()` and `set()`.
- **Restricted Setter Visibility**: Use `private set` and `protected set` on fields and properties.
- **Late-initialized `val`**: Support for `val` fields that are initialized in `init` blocks or class bodies.
- **Transient Members**: Use `@Transient` to exclude members from serialization and equality checks.
- **Named Arguments and Splats**: Improved call-site readability with `name: value` and map-based splats.
- **Refined Visibility**: Improved `protected` access and `closed` modifier for better encapsulation.
## Language Features
### The `return` Statement
You can now exit from the innermost enclosing callable (function or lambda) using `return`. Lyng also supports non-local returns to outer scopes using labels. See [Return Statement](return_statement.md).
```lyng
fun findFirst<T>(list: Iterable<T>, predicate: (T)->Bool): T? {
list.forEach {
if (predicate(it)) return@findFirst it
}
null
}
```
### Abstract Classes and Interfaces
Lyng now supports the `abstract` modifier for classes and their members. `interface` is introduced as a synonym for `abstract class`, allowing for rich multi-inheritance patterns.
```lyng
interface Shape {
abstract val area: Real
fun describe() = "Area: %g"(area)
}
class Circle(val radius: Real) : Shape {
override val area get = Math.PI * radius * radius
}
```
### Class Properties with Accessors
Properties can now have custom getters and setters. They do not have automatic backing fields, making them perfect for computed values or delegation.
```lyng
class Rectangle(var width: Real, var height: Real) {
val area: Real get() = width * height
var squareSize: Real
get() = area
set(v) {
width = sqrt(v)
height = width
}
}
```
### Singleton Objects
Declare singletons or anonymous objects easily.
```lyng
object Database {
val connection = "connected"
}
val runner = object : Runnable {
override fun run() = println("Running!")
}
```
### Named Arguments and Named Splats
Improve call-site clarity by specifying argument names. You can also expand a Map into named arguments using the splat operator.
```lyng
fun configure(timeout: Int, retry: Int = 3) { ... }
configure(timeout: 5000, retry: 5)
val options = Map("timeout": 1000, "retry": 1)
configure(...options)
```
### Modern Operators
The `?=` operator allows for concise "assign if null" logic.
```lyng
var cache: Map? = null
cache ?= Map("status": "ok") // Only assigns if cache is null
```
## Tooling and IDE
- **IDEA Plugin**: Significant improvements to autocompletion, documentation tooltips, and natural language support (Grazie integration).
- **CLI**: The `lyng fmt` command is now a first-class tool for formatting code with various options like `--check` and `--in-place`.
- **Performance**: Ongoing optimizations in the bytecode VM and compiler for faster execution and smaller footprint.
## Standard Library
- **`with(self, block)`**: Scoped execution with a dedicated `self`.
- **`clamp(value, min, max)`**: Easily restrict values to a range.
## Migration Guide (from 1.3.*)
1. **Check Visibility**: Refined `protected` and `private` rules may catch previously undetected invalid accesses.
2. **Override Keyword**: Ensure all members that override ancestor declarations are marked with the `override` keyword (now mandatory).
3. **Return in Shorthand**: Remember that `return` is forbidden in `=` shorthand functions; use block syntax if you need early exit.
4. **Empty Map Literals**: Use `Map()` or `{:}` for empty maps, as `{}` is now strictly a block/lambda.
## References
- [Object Oriented Programming](OOP.md)
- [Generics](generics.md)
- [Return Statement](return_statement.md)
- [Delegation](delegation.md)
- [Tutorial](tutorial.md)

View File

@ -20,7 +20,7 @@ Files
- Constants: `true`, `false`, `null`, `this`
- Annotations: `@name` (Unicode identifiers supported)
- Labels: `name:` (Unicode identifiers supported)
- Declarations: highlights declared names in `fun|fn name`, `class|enum|interface Name`, `val|var name`
- Declarations: highlights declared names in `fun|fn name`, `class|enum Name`, `val|var name`
- Types: built-ins (`Int|Real|String|Bool|Char|Regex`) and Capitalized identifiers (heuristic)
- Operators including ranges (`..`, `..<`, `...`), null-safe (`?.`, `?[`, `?(`, `?{`, `?:`, `??`), arrows (`->`, `=>`, `::`), match operators (`=~`, `!~`), bitwise, arithmetic, etc.
- Shuttle operator `<=>`

View File

@ -2,7 +2,7 @@
"name": "lyng-textmate",
"displayName": "Lyng",
"description": "TextMate grammar for the Lyng language (for JetBrains IDEs via TextMate Bundles and VS Code).",
"version": "0.1.0",
"version": "0.0.3",
"publisher": "lyng",
"license": "Apache-2.0",
"engines": { "vscode": "^1.0.0" },

View File

@ -42,7 +42,7 @@
{ "name": "constant.numeric.decimal.lyng", "match": "(?<![A-Za-z_])(?:[0-9][0-9_]*)\\.(?:[0-9_]+)(?:[eE][+-]?[0-9_]+)?|(?<![A-Za-z_])(?:[0-9][0-9_]*)(?:[eE][+-]?[0-9_]+)?" }
]
},
"annotations": { "patterns": [ { "name": "entity.name.label.at.lyng", "match": "@[\\p{L}_][\\p{L}\\p{N}_]*" } ] },
"annotations": { "patterns": [ { "name": "entity.name.label.at.lyng", "match": "@[\\p{L}_][\\p{L}\\p{N}_]*:" }, { "name": "storage.modifier.annotation.lyng", "match": "@[\\p{L}_][\\p{L}\\p{N}_]*" } ] },
"mapLiterals": {
"patterns": [
{
@ -74,11 +74,11 @@
}
]
},
"labels": { "patterns": [ { "name": "entity.name.label.lyng", "match": "[\\p{L}_][\\p{L}\\p{N}_]*@" } ] },
"labels": { "patterns": [ { "name": "entity.name.label.lyng", "match": "[\\p{L}_][\\p{L}\\p{N}_]*:" } ] },
"directives": { "patterns": [ { "name": "meta.directive.lyng", "match": "^\\s*#[_A-Za-z][_A-Za-z0-9]*" } ] },
"declarations": { "patterns": [ { "name": "meta.function.declaration.lyng", "match": "\\b(fun|fn)\\s+(?:([\\p{L}_][\\p{L}\\p{N}_]*)\\.)?([\\p{L}_][\\p{L}\\p{N}_]*)", "captures": { "1": { "name": "keyword.declaration.lyng" }, "2": { "name": "entity.name.type.lyng" }, "3": { "name": "entity.name.function.lyng" } } }, { "name": "meta.type.declaration.lyng", "match": "\\b(?:class|enum|interface|object)(?:\\s+([\\p{L}_][\\p{L}\\p{N}_]*))?", "captures": { "1": { "name": "entity.name.type.lyng" } } }, { "name": "meta.variable.declaration.lyng", "match": "\\b(val|var)\\s+(?:([\\p{L}_][\\p{L}\\p{N}_]*)\\.)?([\\p{L}_][\\p{L}\\p{N}_]*)", "captures": { "1": { "name": "keyword.declaration.lyng" }, "2": { "name": "entity.name.type.lyng" }, "3": { "name": "variable.other.declaration.lyng" } } } ] },
"keywords": { "patterns": [ { "name": "keyword.control.lyng", "match": "\\b(?:if|else|when|while|do|for|try|catch|finally|throw|return|break|continue)\\b" }, { "name": "keyword.declaration.lyng", "match": "\\b(?:fun|fn|class|enum|interface|val|var|import|package|constructor|property|abstract|override|open|closed|extern|private|protected|static|get|set|object|init|by)\\b" }, { "name": "keyword.operator.word.lyng", "match": "\\bnot\\s+(?:in|is)\\b" }, { "name": "keyword.operator.word.lyng", "match": "\\b(?:and|or|not|in|is|as|as\\?)\\b" } ] },
"constants": { "patterns": [ { "name": "constant.language.lyng", "match": "(?:\\b(?:true|false|null|this(?:@[\\p{L}_][\\p{L}\\p{N}_]*)?)\\b|π)" } ] },
"declarations": { "patterns": [ { "name": "meta.function.declaration.lyng", "match": "\\b(?:fun|fn)\\s+([\\p{L}_][\\p{L}\\p{N}_]*)", "captures": { "1": { "name": "entity.name.function.lyng" } } }, { "name": "meta.type.declaration.lyng", "match": "\\b(?:class|enum)\\s+([\\p{L}_][\\p{L}\\p{N}_]*)", "captures": { "1": { "name": "entity.name.type.lyng" } } }, { "name": "meta.variable.declaration.lyng", "match": "\\b(?:val|var)\\s+([\\p{L}_][\\p{L}\\p{N}_]*)", "captures": { "1": { "name": "variable.other.declaration.lyng" } } } ] },
"keywords": { "patterns": [ { "name": "keyword.control.lyng", "match": "\\b(?:if|else|when|while|do|for|try|catch|finally|throw|return|break|continue)\\b" }, { "name": "keyword.declaration.lyng", "match": "\\b(?:fun|fn|class|enum|val|var|import|package|constructor|property|open|extern|private|protected|static)\\b" }, { "name": "keyword.operator.word.lyng", "match": "\\bnot\\s+(?:in|is)\\b" }, { "name": "keyword.operator.word.lyng", "match": "\\b(?:and|or|not|in|is|as|as\\?)\\b" } ] },
"constants": { "patterns": [ { "name": "constant.language.lyng", "match": "(?:\\b(?:true|false|null|this)\\b|π)" } ] },
"types": { "patterns": [ { "name": "storage.type.lyng", "match": "\\b(?:Int|Real|String|Bool|Char|Regex)\\b" }, { "name": "entity.name.type.lyng", "match": "\\b[A-Z][A-Za-z0-9_]*\\b(?!\\s*\\()" } ] },
"operators": { "patterns": [ { "name": "keyword.operator.comparison.lyng", "match": "===|!==|==|!=|<=|>=|<|>" }, { "name": "keyword.operator.shuttle.lyng", "match": "<=>" }, { "name": "keyword.operator.arrow.lyng", "match": "=>|->|::" }, { "name": "keyword.operator.range.lyng", "match": "\\.\\.\\.|\\.\\.<|\\.\\." }, { "name": "keyword.operator.nullsafe.lyng", "match": "\\?\\.|\\?\\[|\\?\\(|\\?\\{|\\?:|\\?\\?" }, { "name": "keyword.operator.assignment.lyng", "match": "(?:\\+=|-=|\\*=|/=|%=|=)" }, { "name": "keyword.operator.logical.lyng", "match": "&&|\\|\\|" }, { "name": "keyword.operator.bitwise.lyng", "match": "<<|>>|&|\\||\\^|~" }, { "name": "keyword.operator.match.lyng", "match": "=~|!~" }, { "name": "keyword.operator.arithmetic.lyng", "match": "\\+\\+|--|[+\\-*/%]" }, { "name": "keyword.operator.other.lyng", "match": "[!?]" } ] },
"punctuation": { "patterns": [ { "name": "punctuation.separator.comma.lyng", "match": "," }, { "name": "punctuation.terminator.statement.lyng", "match": ";" }, { "name": "punctuation.section.block.begin.lyng", "match": "[(]{1}|[{]{1}|\\[" }, { "name": "punctuation.section.block.end.lyng", "match": "[)]{1}|[}]{1}|\\]" }, { "name": "punctuation.accessor.dot.lyng", "match": "\\." }, { "name": "punctuation.separator.colon.lyng", "match": ":" } ] }

View File

@ -1,5 +1,5 @@
#
# Copyright 2026 Sergey S. Chernov real.sergeych@gmail.com
# Copyright 2025 Sergey S. Chernov real.sergeych@gmail.com
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
@ -16,7 +16,7 @@
#
#Gradle
org.gradle.jvmargs=-Xmx4096M -Dfile.encoding=UTF-8 -Dkotlin.daemon.jvm.options\="-Xmx2048M"
org.gradle.jvmargs=-Xmx2048M -Dfile.encoding=UTF-8 -Dkotlin.daemon.jvm.options\="-Xmx2048M"
org.gradle.caching=true
org.gradle.configuration-cache=true
#Kotlin

View File

@ -1,12 +1,11 @@
[versions]
agp = "8.5.2"
clikt = "5.0.3"
kotlin = "2.3.0"
kotlin = "2.2.21"
android-minSdk = "24"
android-compileSdk = "34"
kotlinx-coroutines = "1.10.2"
kotlinx-datetime = "0.6.1"
mp_bintools = "0.3.2"
mp_bintools = "0.1.12"
firebaseCrashlyticsBuildtools = "3.0.3"
okioVersion = "3.10.2"
compiler = "3.2.0-alpha11"
@ -17,7 +16,6 @@ clikt-markdown = { module = "com.github.ajalt.clikt:clikt-markdown", version.ref
kotlin-test = { module = "org.jetbrains.kotlin:kotlin-test", version.ref = "kotlin" }
kotlinx-coroutines-test = { module = "org.jetbrains.kotlinx:kotlinx-coroutines-test", version.ref = "kotlinx-coroutines" }
kotlinx-coroutines-core = { module = "org.jetbrains.kotlinx:kotlinx-coroutines-core", version.ref = "kotlinx-coroutines" }
kotlinx-datetime = { module = "org.jetbrains.kotlinx:kotlinx-datetime", version.ref = "kotlinx-datetime" }
mp_bintools = { module = "net.sergeych:mp_bintools", version.ref = "mp_bintools" }
firebase-crashlytics-buildtools = { group = "com.google.firebase", name = "firebase-crashlytics-buildtools", version.ref = "firebaseCrashlyticsBuildtools" }
okio = { module = "com.squareup.okio:okio", version.ref = "okioVersion" }

View File

@ -1,5 +1,5 @@
/*
* Copyright 2026 Sergey S. Chernov real.sergeych@gmail.com
* Copyright 2025 Sergey S. Chernov real.sergeych@gmail.com
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
@ -17,11 +17,11 @@
plugins {
kotlin("jvm")
id("org.jetbrains.intellij") version "1.17.4"
id("org.jetbrains.intellij") version "1.17.3"
}
group = "net.sergeych.lyng"
version = "0.0.5-SNAPSHOT"
version = "0.0.3-SNAPSHOT"
kotlin {
jvmToolchain(17)
@ -45,14 +45,12 @@ dependencies {
// Tests for IntelliJ Platform fixtures rely on JUnit 3/4 API (junit.framework.TestCase)
// Add JUnit 4 which contains the JUnit 3 compatibility classes used by BasePlatformTestCase/UsefulTestCase
testImplementation("junit:junit:4.13.2")
testRuntimeOnly("org.junit.vintage:junit-vintage-engine:5.10.2")
testImplementation("org.opentest4j:opentest4j:1.3.0")
}
intellij {
type.set("IC")
// Build against a modern baseline. Install range is controlled by since/until below.
version.set("2024.1.6")
version.set("2024.3.1")
// We manage <idea-version> ourselves in plugin.xml to keep it open-ended (no upper cap)
updateSinceUntilBuild.set(false)
// Include only available bundled plugins for this IDE build

View File

@ -1,33 +0,0 @@
/*
* Copyright 2026 Sergey S. Chernov real.sergeych@gmail.com
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*
*/
package net.sergeych.lyng.idea
import com.intellij.openapi.fileTypes.FileTypeConsumer
import com.intellij.openapi.fileTypes.FileTypeFactory
import com.intellij.openapi.fileTypes.WildcardFileNameMatcher
/**
* Legacy way to register file type matchers, used here to robustly match *.lyng.d
* without conflicting with standard .d extensions from other plugins.
*/
@Suppress("DEPRECATION")
class LyngFileTypeFactory : FileTypeFactory() {
override fun createFileTypes(consumer: FileTypeConsumer) {
// Register the multi-dot pattern explicitly
consumer.consume(LyngFileType, WildcardFileNameMatcher("*.lyng.d"))
}
}

View File

@ -1,109 +0,0 @@
/*
* Copyright 2026 Sergey S. Chernov real.sergeych@gmail.com
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*
*/
package net.sergeych.lyng.idea.actions
import com.intellij.execution.filters.TextConsoleBuilderFactory
import com.intellij.execution.ui.ConsoleView
import com.intellij.execution.ui.ConsoleViewContentType
import com.intellij.openapi.actionSystem.AnAction
import com.intellij.openapi.actionSystem.AnActionEvent
import com.intellij.openapi.actionSystem.CommonDataKeys
import com.intellij.openapi.project.Project
import com.intellij.openapi.wm.ToolWindow
import com.intellij.openapi.wm.ToolWindowAnchor
import com.intellij.openapi.wm.ToolWindowId
import com.intellij.openapi.wm.ToolWindowManager
import com.intellij.psi.PsiFile
import com.intellij.psi.PsiManager
import com.intellij.ui.content.ContentFactory
import kotlinx.coroutines.CoroutineScope
import kotlinx.coroutines.Dispatchers
import kotlinx.coroutines.SupervisorJob
import kotlinx.coroutines.launch
import net.sergeych.lyng.idea.LyngIcons
class RunLyngScriptAction : AnAction(LyngIcons.FILE) {
private val scope = CoroutineScope(Dispatchers.Default + SupervisorJob())
private fun getPsiFile(e: AnActionEvent): PsiFile? {
val project = e.project ?: return null
return e.getData(CommonDataKeys.PSI_FILE) ?: run {
val vf = e.getData(CommonDataKeys.VIRTUAL_FILE)
if (vf != null) PsiManager.getInstance(project).findFile(vf) else null
}
}
override fun update(e: AnActionEvent) {
val psiFile = getPsiFile(e)
val isLyng = psiFile?.name?.endsWith(".lyng") == true
e.presentation.isEnabledAndVisible = isLyng
if (isLyng) {
e.presentation.isEnabled = false
e.presentation.text = "Run '${psiFile.name}' (disabled)"
e.presentation.description = "Running scripts from the IDE is disabled; use the CLI."
} else {
e.presentation.text = "Run Lyng Script"
}
}
override fun actionPerformed(e: AnActionEvent) {
val project = e.project ?: return
val psiFile = getPsiFile(e) ?: return
val fileName = psiFile.name
val (console, toolWindow) = getConsoleAndToolWindow(project)
console.clear()
toolWindow.show {
scope.launch {
console.print("--- Run is disabled ---\n", ConsoleViewContentType.SYSTEM_OUTPUT)
console.print("Lyng now runs in bytecode-only mode; the IDE no longer evaluates scripts.\n", ConsoleViewContentType.NORMAL_OUTPUT)
console.print("Use the CLI to run scripts, e.g. `lyng run $fileName`.\n", ConsoleViewContentType.NORMAL_OUTPUT)
}
}
}
private fun getConsoleAndToolWindow(project: Project): Pair<ConsoleView, ToolWindow> {
val toolWindowManager = ToolWindowManager.getInstance(project)
var toolWindow = toolWindowManager.getToolWindow(ToolWindowId.RUN)
if (toolWindow == null) {
toolWindow = toolWindowManager.getToolWindow(ToolWindowId.MESSAGES_WINDOW)
}
if (toolWindow == null) {
toolWindow = toolWindowManager.getToolWindow("Lyng")
}
val actualToolWindow = toolWindow ?: run {
@Suppress("DEPRECATION")
toolWindowManager.registerToolWindow("Lyng", true, ToolWindowAnchor.BOTTOM)
}
val contentManager = actualToolWindow.contentManager
val existingContent = contentManager.findContent("Lyng Run")
if (existingContent != null) {
val console = existingContent.component as ConsoleView
contentManager.setSelectedContent(existingContent)
return console to actualToolWindow
}
val console = TextConsoleBuilderFactory.getInstance().createBuilder(project).console
val content = ContentFactory.getInstance().createContent(console.component, "Lyng Run", false)
contentManager.addContent(content)
contentManager.setSelectedContent(content)
return console to actualToolWindow
}
}

View File

@ -1,5 +1,5 @@
/*
* Copyright 2026 Sergey S. Chernov real.sergeych@gmail.com
* Copyright 2025 Sergey S. Chernov real.sergeych@gmail.com
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
@ -25,135 +25,314 @@ import com.intellij.openapi.progress.ProgressManager
import com.intellij.openapi.util.Key
import com.intellij.openapi.util.TextRange
import com.intellij.psi.PsiFile
import kotlinx.coroutines.runBlocking
import net.sergeych.lyng.Compiler
import net.sergeych.lyng.ScriptError
import net.sergeych.lyng.Source
import net.sergeych.lyng.binding.Binder
import net.sergeych.lyng.binding.SymbolKind
import net.sergeych.lyng.highlight.HighlightKind
import net.sergeych.lyng.highlight.SimpleLyngHighlighter
import net.sergeych.lyng.highlight.offsetOf
import net.sergeych.lyng.idea.highlight.LyngHighlighterColors
import net.sergeych.lyng.idea.util.LyngAstManager
import net.sergeych.lyng.tools.LyngDiagnosticSeverity
import net.sergeych.lyng.tools.LyngLanguageTools
import net.sergeych.lyng.tools.LyngSemanticKind
import net.sergeych.lyng.idea.util.IdeLenientImportProvider
import net.sergeych.lyng.miniast.*
/**
* ExternalAnnotator that runs Lyng MiniAst on the document text in background
* and applies semantic highlighting comparable with the web highlighter.
*/
class LyngExternalAnnotator : ExternalAnnotator<LyngExternalAnnotator.Input, LyngExternalAnnotator.Result>() {
data class Input(val text: String, val modStamp: Long, val previousSpans: List<Span>?, val file: PsiFile)
data class Input(val text: String, val modStamp: Long, val previousSpans: List<Span>?)
data class Span(val start: Int, val end: Int, val key: com.intellij.openapi.editor.colors.TextAttributesKey)
data class Diag(val start: Int, val end: Int, val message: String, val severity: HighlightSeverity)
data class Result(val modStamp: Long, val spans: List<Span>, val diagnostics: List<Diag> = emptyList())
data class Error(val start: Int, val end: Int, val message: String)
data class Result(val modStamp: Long, val spans: List<Span>, val error: Error? = null,
val spellIdentifiers: List<IntRange> = emptyList(),
val spellComments: List<IntRange> = emptyList(),
val spellStrings: List<IntRange> = emptyList())
override fun collectInformation(file: PsiFile): Input? {
val doc: Document = file.viewProvider.document ?: return null
val cached = file.getUserData(CACHE_KEY)
val combinedStamp = LyngAstManager.getCombinedStamp(file)
val prev = if (cached != null && cached.modStamp == combinedStamp) cached.spans else null
return Input(doc.text, combinedStamp, prev, file)
// Fast fix (1): reuse cached spans only if they were computed for the same modification stamp
val prev = if (cached != null && cached.modStamp == doc.modificationStamp) cached.spans else null
return Input(doc.text, doc.modificationStamp, prev)
}
override fun doAnnotate(collectedInfo: Input?): Result? {
if (collectedInfo == null) return null
ProgressManager.checkCanceled()
val text = collectedInfo.text
val analysis = LyngAstManager.getAnalysis(collectedInfo.file)
?: return Result(collectedInfo.modStamp, collectedInfo.previousSpans ?: emptyList())
val mini = analysis.mini
// Build Mini-AST using the same mechanism as web highlighter
val sink = MiniAstBuilder()
val source = Source("<ide>", text)
try {
// Call suspend API from blocking context
val provider = IdeLenientImportProvider.create()
runBlocking { Compiler.compileWithMini(source, provider, sink) }
} catch (e: Throwable) {
if (e is com.intellij.openapi.progress.ProcessCanceledException) throw e
// On script parse error: keep previous spans and report the error location
if (e is ScriptError) {
val off = try { source.offsetOf(e.pos) } catch (_: Throwable) { -1 }
val start0 = off.coerceIn(0, text.length.coerceAtLeast(0))
val (start, end) = expandErrorRange(text, start0)
// Fast fix (5): clear cached highlighting after the error start position
val trimmed = collectedInfo.previousSpans?.filter { it.end <= start } ?: emptyList()
return Result(
collectedInfo.modStamp,
trimmed,
Error(start, end, e.errorMessage)
)
}
// Other failures: keep previous spans without error
return Result(collectedInfo.modStamp, collectedInfo.previousSpans ?: emptyList(), null)
}
ProgressManager.checkCanceled()
val mini = sink.build() ?: return Result(collectedInfo.modStamp, collectedInfo.previousSpans ?: emptyList())
val out = ArrayList<Span>(256)
val diags = ArrayList<Diag>()
fun putRange(start: Int, end: Int, key: com.intellij.openapi.editor.colors.TextAttributesKey) {
if (start in 0..end && end <= text.length && start < end) out += Span(start, end, key)
}
fun keyForKind(kind: LyngSemanticKind): com.intellij.openapi.editor.colors.TextAttributesKey? = when (kind) {
LyngSemanticKind.Function -> LyngHighlighterColors.FUNCTION
LyngSemanticKind.Class, LyngSemanticKind.Enum, LyngSemanticKind.TypeAlias -> LyngHighlighterColors.TYPE
LyngSemanticKind.Value -> LyngHighlighterColors.VALUE
LyngSemanticKind.Variable -> LyngHighlighterColors.VARIABLE
LyngSemanticKind.Parameter -> LyngHighlighterColors.PARAMETER
LyngSemanticKind.TypeRef -> LyngHighlighterColors.TYPE
LyngSemanticKind.EnumConstant -> LyngHighlighterColors.ENUM_CONSTANT
fun putName(startPos: net.sergeych.lyng.Pos, name: String, key: com.intellij.openapi.editor.colors.TextAttributesKey) {
val s = source.offsetOf(startPos)
putRange(s, (s + name.length).coerceAtMost(text.length), key)
}
fun putMiniRange(r: MiniRange, key: com.intellij.openapi.editor.colors.TextAttributesKey) {
val s = source.offsetOf(r.start)
val e = source.offsetOf(r.end)
putRange(s, e, key)
}
// Semantic highlights from shared tooling
LyngLanguageTools.semanticHighlights(analysis).forEach { span ->
keyForKind(span.kind)?.let { putRange(span.range.start, span.range.endExclusive, it) }
}
// Imports: each segment as namespace/path
mini?.imports?.forEach { imp ->
imp.segments.forEach { seg ->
val start = analysis.source.offsetOf(seg.range.start)
val end = analysis.source.offsetOf(seg.range.end)
putRange(start, end, LyngHighlighterColors.NAMESPACE)
// Declarations
for (d in mini.declarations) {
when (d) {
is MiniFunDecl -> putName(d.nameStart, d.name, LyngHighlighterColors.FUNCTION_DECLARATION)
is MiniClassDecl -> putName(d.nameStart, d.name, LyngHighlighterColors.TYPE)
is MiniValDecl -> putName(
d.nameStart,
d.name,
if (d.mutable) LyngHighlighterColors.VARIABLE else LyngHighlighterColors.VALUE
)
}
}
// Add annotation/label coloring using token highlighter
run {
analysis.lexicalHighlights.forEach { s ->
if (s.kind == HighlightKind.Label) {
val start = s.range.start
val end = s.range.endExclusive
if (start in 0..end && end <= text.length && start < end) {
val lexeme = try {
text.substring(start, end)
} catch (_: Throwable) {
null
}
if (lexeme != null) {
// Heuristic: if it starts with @ and follows a control keyword, it's likely a label
// Otherwise if it starts with @ it's an annotation.
// If it ends with @ it's a loop label.
when {
lexeme.endsWith("@") -> putRange(start, end, LyngHighlighterColors.LABEL)
lexeme.startsWith("@") -> {
// Try to see if it's an exit label
val prevNonWs = prevNonWs(text, start)
val prevWord = if (prevNonWs >= 0) {
var wEnd = prevNonWs + 1
var wStart = prevNonWs
while (wStart > 0 && text[wStart - 1].isLetter()) wStart--
text.substring(wStart, wEnd)
} else null
// Imports: each segment as namespace/path
for (imp in mini.imports) {
for (seg in imp.segments) putMiniRange(seg.range, LyngHighlighterColors.NAMESPACE)
}
if (prevWord in setOf("return", "break", "continue")) {
putRange(start, end, LyngHighlighterColors.LABEL)
} else {
putRange(start, end, LyngHighlighterColors.ANNOTATION)
}
}
}
}
// Parameters
for (fn in mini.declarations.filterIsInstance<MiniFunDecl>()) {
for (p in fn.params) putName(p.nameStart, p.name, LyngHighlighterColors.PARAMETER)
}
// Type name segments (including generics base & args)
fun addTypeSegments(t: MiniTypeRef?) {
when (t) {
is MiniTypeName -> t.segments.forEach { seg ->
val s = source.offsetOf(seg.range.start)
putRange(s, (s + seg.name.length).coerceAtMost(text.length), LyngHighlighterColors.TYPE)
}
is MiniGenericType -> {
addTypeSegments(t.base)
t.args.forEach { addTypeSegments(it) }
}
is MiniFunctionType -> {
t.receiver?.let { addTypeSegments(it) }
t.params.forEach { addTypeSegments(it) }
addTypeSegments(t.returnType)
}
is MiniTypeVar -> { /* name is in range; could be highlighted as TYPE as well */
putMiniRange(t.range, LyngHighlighterColors.TYPE)
}
null -> {}
}
}
for (d in mini.declarations) {
when (d) {
is MiniFunDecl -> {
addTypeSegments(d.returnType)
d.params.forEach { addTypeSegments(it.type) }
}
is MiniValDecl -> addTypeSegments(d.type)
is MiniClassDecl -> {
d.ctorFields.forEach { addTypeSegments(it.type) }
d.classFields.forEach { addTypeSegments(it.type) }
}
}
}
ProgressManager.checkCanceled()
// Semantic usages via Binder (best-effort)
try {
val binding = Binder.bind(text, mini)
// Map declaration ranges to avoid duplicating them as usages
val declKeys = HashSet<Pair<Int, Int>>(binding.symbols.size * 2)
for (sym in binding.symbols) declKeys += (sym.declStart to sym.declEnd)
fun keyForKind(k: SymbolKind) = when (k) {
SymbolKind.Function -> LyngHighlighterColors.FUNCTION
SymbolKind.Class, SymbolKind.Enum -> LyngHighlighterColors.TYPE
SymbolKind.Param -> LyngHighlighterColors.PARAMETER
SymbolKind.Val -> LyngHighlighterColors.VALUE
SymbolKind.Var -> LyngHighlighterColors.VARIABLE
}
// Track covered ranges to not override later heuristics
val covered = HashSet<Pair<Int, Int>>()
for (ref in binding.references) {
val key = ref.start to ref.end
if (declKeys.contains(key)) continue
val sym = binding.symbols.firstOrNull { it.id == ref.symbolId } ?: continue
val color = keyForKind(sym.kind)
putRange(ref.start, ref.end, color)
covered += key
}
// Heuristics on top of binder: function call-sites and simple name-based roles
ProgressManager.checkCanceled()
val tokens = try { SimpleLyngHighlighter().highlight(text) } catch (_: Throwable) { emptyList() }
fun isFollowedByParenOrBlock(rangeEnd: Int): Boolean {
var i = rangeEnd
while (i < text.length) {
val ch = text[i]
if (ch == ' ' || ch == '\t' || ch == '\r' || ch == '\n') { i++; continue }
return ch == '(' || ch == '{'
}
return false
}
// Build simple name -> role map for top-level vals/vars and parameters
val nameRole = HashMap<String, com.intellij.openapi.editor.colors.TextAttributesKey>(8)
for (d in mini.declarations) when (d) {
is MiniValDecl -> nameRole[d.name] = if (d.mutable) LyngHighlighterColors.VARIABLE else LyngHighlighterColors.VALUE
is MiniFunDecl -> d.params.forEach { p -> nameRole[p.name] = LyngHighlighterColors.PARAMETER }
else -> {}
}
for (s in tokens) if (s.kind == HighlightKind.Identifier) {
val start = s.range.start
val end = s.range.endExclusive
val key = start to end
if (key in covered || key in declKeys) continue
// Call-site detection first so it wins over var/param role
if (isFollowedByParenOrBlock(end)) {
putRange(start, end, LyngHighlighterColors.FUNCTION)
covered += key
continue
}
// Simple role by known names
val ident = try { text.substring(start, end) } catch (_: Throwable) { null }
if (ident != null) {
val roleKey = nameRole[ident]
if (roleKey != null) {
putRange(start, end, roleKey)
covered += key
}
}
}
} catch (e: Throwable) {
// Must rethrow cancellation; otherwise ignore binder failures (best-effort)
if (e is com.intellij.openapi.progress.ProcessCanceledException) throw e
}
// Add annotation coloring using token highlighter (treat @Label as annotation)
run {
val tokens = try { SimpleLyngHighlighter().highlight(text) } catch (_: Throwable) { emptyList() }
for (s in tokens) if (s.kind == HighlightKind.Label) {
val start = s.range.start
val end = s.range.endExclusive
if (start in 0..end && end <= text.length && start < end) {
val lexeme = try { text.substring(start, end) } catch (_: Throwable) { null }
if (lexeme != null && lexeme.startsWith("@")) {
putRange(start, end, LyngHighlighterColors.ANNOTATION)
}
}
}
}
analysis.diagnostics.forEach { d ->
val range = d.range ?: return@forEach
val severity = if (d.severity == LyngDiagnosticSeverity.Warning) HighlightSeverity.WARNING else HighlightSeverity.ERROR
diags += Diag(range.start, range.endExclusive, d.message, severity)
// Map Enum constants from token highlighter to IDEA enum constant color
run {
val tokens = try { SimpleLyngHighlighter().highlight(text) } catch (_: Throwable) { emptyList() }
for (s in tokens) if (s.kind == HighlightKind.EnumConstant) {
val start = s.range.start
val end = s.range.endExclusive
if (start in 0..end && end <= text.length && start < end) {
putRange(start, end, LyngHighlighterColors.ENUM_CONSTANT)
}
}
}
return Result(collectedInfo.modStamp, out, diags)
}
// Build spell index payload: identifiers from symbols + references; comments/strings from simple highlighter
val idRanges = mutableSetOf<IntRange>()
try {
val binding = Binder.bind(text, mini)
for (sym in binding.symbols) {
val s = sym.declStart; val e = sym.declEnd
if (s in 0..e && e <= text.length && s < e) idRanges += (s until e)
}
for (ref in binding.references) {
val s = ref.start; val e = ref.end
if (s in 0..e && e <= text.length && s < e) idRanges += (s until e)
}
} catch (_: Throwable) {
// Best-effort; no identifiers if binder fails
}
val tokens = try { SimpleLyngHighlighter().highlight(text) } catch (_: Throwable) { emptyList() }
val commentRanges = tokens.filter { it.kind == HighlightKind.Comment }.map { it.range.start until it.range.endExclusive }
val stringRanges = tokens.filter { it.kind == HighlightKind.String }.map { it.range.start until it.range.endExclusive }
return Result(collectedInfo.modStamp, out, null,
spellIdentifiers = idRanges.toList(),
spellComments = commentRanges,
spellStrings = stringRanges)
}
override fun apply(file: PsiFile, annotationResult: Result?, holder: AnnotationHolder) {
if (annotationResult == null) return
// Skip if cache is up-to-date
val combinedStamp = LyngAstManager.getCombinedStamp(file)
val doc = file.viewProvider.document
val currentStamp = doc?.modificationStamp
val cached = file.getUserData(CACHE_KEY)
val result = if (cached != null && cached.modStamp == combinedStamp) cached else annotationResult
val result = if (cached != null && currentStamp != null && cached.modStamp == currentStamp) cached else annotationResult
file.putUserData(CACHE_KEY, result)
val doc = file.viewProvider.document
// Store spell index for spell/grammar engines to consume (suspend until ready)
val ids = result.spellIdentifiers.map { TextRange(it.first, it.last + 1) }
val coms = result.spellComments.map { TextRange(it.first, it.last + 1) }
val strs = result.spellStrings.map { TextRange(it.first, it.last + 1) }
net.sergeych.lyng.idea.spell.LyngSpellIndex.store(file,
net.sergeych.lyng.idea.spell.LyngSpellIndex.Data(
modStamp = result.modStamp,
identifiers = ids,
comments = coms,
strings = strs
)
)
// Optional diagnostic overlay: visualize the ranges we will feed to spellcheckers
val settings = net.sergeych.lyng.idea.settings.LyngFormatterSettings.getInstance(file.project)
if (settings.debugShowSpellFeed) {
fun paint(r: TextRange, label: String) {
holder.newAnnotation(HighlightSeverity.WEAK_WARNING, "spell-feed: $label")
.range(r)
.create()
}
ids.forEach { paint(it, "id") }
coms.forEach { paint(it, "comment") }
if (settings.spellCheckStringLiterals) strs.forEach { paint(it, "string") }
}
for (s in result.spans) {
holder.newSilentAnnotation(HighlightSeverity.INFORMATION)
@ -162,12 +341,13 @@ class LyngExternalAnnotator : ExternalAnnotator<LyngExternalAnnotator.Input, Lyn
.create()
}
// Show errors and warnings
result.diagnostics.forEach { d ->
val start = d.start.coerceIn(0, (doc?.textLength ?: 0))
val end = d.end.coerceIn(start, (doc?.textLength ?: start))
// Show syntax error if present
val err = result.error
if (err != null) {
val start = err.start.coerceIn(0, (doc?.textLength ?: 0))
val end = err.end.coerceIn(start, (doc?.textLength ?: start))
if (end > start) {
holder.newAnnotation(d.severity, d.message)
holder.newAnnotation(HighlightSeverity.ERROR, err.message)
.range(TextRange(start, end))
.create()
}
@ -178,15 +358,30 @@ class LyngExternalAnnotator : ExternalAnnotator<LyngExternalAnnotator.Input, Lyn
private val CACHE_KEY: Key<Result> = Key.create("LYNG_SEMANTIC_CACHE")
}
private fun prevNonWs(text: String, idxExclusive: Int): Int {
var i = idxExclusive - 1
while (i >= 0) {
val ch = text[i]
if (ch != ' ' && ch != '\t' && ch != '\n' && ch != '\r') return i
i--
}
return -1
}
/**
* Make the error highlight a bit wider than a single character so it is easier to see and click.
* Strategy:
* - If the offset points inside an identifier-like token (letters/digits/underscore), expand to the full token.
* - Otherwise select a small range starting at the offset with a minimum width, but not crossing the line end.
*/
private fun expandErrorRange(text: String, rawStart: Int): Pair<Int, Int> {
if (text.isEmpty()) return 0 to 0
val len = text.length
val start = rawStart.coerceIn(0, len)
fun isWord(ch: Char) = ch == '_' || ch.isLetterOrDigit()
if (start < len && isWord(text[start])) {
var s = start
var e = start
while (s > 0 && isWord(text[s - 1])) s--
while (e < len && isWord(text[e])) e++
return s to e
}
// Not inside a word: select a short, visible range up to EOL
val lineEnd = text.indexOf('\n', start).let { if (it == -1) len else it }
val minWidth = 4
val end = (start + minWidth).coerceAtMost(lineEnd).coerceAtLeast((start + 1).coerceAtMost(lineEnd))
return start to end
}
}

View File

@ -1,20 +1,3 @@
/*
* Copyright 2026 Sergey S. Chernov real.sergeych@gmail.com
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*
*/
/*
* Lightweight BASIC completion for Lyng, MVP version.
* Uses MiniAst (best-effort) + BuiltinDocRegistry to suggest symbols.
@ -26,15 +9,18 @@ import com.intellij.codeInsight.lookup.LookupElementBuilder
import com.intellij.icons.AllIcons
import com.intellij.openapi.diagnostic.Logger
import com.intellij.openapi.editor.Document
import com.intellij.openapi.util.Key
import com.intellij.patterns.PlatformPatterns
import com.intellij.psi.PsiFile
import com.intellij.util.ProcessingContext
import kotlinx.coroutines.runBlocking
import net.sergeych.lyng.Compiler
import net.sergeych.lyng.Source
import net.sergeych.lyng.highlight.offsetOf
import net.sergeych.lyng.idea.LyngLanguage
import net.sergeych.lyng.idea.highlight.LyngTokenTypes
import net.sergeych.lyng.idea.settings.LyngFormatterSettings
import net.sergeych.lyng.idea.util.DocsBootstrap
import net.sergeych.lyng.idea.util.LyngAstManager
import net.sergeych.lyng.idea.util.IdeLenientImportProvider
import net.sergeych.lyng.idea.util.TextCtx
import net.sergeych.lyng.miniast.*
@ -62,12 +48,6 @@ class LyngCompletionContributor : CompletionContributor() {
StdlibDocsBootstrap.ensure()
val file: PsiFile = parameters.originalFile
if (file.language != LyngLanguage) return
// Disable completion inside comments
val pos = parameters.position
val et = pos.node.elementType
if (et == LyngTokenTypes.LINE_COMMENT || et == LyngTokenTypes.BLOCK_COMMENT) return
// Feature toggle: allow turning completion off from settings
val settings = LyngFormatterSettings.getInstance(file.project)
if (!settings.enableLyngCompletionExperimental) return
@ -96,14 +76,12 @@ class LyngCompletionContributor : CompletionContributor() {
log.info("[LYNG_DEBUG] Completion: caret=$caret prefix='${prefix}' memberDotPos=${memberDotPos} file='${file.name}'")
}
// Build analysis (cached) for both global and member contexts to enable local class/val inference
val analysis = LyngAstManager.getAnalysis(file)
val mini = analysis?.mini
val binding = analysis?.binding
// Build MiniAst (cached) for both global and member contexts to enable local class/val inference
val mini = buildMiniAstCached(file, text)
// Delegate computation to the shared engine to keep behavior in sync with tests
val engineItems = try {
runBlocking { CompletionEngineLight.completeSuspend(text, caret, mini, binding) }
runBlocking { CompletionEngineLight.completeSuspend(text, caret) }
} catch (t: Throwable) {
if (DEBUG_COMPLETION) log.warn("[LYNG_DEBUG] Engine completion failed: ${t.message}")
emptyList()
@ -117,27 +95,26 @@ class LyngCompletionContributor : CompletionContributor() {
if (memberDotPos != null && engineItems.isEmpty()) {
if (DEBUG_COMPLETION) log.info("[LYNG_DEBUG] Fallback: engine returned 0 in member context; trying local inference")
// Build imported modules from text (lenient) + stdlib; avoid heavy MiniAst here
val fromText = DocLookupUtils.extractImportsFromText(text)
val fromText = extractImportsFromText(text)
val imported = LinkedHashSet<String>().apply {
fromText.forEach { add(it) }
add("lyng.stdlib")
}.toList()
val staticOnly = DocLookupUtils.isStaticReceiver(mini, text, memberDotPos, imported, binding)
// Try inferring return/receiver class around the dot
val inferred =
// Prefer MiniAst-based inference (return type from member call or receiver type)
DocLookupUtils.guessReturnClassFromMemberCallBeforeMini(mini, text, memberDotPos, imported, binding)
?: DocLookupUtils.guessReceiverClassViaMini(mini, text, memberDotPos, imported, binding)
guessReturnClassFromMemberCallBeforeMini(mini, text, memberDotPos, imported)
?: guessReceiverClassViaMini(mini, text, memberDotPos, imported)
?:
DocLookupUtils.guessReturnClassFromMemberCallBefore(text, memberDotPos, imported, mini)
?: DocLookupUtils.guessReturnClassFromTopLevelCallBefore(text, memberDotPos, imported, mini)
?: DocLookupUtils.guessReturnClassAcrossKnownCallees(text, memberDotPos, imported, mini)
?: DocLookupUtils.guessReceiverClass(text, memberDotPos, imported, mini)
guessReturnClassFromMemberCallBefore(text, memberDotPos, imported)
?: guessReturnClassFromTopLevelCallBefore(text, memberDotPos, imported)
?: guessReturnClassAcrossKnownCallees(text, memberDotPos, imported)
?: guessReceiverClass(text, memberDotPos, imported)
if (inferred != null) {
if (DEBUG_COMPLETION) log.info("[LYNG_DEBUG] Fallback inferred receiver/return class='$inferred' — offering its members")
offerMembers(emit, imported, inferred, staticOnly = staticOnly, sourceText = text, mini = mini)
offerMembers(emit, imported, inferred, sourceText = text, mini = mini)
return
} else {
if (DEBUG_COMPLETION) log.info("[LYNG_DEBUG] Fallback could not infer class; keeping list empty (no globals after dot)")
@ -145,6 +122,11 @@ class LyngCompletionContributor : CompletionContributor() {
}
}
// In global context, add params in scope first (engine does not include them)
if (memberDotPos == null && mini != null) {
offerParamsInScope(emit, mini, text, caret)
}
// Render engine items
for (ci in engineItems) {
val builder = when (ci.kind) {
@ -160,84 +142,57 @@ class LyngCompletionContributor : CompletionContributor() {
.withInsertHandler(ParenInsertHandler)
Kind.Class_ -> LookupElementBuilder.create(ci.name)
.withIcon(AllIcons.Nodes.Class)
Kind.Enum -> LookupElementBuilder.create(ci.name)
.withIcon(AllIcons.Nodes.Enum)
Kind.TypeAlias -> LookupElementBuilder.create(ci.name)
.withIcon(AllIcons.Nodes.Class)
Kind.Value -> LookupElementBuilder.create(ci.name)
.withIcon(AllIcons.Nodes.Variable)
.withIcon(AllIcons.Nodes.Field)
.let { b -> if (!ci.typeText.isNullOrBlank()) b.withTypeText(ci.typeText, true) else b }
Kind.Field -> LookupElementBuilder.create(ci.name)
.withIcon(AllIcons.Nodes.Field)
.let { b -> if (!ci.typeText.isNullOrBlank()) b.withTypeText(ci.typeText, true) else b }
}
if (ci.priority != 0.0) {
emit(PrioritizedLookupElement.withPriority(builder, ci.priority))
} else {
emit(builder)
}
emit(builder)
}
// In member context, ensure stdlib extension-like methods (e.g., String.re) are present
if (memberDotPos != null) {
val existing = engineItems.map { it.name }.toMutableSet()
val fromText = DocLookupUtils.extractImportsFromText(text)
val fromText = extractImportsFromText(text)
val imported = LinkedHashSet<String>().apply {
fromText.forEach { add(it) }
add("lyng.stdlib")
}.toList()
val inferredClass =
DocLookupUtils.guessReturnClassFromMemberCallBeforeMini(mini, text, memberDotPos, imported, binding)
?: DocLookupUtils.guessReceiverClassViaMini(mini, text, memberDotPos, imported, binding)
?: DocLookupUtils.guessReturnClassFromMemberCallBefore(text, memberDotPos, imported, mini)
?: DocLookupUtils.guessReturnClassFromTopLevelCallBefore(text, memberDotPos, imported, mini)
?: DocLookupUtils.guessReturnClassAcrossKnownCallees(text, memberDotPos, imported, mini)
?: DocLookupUtils.guessReceiverClass(text, memberDotPos, imported, mini)
guessReturnClassFromMemberCallBeforeMini(mini, text, memberDotPos, imported)
?: guessReceiverClassViaMini(mini, text, memberDotPos, imported)
?: guessReturnClassFromMemberCallBefore(text, memberDotPos, imported)
?: guessReturnClassFromTopLevelCallBefore(text, memberDotPos, imported)
?: guessReturnClassAcrossKnownCallees(text, memberDotPos, imported)
?: guessReceiverClass(text, memberDotPos, imported)
if (!inferredClass.isNullOrBlank()) {
val ext = DocLookupUtils.collectExtensionMemberNames(imported, inferredClass, mini)
if (DEBUG_COMPLETION) log.info("[LYNG_DEBUG] Post-engine extension check for $inferredClass: ${ext}")
val ext = BuiltinDocRegistry.extensionMethodNamesFor(inferredClass)
if (DEBUG_COMPLETION) log.info("[LYNG_DEBUG] Post-engine extension check for $inferredClass: ${'$'}{ext}")
for (name in ext) {
if (existing.contains(name)) continue
val resolved = DocLookupUtils.resolveMemberWithInheritance(imported, inferredClass, name, mini)
val resolved = DocLookupUtils.resolveMemberWithInheritance(imported, inferredClass, name)
if (resolved != null) {
val m = resolved.second
val builder = when (m) {
when (val member = resolved.second) {
is MiniMemberFunDecl -> {
val params = m.params.joinToString(", ") { it.name }
val ret = typeOf(m.returnType)
LookupElementBuilder.create(name)
val params = member.params.joinToString(", ") { it.name }
val ret = typeOf(member.returnType)
val builder = LookupElementBuilder.create(name)
.withIcon(AllIcons.Nodes.Method)
.withTailText("($params)", true)
.withTypeText(ret, true)
.withInsertHandler(ParenInsertHandler)
}
is MiniFunDecl -> {
val params = m.params.joinToString(", ") { it.name }
val ret = typeOf(m.returnType)
LookupElementBuilder.create(name)
.withIcon(AllIcons.Nodes.Method)
.withTailText("($params)", true)
.withTailText("(${ '$' }params)", true)
.withTypeText(ret, true)
.withInsertHandler(ParenInsertHandler)
emit(builder)
existing.add(name)
}
is MiniMemberValDecl -> {
LookupElementBuilder.create(name)
.withIcon(if (m.mutable) AllIcons.Nodes.Variable else AllIcons.Nodes.Field)
.withTypeText(typeOf(m.type), true)
}
is MiniValDecl -> {
LookupElementBuilder.create(name)
.withIcon(if (m.mutable) AllIcons.Nodes.Variable else AllIcons.Nodes.Field)
.withTypeText(typeOf(m.type), true)
}
else -> {
LookupElementBuilder.create(name)
.withIcon(AllIcons.Nodes.Method)
.withTailText("()", true)
.withInsertHandler(ParenInsertHandler)
val builder = LookupElementBuilder.create(name)
.withIcon(if (member.mutable) AllIcons.Nodes.Variable else AllIcons.Nodes.Field)
.withTypeText(typeOf(member.type), true)
emit(builder)
existing.add(name)
}
}
emit(builder)
existing.add(name)
} else {
// Fallback: emit simple method name without detailed types
val builder = LookupElementBuilder.create(name)
@ -253,18 +208,18 @@ class LyngCompletionContributor : CompletionContributor() {
// If in member context and engine items are suspiciously sparse, try to enrich via local inference + offerMembers
if (memberDotPos != null && engineItems.size < 3) {
if (DEBUG_COMPLETION) log.info("[LYNG_DEBUG] Engine produced only ${engineItems.size} items in member context — trying enrichment")
val fromText = DocLookupUtils.extractImportsFromText(text)
val fromText = extractImportsFromText(text)
val imported = LinkedHashSet<String>().apply {
fromText.forEach { add(it) }
add("lyng.stdlib")
}.toList()
val inferred =
DocLookupUtils.guessReturnClassFromMemberCallBeforeMini(mini, text, memberDotPos, imported, binding)
?: DocLookupUtils.guessReceiverClassViaMini(mini, text, memberDotPos, imported, binding)
?: DocLookupUtils.guessReturnClassFromMemberCallBefore(text, memberDotPos, imported, mini)
?: DocLookupUtils.guessReturnClassFromTopLevelCallBefore(text, memberDotPos, imported, mini)
?: DocLookupUtils.guessReturnClassAcrossKnownCallees(text, memberDotPos, imported, mini)
?: DocLookupUtils.guessReceiverClass(text, memberDotPos, imported, mini)
guessReturnClassFromMemberCallBeforeMini(mini, text, memberDotPos, imported)
?: guessReceiverClassViaMini(mini, text, memberDotPos, imported)
?: guessReturnClassFromMemberCallBefore(text, memberDotPos, imported)
?: guessReturnClassFromTopLevelCallBefore(text, memberDotPos, imported)
?: guessReturnClassAcrossKnownCallees(text, memberDotPos, imported)
?: guessReceiverClass(text, memberDotPos, imported)
if (inferred != null) {
if (DEBUG_COMPLETION) log.info("[LYNG_DEBUG] Enrichment inferred class='$inferred' — offering its members")
offerMembers(emit, imported, inferred, sourceText = text, mini = mini)
@ -294,11 +249,7 @@ class LyngCompletionContributor : CompletionContributor() {
.withIcon(kindIcon)
.withTypeText(typeOf(d.type), true)
}
is MiniEnumDecl -> LookupElementBuilder.create(name)
.withIcon(AllIcons.Nodes.Enum)
is MiniTypeAliasDecl -> LookupElementBuilder.create(name)
.withIcon(AllIcons.Nodes.Class)
.withTypeText(typeOf(d.target), true)
else -> LookupElementBuilder.create(name)
}
emit(builder)
}
@ -325,7 +276,9 @@ class LyngCompletionContributor : CompletionContributor() {
, sourceText: String,
mini: MiniScript? = null
) {
val classes = DocLookupUtils.aggregateClasses(imported, mini)
// Ensure modules are seeded in the registry (triggers lazy stdlib build too)
for (m in imported) BuiltinDocRegistry.docsForModule(m)
val classes = DocLookupUtils.aggregateClasses(imported)
if (DEBUG_COMPLETION) {
val keys = classes.keys.joinToString(", ")
log.info("[LYNG_DEBUG] offerMembers: imported=${imported} classes=[${keys}] target=${className}")
@ -344,7 +297,7 @@ class LyngCompletionContributor : CompletionContributor() {
}
// If MiniAst didn't populate members (empty), try to scan class body text for member signatures
if (localClass.members.isEmpty()) {
val scanned = DocLookupUtils.scanLocalClassMembersFromText(mini, text = sourceText, cls = localClass)
val scanned = scanLocalClassMembersFromText(mini, text = sourceText, cls = localClass)
if (DEBUG_COMPLETION) log.info("[LYNG_DEBUG] Local scan for class ${localClass.name}: found ${scanned.size} members -> ${scanned.keys}")
for ((name, sig) in scanned) {
when (sig.kind) {
@ -376,8 +329,6 @@ class LyngCompletionContributor : CompletionContributor() {
when (m) {
is MiniMemberFunDecl -> if (!m.isStatic) continue
is MiniMemberValDecl -> if (!m.isStatic) continue
is MiniMemberTypeAliasDecl -> if (!m.isStatic) continue
is MiniInitDecl -> continue
}
}
val list = target.getOrPut(m.name) { mutableListOf() }
@ -413,23 +364,18 @@ class LyngCompletionContributor : CompletionContributor() {
}
supplementPreferredBases(className)
fun emitGroup(map: LinkedHashMap<String, MutableList<MiniMemberDecl>>, groupPriority: Double) {
fun emitGroup(map: LinkedHashMap<String, MutableList<MiniMemberDecl>>) {
val keys = map.keys.sortedBy { it.lowercase() }
for (name in keys) {
val list = map[name] ?: continue
// Choose a representative for display:
// 1) Prefer a method with return type AND parameters
// 2) Prefer a method with parameters
// 3) Prefer a method with return type
// 4) Else any method
// 5) Else the first variant
// 1) Prefer a method with a known return type
// 2) Else any method
// 3) Else the first variant
val rep =
list.asSequence().filterIsInstance<MiniMemberFunDecl>()
.firstOrNull { it.returnType != null && it.params.isNotEmpty() }
?: list.asSequence().filterIsInstance<MiniMemberFunDecl>()
.firstOrNull { it.params.isNotEmpty() }
?: list.asSequence().filterIsInstance<MiniMemberFunDecl>()
.firstOrNull { it.returnType != null }
list.asSequence()
.filterIsInstance<MiniMemberFunDecl>()
.firstOrNull { it.returnType != null }
?: list.firstOrNull { it is MiniMemberFunDecl }
?: list.first()
when (rep) {
@ -445,11 +391,7 @@ class LyngCompletionContributor : CompletionContributor() {
.withTailText(tail, true)
.withTypeText(ret, true)
.withInsertHandler(ParenInsertHandler)
if (groupPriority != 0.0) {
emit(PrioritizedLookupElement.withPriority(builder, groupPriority))
} else {
emit(builder)
}
emit(builder)
}
is MiniMemberValDecl -> {
val icon = if (rep.mutable) AllIcons.Nodes.Variable else AllIcons.Nodes.Field
@ -459,31 +401,16 @@ class LyngCompletionContributor : CompletionContributor() {
.firstOrNull { it.type != null } ?: rep
val builder = LookupElementBuilder.create(name)
.withIcon(icon)
.withTypeText(typeOf(chosen.type), true)
if (groupPriority != 0.0) {
emit(PrioritizedLookupElement.withPriority(builder, groupPriority))
} else {
emit(builder)
}
.withTypeText(typeOf((chosen as MiniMemberValDecl).type), true)
emit(builder)
}
is MiniMemberTypeAliasDecl -> {
val builder = LookupElementBuilder.create(name)
.withIcon(AllIcons.Nodes.Class)
.withTypeText(typeOf(rep.target), true)
if (groupPriority != 0.0) {
emit(PrioritizedLookupElement.withPriority(builder, groupPriority))
} else {
emit(builder)
}
}
is MiniInitDecl -> {}
}
}
}
// Emit what we have first
emitGroup(directMap, 100.0)
emitGroup(inheritedMap, 0.0)
emitGroup(directMap)
emitGroup(inheritedMap)
// If suggestions are suspiciously sparse for known container classes,
// try to conservatively supplement using a curated list resolved via docs registry.
@ -505,47 +432,29 @@ class LyngCompletionContributor : CompletionContributor() {
for (name in common) {
if (name in already) continue
// Try resolve across classes first to get types/params; if it fails, emit a synthetic safe suggestion.
val resolved = DocLookupUtils.findMemberAcrossClasses(imported, name, mini)
val resolved = DocLookupUtils.findMemberAcrossClasses(imported, name)
if (resolved != null) {
val member = resolved.second
val builder = when (member) {
when (member) {
is MiniMemberFunDecl -> {
val params = member.params.joinToString(", ") { it.name }
val ret = typeOf(member.returnType)
LookupElementBuilder.create(name)
val builder = LookupElementBuilder.create(name)
.withIcon(AllIcons.Nodes.Method)
.withTailText("($params)", true)
.withTypeText(ret, true)
.withInsertHandler(ParenInsertHandler)
}
is MiniFunDecl -> {
val params = member.params.joinToString(", ") { it.name }
val ret = typeOf(member.returnType)
LookupElementBuilder.create(name)
.withIcon(AllIcons.Nodes.Method)
.withTailText("($params)", true)
.withTailText("(${params})", true)
.withTypeText(ret, true)
.withInsertHandler(ParenInsertHandler)
emit(builder)
already.add(name)
}
is MiniMemberValDecl -> {
LookupElementBuilder.create(name)
.withIcon(if (member.mutable) AllIcons.Nodes.Variable else AllIcons.Nodes.Field)
val builder = LookupElementBuilder.create(name)
.withIcon(AllIcons.Nodes.Field)
.withTypeText(typeOf(member.type), true)
}
is MiniValDecl -> {
LookupElementBuilder.create(name)
.withIcon(if (member.mutable) AllIcons.Nodes.Variable else AllIcons.Nodes.Field)
.withTypeText(typeOf(member.type), true)
}
else -> {
LookupElementBuilder.create(name)
.withIcon(AllIcons.Nodes.Method)
.withTailText("()", true)
.withInsertHandler(ParenInsertHandler)
emit(builder)
already.add(name)
}
}
emit(PrioritizedLookupElement.withPriority(builder, 50.0))
already.add(name)
} else {
// Synthetic fallback: method without detailed params/types to improve UX in absence of docs
val isProperty = name in setOf("size", "length")
@ -558,78 +467,468 @@ class LyngCompletionContributor : CompletionContributor() {
.withTailText("()", true)
.withInsertHandler(ParenInsertHandler)
}
emit(PrioritizedLookupElement.withPriority(builder, 50.0))
emit(builder)
already.add(name)
}
}
}
// Supplement with stdlib extension members defined in root.lyng (e.g., fun String.trim(...))
// Supplement with stdlib extension-like methods defined in root.lyng (e.g., fun String.trim(...))
run {
val already = (directMap.keys + inheritedMap.keys).toMutableSet()
val ext = BuiltinDocRegistry.extensionMemberNamesFor(className)
val ext = BuiltinDocRegistry.extensionMethodNamesFor(className)
if (DEBUG_COMPLETION) log.info("[LYNG_DEBUG] Extensions for $className: count=${ext.size} -> ${ext}")
for (name in ext) {
if (already.contains(name)) continue
// Try to resolve full signature via registry first to get params and return type
val resolved = DocLookupUtils.resolveMemberWithInheritance(imported, className, name, mini)
val resolved = DocLookupUtils.resolveMemberWithInheritance(imported, className, name)
if (resolved != null) {
val m = resolved.second
val builder = when (m) {
when (val member = resolved.second) {
is MiniMemberFunDecl -> {
val params = m.params.joinToString(", ") { it.name }
val ret = typeOf(m.returnType)
LookupElementBuilder.create(name)
val params = member.params.joinToString(", ") { it.name }
val ret = typeOf(member.returnType)
val builder = LookupElementBuilder.create(name)
.withIcon(AllIcons.Nodes.Method)
.withTailText("($params)", true)
.withTypeText(ret, true)
.withInsertHandler(ParenInsertHandler)
}
is MiniFunDecl -> {
val params = m.params.joinToString(", ") { it.name }
val ret = typeOf(m.returnType)
LookupElementBuilder.create(name)
.withIcon(AllIcons.Nodes.Method)
.withTailText("($params)", true)
.withTailText("(${params})", true)
.withTypeText(ret, true)
.withInsertHandler(ParenInsertHandler)
emit(builder)
already.add(name)
continue
}
is MiniMemberValDecl -> {
LookupElementBuilder.create(name)
.withIcon(if (m.mutable) AllIcons.Nodes.Variable else AllIcons.Nodes.Field)
.withTypeText(typeOf(m.type), true)
}
is MiniValDecl -> {
LookupElementBuilder.create(name)
.withIcon(if (m.mutable) AllIcons.Nodes.Variable else AllIcons.Nodes.Field)
.withTypeText(typeOf(m.type), true)
}
else -> {
LookupElementBuilder.create(name)
.withIcon(AllIcons.Nodes.Method)
.withTailText("()", true)
.withInsertHandler(ParenInsertHandler)
val builder = LookupElementBuilder.create(name)
.withIcon(if (member.mutable) AllIcons.Nodes.Variable else AllIcons.Nodes.Field)
.withTypeText(typeOf(member.type), true)
emit(builder)
already.add(name)
continue
}
}
emit(PrioritizedLookupElement.withPriority(builder, 50.0))
already.add(name)
continue
}
// Fallback: emit without detailed types if we couldn't resolve
val builder = LookupElementBuilder.create(name)
.withIcon(AllIcons.Nodes.Method)
.withTailText("()", true)
.withInsertHandler(ParenInsertHandler)
emit(PrioritizedLookupElement.withPriority(builder, 50.0))
emit(builder)
already.add(name)
}
}
}
// --- MiniAst-based inference helpers ---
private fun previousIdentifierBeforeDot(text: String, dotPos: Int): String? {
var i = dotPos - 1
// skip whitespace
while (i >= 0 && text[i].isWhitespace()) i--
val end = i + 1
while (i >= 0 && TextCtx.isIdentChar(text[i])) i--
val start = i + 1
return if (start < end) text.substring(start, end) else null
}
private fun guessReceiverClassViaMini(mini: MiniScript?, text: String, dotPos: Int, imported: List<String>): String? {
if (mini == null) return null
val ident = previousIdentifierBeforeDot(text, dotPos) ?: return null
// 1) Local val/var in the file
val valDecl = mini.declarations.filterIsInstance<MiniValDecl>().firstOrNull { it.name == ident }
val typeFromVal = valDecl?.type?.let { simpleClassNameOf(it) }
if (!typeFromVal.isNullOrBlank()) return typeFromVal
// If initializer exists, try to sniff ClassName(
val initR = valDecl?.initRange
if (initR != null) {
val src = mini.range.start.source
val s = src.offsetOf(initR.start)
val e = src.offsetOf(initR.end).coerceAtMost(text.length)
if (s in 0..e && e <= text.length) {
val init = text.substring(s, e)
Regex("([A-Za-z_][A-Za-z0-9_]*)\\s*\\(").find(init)?.let { m ->
val cls = m.groupValues[1]
return cls
}
}
}
// 2) Parameters in any function (best-effort without scope mapping)
val paramType = mini.declarations.filterIsInstance<MiniFunDecl>()
.asSequence()
.flatMap { it.params.asSequence() }
.firstOrNull { it.name == ident }?.type
val typeFromParam = simpleClassNameOf(paramType)
if (!typeFromParam.isNullOrBlank()) return typeFromParam
return null
}
private fun guessReturnClassFromMemberCallBeforeMini(mini: MiniScript?, text: String, dotPos: Int, imported: List<String>): String? {
if (mini == null) return null
var i = TextCtx.prevNonWs(text, dotPos - 1)
if (i < 0 || text[i] != ')') return null
// back to matching '('
i--
var depth = 0
while (i >= 0) {
when (text[i]) {
')' -> depth++
'(' -> if (depth == 0) break else depth--
}
i--
}
if (i < 0 || text[i] != '(') return null
var j = i - 1
while (j >= 0 && text[j].isWhitespace()) j--
val end = j + 1
while (j >= 0 && TextCtx.isIdentChar(text[j])) j--
val start = j + 1
if (start >= end) return null
val callee = text.substring(start, end)
// Ensure member call: dot before callee
var k = start - 1
while (k >= 0 && text[k].isWhitespace()) k--
if (k < 0 || text[k] != '.') return null
val prevDot = k
// Resolve receiver class via MiniAst (ident like `x`)
val receiverClass = guessReceiverClassViaMini(mini, text, prevDot, imported) ?: return null
// If receiver class is a locally declared class, resolve member on it
val localClass = mini.declarations.filterIsInstance<MiniClassDecl>().firstOrNull { it.name == receiverClass }
if (localClass != null) {
val mm = localClass.members.firstOrNull { it.name == callee }
if (mm != null) {
val rt = when (mm) {
is MiniMemberFunDecl -> mm.returnType
is MiniMemberValDecl -> mm.type
else -> null
}
return simpleClassNameOf(rt)
} else {
// Try to scan class body text for method signature and extract return type
val sigs = scanLocalClassMembersFromText(mini, text, localClass)
if (DEBUG_COMPLETION) log.info("[LYNG_DEBUG] Local scan for return type in ${receiverClass}.${callee}: candidates=${sigs.keys}")
val sig = sigs[callee]
if (sig != null && sig.typeText != null) return sig.typeText
}
}
// Else fallback to registry-based resolution (covers imported classes)
return DocLookupUtils.resolveMemberWithInheritance(imported, receiverClass, callee)?.second?.let { m ->
val rt = when (m) {
is MiniMemberFunDecl -> m.returnType
is MiniMemberValDecl -> m.type
}
simpleClassNameOf(rt)
}
}
private data class ScannedSig(val kind: String, val params: List<String>?, val typeText: String?)
private fun scanLocalClassMembersFromText(mini: MiniScript, text: String, cls: MiniClassDecl): Map<String, ScannedSig> {
val src = mini.range.start.source
val start = src.offsetOf(cls.bodyRange?.start ?: cls.range.start)
val end = src.offsetOf(cls.bodyRange?.end ?: cls.range.end).coerceAtMost(text.length)
if (start !in 0..end) return emptyMap()
val body = text.substring(start, end)
val map = LinkedHashMap<String, ScannedSig>()
// fun name(params): Type
val funRe = Regex("(?m)^\\s*fun\\s+([A-Za-z_][A-Za-z0-9_]*)\\s*\\(([^)]*)\\)\\s*(?::\\s*([A-Za-z_][A-Za-z0-9_]*))?")
for (m in funRe.findAll(body)) {
val name = m.groupValues.getOrNull(1) ?: continue
val params = m.groupValues.getOrNull(2)?.split(',')?.mapNotNull { it.trim().takeIf { it.isNotEmpty() } } ?: emptyList()
val type = m.groupValues.getOrNull(3)?.takeIf { it.isNotBlank() }
map[name] = ScannedSig("fun", params, type)
}
// val/var name: Type
val valRe = Regex("(?m)^\\s*(val|var)\\s+([A-Za-z_][A-Za-z0-9_]*)\\s*(?::\\s*([A-Za-z_][A-Za-z0-9_]*))?")
for (m in valRe.findAll(body)) {
val kind = m.groupValues.getOrNull(1) ?: continue
val name = m.groupValues.getOrNull(2) ?: continue
val type = m.groupValues.getOrNull(3)?.takeIf { it.isNotBlank() }
map.putIfAbsent(name, ScannedSig(kind, null, type))
}
return map
}
private fun guessReceiverClass(text: String, dotPos: Int, imported: List<String>): String? {
// 1) Try call-based: ClassName(...).
DocLookupUtils.guessClassFromCallBefore(text, dotPos, imported)?.let { return it }
// 2) Literal heuristics based on the immediate char before '.'
var i = TextCtx.prevNonWs(text, dotPos - 1)
if (i >= 0) {
when (text[i]) {
'"' -> {
// Either regular or triple-quoted string; both map to String
return "String"
}
']' -> return "List" // very rough heuristic
'}' -> return "Dict" // map/dictionary literal heuristic
')' -> {
// Parenthesized expression: walk back to matching '(' and inspect inner expression
var j = i - 1
var depth = 0
while (j >= 0) {
when (text[j]) {
')' -> depth++
'(' -> if (depth == 0) break else depth--
}
j--
}
if (j >= 0 && text[j] == '(') {
val innerS = (j + 1).coerceAtLeast(0)
val innerE = i.coerceAtMost(text.length)
if (innerS < innerE) {
val inner = text.substring(innerS, innerE).trim()
if (inner.startsWith('"') && inner.endsWith('"')) return "String"
if (inner.startsWith('[') && inner.endsWith(']')) return "List"
if (inner.startsWith('{') && inner.endsWith('}')) return "Dict"
}
}
}
}
// Numeric literal: support decimal, hex (0x..), and scientific notation (1e-3)
var j = i
var hasDigits = false
var hasDot = false
var hasExp = false
// Walk over digits, letters for hex, dots, and exponent markers
while (j >= 0) {
val ch = text[j]
if (ch.isDigit()) { hasDigits = true; j-- ; continue }
if (ch == '.') { hasDot = true; j-- ; continue }
if (ch == 'e' || ch == 'E') { hasExp = true; j-- ; // optional sign directly before digits
if (j >= 0 && (text[j] == '+' || text[j] == '-')) j--
continue
}
if (ch in listOf('x','X')) { // part of 0x prefix
j--
continue
}
if (ch == 'a' || ch == 'b' || ch == 'c' || ch == 'd' || ch == 'f' ||
ch == 'A' || ch == 'B' || ch == 'C' || ch == 'D' || ch == 'F') {
// hex digit in 0x...
j--
continue
}
break
}
// Now check for 0x/0X prefix
val k = j
val isHex = k >= 1 && text[k] == '0' && (text[k+1] == 'x' || text[k+1] == 'X')
if (hasDigits) {
return if (isHex) "Int" else if (hasDot || hasExp) "Real" else "Int"
}
}
return null
}
/**
* Try to infer the class of the return value of the member call immediately before the dot.
* Example: `Path(".." ).lines().<caret>` detects `lines()` on receiver class `Path` and returns `Iterator`.
*/
private fun guessReturnClassFromMemberCallBefore(text: String, dotPos: Int, imported: List<String>): String? {
var i = TextCtx.prevNonWs(text, dotPos - 1)
if (i < 0) return null
// We expect a call just before the dot, i.e., ')' ... '.'
if (text[i] != ')') return null
// Walk back to matching '('
i--
var depth = 0
while (i >= 0) {
val ch = text[i]
when (ch) {
')' -> depth++
'(' -> if (depth == 0) break else depth--
}
i--
}
if (i < 0 || text[i] != '(') return null
// Identify callee identifier just before '('
var j = i - 1
while (j >= 0 && text[j].isWhitespace()) j--
val end = j + 1
while (j >= 0 && TextCtx.isIdentChar(text[j])) j--
val start = j + 1
if (start >= end) return null
val callee = text.substring(start, end)
// Ensure it's a member call (there must be a dot immediately before the callee, ignoring spaces)
var k = start - 1
while (k >= 0 && text[k].isWhitespace()) k--
if (k < 0 || text[k] != '.') return null
val prevDot = k
// Infer receiver class at the previous dot
val receiverClass = guessReceiverClass(text, prevDot, imported) ?: return null
// Resolve the callee as a member of receiver class, including inheritance
val resolved = DocLookupUtils.resolveMemberWithInheritance(imported, receiverClass, callee) ?: return null
val member = resolved.second
val returnType = when (member) {
is MiniMemberFunDecl -> member.returnType
is MiniMemberValDecl -> member.type
}
return simpleClassNameOf(returnType)
}
/**
* Infer return class of a top-level call right before the dot: e.g., `files().<caret>`.
* We extract callee name and resolve it among imported modules' top-level functions.
*/
private fun guessReturnClassFromTopLevelCallBefore(text: String, dotPos: Int, imported: List<String>): String? {
var i = TextCtx.prevNonWs(text, dotPos - 1)
if (i < 0 || text[i] != ')') return null
// Walk back to matching '('
i--
var depth = 0
while (i >= 0) {
val ch = text[i]
when (ch) {
')' -> depth++
'(' -> if (depth == 0) break else depth--
}
i--
}
if (i < 0 || text[i] != '(') return null
// Extract callee ident before '('
var j = i - 1
while (j >= 0 && text[j].isWhitespace()) j--
val end = j + 1
while (j >= 0 && TextCtx.isIdentChar(text[j])) j--
val start = j + 1
if (start >= end) return null
val callee = text.substring(start, end)
// If it's a member call, bail out (handled in member-call inference)
var k = start - 1
while (k >= 0 && text[k].isWhitespace()) k--
if (k >= 0 && text[k] == '.') return null
// Resolve top-level function in imported modules
for (mod in imported) {
val decls = BuiltinDocRegistry.docsForModule(mod)
val fn = decls.asSequence().filterIsInstance<MiniFunDecl>().firstOrNull { it.name == callee }
if (fn != null) return simpleClassNameOf(fn.returnType)
}
return null
}
/**
* Fallback: if we can at least extract a callee name before the dot and it exists across common classes,
* derive its return type using cross-class lookup (Iterable/Iterator/List preference). This ignores the receiver.
* Example: `something.lines().<caret>` where `something` type is unknown, but `lines()` commonly returns Iterator<String>.
*/
private fun guessReturnClassAcrossKnownCallees(text: String, dotPos: Int, imported: List<String>): String? {
var i = TextCtx.prevNonWs(text, dotPos - 1)
if (i < 0 || text[i] != ')') return null
// Walk back to matching '('
i--
var depth = 0
while (i >= 0) {
val ch = text[i]
when (ch) {
')' -> depth++
'(' -> if (depth == 0) break else depth--
}
i--
}
if (i < 0 || text[i] != '(') return null
// Extract callee ident before '('
var j = i - 1
while (j >= 0 && text[j].isWhitespace()) j--
val end = j + 1
while (j >= 0 && TextCtx.isIdentChar(text[j])) j--
val start = j + 1
if (start >= end) return null
val callee = text.substring(start, end)
// Try cross-class resolution
val resolved = DocLookupUtils.findMemberAcrossClasses(imported, callee) ?: return null
val member = resolved.second
val returnType = when (member) {
is MiniMemberFunDecl -> member.returnType
is MiniMemberValDecl -> member.type
}
return simpleClassNameOf(returnType)
}
/** Convert a MiniTypeRef to a simple class name as used by docs (e.g., Iterator from Iterator<String>). */
private fun simpleClassNameOf(t: MiniTypeRef?): String? = when (t) {
null -> null
is MiniTypeName -> t.segments.lastOrNull()?.name
is MiniGenericType -> simpleClassNameOf(t.base)
is MiniFunctionType -> null
is MiniTypeVar -> null
}
private fun buildMiniAst(text: String): MiniScript? {
return try {
val sink = MiniAstBuilder()
val provider = IdeLenientImportProvider.create()
val src = Source("<ide>", text)
runBlocking { Compiler.compileWithMini(src, provider, sink) }
sink.build()
} catch (_: Throwable) {
null
}
}
// Cached per PsiFile by document modification stamp
private val MINI_KEY = Key.create<MiniScript>("lyng.mini.cache")
private val STAMP_KEY = Key.create<Long>("lyng.mini.cache.stamp")
private fun buildMiniAstCached(file: PsiFile, text: String): MiniScript? {
val doc = file.viewProvider.document ?: return null
val stamp = doc.modificationStamp
val prevStamp = file.getUserData(STAMP_KEY)
val cached = file.getUserData(MINI_KEY)
if (cached != null && prevStamp != null && prevStamp == stamp) return cached
val built = buildMiniAst(text)
// Cache even null? avoid caching failures; only cache non-null
if (built != null) {
file.putUserData(MINI_KEY, built)
file.putUserData(STAMP_KEY, stamp)
}
return built
}
private fun offerParamsInScope(emit: (com.intellij.codeInsight.lookup.LookupElement) -> Unit, mini: MiniScript, text: String, caret: Int) {
val src = mini.range.start.source
// Find function whose body contains caret or whose whole range contains caret
val fns = mini.declarations.filterIsInstance<MiniFunDecl>()
for (fn in fns) {
val start = src.offsetOf(fn.range.start)
val end = src.offsetOf(fn.range.end).coerceAtMost(text.length)
if (caret in start..end) {
for (p in fn.params) {
val builder = LookupElementBuilder.create(p.name)
.withIcon(AllIcons.Nodes.Variable)
.withTypeText(typeOf(p.type), true)
emit(builder)
}
return
}
}
}
// Lenient textual import extractor (duplicated from QuickDoc privately)
private fun extractImportsFromText(text: String): List<String> {
val result = LinkedHashSet<String>()
val re = Regex("(?m)^\\s*import\\s+([a-zA-Z_][a-zA-Z0-9_]*(?:\\.[a-zA-Z_][a-zA-Z0-9_]*)*)")
re.findAll(text).forEach { m ->
val raw = m.groupValues.getOrNull(1)?.trim().orEmpty()
if (raw.isNotEmpty()) {
val canon = if (raw.startsWith("lyng.")) raw else "lyng.$raw"
result.add(canon)
}
}
return result.toList()
}
private fun typeOf(t: MiniTypeRef?): String {
val s = DocLookupUtils.typeOf(t)
return if (s.isEmpty()) "" else ": $s"
return when (t) {
null -> ""
is MiniTypeName -> t.segments.lastOrNull()?.name?.let { ": $it" } ?: ""
is MiniGenericType -> {
val base = typeOf(t.base).removePrefix(": ")
val args = t.args.joinToString(",") { typeOf(it).removePrefix(": ") }
": ${base}<${args}>"
}
is MiniFunctionType -> ": (fn)"
is MiniTypeVar -> ": ${t.name}"
}
}
}
}

View File

@ -1,5 +1,5 @@
/*
* Copyright 2026 Sergey S. Chernov real.sergeych@gmail.com
* Copyright 2025 Sergey S. Chernov real.sergeych@gmail.com
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
@ -25,13 +25,14 @@ import com.intellij.openapi.util.TextRange
import com.intellij.psi.PsiElement
import com.intellij.psi.PsiFile
import kotlinx.coroutines.runBlocking
import net.sergeych.lyng.Compiler
import net.sergeych.lyng.Pos
import net.sergeych.lyng.Source
import net.sergeych.lyng.highlight.offsetOf
import net.sergeych.lyng.idea.LyngLanguage
import net.sergeych.lyng.idea.util.LyngAstManager
import net.sergeych.lyng.idea.util.IdeLenientImportProvider
import net.sergeych.lyng.idea.util.TextCtx
import net.sergeych.lyng.miniast.*
import net.sergeych.lyng.tools.LyngLanguageTools
import net.sergeych.lyng.tools.LyngSymbolInfo
/**
* Quick Docs backed by MiniAst: when caret is on an identifier that corresponds
@ -68,244 +69,59 @@ class LyngDocumentationProvider : AbstractDocumentationProvider() {
val ident = text.substring(idRange.startOffset, idRange.endOffset)
if (DEBUG_LOG) log.info("[LYNG_DEBUG] QuickDoc: ident='$ident' at ${idRange.startOffset}..${idRange.endOffset} in ${file.name}")
// 1. Get merged mini-AST from Manager (handles local + .lyng.d merged declarations)
val analysis = LyngAstManager.getAnalysis(file) ?: return null
val mini = analysis.mini ?: return null
val miniSource = mini.range.start.source
val imported = analysis.importedModules.ifEmpty { DocLookupUtils.canonicalImportedModules(mini, text) }
// Single-source quick doc lookup
LyngLanguageTools.docAt(analysis, offset)?.let { info ->
val enriched = if (info.doc == null) {
findDocInDeclarationFiles(file, info.target.containerName, info.target.name)
?.let { info.copy(doc = it) } ?: info
} else {
info
}
renderDocFromInfo(enriched)?.let { return it }
// Build MiniAst for this file (fast and resilient). Best-effort; on failure continue with registry lookup only.
val sink = MiniAstBuilder()
// Use lenient import provider so unresolved imports (e.g., lyng.io.fs) don't break docs
val provider = IdeLenientImportProvider.create()
val src = Source("<ide>", text)
var mini: MiniScript? = try {
runBlocking { Compiler.compileWithMini(src, provider, sink) }
sink.build()
} catch (t: Throwable) {
// Do not bail out completely: we still can resolve built-in and imported docs (e.g., println)
if (DEBUG_LOG) log.warn("[LYNG_DEBUG] QuickDoc: compileWithMini failed: ${t.message}")
null
}
// Fallback: resolve references against merged MiniAst (including .lyng.d) when binder cannot
run {
val dotPos = DocLookupUtils.findDotLeft(text, idRange.startOffset)
if (dotPos != null) {
val receiverClass = DocLookupUtils.guessReceiverClassViaMini(mini, text, dotPos, imported, analysis.binding)
?: DocLookupUtils.guessReceiverClass(text, dotPos, imported, mini)
if (receiverClass != null) {
val resolved = DocLookupUtils.resolveMemberWithInheritance(imported, receiverClass, ident, mini)
if (resolved != null) {
val owner = resolved.first
val member = resolved.second
val withDoc = if (member.doc == null) {
findDocInDeclarationFiles(file, owner, member.name)?.let { doc ->
when (member) {
is MiniMemberFunDecl -> member.copy(doc = doc)
is MiniMemberValDecl -> member.copy(doc = doc)
is MiniMemberTypeAliasDecl -> member.copy(doc = doc)
else -> member
}
} ?: member
} else {
member
}
return when (withDoc) {
is MiniMemberFunDecl -> renderMemberFunDoc(owner, withDoc)
is MiniMemberValDecl -> renderMemberValDoc(owner, withDoc)
is MiniMemberTypeAliasDecl -> renderMemberTypeAliasDoc(owner, withDoc)
else -> null
}
}
}
} else {
mini.declarations.firstOrNull { it.name == ident }?.let { decl ->
return renderDeclDoc(decl, text, mini, imported)
}
}
val haveMini = mini != null
if (mini == null) {
// Ensure we have a dummy script object to avoid NPE in downstream helpers that expect a MiniScript
mini = MiniScript(MiniRange(Pos(src, 1, 1), Pos(src, 1, 1)))
}
val source = src
// Try resolve to: function param at position, function/class/val declaration at position
// 1) Use unified declaration detection
DocLookupUtils.findDeclarationAt(mini, offset, ident)?.let { (name, kind) ->
if (DEBUG_LOG) log.info("[LYNG_DEBUG] QuickDoc: matched declaration '$name' kind=$kind")
// Find the actual declaration object to render
mini.declarations.forEach { d ->
if (d.name == name) {
val s: Int = miniSource.offsetOf(d.nameStart)
if (s <= offset && s + d.name.length > offset) {
return renderDeclDoc(d, text, mini, imported)
}
}
// Handle members if it was a member
if (d is MiniClassDecl) {
d.members.forEach { m ->
if (m.name == name) {
val s: Int = miniSource.offsetOf(m.nameStart)
if (s <= offset && s + m.name.length > offset) {
return when (m) {
is MiniMemberFunDecl -> renderMemberFunDoc(d.name, m)
is MiniMemberValDecl -> renderMemberValDoc(d.name, m)
is MiniMemberTypeAliasDecl -> renderMemberTypeAliasDoc(d.name, m)
else -> null
}
}
}
}
d.ctorFields.forEach { cf ->
if (cf.name == name) {
val s: Int = miniSource.offsetOf(cf.nameStart)
if (s <= offset && s + cf.name.length > offset) {
// Render as a member val
val mv = MiniMemberValDecl(
range = MiniRange(cf.nameStart, cf.nameStart), // dummy
name = cf.name,
mutable = cf.mutable,
type = cf.type,
initRange = null,
doc = null,
nameStart = cf.nameStart
)
return renderMemberValDoc(d.name, mv)
}
}
}
d.classFields.forEach { cf ->
if (cf.name == name) {
val s: Int = miniSource.offsetOf(cf.nameStart)
if (s <= offset && s + cf.name.length > offset) {
// Render as a member val
val mv = MiniMemberValDecl(
range = MiniRange(cf.nameStart, cf.nameStart), // dummy
name = cf.name,
mutable = cf.mutable,
type = cf.type,
initRange = null,
doc = null,
nameStart = cf.nameStart
)
return renderMemberValDoc(d.name, mv)
}
}
}
}
if (d is MiniEnumDecl) {
if (d.entries.contains(name)) {
val s: Int = miniSource.offsetOf(d.range.start)
val e: Int = miniSource.offsetOf(d.range.end)
if (offset >= s && offset <= e) {
// For enum constant, we don't have detailed docs in MiniAst yet, but we can render a title
return renderTitle("enum constant ${d.name}.${name}")
}
}
}
// 1) Check declarations whose name range contains offset
if (haveMini) for (d in mini.declarations) {
val s = source.offsetOf(d.nameStart)
val e = (s + d.name.length).coerceAtMost(text.length)
if (offset in s until e) {
if (DEBUG_LOG) log.info("[LYNG_DEBUG] QuickDoc: matched decl '${d.name}' kind=${d::class.simpleName}")
return renderDeclDoc(d)
}
// Check parameters
mini.declarations.filterIsInstance<MiniFunDecl>().forEach { fn ->
fn.params.forEach { p ->
if (p.name == name) {
val s: Int = miniSource.offsetOf(p.nameStart)
if (s <= offset && s + p.name.length > offset) {
return renderParamDoc(fn, p)
}
}
}
// 2) Check parameters of functions
if (haveMini) for (fn in mini.declarations.filterIsInstance<MiniFunDecl>()) {
for (p in fn.params) {
val s = source.offsetOf(p.nameStart)
val e = (s + p.name.length).coerceAtMost(text.length)
if (offset in s until e) {
if (DEBUG_LOG) log.info("[LYNG_DEBUG] QuickDoc: matched param '${p.name}' in fun '${fn.name}'")
return renderParamDoc(fn, p)
}
}
}
// 3) usages in current file via Binder (resolves local variables, parameters, and classes)
try {
val binding = net.sergeych.lyng.binding.Binder.bind(text, mini)
val ref = binding.references.firstOrNull { offset in it.start until it.end }
if (ref != null) {
val sym = binding.symbols.firstOrNull { it.id == ref.symbolId }
if (sym != null) {
// Find local declaration that matches this symbol
var dsFound: MiniDecl? = null
mini.declarations.forEach { decl ->
if (decl.name == sym.name) {
val sOffset: Int = miniSource.offsetOf(decl.nameStart)
if (sOffset == sym.declStart) {
dsFound = decl
}
}
}
if (dsFound != null) return renderDeclDoc(dsFound, text, mini, imported)
// Check parameters
mini.declarations.filterIsInstance<MiniFunDecl>().forEach { fn ->
fn.params.forEach { p ->
if (p.name == sym.name) {
val sOffset: Int = miniSource.offsetOf(p.nameStart)
if (sOffset == sym.declStart) {
return renderParamDoc(fn, p)
}
}
}
}
// Check class members (fields/functions)
mini.declarations.filterIsInstance<MiniClassDecl>().forEach { cls ->
cls.members.forEach { m ->
if (m.name == sym.name) {
val sOffset: Int = miniSource.offsetOf(m.nameStart)
if (sOffset == sym.declStart) {
return when (m) {
is MiniMemberFunDecl -> renderMemberFunDoc(cls.name, m)
is MiniMemberValDecl -> renderMemberValDoc(cls.name, m)
is MiniMemberTypeAliasDecl -> renderMemberTypeAliasDoc(cls.name, m)
else -> null
}
}
}
}
cls.ctorFields.forEach { cf ->
if (cf.name == sym.name) {
val sOffset: Int = miniSource.offsetOf(cf.nameStart)
if (sOffset == sym.declStart) {
// Render as a member val
val mv = MiniMemberValDecl(
range = MiniRange(cf.nameStart, cf.nameStart), // dummy
name = cf.name,
mutable = cf.mutable,
type = cf.type,
initRange = null,
doc = null,
nameStart = cf.nameStart
)
return renderMemberValDoc(cls.name, mv)
}
}
}
cls.classFields.forEach { cf ->
if (cf.name == sym.name) {
val sOffset: Int = miniSource.offsetOf(cf.nameStart)
if (sOffset == sym.declStart) {
// Render as a member val
val mv = MiniMemberValDecl(
range = MiniRange(cf.nameStart, cf.nameStart), // dummy
name = cf.name,
mutable = cf.mutable,
type = cf.type,
initRange = null,
doc = null,
nameStart = cf.nameStart
)
return renderMemberValDoc(cls.name, mv)
}
}
}
}
}
}
} catch (e: Throwable) {
if (DEBUG_LOG) log.warn("[LYNG_DEBUG] QuickDoc: local binder resolution failed: ${e.message}")
}
// 4) Member-context resolution first (dot immediately before identifier): handle literals and calls
// 3) Member-context resolution first (dot immediately before identifier): handle literals and calls
run {
val dotPos = TextCtx.findDotLeft(text, idRange.startOffset)
?: TextCtx.findDotLeft(text, offset)
if (dotPos != null) {
// Build imported modules (MiniAst-derived if available, else lenient from text) and ensure stdlib is present
val importedModules = DocLookupUtils.canonicalImportedModules(mini, text)
var importedModules = if (haveMini) DocLookupUtils.canonicalImportedModules(mini) else emptyList()
if (importedModules.isEmpty()) {
val fromText = extractImportsFromText(text)
importedModules = if (fromText.isEmpty()) listOf("lyng.stdlib") else fromText
}
if (!importedModules.contains("lyng.stdlib")) importedModules = importedModules + "lyng.stdlib"
// Try literal and call-based receiver inference around the dot
val i = TextCtx.prevNonWs(text, dotPos - 1)
@ -338,43 +154,15 @@ class LyngDocumentationProvider : AbstractDocumentationProvider() {
} else null
} else null
}
else -> {
DocLookupUtils.guessReceiverClassViaMini(mini, text, dotPos, importedModules)
?: DocLookupUtils.guessClassFromCallBefore(text, dotPos, importedModules, mini)
?: run {
// handle this@Type or as Type
val i2 = TextCtx.prevNonWs(text, dotPos - 1)
if (i2 >= 0) {
val identRange = TextCtx.wordRangeAt(text, i2 + 1)
if (identRange != null) {
val id = text.substring(identRange.startOffset, identRange.endOffset)
val k = TextCtx.prevNonWs(text, identRange.startOffset - 1)
if (k >= 1 && text[k] == 's' && text[k - 1] == 'a' && (k - 1 == 0 || !text[k - 2].isLetterOrDigit())) {
id
} else if (k >= 0 && text[k] == '@') {
val k2 = TextCtx.prevNonWs(text, k - 1)
if (k2 >= 3 && text.substring(k2 - 3, k2 + 1) == "this") id else null
} else null
} else null
} else null
}
}
else -> DocLookupUtils.guessClassFromCallBefore(text, dotPos, importedModules)
}
if (DEBUG_LOG) log.info("[LYNG_DEBUG] QuickDoc: memberCtx dotPos=${dotPos} chBeforeDot='${if (dotPos > 0) text[dotPos - 1] else ' '}' classGuess=${className} imports=${importedModules}")
if (DEBUG_LOG) log.info("[LYNG_DEBUG] QuickDoc: memberCtx dotPos=${dotPos} chBeforeDot='${if (dotPos>0) text[dotPos-1] else ' '}' classGuess=${className} imports=${importedModules}")
if (className != null) {
val staticOnly = DocLookupUtils.isStaticReceiver(mini, text, dotPos, importedModules, analysis.binding)
DocLookupUtils.resolveMemberWithInheritance(importedModules, className, ident, mini, staticOnly = staticOnly)?.let { (owner, member) ->
DocLookupUtils.resolveMemberWithInheritance(importedModules, className, ident)?.let { (owner, member) ->
if (DEBUG_INHERITANCE) log.info("[LYNG_DEBUG] QuickDoc: literal/call '$ident' resolved to $owner.${member.name}")
return when (member) {
is MiniMemberFunDecl -> renderMemberFunDoc(owner, member)
is MiniMemberValDecl -> renderMemberValDoc(owner, member)
is MiniMemberTypeAliasDecl -> renderMemberTypeAliasDoc(owner, member)
is MiniInitDecl -> null
is MiniFunDecl -> renderDeclDoc(member, text, mini, importedModules)
is MiniValDecl -> renderDeclDoc(member, text, mini, importedModules)
is MiniClassDecl -> renderDeclDoc(member, text, mini, importedModules)
is MiniEnumDecl -> renderDeclDoc(member, text, mini, importedModules)
is MiniTypeAliasDecl -> renderDeclDoc(member, text, mini, importedModules)
}
}
log.info("[LYNG_DEBUG] QuickDoc: resolve failed for ${className}.${ident}")
@ -383,14 +171,21 @@ class LyngDocumentationProvider : AbstractDocumentationProvider() {
}
// 4) As a fallback, if the caret is on an identifier text that matches any declaration name, show that
mini.declarations.firstOrNull { it.name == ident }?.let {
if (haveMini) mini.declarations.firstOrNull { it.name == ident }?.let {
log.info("[LYNG_DEBUG] QuickDoc: fallback by name '${it.name}' kind=${it::class.simpleName}")
return renderDeclDoc(it, text, mini, imported)
return renderDeclDoc(it)
}
// 4) Consult BuiltinDocRegistry for imported modules (top-level and class members)
// Canonicalize import names using ImportManager, as users may write shortened names (e.g., "io.fs")
var importedModules = DocLookupUtils.canonicalImportedModules(mini, text)
var importedModules = if (haveMini) DocLookupUtils.canonicalImportedModules(mini) else emptyList()
// If MiniAst failed or captured no imports, try a lightweight textual import scan
if (importedModules.isEmpty()) {
val fromText = extractImportsFromText(text)
if (fromText.isNotEmpty()) {
importedModules = fromText
}
}
// Always include stdlib as a fallback context
if (!importedModules.contains("lyng.stdlib")) importedModules = importedModules + "lyng.stdlib"
// 4a) try top-level decls
@ -405,14 +200,12 @@ class LyngDocumentationProvider : AbstractDocumentationProvider() {
if (arity != null && chosen.params.size != arity && matches.size > 1) {
return renderOverloads(ident, matches)
}
return renderDeclDoc(chosen, text, mini, imported)
return renderDeclDoc(chosen)
}
// Also allow values/consts
docs.filterIsInstance<MiniValDecl>().firstOrNull { it.name == ident }?.let { return renderDeclDoc(it, text, mini, imported) }
// And classes/enums
docs.filterIsInstance<MiniClassDecl>().firstOrNull { it.name == ident }?.let { return renderDeclDoc(it, text, mini, imported) }
docs.filterIsInstance<MiniEnumDecl>().firstOrNull { it.name == ident }?.let { return renderDeclDoc(it, text, mini, imported) }
docs.filterIsInstance<MiniTypeAliasDecl>().firstOrNull { it.name == ident }?.let { return renderDeclDoc(it, text, mini, imported) }
docs.filterIsInstance<MiniValDecl>().firstOrNull { it.name == ident }?.let { return renderDeclDoc(it) }
// And classes
docs.filterIsInstance<MiniClassDecl>().firstOrNull { it.name == ident }?.let { return renderDeclDoc(it) }
}
// Defensive fallback: if nothing found and it's a well-known stdlib function, render minimal inline docs
if (ident == "println" || ident == "print") {
@ -420,26 +213,17 @@ class LyngDocumentationProvider : AbstractDocumentationProvider() {
"Print values to the standard output and append a newline. Accepts any number of arguments." else
"Print values to the standard output without a trailing newline. Accepts any number of arguments."
val title = "function $ident(values)"
return renderTitle(title) + styledMarkdown(htmlEscape(fallback))
return "<div class='doc-title'>${htmlEscape(title)}</div>" + styledMarkdown(htmlEscape(fallback))
}
// 4b) try class members like ClassName.member with inheritance fallback
val lhs = previousWordBefore(text, idRange.startOffset)
if (lhs != null && hasDotBetween(text, lhs.endOffset, idRange.startOffset)) {
val className = text.substring(lhs.startOffset, lhs.endOffset)
val dotPos = findDotLeft(text, idRange.startOffset)
val staticOnly = dotPos?.let { DocLookupUtils.isStaticReceiver(mini, text, it, importedModules, analysis.binding) } ?: false
DocLookupUtils.resolveMemberWithInheritance(importedModules, className, ident, mini, staticOnly = staticOnly)?.let { (owner, member) ->
DocLookupUtils.resolveMemberWithInheritance(importedModules, className, ident)?.let { (owner, member) ->
if (DEBUG_INHERITANCE) log.info("[LYNG_DEBUG] Inheritance resolved $className.$ident to $owner.${member.name}")
return when (member) {
is MiniMemberFunDecl -> renderMemberFunDoc(owner, member)
is MiniMemberValDecl -> renderMemberValDoc(owner, member)
is MiniMemberTypeAliasDecl -> renderMemberTypeAliasDoc(owner, member)
is MiniInitDecl -> null
is MiniFunDecl -> renderDeclDoc(member, text, mini, importedModules)
is MiniValDecl -> renderDeclDoc(member, text, mini, importedModules)
is MiniClassDecl -> renderDeclDoc(member, text, mini, importedModules)
is MiniEnumDecl -> renderDeclDoc(member, text, mini, importedModules)
is MiniTypeAliasDecl -> renderDeclDoc(member, text, mini, importedModules)
}
}
} else {
@ -450,22 +234,14 @@ class LyngDocumentationProvider : AbstractDocumentationProvider() {
if (dotPos != null) {
val guessed = when {
looksLikeListLiteralBefore(text, dotPos) -> "List"
else -> DocLookupUtils.guessClassFromCallBefore(text, dotPos, importedModules, mini)
else -> DocLookupUtils.guessClassFromCallBefore(text, dotPos, importedModules)
}
if (guessed != null) {
val staticOnly = DocLookupUtils.isStaticReceiver(mini, text, dotPos, importedModules, analysis.binding)
DocLookupUtils.resolveMemberWithInheritance(importedModules, guessed, ident, mini, staticOnly = staticOnly)?.let { (owner, member) ->
DocLookupUtils.resolveMemberWithInheritance(importedModules, guessed, ident)?.let { (owner, member) ->
if (DEBUG_INHERITANCE) log.info("[LYNG_DEBUG] Heuristic '$guessed.$ident' resolved via inheritance to $owner.${member.name}")
return when (member) {
is MiniMemberFunDecl -> renderMemberFunDoc(owner, member)
is MiniMemberValDecl -> renderMemberValDoc(owner, member)
is MiniMemberTypeAliasDecl -> renderMemberTypeAliasDoc(owner, member)
is MiniInitDecl -> null
is MiniFunDecl -> renderDeclDoc(member, text, mini, importedModules)
is MiniValDecl -> renderDeclDoc(member, text, mini, importedModules)
is MiniClassDecl -> renderDeclDoc(member, text, mini, importedModules)
is MiniEnumDecl -> renderDeclDoc(member, text, mini, importedModules)
is MiniTypeAliasDecl -> renderDeclDoc(member, text, mini, importedModules)
}
}
} else {
@ -473,26 +249,18 @@ class LyngDocumentationProvider : AbstractDocumentationProvider() {
run {
val candidates = listOf("String", "Iterable", "Iterator", "List", "Collection", "Array", "Dict", "Regex")
for (c in candidates) {
val staticOnly = DocLookupUtils.isStaticReceiver(mini, text, dotPos, importedModules, analysis.binding)
DocLookupUtils.resolveMemberWithInheritance(importedModules, c, ident, mini, staticOnly = staticOnly)?.let { (owner, member) ->
DocLookupUtils.resolveMemberWithInheritance(importedModules, c, ident)?.let { (owner, member) ->
if (DEBUG_INHERITANCE) log.info("[LYNG_DEBUG] Candidate '$c.$ident' resolved via inheritance to $owner.${member.name}")
return when (member) {
is MiniMemberFunDecl -> renderMemberFunDoc(owner, member)
is MiniMemberValDecl -> renderMemberValDoc(owner, member)
is MiniMemberTypeAliasDecl -> renderMemberTypeAliasDoc(owner, member)
is MiniInitDecl -> null
is MiniFunDecl -> renderDeclDoc(member, text, mini, importedModules)
is MiniValDecl -> renderDeclDoc(member, text, mini, importedModules)
is MiniClassDecl -> renderDeclDoc(member, text, mini, importedModules)
is MiniEnumDecl -> renderDeclDoc(member, text, mini, importedModules)
is MiniTypeAliasDecl -> renderDeclDoc(member, text, mini, importedModules)
}
}
}
}
// As a last resort try aggregated String members (extensions from stdlib text)
run {
val classes = DocLookupUtils.aggregateClasses(importedModules, mini)
val classes = DocLookupUtils.aggregateClasses(importedModules)
val stringCls = classes["String"]
val m = stringCls?.members?.firstOrNull { it.name == ident }
if (m != null) {
@ -500,24 +268,15 @@ class LyngDocumentationProvider : AbstractDocumentationProvider() {
return when (m) {
is MiniMemberFunDecl -> renderMemberFunDoc("String", m)
is MiniMemberValDecl -> renderMemberValDoc("String", m)
is MiniMemberTypeAliasDecl -> renderMemberTypeAliasDoc("String", m)
is MiniInitDecl -> null
}
}
}
// Search across classes; prefer Iterable, then Iterator, then List for common ops
DocLookupUtils.findMemberAcrossClasses(importedModules, ident, mini)?.let { (owner, member) ->
DocLookupUtils.findMemberAcrossClasses(importedModules, ident)?.let { (owner, member) ->
if (DEBUG_INHERITANCE) log.info("[LYNG_DEBUG] Cross-class '$ident' resolved to $owner.${member.name}")
return when (member) {
is MiniMemberFunDecl -> renderMemberFunDoc(owner, member)
is MiniMemberValDecl -> renderMemberValDoc(owner, member)
is MiniMemberTypeAliasDecl -> renderMemberTypeAliasDoc(owner, member)
is MiniInitDecl -> null
is MiniFunDecl -> renderDeclDoc(member, text, mini, importedModules)
is MiniValDecl -> renderDeclDoc(member, text, mini, importedModules)
is MiniClassDecl -> renderDeclDoc(member, text, mini, importedModules)
is MiniEnumDecl -> renderDeclDoc(member, text, mini, importedModules)
is MiniTypeAliasDecl -> renderDeclDoc(member, text, mini, importedModules)
}
}
}
@ -528,44 +287,49 @@ class LyngDocumentationProvider : AbstractDocumentationProvider() {
return null
}
/**
* Very lenient import extractor for cases when MiniAst is unavailable.
* Looks for lines like `import xxx.yyy` and returns canonical module names
* (prefixing with `lyng.` if missing).
*/
private fun extractImportsFromText(text: String): List<String> {
val result = LinkedHashSet<String>()
val re = Regex("(?m)^\\s*import\\s+([a-zA-Z_][a-zA-Z0-9_]*(?:\\.[a-zA-Z_][a-zA-Z0-9_]*)*)")
re.findAll(text).forEach { m ->
val raw = m.groupValues.getOrNull(1)?.trim().orEmpty()
if (raw.isNotEmpty()) {
val canon = if (raw.startsWith("lyng.")) raw else "lyng.$raw"
result.add(canon)
}
}
return result.toList()
}
// External docs registrars discovery via reflection to avoid hard dependencies on optional modules
private val externalDocsLoaded: Boolean by lazy { tryLoadExternalDocs() }
private fun ensureExternalDocsRegistered() { @Suppress("UNUSED_EXPRESSION") externalDocsLoaded }
private fun tryLoadExternalDocs(): Boolean {
var anyLoaded = false
try {
return try {
// Try known registrars; ignore failures if module is absent
val cls = Class.forName("net.sergeych.lyngio.docs.FsBuiltinDocs")
val m = cls.getMethod("ensure")
m.invoke(null)
log.info("[LYNG_DEBUG] QuickDoc: external docs loaded: net.sergeych.lyngio.docs.FsBuiltinDocs.ensure() OK")
anyLoaded = true
} catch (_: Throwable) {}
try {
val cls = Class.forName("net.sergeych.lyngio.docs.ProcessBuiltinDocs")
val m = cls.getMethod("ensure")
m.invoke(null)
log.info("[LYNG_DEBUG] QuickDoc: external docs loaded: net.sergeych.lyngio.docs.ProcessBuiltinDocs.ensure() OK")
anyLoaded = true
} catch (_: Throwable) {}
if (!anyLoaded) {
true
} catch (_: Throwable) {
// Seed a minimal plugin-local fallback so Path docs still work without lyngio
val seeded = try {
FsDocsFallback.ensureOnce()
ProcessDocsFallback.ensureOnce()
true
} catch (_: Throwable) { false }
if (seeded) {
log.info("[LYNG_DEBUG] QuickDoc: external docs NOT found; seeded plugin fallbacks")
log.info("[LYNG_DEBUG] QuickDoc: external docs NOT found; seeded plugin fallback for lyng.io.fs")
} else {
log.info("[LYNG_DEBUG] QuickDoc: external docs NOT found (lyngio absent on classpath)")
}
return seeded
seeded
}
return true
}
override fun getCustomDocumentationElement(
@ -579,105 +343,25 @@ class LyngDocumentationProvider : AbstractDocumentationProvider() {
return contextElement ?: file.findElementAt(targetOffset)
}
private fun renderDeclDoc(d: MiniDecl, text: String, mini: MiniScript, imported: List<String>): String {
private fun renderDeclDoc(d: MiniDecl): String {
val title = when (d) {
is MiniFunDecl -> "function ${d.name}${signatureOf(d)}"
is MiniClassDecl -> "class ${d.name}"
is MiniEnumDecl -> "enum ${d.name} { ${d.entries.joinToString(", ")} }"
is MiniTypeAliasDecl -> "type ${d.name}${typeAliasSuffix(d)}"
is MiniValDecl -> {
val t = d.type ?: DocLookupUtils.inferTypeRefForVal(d, text, imported, mini)
val typeStr = if (t == null) ": Object?" else typeOf(t)
if (d.mutable) "var ${d.name}${typeStr}" else "val ${d.name}${typeStr}"
}
is MiniValDecl -> if (d.mutable) "var ${d.name}${typeOf(d.type)}" else "val ${d.name}${typeOf(d.type)}"
else -> d.name
}
// Show full detailed documentation, not just the summary
val raw = d.doc?.raw
val doc: String? = if (raw.isNullOrBlank()) null else MarkdownRenderer.render(raw)
val sb = StringBuilder()
sb.append(renderTitle(title))
sb.append(renderDocBody(d.doc))
sb.append("<div class='doc-title'>").append(htmlEscape(title)).append("</div>")
if (!doc.isNullOrBlank()) sb.append(styledMarkdown(doc!!))
return sb.toString()
}
private fun renderDocFromInfo(info: LyngSymbolInfo): String? {
val kind = when (info.target.kind) {
net.sergeych.lyng.binding.SymbolKind.Function -> "function"
net.sergeych.lyng.binding.SymbolKind.Class -> "class"
net.sergeych.lyng.binding.SymbolKind.Enum -> "enum"
net.sergeych.lyng.binding.SymbolKind.TypeAlias -> "type"
net.sergeych.lyng.binding.SymbolKind.Value -> "val"
net.sergeych.lyng.binding.SymbolKind.Variable -> "var"
net.sergeych.lyng.binding.SymbolKind.Parameter -> "parameter"
}
val title = info.signature ?: "$kind ${info.target.name}"
if (title.isBlank() && info.doc == null) return null
val sb = StringBuilder()
sb.append(renderTitle(title))
sb.append(renderDocBody(info.doc))
return sb.toString()
}
private fun findDocInDeclarationFiles(file: PsiFile, container: String?, name: String): MiniDoc? {
val declFiles = LyngAstManager.getDeclarationFiles(file)
if (declFiles.isEmpty()) return null
fun findInMini(mini: MiniScript): MiniDoc? {
if (container == null) {
mini.declarations.firstOrNull { it.name == name }?.let { return it.doc }
return null
}
val cls = mini.declarations.filterIsInstance<MiniClassDecl>().firstOrNull { it.name == container } ?: return null
cls.members.firstOrNull { it.name == name }?.let { return it.doc }
cls.ctorFields.firstOrNull { it.name == name }?.let { return null }
cls.classFields.firstOrNull { it.name == name }?.let { return null }
return null
}
for (df in declFiles) {
val mini = LyngAstManager.getMiniAst(df)
?: run {
try {
val res = runBlocking {
LyngLanguageTools.analyze(df.text, df.name)
}
res.mini
} catch (_: Throwable) {
null
}
}
if (mini != null) {
val doc = findInMini(mini)
if (doc != null) return doc
}
// Text fallback: parse preceding doc comment for the symbol
val parsed = parseDocFromText(df.text, name)
if (parsed != null) return parsed
}
return null
}
private fun parseDocFromText(text: String, name: String): MiniDoc? {
if (text.isBlank()) return null
val pattern = Regex("/\\*\\*([\\s\\S]*?)\\*/\\s*(?:public|private|protected|static|abstract|extern|open|closed|override\\s+)*\\s*(?:fun|val|var|class|interface|enum|type)\\s+$name\\b")
val m = pattern.find(text) ?: return null
val raw = m.groupValues.getOrNull(1)?.trim() ?: return null
if (raw.isBlank()) return null
val src = net.sergeych.lyng.Source("<doc>", raw)
return MiniDoc.parse(MiniRange(src.startPos, src.startPos), raw.lines())
}
private fun renderParamDoc(fn: MiniFunDecl, p: MiniParam): String {
val title = "parameter ${p.name}${typeOf(p.type)} in ${fn.name}${signatureOf(fn)}"
val sb = StringBuilder()
sb.append(renderTitle(title))
// Find matching @param tag
fn.doc?.tags?.get("param")?.forEach { tag ->
val parts = tag.split(Regex("\\s+"), 2)
if (parts.getOrNull(0) == p.name && parts.size > 1) {
sb.append(styledMarkdown(MarkdownRenderer.render(parts[1])))
}
}
return sb.toString()
return "<div class='doc-title'>${htmlEscape(title)}</div>"
}
private fun renderMemberFunDoc(className: String, m: MiniMemberFunDecl): String {
@ -688,45 +372,37 @@ class LyngDocumentationProvider : AbstractDocumentationProvider() {
val ret = typeOf(m.returnType)
val staticStr = if (m.isStatic) "static " else ""
val title = "${staticStr}method $className.${m.name}(${params})${ret}"
val raw = m.doc?.raw
val doc: String? = if (raw.isNullOrBlank()) null else MarkdownRenderer.render(raw)
val sb = StringBuilder()
sb.append(renderTitle(title))
sb.append(renderDocBody(m.doc))
sb.append("<div class='doc-title'>").append(htmlEscape(title)).append("</div>")
if (!doc.isNullOrBlank()) sb.append(styledMarkdown(doc!!))
return sb.toString()
}
private fun renderMemberValDoc(className: String, m: MiniMemberValDecl): String {
val ts = if (m.type == null) ": Object?" else typeOf(m.type)
val ts = typeOf(m.type)
val kind = if (m.mutable) "var" else "val"
val staticStr = if (m.isStatic) "static " else ""
val title = "${staticStr}${kind} $className.${m.name}${ts}"
val raw = m.doc?.raw
val doc: String? = if (raw.isNullOrBlank()) null else MarkdownRenderer.render(raw)
val sb = StringBuilder()
sb.append(renderTitle(title))
sb.append(renderDocBody(m.doc))
sb.append("<div class='doc-title'>").append(htmlEscape(title)).append("</div>")
if (!doc.isNullOrBlank()) sb.append(styledMarkdown(doc!!))
return sb.toString()
}
private fun renderMemberTypeAliasDoc(className: String, m: MiniMemberTypeAliasDecl): String {
val tp = if (m.typeParams.isEmpty()) "" else "<" + m.typeParams.joinToString(", ") + ">"
val body = typeOf(m.target)
val rhs = if (body.isBlank()) "" else " = ${body.removePrefix(": ")}"
val staticStr = if (m.isStatic) "static " else ""
val title = "${staticStr}type $className.${m.name}$tp$rhs"
val sb = StringBuilder()
sb.append(renderTitle(title))
sb.append(renderDocBody(m.doc))
return sb.toString()
}
private fun typeAliasSuffix(d: MiniTypeAliasDecl): String {
val tp = if (d.typeParams.isEmpty()) "" else "<" + d.typeParams.joinToString(", ") + ">"
val body = typeOf(d.target)
val rhs = if (body.isBlank()) "" else " = ${body.removePrefix(": ")}"
return "$tp$rhs"
}
private fun typeOf(t: MiniTypeRef?): String {
val s = DocLookupUtils.typeOf(t)
return if (s.isEmpty()) (if (t == null) ": Object?" else "") else ": $s"
private fun typeOf(t: MiniTypeRef?): String = when (t) {
is MiniTypeName -> ": ${t.segments.joinToString(".") { it.name }}${if (t.nullable) "?" else ""}"
is MiniGenericType -> {
val base = typeOf(t.base).removePrefix(": ")
val args = t.args.joinToString(", ") { typeOf(it).removePrefix(": ") }
": ${base}<${args}>${if (t.nullable) "?" else ""}"
}
is MiniFunctionType -> ": (..) -> ..${if (t.nullable) "?" else ""}"
is MiniTypeVar -> ": ${t.name}${if (t.nullable) "?" else ""}"
null -> ""
}
private fun signatureOf(fn: MiniFunDecl): String {
@ -738,10 +414,6 @@ class LyngDocumentationProvider : AbstractDocumentationProvider() {
return "(${params})${ret}"
}
private fun renderTitle(title: String): String {
return "<div class='doc-title' style='margin-bottom: 0.8em;'>${htmlEscape(title)}</div>"
}
private fun htmlEscape(s: String): String = buildString(s.length) {
for (ch in s) append(
when (ch) {
@ -808,7 +480,7 @@ class LyngDocumentationProvider : AbstractDocumentationProvider() {
private fun renderOverloads(name: String, overloads: List<MiniFunDecl>): String {
val sb = StringBuilder()
sb.append(renderTitle("Overloads for $name"))
sb.append("<div class='doc-title'>Overloads for ").append(htmlEscape(name)).append("</div>")
sb.append("<ul>")
overloads.forEach { fn ->
sb.append("<li><code>")
@ -830,64 +502,11 @@ class LyngDocumentationProvider : AbstractDocumentationProvider() {
return if (e > s) TextRange(s, e) else null
}
private fun renderDocBody(doc: MiniDoc?): String {
if (doc == null) return ""
val sb = StringBuilder()
if (doc.raw.isNotBlank()) {
sb.append(styledMarkdown(MarkdownRenderer.render(doc.raw)))
}
if (doc.tags.isNotEmpty()) {
sb.append(renderTags(doc.tags))
}
return sb.toString()
}
private fun renderTags(tags: Map<String, List<String>>): String {
if (tags.isEmpty()) return ""
val sb = StringBuilder()
sb.append("<table class='sections'>")
fun section(title: String, list: List<String>, isKeyValue: Boolean = false) {
if (list.isEmpty()) return
sb.append("<tr><td valign='top' class='section'><p>").append(htmlEscape(title)).append(":</p></td><td valign='top'>")
list.forEachIndexed { index, item ->
if (index > 0) sb.append("<p>")
if (isKeyValue) {
val parts = item.split(Regex("\\s+"), 2)
sb.append("<code>").append(htmlEscape(parts[0])).append("</code>")
if (parts.size > 1) {
sb.append("").append(MarkdownRenderer.render(parts[1]).removePrefix("<p>").removeSuffix("</p>"))
}
} else {
sb.append(MarkdownRenderer.render(item).removePrefix("<p>").removeSuffix("</p>"))
}
}
sb.append("</td></tr>")
}
section("Parameters", tags["param"] ?: emptyList(), isKeyValue = true)
section("Returns", tags["return"] ?: emptyList())
section("Throws", tags["throws"] ?: emptyList(), isKeyValue = true)
tags.forEach { (name, list) ->
if (name !in listOf("param", "return", "throws")) {
section(name.replaceFirstChar { if (it.isLowerCase()) it.titlecase() else it.toString() }, list)
}
}
sb.append("</table>")
return sb.toString()
}
private fun previousWordBefore(text: String, offset: Int): TextRange? {
// skip spaces and the dot to the left, but stop after hitting a non-identifier boundary
// skip spaces and dots to the left, but stop after hitting a non-identifier or dot boundary
var i = (offset - 1).coerceAtLeast(0)
// skip trailing spaces
while (i >= 0 && text[i].isWhitespace()) i--
// skip the dot if present
if (i >= 0 && text[i] == '.') i--
// skip spaces before the dot
while (i >= 0 && text[i].isWhitespace()) i--
// first, move left past spaces
while (i > 0 && text[i].isWhitespace()) i--
// remember position to check for dot between words
val end = i + 1
// now find the start of the identifier

View File

@ -1,69 +0,0 @@
/*
* Copyright 2026 Sergey S. Chernov real.sergeych@gmail.com
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*
*/
/*
* Minimal fallback docs seeding for `lyng.io.process` used only inside the IDEA plugin
* when external docs module (lyngio) is not present on the classpath.
*/
package net.sergeych.lyng.idea.docs
import net.sergeych.lyng.miniast.BuiltinDocRegistry
import net.sergeych.lyng.miniast.ParamDoc
import net.sergeych.lyng.miniast.type
internal object ProcessDocsFallback {
@Volatile
private var seeded = false
fun ensureOnce(): Boolean {
if (seeded) return true
synchronized(this) {
if (seeded) return true
BuiltinDocRegistry.module("lyng.io.process") {
classDoc(name = "Process", doc = "Process execution and control.") {
method(
name = "execute",
doc = "Execute a process with arguments.",
params = listOf(ParamDoc("executable", type("lyng.String")), ParamDoc("args", type("lyng.List"))),
returns = type("RunningProcess"),
isStatic = true
)
method(
name = "shell",
doc = "Execute a command via system shell.",
params = listOf(ParamDoc("command", type("lyng.String"))),
returns = type("RunningProcess"),
isStatic = true
)
}
classDoc(name = "RunningProcess", doc = "Handle to a running process.") {
method(name = "stdout", doc = "Get standard output stream as a Flow of lines.", returns = type("lyng.Flow"))
method(name = "stderr", doc = "Get standard error stream as a Flow of lines.", returns = type("lyng.Flow"))
method(name = "waitFor", doc = "Wait for the process to exit.", returns = type("lyng.Int"))
method(name = "signal", doc = "Send a signal to the process.", params = listOf(ParamDoc("signal", type("lyng.String"))))
method(name = "destroy", doc = "Forcefully terminate the process.")
}
valDoc(name = "Process", doc = "Process execution and control.", type = type("Process"))
valDoc(name = "RunningProcess", doc = "Handle to a running process.", type = type("RunningProcess"))
}
seeded = true
return true
}
}
}

View File

@ -1,5 +1,5 @@
/*
* Copyright 2026 Sergey S. Chernov real.sergeych@gmail.com
* Copyright 2025 Sergey S. Chernov real.sergeych@gmail.com
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
@ -35,8 +35,6 @@ import com.intellij.psi.codeStyle.CodeStyleManager
import net.sergeych.lyng.format.LyngFormatConfig
import net.sergeych.lyng.format.LyngFormatter
import net.sergeych.lyng.idea.LyngLanguage
import net.sergeych.lyng.idea.util.FormattingUtils.computeDesiredIndent
import net.sergeych.lyng.idea.util.FormattingUtils.findFirstNonWs
class LyngEnterHandler : EnterHandlerDelegate {
private val log = Logger.getInstance(LyngEnterHandler::class.java)
@ -82,39 +80,16 @@ class LyngEnterHandler : EnterHandlerDelegate {
val trimmed = prevText.trimStart()
// consider only code part before // comment
val code = trimmed.substringBefore("//").trim()
if (code == "}" || code == "*/") {
// Adjust indent for the previous line if it's a block or comment closer
val prevStart = doc.getLineStartOffset(prevLine)
if (file.context == null) {
try {
CodeStyleManager.getInstance(project).adjustLineIndent(file, prevStart)
} catch (e: Exception) {
log.warn("Failed to adjust line indent for previous line: ${e.message}")
}
}
// Fallback for previous line: manual application
val desiredPrev = computeDesiredIndent(project, doc, prevLine)
val lineStartPrev = doc.getLineStartOffset(prevLine)
val lineEndPrev = doc.getLineEndOffset(prevLine)
val firstNonWsPrev = findFirstNonWs(doc, lineStartPrev, lineEndPrev)
val currentIndentLenPrev = firstNonWsPrev - lineStartPrev
if (doc.getText(TextRange(lineStartPrev, lineStartPrev + currentIndentLenPrev)) != desiredPrev) {
WriteCommandAction.runWriteCommandAction(project) {
doc.replaceString(lineStartPrev, lineStartPrev + currentIndentLenPrev, desiredPrev)
}
}
if (code == "}") {
// Previously we reindented the enclosed block on Enter after a lone '}'.
// Per new behavior, this action is now bound to typing '}' instead.
// Keep Enter flow limited to indenting the new line only.
}
}
// Adjust indent for the current (new) line
val currentStart = doc.getLineStartOffsetSafe(currentLine)
if (file.context == null) {
try {
CodeStyleManager.getInstance(project).adjustLineIndent(file, currentStart)
} catch (e: Exception) {
log.warn("Failed to adjust line indent for current line: ${e.message}")
}
}
val csm = CodeStyleManager.getInstance(project)
csm.adjustLineIndent(file, currentStart)
// Fallback: if the platform didn't physically insert indentation, compute it from our formatter and apply
val lineStart = doc.getLineStartOffset(currentLine)
@ -184,6 +159,35 @@ class LyngEnterHandler : EnterHandlerDelegate {
caret.moveToOffset(target)
}
private fun computeDesiredIndent(project: Project, doc: Document, line: Int): String {
val options = CodeStyle.getIndentOptions(project, doc)
val start = 0
val end = doc.getLineEndOffset(line)
val snippet = doc.getText(TextRange(start, end))
val isBlankLine = doc.getLineText(line).trim().isEmpty()
val snippetForCalc = if (isBlankLine) snippet + "x" else snippet
val cfg = LyngFormatConfig(
indentSize = options.INDENT_SIZE.coerceAtLeast(1),
useTabs = options.USE_TAB_CHARACTER,
continuationIndentSize = options.CONTINUATION_INDENT_SIZE.coerceAtLeast(options.INDENT_SIZE.coerceAtLeast(1)),
)
val formatted = LyngFormatter.reindent(snippetForCalc, cfg)
val lastNl = formatted.lastIndexOf('\n')
val lastLine = if (lastNl >= 0) formatted.substring(lastNl + 1) else formatted
val wsLen = lastLine.indexOfFirst { it != ' ' && it != '\t' }.let { if (it < 0) lastLine.length else it }
return lastLine.substring(0, wsLen)
}
private fun findFirstNonWs(doc: Document, start: Int, end: Int): Int {
var i = start
val text = doc.charsSequence
while (i < end) {
val ch = text[i]
if (ch != ' ' && ch != '\t') break
i++
}
return i
}
private fun Document.safeLineNumber(offset: Int): Int =
getLineNumber(offset.coerceIn(0, textLength))

View File

@ -1,5 +1,5 @@
/*
* Copyright 2026 Sergey S. Chernov real.sergeych@gmail.com
* Copyright 2025 Sergey S. Chernov real.sergeych@gmail.com
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
@ -32,66 +32,31 @@ import net.sergeych.lyng.format.LyngFormatConfig
import net.sergeych.lyng.format.LyngFormatter
import net.sergeych.lyng.idea.LyngLanguage
import net.sergeych.lyng.idea.settings.LyngFormatterSettings
import net.sergeych.lyng.idea.util.FormattingUtils.computeDesiredIndent
import net.sergeych.lyng.idea.util.FormattingUtils.findFirstNonWs
class LyngTypedHandler : TypedHandlerDelegate() {
private val log = Logger.getInstance(LyngTypedHandler::class.java)
override fun charTyped(c: Char, project: Project, editor: Editor, file: PsiFile): Result {
if (file.language != LyngLanguage) return Result.CONTINUE
if (c == '}') {
val doc = editor.document
PsiDocumentManager.getInstance(project).commitDocument(doc)
if (c != '}') return Result.CONTINUE
val offset = editor.caretModel.offset
val line = doc.getLineNumber((offset - 1).coerceAtLeast(0))
if (line < 0) return Result.CONTINUE
val doc = editor.document
PsiDocumentManager.getInstance(project).commitDocument(doc)
val rawLine = doc.getLineText(line)
val code = rawLine.substringBefore("//").trim()
if (code == "}") {
val settings = LyngFormatterSettings.getInstance(project)
if (settings.reindentClosedBlockOnEnter) {
reindentClosedBlockAroundBrace(project, file, doc, line)
}
// After block reindent, adjust line indent to what platform thinks (no-op in many cases)
val lineStart = doc.getLineStartOffset(line)
if (file.context == null) {
try {
CodeStyleManager.getInstance(project).adjustLineIndent(file, lineStart)
} catch (e: Exception) {
log.warn("Failed to adjust line indent for current line: ${e.message}")
}
}
}
} else if (c == '/') {
val doc = editor.document
val offset = editor.caretModel.offset
if (offset >= 2 && doc.getText(TextRange(offset - 2, offset)) == "*/") {
PsiDocumentManager.getInstance(project).commitDocument(doc)
val line = doc.getLineNumber(offset - 1)
val lineStart = doc.getLineStartOffset(line)
if (file.context == null) {
try {
CodeStyleManager.getInstance(project).adjustLineIndent(file, lineStart)
} catch (e: Exception) {
log.warn("Failed to adjust line indent for comment: ${e.message}")
}
}
// Manual application fallback
val desired = computeDesiredIndent(project, doc, line)
val lineEnd = doc.getLineEndOffset(line)
val firstNonWs = findFirstNonWs(doc, lineStart, lineEnd)
val currentIndentLen = firstNonWs - lineStart
if (doc.getText(TextRange(lineStart, lineStart + currentIndentLen)) != desired) {
WriteCommandAction.runWriteCommandAction(project) {
doc.replaceString(lineStart, lineStart + currentIndentLen, desired)
}
}
val offset = editor.caretModel.offset
val line = doc.getLineNumber((offset - 1).coerceAtLeast(0))
if (line < 0) return Result.CONTINUE
val rawLine = doc.getLineText(line)
val code = rawLine.substringBefore("//").trim()
if (code == "}") {
val settings = LyngFormatterSettings.getInstance(project)
if (settings.reindentClosedBlockOnEnter) {
reindentClosedBlockAroundBrace(project, file, doc, line)
}
// After block reindent, adjust line indent to what platform thinks (no-op in many cases)
val lineStart = doc.getLineStartOffset(line)
CodeStyleManager.getInstance(project).adjustLineIndent(file, lineStart)
}
return Result.CONTINUE
}

View File

@ -1,5 +1,5 @@
/*
* Copyright 2026 Sergey S. Chernov real.sergeych@gmail.com
* Copyright 2025 Sergey S. Chernov real.sergeych@gmail.com
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
@ -42,7 +42,7 @@ private class LineBlocksRootBlock(
private val file: PsiFile,
private val settings: CodeStyleSettings
) : Block {
override fun getTextRange(): TextRange = TextRange(0, file.textLength)
override fun getTextRange(): TextRange = file.textRange
override fun getSubBlocks(): List<Block> = emptyList()
@ -52,7 +52,7 @@ private class LineBlocksRootBlock(
override fun getSpacing(child1: Block?, child2: Block): Spacing? = null
override fun getChildAttributes(newChildIndex: Int): ChildAttributes = ChildAttributes(Indent.getNoneIndent(), null)
override fun isIncomplete(): Boolean = false
override fun isLeaf(): Boolean = true
override fun isLeaf(): Boolean = false
}
// Intentionally no sub-blocks/spacing: indentation is handled by PreFormatProcessor + LineIndentProvider

View File

@ -1,5 +1,5 @@
/*
* Copyright 2026 Sergey S. Chernov real.sergeych@gmail.com
* Copyright 2025 Sergey S. Chernov real.sergeych@gmail.com
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
@ -24,8 +24,9 @@ import com.intellij.openapi.util.TextRange
import com.intellij.psi.PsiDocumentManager
import com.intellij.psi.codeStyle.CommonCodeStyleSettings.IndentOptions
import com.intellij.psi.codeStyle.lineIndent.LineIndentProvider
import net.sergeych.lyng.format.LyngFormatConfig
import net.sergeych.lyng.format.LyngFormatter
import net.sergeych.lyng.idea.LyngLanguage
import net.sergeych.lyng.idea.util.FormattingUtils
/**
* Lightweight indentation provider for Lyng.
@ -44,7 +45,8 @@ class LyngLineIndentProvider : LineIndentProvider {
val options = CodeStyle.getIndentOptions(project, doc)
val line = doc.getLineNumberSafe(offset)
return FormattingUtils.computeDesiredIndent(project, doc, line)
val indent = computeDesiredIndentFromCore(doc, line, options)
return indent
}
override fun isSuitableFor(language: Language?): Boolean = language == null || language == LyngLanguage
@ -77,4 +79,25 @@ class LyngLineIndentProvider : LineIndentProvider {
return spaces / size
}
private fun computeDesiredIndentFromCore(doc: Document, line: Int, options: IndentOptions): String {
// Build a minimal text consisting of all previous lines and the current line.
// Special case: when the current line is blank (newly created by Enter), compute the
// indent as if there was a non-whitespace character at line start (append a sentinel).
val start = 0
val end = doc.getLineEndOffset(line)
val snippet = doc.getText(TextRange(start, end))
val isBlankLine = doc.getLineText(line).trim().isEmpty()
val snippetForCalc = if (isBlankLine) snippet + "x" else snippet
val cfg = LyngFormatConfig(
indentSize = options.INDENT_SIZE.coerceAtLeast(1),
useTabs = options.USE_TAB_CHARACTER,
continuationIndentSize = options.CONTINUATION_INDENT_SIZE.coerceAtLeast(options.INDENT_SIZE.coerceAtLeast(1)),
)
val formatted = LyngFormatter.reindent(snippetForCalc, cfg)
// Grab the last line's leading whitespace as the indent for the current line
val lastNl = formatted.lastIndexOf('\n')
val lastLine = if (lastNl >= 0) formatted.substring(lastNl + 1) else formatted
val wsLen = lastLine.indexOfFirst { it != ' ' && it != '\t' }.let { if (it < 0) lastLine.length else it }
return lastLine.substring(0, wsLen)
}
}

View File

@ -1,5 +1,5 @@
/*
* Copyright 2026 Sergey S. Chernov real.sergeych@gmail.com
* Copyright 2025 Sergey S. Chernov real.sergeych@gmail.com
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
@ -44,67 +44,25 @@ class LyngPreFormatProcessor : PreFormatProcessor {
// When both spacing and wrapping are OFF, still fix indentation for the whole file to
// guarantee visible changes on Reformat Code.
val runFullFileIndent = !settings.enableSpacing && !settings.enableWrapping
// Maintain a working range and a modification flag to avoid stale offsets after replacements
var modified = false
fun fullRange(): TextRange = TextRange(0, doc.textLength)
var workingRange: TextRange = range.intersection(fullRange()) ?: fullRange()
val docW = doc as? com.intellij.injected.editor.DocumentWindow
// The host range of the entire injected fragment (or the whole file if not injected).
fun currentHostRange(): TextRange = if (docW != null) {
TextRange(docW.injectedToHost(0), docW.injectedToHost(doc.textLength))
} else {
file.textRange
}
// The range in 'doc' coordinate system (local 0..len for injections, host offsets for normal files).
fun currentLocalRange(): TextRange = if (docW != null) {
TextRange(0, doc.textLength)
} else {
file.textRange
}
val clr = currentLocalRange()
val chr = currentHostRange()
// Convert the input range to the coordinate system of 'doc'
var workingRangeLocal: TextRange = if (docW != null) {
val hostIntersection = range.intersection(chr)
if (hostIntersection != null) {
try {
val start = docW.hostToInjected(hostIntersection.startOffset)
val end = docW.hostToInjected(hostIntersection.endOffset)
TextRange(start.coerceAtMost(end), end.coerceAtLeast(start))
} catch (e: Exception) {
clr
}
} else {
range.intersection(clr) ?: clr
}
} else {
range.intersection(clr) ?: clr
}
val startLine = if (runFullFileIndent) {
doc.getLineNumber(currentLocalRange().startOffset)
} else {
doc.getLineNumber(workingRangeLocal.startOffset)
}
val endLine = if (runFullFileIndent) {
if (clr.endOffset <= clr.startOffset) doc.getLineNumber(clr.startOffset)
else doc.getLineNumber(clr.endOffset)
} else {
doc.getLineNumber(workingRangeLocal.endOffset.coerceAtMost(doc.textLength))
}
val startLine = if (runFullFileIndent) 0 else doc.getLineNumber(workingRange.startOffset)
val endLine = if (runFullFileIndent) (doc.lineCount - 1).coerceAtLeast(0)
else doc.getLineNumber(workingRange.endOffset.coerceAtMost(doc.textLength))
fun codePart(s: String): String {
val idx = s.indexOf("//")
return if (idx >= 0) s.substring(0, idx) else s
}
// Pre-scan to compute balances up to startLine.
val fragmentStartLine = doc.getLineNumber(currentLocalRange().startOffset)
// Pre-scan to compute balances up to startLine
var blockLevel = 0
var parenBalance = 0
var bracketBalance = 0
for (ln in fragmentStartLine until startLine) {
for (ln in 0 until startLine) {
val text = doc.getText(TextRange(doc.getLineStartOffset(ln), doc.getLineEndOffset(ln)))
for (ch in codePart(text)) when (ch) {
'{' -> blockLevel++
@ -122,13 +80,7 @@ class LyngPreFormatProcessor : PreFormatProcessor {
val lineStart = doc.getLineStartOffset(line)
// adjustLineIndent delegates to our LineIndentProvider which computes
// indentation from scratch; this is safe and idempotent
if (file.context == null) {
try {
CodeStyleManager.getInstance(project).adjustLineIndent(file, lineStart)
} catch (e: Exception) {
// Log as debug because this can be called many times during reformat
}
}
CodeStyleManager.getInstance(project).adjustLineIndent(file, lineStart)
// After indentation, update block/paren/bracket balances using the current line text
val lineEnd = doc.getLineEndOffset(line)
@ -151,14 +103,15 @@ class LyngPreFormatProcessor : PreFormatProcessor {
useTabs = options.USE_TAB_CHARACTER,
continuationIndentSize = options.CONTINUATION_INDENT_SIZE.coerceAtLeast(options.INDENT_SIZE.coerceAtLeast(1)),
)
val r = if (runFullFileIndent) currentLocalRange() else workingRangeLocal.intersection(currentLocalRange()) ?: currentLocalRange()
val full = fullRange()
val r = if (runFullFileIndent) full else workingRange.intersection(full) ?: full
val text = doc.getText(r)
val formatted = LyngFormatter.reindent(text, cfg)
if (formatted != text) {
doc.replaceString(r.startOffset, r.endOffset, formatted)
modified = true
psiDoc.commitDocument(doc)
workingRangeLocal = currentLocalRange()
workingRange = fullRange()
}
}
@ -171,14 +124,14 @@ class LyngPreFormatProcessor : PreFormatProcessor {
applySpacing = true,
applyWrapping = false,
)
val r = if (runFullFileIndent) currentLocalRange() else workingRangeLocal.intersection(currentLocalRange()) ?: currentLocalRange()
val text = doc.getText(r)
val safe = workingRange.intersection(fullRange()) ?: fullRange()
val text = doc.getText(safe)
val formatted = LyngFormatter.format(text, cfg)
if (formatted != text) {
doc.replaceString(r.startOffset, r.endOffset, formatted)
doc.replaceString(safe.startOffset, safe.endOffset, formatted)
modified = true
psiDoc.commitDocument(doc)
workingRangeLocal = currentLocalRange()
workingRange = fullRange()
}
}
// Optionally apply wrapping (after spacing) when enabled
@ -190,19 +143,17 @@ class LyngPreFormatProcessor : PreFormatProcessor {
applySpacing = settings.enableSpacing,
applyWrapping = true,
)
val r = if (runFullFileIndent) currentLocalRange() else workingRangeLocal.intersection(currentLocalRange()) ?: currentLocalRange()
val text = doc.getText(r)
val wrapped = LyngFormatter.format(text, cfg)
if (wrapped != text) {
doc.replaceString(r.startOffset, r.endOffset, wrapped)
val safe2 = workingRange.intersection(fullRange()) ?: fullRange()
val text2 = doc.getText(safe2)
val wrapped = LyngFormatter.format(text2, cfg)
if (wrapped != text2) {
doc.replaceString(safe2.startOffset, safe2.endOffset, wrapped)
modified = true
psiDoc.commitDocument(doc)
workingRangeLocal = currentLocalRange()
workingRange = fullRange()
}
}
// Return a safe range for the formatter to continue with, preventing stale offsets.
// For injected files, ALWAYS return a range in local coordinates.
val finalRange = currentLocalRange()
return if (modified) finalRange else (range.intersection(finalRange) ?: finalRange)
// Return a safe range for the formatter to continue with, preventing stale offsets
return if (modified) fullRange() else (range.intersection(fullRange()) ?: fullRange())
}
}

View File

@ -0,0 +1,43 @@
/*
* Copyright 2025 Sergey S. Chernov real.sergeych@gmail.com
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*
*/
package net.sergeych.lyng.idea.grazie
import com.intellij.codeInsight.daemon.DaemonCodeAnalyzer
import com.intellij.codeInsight.intention.IntentionAction
import com.intellij.openapi.editor.Editor
import com.intellij.openapi.project.Project
import com.intellij.psi.PsiFile
import net.sergeych.lyng.idea.settings.LyngFormatterSettings
/**
* Lightweight quick-fix that adds a word to the per-project Lyng dictionary.
*/
class AddToLyngDictionaryFix(private val word: String) : IntentionAction {
override fun getText(): String = "Add '$word' to Lyng dictionary"
override fun getFamilyName(): String = "Lyng Spelling"
override fun isAvailable(project: Project, editor: Editor?, file: PsiFile?): Boolean = word.isNotBlank()
override fun startInWriteAction(): Boolean = true
override fun invoke(project: Project, editor: Editor?, file: PsiFile?) {
val settings = LyngFormatterSettings.getInstance(project)
val learned = settings.learnedWords
learned.add(word.lowercase())
settings.learnedWords = learned
// Restart daemon to refresh highlights
if (file != null) DaemonCodeAnalyzer.getInstance(project).restart(file)
}
}

View File

@ -0,0 +1,83 @@
/*
* Copyright 2025 Sergey S. Chernov real.sergeych@gmail.com
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*
*/
package net.sergeych.lyng.idea.grazie
import com.intellij.openapi.diagnostic.Logger
import java.io.BufferedReader
import java.io.InputStreamReader
import java.util.zip.GZIPInputStream
/**
* Very simple English dictionary loader for offline suggestions on IC-243.
* It loads a word list from classpath resources. Supports plain text (one word per line)
* and gzipped text if the resource ends with .gz.
*/
object EnglishDictionary {
private val log = Logger.getInstance(EnglishDictionary::class.java)
@Volatile private var loaded = false
@Volatile private var words: Set<String> = emptySet()
/**
* Load dictionary from bundled resources (once).
* If multiple candidates exist, the first found is used.
*/
private fun ensureLoaded() {
if (loaded) return
synchronized(this) {
if (loaded) return
val candidates = listOf(
// preferred large bundles first (add en-basic.txt.gz ~3–5MB here)
"/dictionaries/en-basic.txt.gz",
"/dictionaries/en-large.txt.gz",
// plain text fallbacks
"/dictionaries/en-basic.txt",
"/dictionaries/en-large.txt",
)
val merged = HashSet<String>(128_000)
for (res in candidates) {
try {
val stream = javaClass.getResourceAsStream(res) ?: continue
val reader = if (res.endsWith(".gz"))
BufferedReader(InputStreamReader(GZIPInputStream(stream)))
else
BufferedReader(InputStreamReader(stream))
var loadedCount = 0
reader.useLines { seq -> seq.forEach { line ->
val w = line.trim()
if (w.isNotEmpty() && !w.startsWith("#")) { merged += w.lowercase(); loadedCount++ }
} }
log.info("EnglishDictionary: loaded $loadedCount words from $res (total=${merged.size})")
} catch (t: Throwable) {
log.info("EnglishDictionary: failed to load $res: ${t.javaClass.simpleName}: ${t.message}")
}
}
if (merged.isEmpty()) {
// Fallback minimal set
merged += setOf("comment","comments","error","errors","found","file","not","word","words","count","value","name","class","function","string")
log.info("EnglishDictionary: using minimal built-in set (${merged.size})")
}
words = merged
loaded = true
}
}
fun allWords(): Set<String> {
ensureLoaded()
return words
}
}

View File

@ -0,0 +1,608 @@
/*
* Copyright 2025 Sergey S. Chernov real.sergeych@gmail.com
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*
*/
/*
* Grazie-backed annotator for Lyng files.
*
* It consumes the MiniAst-driven LyngSpellIndex and, when Grazie is present,
* tries to run Grazie checks on the extracted TextContent. Results are painted
* as warnings in the editor. If the Grazie API changes, we use reflection and
* fail softly with INFO logs (no errors shown to users).
*/
package net.sergeych.lyng.idea.grazie
import com.intellij.codeInsight.daemon.DaemonCodeAnalyzer
import com.intellij.grazie.text.TextContent
import com.intellij.grazie.text.TextContent.TextDomain
import com.intellij.ide.plugins.PluginManagerCore
import com.intellij.lang.annotation.AnnotationHolder
import com.intellij.lang.annotation.ExternalAnnotator
import com.intellij.lang.annotation.HighlightSeverity
import com.intellij.openapi.application.ApplicationManager
import com.intellij.openapi.diagnostic.Logger
import com.intellij.openapi.editor.Document
import com.intellij.openapi.editor.colors.TextAttributesKey
import com.intellij.openapi.project.DumbAware
import com.intellij.openapi.util.Key
import com.intellij.openapi.util.TextRange
import com.intellij.psi.PsiFile
import net.sergeych.lyng.idea.settings.LyngFormatterSettings
import net.sergeych.lyng.idea.spell.LyngSpellIndex
class LyngGrazieAnnotator : ExternalAnnotator<LyngGrazieAnnotator.Input, LyngGrazieAnnotator.Result>(), DumbAware {
private val log = Logger.getInstance(LyngGrazieAnnotator::class.java)
companion object {
// Cache GrammarChecker availability to avoid repeated reflection + noisy logs
@Volatile
private var grammarCheckerAvailable: Boolean? = null
@Volatile
private var grammarCheckerMissingLogged: Boolean = false
private fun isGrammarCheckerKnownMissing(): Boolean = (grammarCheckerAvailable == false)
private fun markGrammarCheckerMissingOnce(log: Logger, message: String) {
if (!grammarCheckerMissingLogged) {
// Downgrade to debug to reduce log noise across projects/sessions
log.debug(message)
grammarCheckerMissingLogged = true
}
}
private val RETRY_KEY: Key<Long> = Key.create("LYNG_GRAZIE_ANN_RETRY_STAMP")
}
data class Input(val modStamp: Long)
data class Finding(val range: TextRange, val message: String)
data class Result(val modStamp: Long, val findings: List<Finding>)
override fun collectInformation(file: PsiFile): Input? {
val doc: Document = file.viewProvider.document ?: return null
// Only require Grazie presence; index readiness is checked in apply with a retry.
val grazie = isGrazieInstalled()
if (!grazie) {
log.info("LyngGrazieAnnotator.collectInformation: skip (grazie=false) file='${file.name}'")
return null
}
log.info("LyngGrazieAnnotator.collectInformation: file='${file.name}', modStamp=${doc.modificationStamp}")
return Input(doc.modificationStamp)
}
override fun doAnnotate(collectedInfo: Input?): Result? {
// All heavy lifting is done in apply where we have the file context
return collectedInfo?.let { Result(it.modStamp, emptyList()) }
}
override fun apply(file: PsiFile, annotationResult: Result?, holder: AnnotationHolder) {
if (annotationResult == null || !isGrazieInstalled()) return
val doc = file.viewProvider.document ?: return
val idx = LyngSpellIndex.getUpToDate(file) ?: run {
log.info("LyngGrazieAnnotator.apply: index not ready for '${file.name}', scheduling one-shot restart")
scheduleOneShotRestart(file, annotationResult.modStamp)
return
}
val settings = LyngFormatterSettings.getInstance(file.project)
// Build TextContent fragments for comments/strings/identifiers according to settings
val fragments = mutableListOf<Pair<TextContent, TextRange>>()
try {
fun addFragments(ranges: List<TextRange>, domain: TextDomain) {
for (r in ranges) {
val local = rangeToTextContent(file, domain, r) ?: continue
fragments += local to r
}
}
// Comments always via COMMENTS
addFragments(idx.comments, TextDomain.COMMENTS)
// Strings: LITERALS if requested, else COMMENTS if fallback enabled
if (settings.spellCheckStringLiterals) {
val domain = if (settings.grazieTreatLiteralsAsComments) TextDomain.COMMENTS else TextDomain.LITERALS
addFragments(idx.strings, domain)
}
// Identifiers via COMMENTS to force painting in 243 unless user disables fallback
val idsDomain = if (settings.grazieTreatIdentifiersAsComments) TextDomain.COMMENTS else TextDomain.DOCUMENTATION
addFragments(idx.identifiers, idsDomain)
log.info(
"LyngGrazieAnnotator.apply: file='${file.name}', idxCounts ids=${idx.identifiers.size}, comments=${idx.comments.size}, strings=${idx.strings.size}, builtFragments=${fragments.size}"
)
} catch (e: Throwable) {
log.info("LyngGrazieAnnotator: failed to build TextContent fragments: ${e.javaClass.simpleName}: ${e.message}")
return
}
if (fragments.isEmpty()) return
val findings = mutableListOf<Finding>()
var totalReturned = 0
var chosenEntry: String? = null
for ((content, hostRange) in fragments) {
try {
val (typos, entryNote) = runGrazieChecksWithTracing(file, content)
if (chosenEntry == null) chosenEntry = entryNote
if (typos != null) {
totalReturned += typos.size
for (t in typos) {
val rel = extractRangeFromTypo(t) ?: continue
// Map relative range inside fragment to host file range
val abs = TextRange(hostRange.startOffset + rel.startOffset, hostRange.startOffset + rel.endOffset)
findings += Finding(abs, extractMessageFromTypo(t) ?: "Spelling/Grammar")
}
}
} catch (e: Throwable) {
log.info("LyngGrazieAnnotator: Grazie check failed: ${e.javaClass.simpleName}: ${e.message}")
}
}
log.info("LyngGrazieAnnotator.apply: used=${chosenEntry ?: "<none>"}, totalFindings=$totalReturned, painting=${findings.size}")
// IMPORTANT: Do NOT fallback to the tiny bundled vocabulary on modern IDEs.
// If Grazie/Natural Languages processing returned nothing, we simply exit here
// to avoid low‑quality results from the legacy dictionary.
if (findings.isEmpty()) return
for (f in findings) {
val ab = holder.newAnnotation(HighlightSeverity.INFORMATION, f.message).range(f.range)
applyTypoStyleIfRequested(file, ab)
ab.create()
}
}
private fun scheduleOneShotRestart(file: PsiFile, modStamp: Long) {
try {
val last = file.getUserData(RETRY_KEY)
if (last == modStamp) {
log.info("LyngGrazieAnnotator.restart: already retried for modStamp=$modStamp, skip")
return
}
file.putUserData(RETRY_KEY, modStamp)
ApplicationManager.getApplication().invokeLater({
try {
DaemonCodeAnalyzer.getInstance(file.project).restart(file)
log.info("LyngGrazieAnnotator.restart: daemon restarted for '${file.name}'")
} catch (e: Throwable) {
log.info("LyngGrazieAnnotator.restart failed: ${e.javaClass.simpleName}: ${e.message}")
}
})
} catch (e: Throwable) {
log.info("LyngGrazieAnnotator.scheduleOneShotRestart failed: ${e.javaClass.simpleName}: ${e.message}")
}
}
private fun isGrazieInstalled(): Boolean {
return PluginManagerCore.isPluginInstalled(com.intellij.openapi.extensions.PluginId.getId("com.intellij.grazie")) ||
PluginManagerCore.isPluginInstalled(com.intellij.openapi.extensions.PluginId.getId("tanvd.grazi"))
}
private fun rangeToTextContent(file: PsiFile, domain: TextDomain, range: TextRange): TextContent? {
// Build TextContent via reflection: prefer psiFragment(domain, element)
return try {
// Try to find an element that fully covers the target range
var element = file.findElementAt(range.startOffset) ?: return null
val start = range.startOffset
val end = range.endOffset
while (element.parent != null && (element.textRange.startOffset > start || element.textRange.endOffset < end)) {
element = element.parent
}
if (element.textRange.startOffset > start || element.textRange.endOffset < end) return null
// In many cases, the element may not span the whole range; use file + range via suitable factory
val methods = TextContent::class.java.methods.filter { it.name == "psiFragment" }
val byElementDomain = methods.firstOrNull { it.parameterCount == 2 && it.parameterTypes[0].name.endsWith("PsiElement") }
if (byElementDomain != null) {
@Suppress("UNCHECKED_CAST")
return (byElementDomain.invoke(null, element, domain) as? TextContent)?.let { tc ->
val relStart = start - element.textRange.startOffset
val relEnd = end - element.textRange.startOffset
if (relStart < 0 || relEnd > tc.length || relStart >= relEnd) return null
tc.subText(TextRange(relStart, relEnd))
}
}
val byDomainElement = methods.firstOrNull { it.parameterCount == 2 && it.parameterTypes[0].name.endsWith("TextDomain") }
if (byDomainElement != null) {
@Suppress("UNCHECKED_CAST")
return (byDomainElement.invoke(null, domain, element) as? TextContent)?.let { tc ->
val relStart = start - element.textRange.startOffset
val relEnd = end - element.textRange.startOffset
if (relStart < 0 || relEnd > tc.length || relStart >= relEnd) return null
tc.subText(TextRange(relStart, relEnd))
}
}
null
} catch (e: Throwable) {
log.info("LyngGrazieAnnotator: rangeToTextContent failed: ${e.javaClass.simpleName}: ${e.message}")
null
}
}
private fun runGrazieChecksWithTracing(file: PsiFile, content: TextContent): Pair<Collection<Any>?, String?> {
// Try known entry points via reflection to avoid hard dependencies on Grazie internals
if (isGrammarCheckerKnownMissing()) return null to null
try {
// 1) Static GrammarChecker.check(TextContent)
val checkerCls = try {
Class.forName("com.intellij.grazie.grammar.GrammarChecker").also { grammarCheckerAvailable = true }
} catch (t: Throwable) {
grammarCheckerAvailable = false
markGrammarCheckerMissingOnce(log, "LyngGrazieAnnotator: GrammarChecker class not found: ${t.javaClass.simpleName}: ${t.message}")
null
}
if (checkerCls != null) {
// Diagnostic: list available 'check' methods once
runCatching {
val checks = checkerCls.methods.filter { it.name == "check" }
val sig = checks.joinToString { m ->
val params = m.parameterTypes.joinToString(prefix = "(", postfix = ")") { it.simpleName }
"${m.name}$params static=${java.lang.reflect.Modifier.isStatic(m.modifiers)}"
}
log.info("LyngGrazieAnnotator: GrammarChecker.check candidates: ${if (sig.isEmpty()) "<none>" else sig}")
}
checkerCls.methods.firstOrNull { it.name == "check" && it.parameterCount == 1 && it.parameterTypes[0].name.endsWith("TextContent") }?.let { m ->
@Suppress("UNCHECKED_CAST")
val res = m.invoke(null, content) as? Collection<Any>
return res to "GrammarChecker.check(TextContent) static"
}
// 2) GrammarChecker.getInstance().check(TextContent)
val getInstance = checkerCls.methods.firstOrNull { it.name == "getInstance" && it.parameterCount == 0 }
val inst = getInstance?.invoke(null)
if (inst != null) {
val m = checkerCls.methods.firstOrNull { it.name == "check" && it.parameterCount == 1 && it.parameterTypes[0].name.endsWith("TextContent") }
if (m != null) {
@Suppress("UNCHECKED_CAST")
val res = m.invoke(inst, content) as? Collection<Any>
return res to "GrammarChecker.getInstance().check(TextContent)"
}
}
// 3) GrammarChecker.getDefault().check(TextContent)
val getDefault = checkerCls.methods.firstOrNull { it.name == "getDefault" && it.parameterCount == 0 }
val def = getDefault?.invoke(null)
if (def != null) {
val m = checkerCls.methods.firstOrNull { it.name == "check" && it.parameterCount == 1 && it.parameterTypes[0].name.endsWith("TextContent") }
if (m != null) {
@Suppress("UNCHECKED_CAST")
val res = m.invoke(def, content) as? Collection<Any>
return res to "GrammarChecker.getDefault().check(TextContent)"
}
}
// 4) Service from project/application: GrammarChecker as a service
runCatching {
val app = com.intellij.openapi.application.ApplicationManager.getApplication()
val getService = app::class.java.methods.firstOrNull { it.name == "getService" && it.parameterCount == 1 }
val svc = getService?.invoke(app, checkerCls)
if (svc != null) {
val m = checkerCls.methods.firstOrNull { it.name == "check" && it.parameterCount == 1 && it.parameterTypes[0].name.endsWith("TextContent") }
if (m != null) {
@Suppress("UNCHECKED_CAST")
val res = m.invoke(svc, content) as? Collection<Any>
if (res != null) return res to "Application.getService(GrammarChecker).check(TextContent)"
}
}
}
runCatching {
val getService = file.project::class.java.methods.firstOrNull { it.name == "getService" && it.parameterCount == 1 }
val svc = getService?.invoke(file.project, checkerCls)
if (svc != null) {
val m = checkerCls.methods.firstOrNull { it.name == "check" && it.parameterCount == 1 && it.parameterTypes[0].name.endsWith("TextContent") }
if (m != null) {
@Suppress("UNCHECKED_CAST")
val res = m.invoke(svc, content) as? Collection<Any>
if (res != null) return res to "Project.getService(GrammarChecker).check(TextContent)"
}
}
}
}
// 5) Fallback: search any public method named check that accepts TextContent in any Grazie class (static)
val candidateClasses = listOf(
"com.intellij.grazie.grammar.GrammarChecker",
"com.intellij.grazie.grammar.GrammarRunner",
"com.intellij.grazie.grammar.Grammar" // historical names
)
for (cn in candidateClasses) {
val cls = try { Class.forName(cn) } catch (_: Throwable) { continue }
val m = cls.methods.firstOrNull { it.name == "check" && it.parameterTypes.any { p -> p.name.endsWith("TextContent") } }
if (m != null) {
val args = arrayOfNulls<Any>(m.parameterCount)
// place content to the first TextContent parameter; others left null (common defaults)
for (i in 0 until m.parameterCount) if (m.parameterTypes[i].name.endsWith("TextContent")) { args[i] = content; break }
@Suppress("UNCHECKED_CAST")
val res = m.invoke(null, *args) as? Collection<Any>
if (res != null) return res to "$cn.${m.name}(TextContent)"
}
}
// 6) Kotlin top-level function: GrammarCheckerKt.check(TextContent)
runCatching {
val kt = Class.forName("com.intellij.grazie.grammar.GrammarCheckerKt")
val m = kt.methods.firstOrNull { it.name == "check" && it.parameterTypes.any { p -> p.name.endsWith("TextContent") } }
if (m != null) {
val args = arrayOfNulls<Any>(m.parameterCount)
for (i in 0 until m.parameterCount) if (m.parameterTypes[i].name.endsWith("TextContent")) { args[i] = content; break }
@Suppress("UNCHECKED_CAST")
val res = m.invoke(null, *args) as? Collection<Any>
if (res != null) return res to "GrammarCheckerKt.check(TextContent)"
}
}
} catch (e: Throwable) {
log.info("LyngGrazieAnnotator: runGrazieChecks reflection failed: ${e.javaClass.simpleName}: ${e.message}")
}
return null to null
}
private fun extractRangeFromTypo(typo: Any): TextRange? {
// Try to get a relative range from returned Grazie issue/typo via common accessors
return try {
// Common getters
val m1 = typo.javaClass.methods.firstOrNull { it.name == "getRange" && it.parameterCount == 0 }
val r1 = if (m1 != null) m1.invoke(typo) else null
when (r1) {
is TextRange -> return r1
is IntRange -> return TextRange(r1.first, r1.last + 1)
}
val m2 = typo.javaClass.methods.firstOrNull { it.name == "getHighlightRange" && it.parameterCount == 0 }
val r2 = if (m2 != null) m2.invoke(typo) else null
when (r2) {
is TextRange -> return r2
is IntRange -> return TextRange(r2.first, r2.last + 1)
}
// Separate from/to ints
val fromM = typo.javaClass.methods.firstOrNull { it.name == "getFrom" && it.parameterCount == 0 && it.returnType == Int::class.javaPrimitiveType }
val toM = typo.javaClass.methods.firstOrNull { it.name == "getTo" && it.parameterCount == 0 && it.returnType == Int::class.javaPrimitiveType }
if (fromM != null && toM != null) {
val s = (fromM.invoke(typo) as? Int) ?: return null
val e = (toM.invoke(typo) as? Int) ?: return null
if (e > s) return TextRange(s, e)
}
null
} catch (_: Throwable) { null }
}
private fun extractMessageFromTypo(typo: Any): String? {
return try {
val m = typo.javaClass.methods.firstOrNull { it.name == "getMessage" && it.parameterCount == 0 }
(m?.invoke(typo) as? String)
} catch (_: Throwable) { null }
}
// Fallback that uses legacy SpellCheckerManager (if present) via reflection to validate words in fragments.
// Returns number of warnings painted.
private fun fallbackWithLegacySpellcheckerIfAvailable(
file: PsiFile,
fragments: List<Pair<TextContent, TextRange>>,
holder: AnnotationHolder
): Int {
return try {
val mgrCls = Class.forName("com.intellij.spellchecker.SpellCheckerManager")
val getInstance = mgrCls.methods.firstOrNull { it.name == "getInstance" && it.parameterCount == 1 }
val isCorrect = mgrCls.methods.firstOrNull { it.name == "isCorrect" && it.parameterCount == 1 && it.parameterTypes[0] == String::class.java }
if (getInstance == null || isCorrect == null) {
// No legacy spellchecker API available — fall back to naive painter
return naiveFallbackPaint(file, fragments, holder)
}
val mgr = getInstance.invoke(null, file.project)
if (mgr == null) {
// Legacy manager not present for this project — use naive fallback
return naiveFallbackPaint(file, fragments, holder)
}
var painted = 0
val docText = file.viewProvider.document?.text ?: return 0
val tokenRegex = Regex("[A-Za-z][A-Za-z0-9_']{2,}")
for ((content, hostRange) in fragments) {
val text = try { docText.substring(hostRange.startOffset, hostRange.endOffset) } catch (_: Throwable) { null } ?: continue
var seen = 0
var flagged = 0
for (m in tokenRegex.findAll(text)) {
val token = m.value
if ('%' in token) continue // skip printf fragments defensively
// Split snake_case and camelCase within the token
val parts = splitIdentifier(token)
for (part in parts) {
if (part.length <= 2) continue
if (isAllowedWord(part)) continue
// Quick allowlist for very common words to reduce noise if dictionaries differ
val ok = try { isCorrect.invoke(mgr, part) as? Boolean } catch (_: Throwable) { null }
if (ok == false) {
// Map part back to original token occurrence within this hostRange
val localStart = m.range.first + token.indexOf(part)
val localEnd = localStart + part.length
val abs = TextRange(hostRange.startOffset + localStart, hostRange.startOffset + localEnd)
paintTypoAnnotation(file, holder, abs, part)
painted++
flagged++
}
seen++
}
}
log.info("LyngGrazieAnnotator.fallback: fragment words=$seen, flagged=$flagged")
}
painted
} catch (_: Throwable) {
// If legacy manager is not available, fall back to a very naive heuristic (no external deps)
return naiveFallbackPaint(file, fragments, holder)
}
}
private fun naiveFallbackPaint(
file: PsiFile,
fragments: List<Pair<TextContent, TextRange>>,
holder: AnnotationHolder
): Int {
var painted = 0
val docText = file.viewProvider.document?.text
val tokenRegex = Regex("[A-Za-z][A-Za-z0-9_']{2,}")
val baseWords = setOf(
// small, common vocabulary to catch near-miss typos in typical code/comments
"comment","comments","error","errors","found","file","not","word","words","count","value","name","class","function","string"
)
for ((content, hostRange) in fragments) {
val text: String? = docText?.let { dt ->
try { dt.substring(hostRange.startOffset, hostRange.endOffset) } catch (_: Throwable) { null }
}
if (text.isNullOrBlank()) continue
var seen = 0
var flagged = 0
for (m in tokenRegex.findAll(text)) {
val token = m.value
if ('%' in token) continue
val parts = splitIdentifier(token)
for (part in parts) {
seen++
val lower = part.lowercase()
if (lower.length <= 2 || isAllowedWord(part)) continue
// Heuristic: no vowels OR 3 repeated chars OR ends with unlikely double consonants
val noVowel = lower.none { it in "aeiouy" }
val triple = Regex("(.)\\1\\1").containsMatchIn(lower)
val dblCons = Regex("[bcdfghjklmnpqrstvwxyz]{2}$").containsMatchIn(lower)
var looksWrong = noVowel || triple || dblCons
// Additional: low vowel ratio for length>=4
if (!looksWrong && lower.length >= 4) {
val vowels = lower.count { it in "aeiouy" }
val ratio = if (lower.isNotEmpty()) vowels.toDouble() / lower.length else 1.0
if (ratio < 0.25) looksWrong = true
}
// Additional: near-miss to a small base vocabulary (edit distance 1, or 2 for words >=6)
if (!looksWrong) {
for (bw in baseWords) {
val d = editDistance(lower, bw)
if (d == 1 || (d == 2 && lower.length >= 6)) { looksWrong = true; break }
}
}
if (looksWrong) {
val localStart = m.range.first + token.indexOf(part)
val localEnd = localStart + part.length
val abs = TextRange(hostRange.startOffset + localStart, hostRange.startOffset + localEnd)
paintTypoAnnotation(file, holder, abs, part)
painted++
flagged++
}
}
}
log.info("LyngGrazieAnnotator.fallback(naive): fragment words=$seen, flagged=$flagged")
}
return painted
}
private fun paintTypoAnnotation(file: PsiFile, holder: AnnotationHolder, range: TextRange, word: String) {
val settings = LyngFormatterSettings.getInstance(file.project)
val ab = holder.newAnnotation(HighlightSeverity.INFORMATION, "Possible typo")
.range(range)
applyTypoStyleIfRequested(file, ab)
if (settings.offerLyngTypoQuickFixes) {
// Offer lightweight fixes; for 243 provide Add-to-dictionary always
ab.withFix(net.sergeych.lyng.idea.grazie.AddToLyngDictionaryFix(word))
// Offer "Replace with…" candidates (top 7)
val cands = suggestReplacements(file, word).take(7)
for (c in cands) {
ab.withFix(net.sergeych.lyng.idea.grazie.ReplaceWordFix(range, word, c))
}
}
ab.create()
}
private fun applyTypoStyleIfRequested(file: PsiFile, ab: com.intellij.lang.annotation.AnnotationBuilder) {
val settings = LyngFormatterSettings.getInstance(file.project)
if (!settings.showTyposWithGreenUnderline) return
// Use the standard TYPO text attributes key used by the platform
val TYPO: TextAttributesKey = TextAttributesKey.createTextAttributesKey("TYPO")
try {
ab.textAttributes(TYPO)
} catch (_: Throwable) {
// some IDEs may not allow setting attributes on INFORMATION; ignore gracefully
}
}
private fun suggestReplacements(file: PsiFile, word: String): List<String> {
val lower = word.lowercase()
val fromProject = collectProjectWords(file)
val fromTech = TechDictionary.allWords()
val fromEnglish = EnglishDictionary.allWords()
// Merge with priority: project (p=0), tech (p=1), english (p=2)
val all = LinkedHashSet<String>()
all.addAll(fromProject)
all.addAll(fromTech)
all.addAll(fromEnglish)
data class Cand(val w: String, val d: Int, val p: Int)
val cands = ArrayList<Cand>(32)
for (w in all) {
if (w == lower) continue
if (kotlin.math.abs(w.length - lower.length) > 2) continue
val d = editDistance(lower, w)
val p = when {
w in fromProject -> 0
w in fromTech -> 1
else -> 2
}
cands += Cand(w, d, p)
}
cands.sortWith(compareBy<Cand> { it.d }.thenBy { it.p }.thenBy { it.w })
// Return a larger pool so callers can choose desired display count
return cands.take(16).map { it.w }
}
private fun collectProjectWords(file: PsiFile): Set<String> {
// Simple approach: use current file text; can be extended to project scanning later
val text = file.viewProvider.document?.text ?: return emptySet()
val out = LinkedHashSet<String>()
val tokenRegex = Regex("[A-Za-z][A-Za-z0-9_']{2,}")
for (m in tokenRegex.findAll(text)) {
val parts = splitIdentifier(m.value)
parts.forEach { out += it.lowercase() }
}
// Include learned words
val settings = LyngFormatterSettings.getInstance(file.project)
out.addAll(settings.learnedWords.map { it.lowercase() })
return out
}
private fun splitIdentifier(token: String): List<String> {
// Split on underscores and camelCase boundaries
val unders = token.split('_').filter { it.isNotBlank() }
val out = mutableListOf<String>()
val camelBoundary = Regex("(?<=[a-z])(?=[A-Z])")
for (u in unders) out += u.split(camelBoundary).filter { it.isNotBlank() }
return out
}
private fun isAllowedWord(w: String): Boolean {
val s = w.lowercase()
return s in setOf(
// common code words / language keywords to avoid noise
"val","var","fun","class","enum","type","import","package","return","if","else","when","while","for","try","catch","finally","true","false","null",
// very common English words
"the","and","or","not","with","from","into","this","that","file","found","count","name","value","object"
)
}
private fun editDistance(a: String, b: String): Int {
if (a == b) return 0
if (a.isEmpty()) return b.length
if (b.isEmpty()) return a.length
val dp = IntArray(b.length + 1) { it }
for (i in 1..a.length) {
var prev = dp[0]
dp[0] = i
for (j in 1..b.length) {
val temp = dp[j]
dp[j] = minOf(
dp[j] + 1, // deletion
dp[j - 1] + 1, // insertion
prev + if (a[i - 1] == b[j - 1]) 0 else 1 // substitution
)
prev = temp
}
}
return dp[b.length]
}
}

View File

@ -0,0 +1,139 @@
/*
* Copyright 2025 Sergey S. Chernov real.sergeych@gmail.com
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*
*/
package net.sergeych.lyng.idea.grazie
import com.intellij.grazie.grammar.strategy.GrammarCheckingStrategy
import com.intellij.grazie.grammar.strategy.GrammarCheckingStrategy.TextDomain
import com.intellij.ide.plugins.PluginManagerCore
import com.intellij.openapi.diagnostic.Logger
import com.intellij.openapi.extensions.PluginId
import com.intellij.openapi.util.TextRange
import com.intellij.psi.PsiElement
import net.sergeych.lyng.idea.highlight.LyngTokenTypes
import net.sergeych.lyng.idea.settings.LyngFormatterSettings
import net.sergeych.lyng.idea.spell.LyngSpellIndex
/**
* Grazie/Natural Languages strategy for Lyng.
*
* - Comments: checked as natural language (TextDomain.COMMENTS)
* - String literals: optionally checked (setting), skipping printf-like specifiers via stealth ranges (TextDomain.LITERALS)
* - Identifiers (non-keywords): checked under TextDomain.CODE so "Process code" controls apply
* - Keywords: skipped
*/
class LyngGrazieStrategy : GrammarCheckingStrategy {
private val log = Logger.getInstance(LyngGrazieStrategy::class.java)
@Volatile private var loggedOnce = false
@Volatile private var loggedFirstMatch = false
private val seenTypes: MutableSet<String> = java.util.Collections.synchronizedSet(mutableSetOf())
private fun legacySpellcheckerInstalled(): Boolean =
PluginManagerCore.isPluginInstalled(PluginId.getId("com.intellij.spellchecker"))
// Regex for printf-style specifiers: %[flags][width][.precision][length]type
private val spec = Regex("%(?:[-+ #0]*(?:\\d+)?(?:\\.\\d+)?[a-zA-Z%])")
override fun isMyContextRoot(element: PsiElement): Boolean {
val type = element.node?.elementType
val settings = LyngFormatterSettings.getInstance(element.project)
val legacyPresent = legacySpellcheckerInstalled()
if (type != null && seenTypes.size < 10) {
val name = type.toString()
if (seenTypes.add(name)) {
log.info("LyngGrazieStrategy: saw PSI type=$name")
}
}
if (!loggedOnce) {
loggedOnce = true
log.info("LyngGrazieStrategy activated: legacyPresent=$legacyPresent, preferGrazieForCommentsAndLiterals=${settings.preferGrazieForCommentsAndLiterals}, spellCheckStringLiterals=${settings.spellCheckStringLiterals}, grazieChecksIdentifiers=${settings.grazieChecksIdentifiers}")
}
val file = element.containingFile ?: return false
val index = LyngSpellIndex.getUpToDate(file) ?: return false // Suspend until ready
// To ensure Grazie asks TextExtractor for all leafs, accept any Lyng element once index is ready.
// The extractor will decide per-range/domain what to actually provide.
if (!loggedFirstMatch) {
loggedFirstMatch = true
log.info("LyngGrazieStrategy: enabling Grazie on all Lyng elements (index ready)")
}
return true
}
override fun getContextRootTextDomain(root: PsiElement): TextDomain {
val type = root.node?.elementType
val settings = LyngFormatterSettings.getInstance(root.project)
val file = root.containingFile
val index = if (file != null) LyngSpellIndex.getUpToDate(file) else null
val r = root.textRange
fun overlaps(list: List<TextRange>): Boolean = r != null && list.any { it.intersects(r) }
return when (type) {
LyngTokenTypes.LINE_COMMENT, LyngTokenTypes.BLOCK_COMMENT -> TextDomain.COMMENTS
LyngTokenTypes.STRING -> if (settings.grazieTreatLiteralsAsComments) TextDomain.COMMENTS else TextDomain.LITERALS
LyngTokenTypes.IDENTIFIER -> {
// For Grazie-only reliability in 243, route identifiers via COMMENTS when configured
if (settings.grazieTreatIdentifiersAsComments && index != null && r != null && overlaps(index.identifiers))
TextDomain.COMMENTS
else TextDomain.PLAIN_TEXT
}
else -> TextDomain.PLAIN_TEXT
}
}
// Note: do not override getLanguageSupport to keep compatibility with 243 API
override fun getStealthyRanges(root: PsiElement, text: CharSequence): java.util.LinkedHashSet<IntRange> {
val result = LinkedHashSet<IntRange>()
val type = root.node?.elementType
if (type == LyngTokenTypes.STRING) {
if (!shouldCheckLiterals(root)) {
// Hide the entire string when literals checking is disabled by settings
result += (0 until text.length)
return result
}
// Hide printf-like specifiers in strings
val (start, end) = stripQuotesBounds(text)
if (end > start) {
val content = text.subSequence(start, end)
for (m in spec.findAll(content)) {
val ms = start + m.range.first
val me = start + m.range.last
result += (ms..me)
}
if (result.isNotEmpty()) {
log.debug("LyngGrazieStrategy: hidden ${result.size} printf specifier ranges in string literal")
}
}
}
return result
}
override fun isEnabledByDefault(): Boolean = true
private fun shouldCheckLiterals(root: PsiElement): Boolean =
LyngFormatterSettings.getInstance(root.project).spellCheckStringLiterals
private fun stripQuotesBounds(text: CharSequence): Pair<Int, Int> {
if (text.length < 2) return 0 to text.length
val first = text.first()
val last = text.last()
return if ((first == '"' && last == '"') || (first == '\'' && last == '\''))
1 to (text.length - 1) else (0 to text.length)
}
}

View File

@ -1,5 +1,5 @@
/*
* Copyright 2026 Sergey S. Chernov real.sergeych@gmail.com
* Copyright 2025 Sergey S. Chernov real.sergeych@gmail.com
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
@ -19,29 +19,86 @@ package net.sergeych.lyng.idea.grazie
import com.intellij.grazie.text.TextContent
import com.intellij.grazie.text.TextContent.TextDomain
import com.intellij.grazie.text.TextExtractor
import com.intellij.openapi.diagnostic.Logger
import com.intellij.psi.PsiElement
import net.sergeych.lyng.idea.highlight.LyngTokenTypes
import net.sergeych.lyng.idea.psi.LyngElementTypes
import net.sergeych.lyng.idea.settings.LyngFormatterSettings
import net.sergeych.lyng.idea.spell.LyngSpellIndex
/**
* Simplified TextExtractor for Lyng.
* Designates areas for Natural Languages (Grazie) to check.
* Provides Grazie with extractable text for Lyng PSI elements.
* We return text for identifiers, comments, and (optionally) string literals.
* printf-like specifiers are filtered by the Grammar strategy via stealth ranges.
*/
class LyngTextExtractor : TextExtractor() {
private val log = Logger.getInstance(LyngTextExtractor::class.java)
@Volatile private var loggedOnce = false
private val seen: MutableSet<String> = java.util.Collections.synchronizedSet(mutableSetOf())
override fun buildTextContent(element: PsiElement, allowedDomains: Set<TextDomain>): TextContent? {
val type = element.node?.elementType ?: return null
val domain = when (type) {
LyngTokenTypes.LINE_COMMENT, LyngTokenTypes.BLOCK_COMMENT -> TextDomain.COMMENTS
LyngTokenTypes.STRING -> TextDomain.LITERALS
LyngElementTypes.NAME_IDENTIFIER,
LyngElementTypes.PARAMETER_NAME,
LyngElementTypes.ENUM_CONSTANT_NAME -> TextDomain.COMMENTS
else -> return null
if (!loggedOnce) {
loggedOnce = true
log.info("LyngTextExtractor active; allowedDomains=${allowedDomains.joinToString()}")
}
val settings = LyngFormatterSettings.getInstance(element.project)
val file = element.containingFile
val index = if (file != null) LyngSpellIndex.getUpToDate(file) else null
val r = element.textRange
fun overlaps(list: List<com.intellij.openapi.util.TextRange>): Boolean = r != null && list.any { it.intersects(r) }
// Decide target domain by intersection with our MiniAst-driven index; prefer comments > strings > identifiers
var domain: TextDomain? = null
if (index != null && r != null) {
if (overlaps(index.comments)) domain = TextDomain.COMMENTS
else if (overlaps(index.strings) && settings.spellCheckStringLiterals) domain = TextDomain.LITERALS
else if (overlaps(index.identifiers)) domain = if (settings.grazieTreatIdentifiersAsComments) TextDomain.COMMENTS else TextDomain.DOCUMENTATION
} else {
// Fallback to token type if index is not ready (rare timing), mostly for comments
domain = when (type) {
LyngTokenTypes.LINE_COMMENT, LyngTokenTypes.BLOCK_COMMENT -> TextDomain.COMMENTS
else -> null
}
}
if (domain == null) return null
// If literals aren't requested but fallback is enabled, route strings as COMMENTS
if (domain == TextDomain.LITERALS && !allowedDomains.contains(TextDomain.LITERALS) && settings.grazieTreatLiteralsAsComments) {
domain = TextDomain.COMMENTS
}
if (!allowedDomains.contains(domain)) {
if (seen.add("deny-${domain.name}")) {
log.info("LyngTextExtractor: domain ${domain.name} not in allowedDomains; skipping")
}
return null
}
return try {
// Try common factory names across versions
val methods = TextContent::class.java.methods.filter { it.name == "psiFragment" }
val built: TextContent? = when {
// Try psiFragment(PsiElement, TextDomain)
methods.any { it.parameterCount == 2 && it.parameterTypes[0].name.contains("PsiElement") } -> {
val m = methods.first { it.parameterCount == 2 && it.parameterTypes[0].name.contains("PsiElement") }
@Suppress("UNCHECKED_CAST")
(m.invoke(null, element, domain) as? TextContent)?.also {
if (seen.add("ok-${domain.name}")) log.info("LyngTextExtractor: provided ${domain.name} for ${type} via psiFragment(element, domain)")
}
}
// Try psiFragment(TextDomain, PsiElement)
methods.any { it.parameterCount == 2 && it.parameterTypes[0].name.endsWith("TextDomain") } -> {
val m = methods.first { it.parameterCount == 2 && it.parameterTypes[0].name.endsWith("TextDomain") }
@Suppress("UNCHECKED_CAST")
(m.invoke(null, domain, element) as? TextContent)?.also {
if (seen.add("ok-${domain.name}")) log.info("LyngTextExtractor: provided ${domain.name} for ${type} via psiFragment(domain, element)")
}
}
else -> null
}
built
} catch (e: Throwable) {
log.info("LyngTextExtractor: failed to build TextContent: ${e.javaClass.simpleName}: ${e.message}")
null
}
if (!allowedDomains.contains(domain)) return null
return TextContent.psiFragment(domain, element)
}
}

View File

@ -0,0 +1,86 @@
/*
* Copyright 2025 Sergey S. Chernov real.sergeych@gmail.com
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*
*/
package net.sergeych.lyng.idea.grazie
import com.intellij.codeInsight.daemon.DaemonCodeAnalyzer
import com.intellij.codeInsight.intention.IntentionAction
import com.intellij.openapi.command.WriteCommandAction
import com.intellij.openapi.editor.CaretModel
import com.intellij.openapi.editor.Document
import com.intellij.openapi.editor.Editor
import com.intellij.openapi.project.Project
import com.intellij.openapi.util.TextRange
import com.intellij.psi.PsiFile
/**
* Lightweight quick-fix to replace a misspelled word (subrange) with a suggested alternative.
* Works without the legacy Spell Checker. The replacement is applied directly to the file text.
*/
class ReplaceWordFix(
private val range: TextRange,
private val original: String,
private val replacementRaw: String
) : IntentionAction {
override fun getText(): String = "Replace '$original' with '$replacementRaw'"
override fun getFamilyName(): String = "Lyng Spelling"
override fun isAvailable(project: Project, editor: Editor?, file: PsiFile?): Boolean =
editor != null && file != null && range.startOffset in 0..range.endOffset
override fun startInWriteAction(): Boolean = true
override fun invoke(project: Project, editor: Editor?, file: PsiFile?) {
if (editor == null) return
val doc: Document = editor.document
val safeRange = range.constrainTo(doc)
val current = doc.getText(safeRange)
// Preserve basic case style based on the original token
val replacement = adaptCaseStyle(current, replacementRaw)
WriteCommandAction.runWriteCommandAction(project, "Replace word", null, Runnable {
doc.replaceString(safeRange.startOffset, safeRange.endOffset, replacement)
}, file)
// Move caret to end of replacement for convenience
try {
val caret: CaretModel = editor.caretModel
caret.moveToOffset(safeRange.startOffset + replacement.length)
} catch (_: Throwable) {}
// Restart daemon to refresh highlights
if (file != null) DaemonCodeAnalyzer.getInstance(project).restart(file)
}
private fun TextRange.constrainTo(doc: Document): TextRange {
val start = startOffset.coerceIn(0, doc.textLength)
val end = endOffset.coerceIn(start, doc.textLength)
return TextRange(start, end)
}
private fun adaptCaseStyle(sample: String, suggestion: String): String {
if (suggestion.isEmpty()) return suggestion
return when {
sample.all { it.isUpperCase() } -> suggestion.uppercase()
// PascalCase / Capitalized single word
sample.firstOrNull()?.isUpperCase() == true && sample.drop(1).any { it.isLowerCase() } ->
suggestion.replaceFirstChar { if (it.isLowerCase()) it.titlecase() else it.toString() }
// snake_case -> lower
sample.contains('_') -> suggestion.lowercase()
// camelCase -> lower first
sample.firstOrNull()?.isLowerCase() == true && sample.any { it.isUpperCase() } ->
suggestion.replaceFirstChar { it.lowercase() }
else -> suggestion
}
}
}

View File

@ -0,0 +1,77 @@
/*
* Copyright 2025 Sergey S. Chernov real.sergeych@gmail.com
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*
*/
package net.sergeych.lyng.idea.grazie
import com.intellij.openapi.diagnostic.Logger
import java.io.BufferedReader
import java.io.InputStreamReader
import java.util.zip.GZIPInputStream
/**
* Lightweight technical/Lyng vocabulary dictionary.
* Loaded from classpath resources; supports .txt and .txt.gz. Merged with EnglishDictionary.
*/
object TechDictionary {
private val log = Logger.getInstance(TechDictionary::class.java)
@Volatile private var loaded = false
@Volatile private var words: Set<String> = emptySet()
private fun ensureLoaded() {
if (loaded) return
synchronized(this) {
if (loaded) return
val candidates = listOf(
"/dictionaries/tech-lyng.txt.gz",
"/dictionaries/tech-lyng.txt"
)
val merged = HashSet<String>(8_000)
for (res in candidates) {
try {
val stream = javaClass.getResourceAsStream(res) ?: continue
val reader = if (res.endsWith(".gz"))
BufferedReader(InputStreamReader(GZIPInputStream(stream)))
else
BufferedReader(InputStreamReader(stream))
var n = 0
reader.useLines { seq -> seq.forEach { line ->
val w = line.trim()
if (w.isNotEmpty() && !w.startsWith("#")) { merged += w.lowercase(); n++ }
} }
log.info("TechDictionary: loaded $n words from $res (total=${merged.size})")
} catch (t: Throwable) {
log.info("TechDictionary: failed to load $res: ${t.javaClass.simpleName}: ${t.message}")
}
}
if (merged.isEmpty()) {
merged += setOf(
// minimal Lyng/tech seeding to avoid empty dictionary
"lyng","miniast","binder","printf","specifier","specifiers","regex","token","tokens",
"identifier","identifiers","keyword","keywords","comment","comments","string","strings",
"literal","literals","formatting","formatter","grazie","typo","typos","dictionary","dictionaries"
)
log.info("TechDictionary: using minimal built-in set (${merged.size})")
}
words = merged
loaded = true
}
}
fun allWords(): Set<String> {
ensureLoaded()
return words
}
}

View File

@ -43,15 +43,10 @@ class LyngColorSettingsPage : ColorSettingsPage {
}
var counter = 0
outer@ while (counter < 10) {
if (counter == 5) return@outer
counter = counter + 1
}
counter = counter + 1
""".trimIndent()
override fun getAdditionalHighlightingTagToDescriptorMap(): MutableMap<String, TextAttributesKey> = mutableMapOf(
"label" to LyngHighlighterColors.LABEL
)
override fun getAdditionalHighlightingTagToDescriptorMap(): MutableMap<String, TextAttributesKey>? = null
override fun getAttributeDescriptors(): Array<AttributesDescriptor> = arrayOf(
AttributesDescriptor("Keyword", LyngHighlighterColors.KEYWORD),
@ -63,7 +58,6 @@ class LyngColorSettingsPage : ColorSettingsPage {
AttributesDescriptor("Punctuation", LyngHighlighterColors.PUNCT),
// Semantic
AttributesDescriptor("Annotation (semantic)", LyngHighlighterColors.ANNOTATION),
AttributesDescriptor("Label (semantic)", LyngHighlighterColors.LABEL),
AttributesDescriptor("Variable (semantic)", LyngHighlighterColors.VARIABLE),
AttributesDescriptor("Value (semantic)", LyngHighlighterColors.VALUE),
AttributesDescriptor("Function (semantic)", LyngHighlighterColors.FUNCTION),

View File

@ -82,9 +82,4 @@ object LyngHighlighterColors {
val ENUM_CONSTANT: TextAttributesKey = TextAttributesKey.createTextAttributesKey(
"LYNG_ENUM_CONSTANT", DefaultLanguageHighlighterColors.STATIC_FIELD
)
// Labels (label@ or @label used as exit target)
val LABEL: TextAttributesKey = TextAttributesKey.createTextAttributesKey(
"LYNG_LABEL", DefaultLanguageHighlighterColors.LABEL
)
}

View File

@ -1,5 +1,5 @@
/*
* Copyright 2026 Sergey S. Chernov real.sergeych@gmail.com
* Copyright 2025 Sergey S. Chernov real.sergeych@gmail.com
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
@ -32,11 +32,9 @@ class LyngLexer : LexerBase() {
private var myTokenType: IElementType? = null
private val keywords = setOf(
"fun", "val", "var", "class", "interface", "type", "import", "as",
"abstract", "closed", "override", "static", "extern", "open", "private", "protected",
"fun", "val", "var", "class", "type", "import", "as",
"if", "else", "for", "while", "return", "true", "false", "null",
"when", "in", "is", "break", "continue", "try", "catch", "finally",
"get", "set", "object", "enum", "init", "by", "step", "property", "constructor"
"when", "in", "is", "break", "continue", "try", "catch", "finally"
)
override fun start(buffer: CharSequence, startOffset: Int, endOffset: Int, initialState: Int) {
@ -101,9 +99,8 @@ class LyngLexer : LexerBase() {
return
}
// String "..." or '...' with simple escape handling
if (ch == '"' || ch == '\'') {
val quote = ch
// String "..." with simple escape handling
if (ch == '"') {
i++
while (i < endOffset) {
val c = buffer[i]
@ -111,7 +108,7 @@ class LyngLexer : LexerBase() {
i += 2
continue
}
if (c == quote) { i++; break }
if (c == '"') { i++; break }
i++
}
myTokenEnd = i
@ -121,39 +118,12 @@ class LyngLexer : LexerBase() {
// Number
if (ch.isDigit()) {
// Check for hex: 0x...
if (ch == '0' && i + 1 < endOffset && buffer[i + 1] == 'x') {
i += 2
while (i < endOffset && (buffer[i].isDigit() || buffer[i] in 'a'..'f' || buffer[i] in 'A'..'F')) i++
myTokenEnd = i
myTokenType = LyngTokenTypes.NUMBER
return
}
// Decimal or integer
i++
var hasDot = false
var hasE = false
while (i < endOffset) {
val c = buffer[i]
if (c.isDigit() || c == '_') {
i++
continue
}
if (c == '.' && !hasDot && !hasE) {
// Check if it's a fractional part (must be followed by a digit)
if (i + 1 < endOffset && buffer[i + 1].isDigit()) {
hasDot = true
i++
continue
}
}
if ((c == 'e' || c == 'E') && !hasE) {
hasE = true
i++
if (i < endOffset && (buffer[i] == '+' || buffer[i] == '-')) i++
continue
}
if (c.isDigit()) { i++; continue }
if (c == '.' && !hasDot) { hasDot = true; i++; continue }
break
}
myTokenEnd = i
@ -161,24 +131,10 @@ class LyngLexer : LexerBase() {
return
}
// Labels / Annotations: @label or label@
if (ch == '@') {
i++
while (i < endOffset && (buffer[i].isIdentifierPart())) i++
myTokenEnd = i
myTokenType = LyngTokenTypes.LABEL
return
}
// Identifier / keyword
if (ch.isIdentifierStart()) {
i++
while (i < endOffset && buffer[i].isIdentifierPart()) i++
if (i < endOffset && buffer[i] == '@') {
i++
myTokenEnd = i
myTokenType = LyngTokenTypes.LABEL
return
}
myTokenEnd = i
val text = buffer.subSequence(myTokenStart, myTokenEnd).toString()
myTokenType = if (text in keywords) LyngTokenTypes.KEYWORD else LyngTokenTypes.IDENTIFIER
@ -188,35 +144,6 @@ class LyngLexer : LexerBase() {
// Punctuation
if (isPunct(ch)) {
i++
// Handle common multi-char operators for better highlighting
when (ch) {
'.' -> {
if (i < endOffset && buffer[i] == '.') {
i++
if (i < endOffset && (buffer[i] == '.' || buffer[i] == '<')) i++
}
}
'=' -> {
if (i < endOffset && (buffer[i] == '=' || buffer[i] == '>' || buffer[i] == '~')) {
i++
if (buffer[i - 1] == '=' && i < endOffset && buffer[i] == '=') i++
}
}
'+', '-', '*', '/', '%', '!', '<', '>', '&', '|', '?', ':', '^' -> {
if (i < endOffset) {
val next = buffer[i]
if (next == '=' || next == ch) {
i++
if (ch == '<' && next == '=' && i < endOffset && buffer[i] == '>') i++
if (ch == '!' && next == '=' && i < endOffset && buffer[i] == '=') i++
} else if (ch == '?' && (next == '.' || next == '[' || next == '(' || next == '{' || next == ':' || next == '?')) {
i++
} else if (ch == '-' && next == '>') {
i++
}
}
}
}
myTokenEnd = i
myTokenType = LyngTokenTypes.PUNCT
return
@ -231,5 +158,5 @@ class LyngLexer : LexerBase() {
private fun Char.isDigit(): Boolean = this in '0'..'9'
private fun Char.isIdentifierStart(): Boolean = this == '_' || this.isLetter()
private fun Char.isIdentifierPart(): Boolean = this.isIdentifierStart() || this.isDigit()
private fun isPunct(c: Char): Boolean = c in setOf('(', ')', '{', '}', '[', ']', '.', ',', ';', ':', '+', '-', '*', '/', '%', '=', '<', '>', '!', '?', '&', '|', '^', '~', '@')
private fun isPunct(c: Char): Boolean = c in setOf('(', ')', '{', '}', '[', ']', '.', ',', ';', ':', '+', '-', '*', '/', '%', '=', '<', '>', '!', '?', '&', '|', '^', '~')
}

View File

@ -33,7 +33,6 @@ class LyngSyntaxHighlighter : SyntaxHighlighter {
LyngTokenTypes.BLOCK_COMMENT -> pack(LyngHighlighterColors.BLOCK_COMMENT)
LyngTokenTypes.PUNCT -> pack(LyngHighlighterColors.PUNCT)
LyngTokenTypes.IDENTIFIER -> pack(LyngHighlighterColors.IDENTIFIER)
LyngTokenTypes.LABEL -> pack(LyngHighlighterColors.LABEL)
else -> emptyArray()
}

View File

@ -29,7 +29,6 @@ object LyngTokenTypes {
val NUMBER = LyngTokenType("NUMBER")
val KEYWORD = LyngTokenType("KEYWORD")
val IDENTIFIER = LyngTokenType("IDENTIFIER")
val LABEL = LyngTokenType("LABEL")
val PUNCT = LyngTokenType("PUNCT")
val BAD_CHAR = LyngTokenType("BAD_CHAR")
}

View File

@ -1,109 +0,0 @@
/*
* Copyright 2026 Sergey S. Chernov real.sergeych@gmail.com
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*
*/
package net.sergeych.lyng.idea.navigation
import com.intellij.icons.AllIcons
import com.intellij.navigation.ItemPresentation
import com.intellij.openapi.util.TextRange
import com.intellij.psi.PsiDocumentManager
import com.intellij.psi.PsiElement
import com.intellij.psi.PsiFile
import com.intellij.psi.PsiNameIdentifierOwner
import com.intellij.psi.impl.light.LightElement
import com.intellij.util.IncorrectOperationException
import net.sergeych.lyng.idea.LyngLanguage
import javax.swing.Icon
/**
* A light PSI element representing a Lyng declaration (function, class, enum, or variable).
* Used for navigation and to provide a stable anchor for "Find Usages".
*/
class LyngDeclarationElement(
private val nameElement: PsiElement,
private val name: String,
val kind: String = "declaration"
) : LightElement(nameElement.manager, LyngLanguage), PsiNameIdentifierOwner {
override fun getName(): String = name
override fun setName(name: String): PsiElement {
throw IncorrectOperationException("Renaming is not yet supported")
}
override fun getNameIdentifier(): PsiElement = nameElement
override fun getNavigationElement(): PsiElement = nameElement
override fun getTextRange(): TextRange = nameElement.textRange
override fun getContainingFile(): PsiFile = nameElement.containingFile
override fun isValid(): Boolean = nameElement.isValid
override fun getPresentation(): ItemPresentation {
return object : ItemPresentation {
override fun getPresentableText(): String = name
override fun getLocationString(): String {
val file = containingFile
val document = PsiDocumentManager.getInstance(file.project).getDocument(file)
val line = if (document != null) document.getLineNumber(textRange.startOffset) + 1 else "?"
val column = if (document != null) {
val lineStart = document.getLineStartOffset(document.getLineNumber(textRange.startOffset))
textRange.startOffset - lineStart + 1
} else "?"
return "${file.name}:$line:$column"
}
override fun getIcon(unused: Boolean): Icon {
return when (kind) {
"Function" -> AllIcons.Nodes.Function
"Class" -> AllIcons.Nodes.Class
"Enum" -> AllIcons.Nodes.Enum
"EnumConstant" -> AllIcons.Nodes.Enum
"Variable" -> AllIcons.Nodes.Variable
"Value" -> AllIcons.Nodes.Field
"Parameter" -> AllIcons.Nodes.Parameter
"Initializer" -> AllIcons.Nodes.Method
else -> AllIcons.Nodes.Property
}
}
}
}
override fun toString(): String = "$kind:$name"
override fun equals(other: Any?): Boolean {
if (this === other) return true
if (other !is LyngDeclarationElement) return false
return name == other.name && nameElement == other.nameElement
}
override fun isEquivalentTo(another: PsiElement?): Boolean {
if (this === another) return true
if (another == nameElement) return true
if (another is LyngDeclarationElement) {
return name == another.name && nameElement == another.nameElement
}
return super.isEquivalentTo(another)
}
override fun hashCode(): Int {
var result = nameElement.hashCode()
result = 31 * result + name.hashCode()
return result
}
}

View File

@ -1,92 +0,0 @@
/*
* Copyright 2026 Sergey S. Chernov real.sergeych@gmail.com
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*
*/
package net.sergeych.lyng.idea.navigation
import com.intellij.lang.cacheBuilder.DefaultWordsScanner
import com.intellij.lang.cacheBuilder.WordsScanner
import com.intellij.lang.findUsages.FindUsagesProvider
import com.intellij.psi.PsiDocumentManager
import com.intellij.psi.PsiElement
import com.intellij.psi.tree.TokenSet
import net.sergeych.lyng.idea.highlight.LyngLexer
import net.sergeych.lyng.idea.highlight.LyngTokenTypes
import net.sergeych.lyng.idea.util.LyngAstManager
import net.sergeych.lyng.miniast.DocLookupUtils
class LyngFindUsagesProvider : FindUsagesProvider {
override fun getWordsScanner(): WordsScanner {
return DefaultWordsScanner(
LyngLexer(),
TokenSet.create(LyngTokenTypes.IDENTIFIER),
TokenSet.create(LyngTokenTypes.LINE_COMMENT, LyngTokenTypes.BLOCK_COMMENT),
TokenSet.create(LyngTokenTypes.STRING)
)
}
override fun canFindUsagesFor(psiElement: PsiElement): Boolean {
return psiElement is LyngDeclarationElement || isDeclaration(psiElement)
}
private fun isDeclaration(element: PsiElement): Boolean {
val file = element.containingFile ?: return false
val mini = LyngAstManager.getMiniAst(file) ?: return false
val offset = element.textRange.startOffset
val name = element.text ?: ""
return DocLookupUtils.findDeclarationAt(mini, offset, name) != null
}
override fun getHelpId(psiElement: PsiElement): String? = null
override fun getType(element: PsiElement): String {
if (element is LyngDeclarationElement) return element.kind
val file = element.containingFile ?: return "Lyng declaration"
val mini = LyngAstManager.getMiniAst(file) ?: return "Lyng declaration"
val info = DocLookupUtils.findDeclarationAt(mini, element.textRange.startOffset, element.text ?: "")
return info?.second ?: "Lyng declaration"
}
override fun getDescriptiveName(element: PsiElement): String {
if (element is LyngDeclarationElement) {
val file = element.containingFile
val document = PsiDocumentManager.getInstance(file.project).getDocument(file)
val line = if (document != null) document.getLineNumber(element.textRange.startOffset) + 1 else "?"
val column = if (document != null) {
val lineStart = document.getLineStartOffset(document.getLineNumber(element.textRange.startOffset))
element.textRange.startOffset - lineStart + 1
} else "?"
return "${element.name} (${file.name}:$line:$column)"
}
val file = element.containingFile ?: return element.text ?: "unknown"
val mini = LyngAstManager.getMiniAst(file) ?: return element.text ?: "unknown"
val info = DocLookupUtils.findDeclarationAt(mini, element.textRange.startOffset, element.text ?: "")
val document = PsiDocumentManager.getInstance(file.project).getDocument(file)
val line = if (document != null) document.getLineNumber(element.textRange.startOffset) + 1 else "?"
val column = if (document != null) {
val lineStart = document.getLineStartOffset(document.getLineNumber(element.textRange.startOffset))
element.textRange.startOffset - lineStart + 1
} else "?"
val name = info?.first ?: element.text ?: "unknown"
return "$name (${file.name}:$line:$column)"
}
override fun getNodeText(element: PsiElement, useFullName: Boolean): String {
return (element as? LyngDeclarationElement)?.name ?: element.text ?: "unknown"
}
}

View File

@ -1,59 +0,0 @@
/*
* Copyright 2026 Sergey S. Chernov real.sergeych@gmail.com
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*
*/
package net.sergeych.lyng.idea.navigation
import com.intellij.codeInsight.navigation.actions.GotoDeclarationHandler
import com.intellij.openapi.editor.Editor
import com.intellij.psi.PsiElement
import net.sergeych.lyng.idea.LyngLanguage
/**
* Ensures Ctrl+B (Go to Definition) works on Lyng identifiers by resolving through LyngPsiReference.
*/
class LyngGotoDeclarationHandler : GotoDeclarationHandler {
override fun getGotoDeclarationTargets(sourceElement: PsiElement?, offset: Int, editor: Editor?): Array<PsiElement>? {
if (sourceElement == null || sourceElement.language != LyngLanguage) return null
val allTargets = mutableListOf<PsiElement>()
// Find reference at the element or its parent (sometimes the identifier token is wrapped)
val ref = sourceElement.reference ?: sourceElement.parent?.reference
if (ref is LyngPsiReference) {
val resolved = ref.multiResolve(false)
allTargets.addAll(resolved.mapNotNull { it.element })
} else {
// Manual check if not picked up by reference (e.g. if contributor didn't run yet)
val manualRef = LyngPsiReference(sourceElement)
val manualResolved = manualRef.multiResolve(false)
allTargets.addAll(manualResolved.mapNotNull { it.element })
}
if (allTargets.isEmpty()) return null
// If there is only one target and it's equivalent to the source, return null.
// This allows IDEA to treat it as a declaration site and trigger "Show Usages".
if (allTargets.size == 1) {
val target = allTargets[0]
if (target == sourceElement || target.isEquivalentTo(sourceElement)) {
return null
}
}
return allTargets.toTypedArray()
}
}

View File

@ -1,48 +0,0 @@
/*
* Copyright 2026 Sergey S. Chernov real.sergeych@gmail.com
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*
*/
package net.sergeych.lyng.idea.navigation
import com.intellij.icons.AllIcons
import com.intellij.ide.IconProvider
import com.intellij.psi.PsiElement
import net.sergeych.lyng.idea.util.LyngAstManager
import net.sergeych.lyng.miniast.DocLookupUtils
import javax.swing.Icon
class LyngIconProvider : IconProvider() {
override fun getIcon(element: PsiElement, flags: Int): Icon? {
val file = element.containingFile ?: return null
val mini = LyngAstManager.getMiniAst(file) ?: return null
val info = DocLookupUtils.findDeclarationAt(mini, element.textRange.startOffset, element.text ?: "")
if (info != null) {
return when (info.second) {
"Function" -> AllIcons.Nodes.Function
"Class" -> AllIcons.Nodes.Class
"Enum" -> AllIcons.Nodes.Enum
"EnumConstant" -> AllIcons.Nodes.Enum
"Variable" -> AllIcons.Nodes.Variable
"Value" -> AllIcons.Nodes.Field
"Parameter" -> AllIcons.Nodes.Parameter
"Initializer" -> AllIcons.Nodes.Method
else -> null
}
}
return null
}
}

View File

@ -1,291 +0,0 @@
/*
* Copyright 2026 Sergey S. Chernov real.sergeych@gmail.com
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*
*/
package net.sergeych.lyng.idea.navigation
import com.intellij.openapi.project.Project
import com.intellij.openapi.util.TextRange
import com.intellij.psi.*
import com.intellij.psi.search.FileTypeIndex
import com.intellij.psi.search.FilenameIndex
import com.intellij.psi.search.GlobalSearchScope
import kotlinx.coroutines.runBlocking
import net.sergeych.lyng.highlight.offsetOf
import net.sergeych.lyng.idea.LyngFileType
import net.sergeych.lyng.idea.util.LyngAstManager
import net.sergeych.lyng.idea.util.TextCtx
import net.sergeych.lyng.miniast.*
import net.sergeych.lyng.tools.IdeLenientImportProvider
import net.sergeych.lyng.tools.LyngAnalysisRequest
import net.sergeych.lyng.tools.LyngLanguageTools
class LyngPsiReference(element: PsiElement) : PsiPolyVariantReferenceBase<PsiElement>(element, TextRange(0, element.textLength)) {
override fun multiResolve(incompleteCode: Boolean): Array<ResolveResult> {
val file = element.containingFile
val text = file.text
val offset = element.textRange.startOffset
val name = element.text ?: ""
val results = mutableListOf<ResolveResult>()
val analysis = LyngAstManager.getAnalysis(file) ?: return emptyArray()
val mini = analysis.mini ?: return emptyArray()
val binding = analysis.binding
val imported = analysis.importedModules.toSet()
val currentPackage = getPackageName(file)
val allowedPackages = if (currentPackage != null) imported + currentPackage else imported
// 1. Member resolution (obj.member)
val dotPos = TextCtx.findDotLeft(text, offset)
if (dotPos != null) {
val receiverClass = DocLookupUtils.guessReceiverClassViaMini(mini, text, dotPos, imported.toList(), binding)
?: DocLookupUtils.guessReceiverClass(text, dotPos, imported.toList(), mini)
val staticOnly = DocLookupUtils.isStaticReceiver(mini, text, dotPos, imported.toList(), binding)
if (receiverClass != null) {
val resolved = DocLookupUtils.resolveMemberWithInheritance(imported.toList(), receiverClass, name, mini, staticOnly = staticOnly)
if (resolved != null) {
val owner = resolved.first
val member = resolved.second
// We need to find the actual PSI element for this member
val targetFile = findFileForClass(file.project, owner) ?: file
val targetMini = loadMini(targetFile)
if (targetMini != null) {
val targetSrc = targetMini.range.start.source
val off = targetSrc.offsetOf(member.nameStart)
targetFile.findElementAt(off)?.let {
val kind = when(member) {
is MiniMemberFunDecl -> "Function"
is MiniMemberValDecl -> if (member.mutable) "Variable" else "Value"
is MiniMemberTypeAliasDecl -> "TypeAlias"
is MiniInitDecl -> "Initializer"
is MiniFunDecl -> "Function"
is MiniValDecl -> if (member.mutable) "Variable" else "Value"
is MiniClassDecl -> "Class"
is MiniEnumDecl -> "Enum"
is MiniTypeAliasDecl -> "TypeAlias"
}
results.add(PsiElementResolveResult(LyngDeclarationElement(it, member.name, kind)))
}
}
}
}
// If we couldn't resolve exactly, we might still want to search globally but ONLY for members
if (results.isEmpty()) {
results.addAll(resolveGlobally(file.project, name, membersOnly = true, allowedPackages = allowedPackages))
}
} else {
// 2. Local resolution via Binder
if (binding != null) {
val ref = binding.references.firstOrNull { offset >= it.start && offset < it.end }
if (ref != null) {
val sym = binding.symbols.firstOrNull { it.id == ref.symbolId }
if (sym != null && sym.declStart >= 0) {
file.findElementAt(sym.declStart)?.let {
results.add(PsiElementResolveResult(LyngDeclarationElement(it, sym.name, sym.kind.name)))
}
}
}
}
// 3. Global project scan
// Only search globally if we haven't found a strong local match
if (results.isEmpty()) {
results.addAll(resolveGlobally(file.project, name, allowedPackages = allowedPackages))
}
}
// 4. Filter results to exclude duplicates
// Use a more robust de-duplication that prefers the raw element if multiple refer to the same thing
val filtered = mutableListOf<ResolveResult>()
for (res in results) {
val el = res.element ?: continue
val nav = if (el is LyngDeclarationElement) el.navigationElement else el
if (filtered.none { existing ->
val exEl = existing.element
val exNav = if (exEl is LyngDeclarationElement) exEl.navigationElement else exEl
exNav == nav || (exNav != null && exNav.isEquivalentTo(nav))
}) {
filtered.add(res)
}
}
return filtered.toTypedArray()
}
private fun findFileForClass(project: Project, className: String): PsiFile? {
// 1. Try file with matching name first (optimization)
val scope = GlobalSearchScope.projectScope(project)
val psiManager = PsiManager.getInstance(project)
val matchingFiles = FileTypeIndex.getFiles(LyngFileType, scope)
.asSequence()
.filter { it.name == "$className.lyng" }
.mapNotNull { psiManager.findFile(it) }
.toList()
val matchingDeclFiles = FileTypeIndex.getFiles(LyngFileType, scope)
.asSequence()
.filter { it.name == "$className.lyng.d" }
.mapNotNull { psiManager.findFile(it) }
.toList()
for (file in matchingFiles) {
val mini = loadMini(file) ?: continue
if (mini.declarations.any { isLocalDecl(mini, it) && ((it is MiniClassDecl && it.name == className) || (it is MiniEnumDecl && it.name == className)) }) {
return file
}
}
for (file in matchingDeclFiles) {
val mini = loadMini(file) ?: continue
if (mini.declarations.any { isLocalDecl(mini, it) && ((it is MiniClassDecl && it.name == className) || (it is MiniEnumDecl && it.name == className)) }) {
return file
}
}
// 2. Fallback to full project scan
for (file in collectLyngFiles(project)) {
if (matchingFiles.contains(file) || matchingDeclFiles.contains(file)) continue // already checked
val mini = loadMini(file) ?: continue
if (mini.declarations.any { isLocalDecl(mini, it) && ((it is MiniClassDecl && it.name == className) || (it is MiniEnumDecl && it.name == className)) }) {
return file
}
}
return null
}
private fun getPackageName(file: PsiFile): String? {
val mini = loadMini(file) ?: return null
return try {
val pkg = mini.range.start.source.extractPackageName()
if (pkg.startsWith("lyng.")) pkg else "lyng.$pkg"
} catch (e: Exception) {
null
}
}
override fun resolve(): PsiElement? {
val results = multiResolve(false)
if (results.isEmpty()) return null
val target = results[0].element ?: return null
// If the target is equivalent to our source element, return the source element itself.
// This is crucial for IDEA to recognize we are already at the declaration site
// and trigger "Show Usages" instead of performing a no-op navigation.
if (target == element || target.isEquivalentTo(element)) {
return element
}
return target
}
private fun resolveGlobally(project: Project, name: String, membersOnly: Boolean = false, allowedPackages: Set<String>? = null): List<ResolveResult> {
val results = mutableListOf<ResolveResult>()
val psiManager = PsiManager.getInstance(project)
for (file in collectLyngFiles(project)) {
// Filter by package if requested
if (allowedPackages != null) {
val pkg = getPackageName(file)
if (pkg == null) {
if (!file.name.endsWith(".lyng.d")) continue
} else if (pkg !in allowedPackages) continue
}
val mini = loadMini(file) ?: continue
val src = mini.range.start.source
fun addIfMatch(dName: String, nameStart: net.sergeych.lyng.Pos, dKind: String) {
if (dName == name) {
val off = src.offsetOf(nameStart)
file.findElementAt(off)?.let {
results.add(PsiElementResolveResult(LyngDeclarationElement(it, dName, dKind)))
}
}
}
for (d in mini.declarations) {
if (!isLocalDecl(mini, d)) continue
if (!membersOnly) {
val dKind = when(d) {
is net.sergeych.lyng.miniast.MiniFunDecl -> "Function"
is net.sergeych.lyng.miniast.MiniClassDecl -> "Class"
is net.sergeych.lyng.miniast.MiniEnumDecl -> "Enum"
is net.sergeych.lyng.miniast.MiniValDecl -> if (d.mutable) "Variable" else "Value"
is net.sergeych.lyng.miniast.MiniTypeAliasDecl -> "TypeAlias"
}
addIfMatch(d.name, d.nameStart, dKind)
}
// Check members of classes and enums
val members = when(d) {
is MiniClassDecl -> d.members
is MiniEnumDecl -> DocLookupUtils.enumToSyntheticClass(d).members
else -> emptyList()
}
for (m in members) {
if (m.range.start.source != src) continue
val mKind = when(m) {
is net.sergeych.lyng.miniast.MiniMemberFunDecl -> "Function"
is net.sergeych.lyng.miniast.MiniMemberValDecl -> if (m.mutable) "Variable" else "Value"
is net.sergeych.lyng.miniast.MiniMemberTypeAliasDecl -> "TypeAlias"
is net.sergeych.lyng.miniast.MiniInitDecl -> "Initializer"
}
addIfMatch(m.name, m.nameStart, mKind)
}
}
}
return results
}
private fun collectLyngFiles(project: Project): List<PsiFile> {
val scope = GlobalSearchScope.projectScope(project)
val psiManager = PsiManager.getInstance(project)
val out = LinkedHashSet<PsiFile>()
val lyngFiles = FilenameIndex.getAllFilesByExt(project, "lyng", scope)
for (vFile in lyngFiles) {
psiManager.findFile(vFile)?.let { out.add(it) }
}
// Include declaration files (*.lyng.d) which are indexed as extension "d".
val dFiles = FilenameIndex.getAllFilesByExt(project, "d", scope)
for (vFile in dFiles) {
if (!vFile.name.endsWith(".lyng.d")) continue
psiManager.findFile(vFile)?.let { out.add(it) }
}
return out.toList()
}
private fun loadMini(file: PsiFile): MiniScript? {
LyngAstManager.getMiniAst(file)?.let { return it }
return try {
val provider = IdeLenientImportProvider.create()
runBlocking {
LyngLanguageTools.analyze(
LyngAnalysisRequest(text = file.text, fileName = file.name, importProvider = provider)
)
}.mini
} catch (_: Throwable) {
null
}
}
private fun isLocalDecl(mini: MiniScript, decl: MiniDecl): Boolean =
decl.range.start.source == mini.range.start.source
override fun getVariants(): Array<Any> = emptyArray()
}

View File

@ -1,54 +0,0 @@
/*
* Copyright 2026 Sergey S. Chernov real.sergeych@gmail.com
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*
*/
package net.sergeych.lyng.idea.navigation
import com.intellij.patterns.PlatformPatterns
import com.intellij.psi.*
import com.intellij.util.ProcessingContext
import net.sergeych.lyng.idea.LyngLanguage
import net.sergeych.lyng.idea.highlight.LyngTokenTypes
import net.sergeych.lyng.idea.util.LyngAstManager
import net.sergeych.lyng.miniast.DocLookupUtils
class LyngPsiReferenceContributor : PsiReferenceContributor() {
override fun registerReferenceProviders(registrar: PsiReferenceRegistrar) {
registrar.registerReferenceProvider(
PlatformPatterns.psiElement().withLanguage(LyngLanguage),
object : PsiReferenceProvider() {
override fun getReferencesByElement(
element: PsiElement,
context: ProcessingContext
): Array<PsiReference> {
if (element.node.elementType == LyngTokenTypes.IDENTIFIER) {
val file = element.containingFile
val mini = LyngAstManager.getMiniAst(file)
if (mini != null) {
val offset = element.textRange.startOffset
val name = element.text ?: ""
if (DocLookupUtils.findDeclarationAt(mini, offset, name) != null) {
return PsiReference.EMPTY_ARRAY
}
}
return arrayOf(LyngPsiReference(element))
}
return PsiReference.EMPTY_ARRAY
}
}
)
}
}

Some files were not shown because too many files have changed in this diff Show More