Compare commits

..

1 Commits

Author SHA1 Message Date
aec4a3766e wasm generation bug workaround, docs and debugging tips 2026-01-24 18:10:49 +03:00
667 changed files with 12869 additions and 106742 deletions

10
.gitignore vendored
View File

@ -21,13 +21,3 @@ xcuserdata
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
/site/src/version-template/lyng-version.js
/bugcontents.db
/bugs/
contents.db

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

@ -1,43 +0,0 @@
# AI Agent Notes
## Canonical AI References
- Use `docs/ai_language_reference.md` as the primary, compiler-verified Lyng language reference for code generation.
- For generics-heavy code generation, follow `docs/ai_language_reference.md` section `7.1 Generics Runtime Model and Bounds` and `7.2 Differences vs Java / Kotlin / Scala`.
- Use `docs/ai_stdlib_reference.md` for default runtime/module APIs and stdlib surface.
- Treat `LYNG_AI_SPEC.md` and older docs as secondary if they conflict with the two files above.
- Prefer the shortest clear loop: use `for` for straightforward iteration/ranges; use `while` only when loop state/condition is irregular or changes in ways `for` cannot express cleanly.
- In Lyng code, slice strings with range indexing (`text[a..<b]`, `text[..<n]`, `text[n..]`) and avoid Java/Kotlin-style `substring(...)`.
## Lyng-First API Declarations
- Use `.lyng` declarations as the single source of truth for Lyng-facing API docs and types (especially module extern declarations).
- Prefer defining Lyng entities (enums/classes/type shapes) in `.lyng` files; only define them in Kotlin when there is Kotlin/platform-specific implementation detail that cannot be expressed in Lyng.
- Avoid hardcoding Lyng API documentation in Kotlin registrars when it can be declared in `.lyng`; Kotlin-side docs should be fallback/bridge only.
- For mixed pluggable modules (Lyng + Kotlin), embed module `.lyng` sources as generated Kotlin string literals, evaluate them into module scope during registration, then attach Kotlin implementations/bindings.
- When a change adds or changes Lyng-visible runtime/module behavior, update the corresponding `.lyng` declaration in the same change, including declaration-level docs/comments for new API surface.
## 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.
- For any code in `commonMain`, verify it is Kotlin Multiplatform compatible before finishing. Do not use JVM-only APIs or Java-backed convenience methods such as `Map.putIfAbsent`; prefer stdlib/common equivalents and run at least the relevant compile/test task that exercises the `commonMain` source set.
- 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.
## ABI proposal notes
- Runtime generic metadata for generic extern classes is tracked in `proposals/extern_generic_runtime_abi.md`.
- Keep this design `Obj`-centric: do not assume extern-class values are `ObjInstance`; collection must be enabled on `ObjClass`.

View File

@ -1,152 +1,119 @@
## Changelog
### Unreleased
- 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".
- 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
- 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.
- IDEA plugin: Lightweight autocompletion (experimental)
- Global completion: local declarations, in‑scope parameters, imported modules, and stdlib symbols.
- Member completion: after a dot, suggests only members of the inferred receiver type (incl. chained calls like `Path(".." ).lines().``Iterator` methods). No global identifiers appear after a dot.
- Inheritance-aware: direct class members first, then inherited (e.g., `List` includes `Collection`/`Iterable` methods).
- Heuristics: handles literals (`"…"``String`, numbers → `Int/Real`, `[...]``List`, `{...}``Dict`) and static `Namespace.` members.
- Performance: capped results, early prefix filtering, per‑document MiniAst cache, cancellation checks.
- Toggle: Settings | Lyng Formatter → "Enable Lyng autocompletion (experimental)" (default ON).
- Stabilization: DEBUG completion/Quick Doc logs are OFF by default; behavior aligned between IDE and isolated engine tests.
- Language: Named arguments and named splats
- New call-site syntax for named arguments using colon: `name: value`.
- Positional arguments must come before named; positionals after a named argument inside parentheses are rejected.
- Trailing-lambda interaction: if the last parameter is already assigned by name (or via a named splat), a trailing `{ ... }` block is illegal.
- Named splats: `...` can now expand a Map into named arguments.
- Only string keys are allowed; non-string keys raise a clear error.
- Duplicate assignment across named args and named splats is an error.
- Ellipsis (variadic) parameters remain positional-only and cannot be named.
- Rationale: `=` is assignment and an expression in Lyng; `:` at call sites avoids ambiguity. Declarations keep `name: Type`; call-site casts continue to use `as` / `as?`.
- Documentation updated: proposals and declaring-arguments sections now cover named args/splats and error cases.
- Tests added covering success cases and errors for named args/splats and trailing-lambda interactions.
- Tooling: Highlighters and TextMate bundle updated for named args
- Website/editor highlighter (lyngweb + site) works with `name: value` and `...Map("k" => v)`; added JS tests covering punctuation/operator spans for `:` and `...`.
- TextMate grammar updated to recognize named call arguments: `name: value` after `(` or `,` with `name` highlighted as `variable.parameter.named.lyng` and `:` as punctuation; excludes `::`.
- TextMate bundle version bumped to 0.0.3; README updated with details and guidance.
- Multiple Inheritance (MI) completed and enabled by default:
- Active C3 Method Resolution Order (MRO) for deterministic, monotonic lookup across complex hierarchies and diamonds.
- Qualified dispatch:
- `this@Type.member(...)` inside class bodies starts lookup at the specified ancestor.
- Cast-based disambiguation: `(expr as Type).member(...)`, `(expr as? Type)?.member(...)` (works with existing safe-call `?.`).
- Field inheritance (`val`/`var`) under MI:
- Instance storage is disambiguated per declaring class; unqualified read/write resolves to the first match in MRO.
- Qualified read/write targets the chosen ancestor’s storage.
- Constructors and initialization:
- Direct bases are initialized left-to-right; each ancestor is initialized at most once (diamond-safe de-duplication).
- 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 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.
- `as`/`as?` cast errors include actual and target type names.
- 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
This file tracks user-visible Lyng language/runtime/tooling changes.
History note:
- The project had periods where changelog maintenance lagged behind commits.
- Entries below are synchronized and curated for `1.5.x`.
- Earlier history may be incomplete and should be cross-checked with git tags/commits when needed.
All notable changes to this project will be documented in this file.
## Unreleased
- No unreleased entries yet.
- CLI: Added `fmt` as a first-class Clikt subcommand.
- Default behavior: formats files to stdout (no in-place edits by default).
- Options:
- `--check`: check only; print files that would change; exit with code 2 if any changes are needed.
- `-i, --in-place`: write formatted result back to files.
- `--spacing`: apply spacing normalization.
- `--wrap`, `--wrapping`: enable line wrapping.
- 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).
## 1.5.5 (2026-04-23)
- CLI: Preserved legacy script invocation fast-paths:
- `lyng script.lyng [args...]` executes the script directly.
- `lyng -- -file.lyng [args...]` executes a script whose name begins with `-`.
### Concurrency and collections
- Added coroutine coordination primitives and helpers for everyday parallel code:
- `Channel` for coroutine-to-coroutine communication
- `LaunchPool` for bounded-concurrency task execution
- `Iterable<Deferred>.joinAll()` to await a whole collection of deferreds in input order
- `CompletableDeferred.completeExceptionally(...)` and `Deferred.cancelAndJoin()`
- Added docs and examples for the new concurrency APIs, including `joinAll()` coverage in iterable and parallelism references.
### Database and time APIs
- Added the portable `lyng.io.db` SQL contract and the first concrete providers:
- `lyng.io.db.sqlite` on JVM and Linux Native
- `lyng.io.db.jdbc` on JVM
- Added SQLite/JDBC release hardening:
- nested transactions via savepoints
- detached materialized rows
- generated-key support through `ExecutionResult.getGeneratedKeys()`
- schema-driven value conversion for `Bool`, `Decimal`, `Date`, `DateTime`, and `Instant`
- portable SQLite linker/deployment fixes and documented runtime options
- Added `Date` to `lyng.time` and the core runtime as a first-class calendar-date type, plus conversions and arithmetic across `Instant`, `DateTime`, and `Date`.
### Language, stdlib, and tooling
- Added extensions on singleton `object` declarations, including object-scoped indexer overrides for bracket syntax.
- Added backtick string literals and formatter support.
- Added `lyng.legacy_digest` for SHA-1 compatibility work, `String.replace`, and `buffer.base64std`.
- Improved CLI/runtime behavior with `atExit` shutdown handlers, native release-binary work, and follow-up CLI packaging/import fixes.
- Expanded docs across the tutorial, stdlib references, database docs, networking docs, and release notes.
### Runtime/compiler stability and performance
- Extended exact-call and higher-order lambda inlining through the bytecode compiler, including compiled fast paths for simple lambdas, wrappers, captures, and common higher-order helpers.
- Fixed import caching and class/object bytecode dispatch on JVM.
- Fixed immutable `val` compound assignments so true mutating `*Assign` operations continue to work while fallback reassignments report the correct read-only error.
- Fixed closure/capture and import regressions across launched loops, singleton/object extensions, aliasing, transitive re-exports, and immutable capture escaping.
- Improved list-fill/list-append fast paths, nullable-let inference, Decimal/Complex interop, and related regression coverage.
### Release notes
- Release metadata, homepage samples, docs, and README now point to `1.5.5`.
## 1.5.4 (2026-04-03)
### Runtime and compiler stability
- Stabilized the recent `piSpigot` benchmark/compiler work for release.
- Fixed numeric-mix regressions introduced by overly broad int-coercion in bytecode compilation.
- Restored correct behavior for decimal arithmetic, mixed real/int flows, list literals, list size checks, and national-character script cases.
- Fixed plain-list index fast paths so they no longer bypass subclass behavior such as `ObservableList` hooks and flow notifications.
- Hardened local numeric compare fast paths to correctly handle primitive-coded frame slots.
### Performance and examples
- Added `piSpigot` benchmark/example coverage:
- `examples/pi-test.lyng`
- `examples/pi-bench.lyng`
- JVM benchmark test for release-baseline verification
- Kept the safe list/index/runtime wins that improve the optimized `piSpigot` path without reintroducing type-unsound coercions.
- Changed the default `RVAL_FASTPATH` setting off on JVM/Android and in the benchmark preset after verification that it no longer helps the stabilized `piSpigot` workload.
### Release notes
- Full JVM and wasm test gates pass on the release tree.
- Benchmark findings and remaining post-release optimization targets are documented in `notes/pi_spigot_benchmark_baseline_2026-04-03.md`.
## 1.5.1 (2026-03-25)
### Language
- Added string interpolation:
- `"$name"` identifier interpolation.
- `"${expr}"` expression interpolation.
- Added literal-dollar forms in strings:
- `"\$"` -> `$`
- `"$$"` -> `$`
- `\\$x` is parsed as backslash + interpolation of `x`.
- Added per-file interpolation opt-out via leading directive comment:
- `// feature: interpolation: off`
### Docs and AI references
- Updated compiler-accurate AI language docs:
- interpolation syntax and escaping
- per-file feature switch behavior
- Refreshed tutorial examples and doctests to reflect new interpolation semantics.
- Added/reworked current proposal/reference materials for Lyng common-platform guidance.
### Compatibility notes
- Interpolation is enabled by default for normal string literals.
- Existing code that intentionally used `$name` as literal text should use `\$name`, `$$name`, or the file directive `// feature: interpolation: off`.
## 1.5.0 (2026-03-22)
### Major runtime/compiler direction
- Completed migration to bytecode-first/bytecode-only execution paths.
- Removed interpreter fallback behavior in core execution hot paths.
- Continued frame-slot-first local/capture model improvements and related diagnostics.
### Language features and semantics
- Added/finished `return` semantics including labeled non-local forms (`return@label`).
- Added abstract classes/members and `interface` support (as abstract-class-style construct).
- Completed and enabled multiple inheritance with C3 MRO by default.
- Added class properties with accessors (`get`/`set`) and restricted setter visibility (`private set`, `protected set`).
- Added late-initialized class `val` support with `Unset` protection rules.
- Added named arguments (`name: value`) and named splats (`...map`) with stricter validation.
- Added assign-if-null operator `?=`.
- Improved nullable/type-checking behavior (including `T is nullable` and related type checks).
- Added variadic function types (`...` in function type declarations) and tighter lambda type checks.
### Type system and collections
- Added immutable collections hierarchy (`ImmutableList`, `ImmutableSet`, `ImmutableMap`).
- Improved generic runtime binding/checking for explicit type arguments and bounds.
- Added smarter type-aware collection ops (`+=`, `-=`) and stronger declared-member type checks.
### Extern/Kotlin bridge
- Tightened extern declaration rules:
- explicit extern members are required for extern class/object declarations.
- Improved extern generic class behavior and diagnostics.
- Extended bridge APIs for binding global functions/variables and object/member interop scenarios.
### Standard library and modules
- Added `lyng.observable` improvements (`ObservableList` hooks/events).
- Added `Random` stdlib API used by updated samples.
- Added/extended `lyngio.console` support and CLI integration for console interaction.
- Migrated time APIs to `kotlin.time` (`Instant` migration and related docs/tests).
### CLI, IDE, and docs/tooling
- CLI:
- added first-class `fmt` command
- preserved direct script fast-path invocation
- improved command help/dispatch behavior
- IntelliJ plugin:
- improved lightweight completion and documentation/inspection behavior
- continued highlighter and Grazie/spellchecking integration work
- Docs:
- substantial updates across tutorial/OOP/type/runtime references
- expanded bytecode and advanced topics coverage
## Migration checklist for 1.5.x
- If you rely on literal `$...` strings:
- replace with `\$...` or `$$...`, or
- add `// feature: interpolation: off` at file top.
- Review any code relying on interpreter-era fallback behavior; 1.5.x assumes bytecode-first execution.
- For extern declarations, ensure members are explicitly declared where required.
- For named arguments/splats, verify call sites follow stricter ordering/duplication rules.
- CLI: Fixed a regression where the root help banner could print before subcommands.
- Root command no longer prints help when a subcommand (e.g., `fmt`) is invoked.

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.

149
README.md
View File

@ -1,14 +1,6 @@
# 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:
@ -25,48 +17,45 @@ Point(x:, y:).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
- [Latest release notes (1.5.5)](docs/whats_new.md)
- [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)
- [SQL Databases (lyng.io.db)](docs/lyng.io.db.md)
- [Time and Calendar Types](docs/time.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
### Add dependency to your project
```kotlin
val lyngVersion = "1.5.5"
// update to current please:
val lyngVersion = "0.6.1-SNAPSHOT"
repositories {
// ...
@ -90,54 +79,49 @@ 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
runBlocking {
val session = EvalSession()
assert(5 == session.eval(""" 3*3 - 4 """).toInt())
session.eval(""" println("Hello, Lyng!") """)
assert(5 == eval(""" 3*3 - 4 """).toInt())
eval(""" println("Hello, Lyng!") """)
}
```
### Exchanging information
The preferred host runtime is `EvalSession`. It owns the script scope and any coroutines
started with `launch { ... }`. Create a session, grab its scope when you need low-level
binding APIs, then execute scripts through the session:
Script is executed over some `Scope`. Create instance,
add your specific vars and functions to it, and call:
```kotlin
import net.sergeych.lyng.*
runBlocking {
val session = EvalSession()
val scope = session.getScope().apply {
// simple function
addFn("sumOf") {
var sum = 0.0
for (a in args) sum += a.toDouble()
ObjReal(sum)
}
addConst("LIGHT_SPEED", ObjReal(299_792_458.0))
import com.sun.source.tree.Scope
import new.sergeych.lyng.*
// callback back to kotlin to some suspend fn, for example::
// suspend fun doSomeWork(text: String): Int
addFn("doSomeWork") {
// this _is_ a suspend lambda, we can call suspend function,
// and it won't consume the thread.
// note that in kotlin handler, `args` is a list of `Obj` arguments
// and return value from this lambda should be Obj too:
doSomeWork(args[0]).toObj()
}
// simple function
val scope = Script.newScope().apply {
addFn("sumOf") {
var sum = 0.0
for (a in args) sum += a.toDouble()
ObjReal(sum)
}
addConst("LIGHT_SPEED", ObjReal(299_792_458.0))
// execute through the session:
session.eval("sumOf(1,2,3)") // <- 6
// callback back to kotlin to some suspend fn, for example::
// suspend fun doSomeWork(text: String): Int
addFn("doSomeWork") {
// this _is_ a suspend lambda, we can call suspend function,
// and it won't consume the thread.
// note that in kotlin handler, `args` is a list of `Obj` arguments
// and return value from this lambda should be Obj too:
doSomeWork(args[0]).toObj()
}
}
// adding constant:
scope.eval("sumOf(1,2,3)") // <- 6
```
Note that the session reuses one scope, so state persists across `session.eval(...)` calls.
Use raw `Scope.eval(...)` only when you intentionally want low-level control without session-owned coroutine lifecycle.
Note that the scope stores all changes in it so you can make calls on a single scope to preserve state between calls.
## IntelliJ IDEA plugin: Lightweight autocompletion (experimental)
@ -161,12 +145,6 @@ 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/++.
@ -186,7 +164,8 @@ Designed to add scripting to kotlin multiplatform application in easy and effici
# Language Roadmap
The current stable release is **v1.5.5**: the 1.5 cycle now includes the database/date/concurrency additions as well as the latest compiler/runtime stabilization work, and the language, tooling, and site are aligned around this release.
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:
@ -197,7 +176,6 @@ 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?)
@ -213,31 +191,22 @@ Ready features:
- [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 on the [language site](https://lynglang.com) and locally in [docs/tutorial.md](docs/tutorial.md). The site reflects the current release, while development snapshots continue 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
- [x] type specifications
- [ ] type specifications
- [x] Textmate Bundle
- [x] IDEA plugin
- [x] source docs and maybe lyng.md to a standard
- [ ] source docs and maybe lyng.md to a standard
- [ ] metadata first class access from lyng
- [x] aggressive optimizations
- [ ] compile to JVM bytecode optimization
## After 1.5 "Ideal scripting"
* __we are here now ;)__
Estimated summer 2026
- propose your feature!
@ -245,12 +214,8 @@ All of this is documented on the [language site](https://lynglang.com) and local
@-links are for contacting authors on [project home](https://gitea.sergeych.net/SergeychWorks/lyng): this simplest s to open issue for the person you need to convey any information about this project.
<img src="https://www.gravatar.com/avatar/7e3a56ff8a090fc9ffbd1909dea94904?s=32&d=identicon" alt="Sergey Chernov" width="32" height="32" style="vertical-align: middle; margin-right: 0.5em;" /> <b>Sergey Chernov</b> @sergeych, real.sergeych@gmail.com: Initial idea and architecture, language concept, design, implementation.
__Sergey Chernov__ @sergeych: Initial idea and architecture, language concept, design, implementation.
<br/>
__Yulia Nezhinskaya__ @AlterEgoJuliaN: System analysis, math and features design.
<img src="https://www.gravatar.com/avatar/53a90bca30c85a81db8f0c0d8dea43a1?s=32&d=identicon" alt="Yulia Nezhinskaya" width="32" height="32" style="vertical-align: middle; margin-right: 0.5em;" /> <b>Yulia Nezhinskaya</b> @AlterEgoJuliaN, neleka88@gmail.com: System analysis, math and feature 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

@ -1,39 +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 site:jsBrowserDistribution publish buildInstallablePlugin :lyng:linkReleaseExecutableLinuxX64 :lyng:installJvmDist --parallel --no-configuration-cache
#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.
@ -17,16 +17,6 @@
#
#
upload_only=false
target=vps # default: new server; use --old for d.lynglang.com
for arg in "$@"; do
if [[ "$arg" == "-u" || "$arg" == "--upload-only" ]]; then
upload_only=true
elif [[ "$arg" == "--old" ]]; then
target=com
fi
done
function checkState() {
if [[ $? != 0 ]]; then
echo
@ -34,10 +24,9 @@ function checkState() {
echo
exit 100
fi
}
# Update docs/idea_plugin.md to point to the latest built IDEA plugin zip
# from ./distributables before building the site. The change is temporary and
# the original file is restored right after the build.
@ -90,20 +79,19 @@ function updateIdeaPluginDownloadLink() {
fi
}
# target settings (-t com | -t vps)
case "$target" in
# default target settings
case "com" in
com)
SSH_HOST=sergeych@d.lynglang.com
SSH_PORT=22
ROOT=/bigstore/sergeych_pub/lyng
;;
vps)
SSH_HOST=sergeych@94.130.36.94
SSH_PORT=22
ROOT=/var/www/lynglang
SSH_HOST=sergeych@d.lynglang.com # host to deploy to
SSH_PORT=22 # ssh port on it
ROOT=/bigstore/sergeych_pub/lyng # directory to rsync to
;;
# com)
# SSH_HOST=vvk@front-01.neurodatalab.com
# ROOT=/home/vvk
# ;;
*)
echo "*** ERROR: unknown target '$target' (use -t com | -t vps)"
echo "*** ERROR: target not specified (use deploy com | dev)"
echo "*** stop"
exit 101
esac
@ -119,28 +107,28 @@ function refreshTextmateZip() {
(cd editors && zip -rq ../distributables/lyng-textmate.zip .)
}
# Update the IDEA plugin download link in docs (temporarily), then build, then restore the doc
refreshTextmateZip
updateIdeaPluginDownloadLink || echo "WARN: proceeding without updating IDEA plugin download link"
if [[ "$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=$?
./gradlew site:jsBrowserDistribution
BUILD_RC=$?
# 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}
@ -155,15 +143,12 @@ 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

@ -1,7 +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.
@ -19,15 +19,13 @@
set -e
archive=./lyng/build/distributions/lyng-jvm.zip
install_root="$HOME/bin/jlyng-jvm"
launcher="$install_root/lyng-jvm/bin/lyng"
root=./lyng/build/install/lyng-jvm/
./gradlew :lyng:jvmDistZip
mkdir -p ./distributables
cp "$archive" ./distributables/lyng-jvm.zip
rm -rf "$install_root" || true
rm "$HOME/bin/jlyng" 2>/dev/null || true
mkdir -p "$install_root"
unzip -q ./distributables/lyng-jvm.zip -d "$install_root"
ln -s "$launcher" "$HOME/bin/jlyng"
./gradlew :lyng:installJvmDist
#strip $file
#upx $file
rm -rf ~/bin/jlyng-jvm || true
rm ~/bin/jlyng 2>/dev/null || true
mkdir -p ~/bin/jlyng-jvm
cp -R $root ~/bin/jlyng-jvm
ln -s ~/bin/jlyng-jvm/lyng-jvm/bin/lyng ~/bin/jlyng

View File

@ -35,27 +35,3 @@ tasks.register<Exec>("generateDocs") {
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

@ -1,15 +1,8 @@
# Array
It's an interface if the [Collection] that provides indexing access, like `array[3] = 0`.
Array therefore implements [Iterable] too. Well known implementations of `Array` are
[List] and [ImmutableList].
The language-level bracket syntax supports one or more selectors:
- `value[i]`
- `value[i, j]`
Concrete array-like types decide what selectors they accept. Built-in list-like arrays use one selector at a time; custom types such as matrices may interpret multiple selectors.
Array therefore implements [Iterable] too. The well known implementatino of the `Array` is
[List].
Array adds the following methods:
@ -42,4 +35,3 @@ To pre-sort and array use `Iterable.sorted*` or in-place `List.sort*` families,
[Collection]: Collection.md
[Iterable]: Iterable.md
[List]: List.md
[ImmutableList]: ImmutableList.md

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)
@ -120,18 +120,17 @@ which is used in `toString`) and hex encoding:
## Members
| name | meaning | type |
|----------------------------|------------------------------------------------|---------------|
| `size` | size | Int |
| `decodeUtf8` | decode to String using UTF8 rules | Any |
| `+` | buffer concatenation | Any |
| `toMutable()` | create a mutable copy | MutableBuffer |
| `hex` | encode to hex strign | String |
| `Buffer.decodeHex(hexStr) | decode hex string | Buffer |
| `base64` | encode to base64 (url flavor) (2) | String |
| `base64std` | encode to base64 (default vocabulary, filling) | String |
| `Buffer.decodeBase64(str)` | decode base64 to new Buffer (2) | Buffer |
| `toBitInput()` | create bit input from a byte buffer (3) | |
| name | meaning | type |
|----------------------------|-----------------------------------------|---------------|
| `size` | size | Int |
| `decodeUtf8` | decode to String using UTF8 rules | Any |
| `+` | buffer concatenation | Any |
| `toMutable()` | create a mutable copy | MutableBuffer |
| `hex` | encode to hex strign | String |
| `Buffer.decodeHex(hexStr) | decode hex string | Buffer |
| `base64` | encode to base64 (url flavor) (2) | String |
| `Buffer.decodeBase64(str)` | decode base64 to new Buffer (2) | Buffer |
| `toBitInput()` | create bit input from a byte buffer (3) | |
(1)
: optimized implementation that override `Iterable` one

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

@ -1,211 +0,0 @@
# Channel
A `Channel` is a **hot, bidirectional pipe** for passing values between concurrently running coroutines.
Unlike a [Flow], which is cold and replayed on every collection, a `Channel` is stateful: each value
sent is consumed by exactly one receiver.
Channels model the classic _producer / consumer_ pattern and are the right tool when:
- two or more coroutines need to exchange individual values at their own pace;
- you want back-pressure (rendezvous) or explicit buffering control;
- you need a push-based, hot data source (opposite of the pull-based, cold [Flow]).
## Constructors
```
Channel() // rendezvous — sender and receiver must meet
Channel(n: Int) // buffered — sender may run n items ahead of the receiver
Channel(Channel.UNLIMITED) // no limit on buffered items
```
**Rendezvous** (`Channel()`, capacity 0): `send` suspends until a matching `receive` is ready,
and vice-versa. This gives the tightest synchronisation and the smallest memory footprint.
**Buffered** (`Channel(n)`): `send` only suspends when the internal buffer is full. Allows the
producer to get up to _n_ items ahead of the consumer.
**Unlimited** (`Channel(Channel.UNLIMITED)`): `send` never suspends. Useful when the producer is
bursty and you do not want it blocked, but be careful not to grow the buffer without bound.
## Sending and receiving
```lyng
val ch = Channel() // rendezvous channel
val producer = launch {
ch.send("hello") // suspends until the receiver is ready
ch.send("world")
ch.close() // signal: no more values
}
val a = ch.receive() // suspends until "hello" arrives
val b = ch.receive() // suspends until "world" arrives
val c = ch.receive() // channel is closed and drained → null
assertEquals("hello", a)
assertEquals("world", b)
assertEquals(null, c)
```
`receive()` returns `null` when the channel is both **closed** _and_ **fully drained** — that is
the idiomatic loop termination condition:
```lyng
val ch = Channel(4)
launch {
for (i in 1..5) ch.send(i)
ch.close()
}
var item = ch.receive()
while (item != null) {
println(item)
item = ch.receive()
}
```
## Non-suspending poll
`tryReceive()` never suspends. It returns the next buffered value, or `null` if the buffer is
empty or the channel is closed.
```lyng
val ch = Channel(8)
ch.send(42)
println(ch.tryReceive()) // 42
println(ch.tryReceive()) // null — nothing buffered right now
```
Use `tryReceive` for _polling_ patterns where blocking would be unacceptable, for example when
combining channel checks with other work inside a coroutine loop.
## Closing a channel
`close()` marks the channel so that no further `send` calls are accepted. Any items already in the
buffer can still be received. Once the buffer is drained, `receive()` returns `null` and
`isClosedForReceive` becomes `true`.
```lyng
val ch = Channel(2)
ch.send(1)
ch.send(2)
ch.close()
assert(ch.isClosedForSend)
assert(!ch.isClosedForReceive) // still has 2 buffered items
ch.receive() // 1
ch.receive() // 2
assert(ch.isClosedForReceive) // drained
```
Calling `send` after `close()` throws `IllegalStateException`.
## Properties
| property | type | description |
|---------------------|--------|----------------------------------------------------------|
| `isClosedForSend` | `Bool` | `true` after `close()` is called |
| `isClosedForReceive`| `Bool` | `true` when closed _and_ every buffered item is consumed |
## Methods
| method | suspends | description |
|-----------------|----------|----------------------------------------------------------------------------------|
| `send(value)` | yes | send a value; suspends when buffer full (rendezvous: always until partner ready) |
| `receive()` | yes | receive next value; suspends when empty; returns `null` when closed + drained |
| `tryReceive()` | no | return next buffered value or `null`; never suspends |
| `close()` | no | signal end of production; existing buffer items are still receivable |
## Static constants
| constant | value | description |
|---------------------|------------------|-------------------------------------|
| `Channel.UNLIMITED` | `Int.MAX_VALUE` | capacity for an unlimited-buffer channel |
## Common patterns
### Producer / consumer
```lyng
val ch = Channel()
val results = []
val mu = Mutex()
val consumer = launch {
var item = ch.receive()
while (item != null) {
mu.withLock { results += item }
item = ch.receive()
}
}
launch {
for (i in 1..5) ch.send("msg:$i")
ch.close()
}.await()
consumer.await()
println(results)
```
### Fan-out: one channel, many consumers
```lyng
val ch = Channel(16)
// multiple consumers
val workers = (1..4).map { id ->
launch {
var task = ch.receive()
while (task != null) {
println("worker $id handles $task")
task = ch.receive()
}
}
}
// single producer
for (i in 1..20) ch.send(i)
ch.close()
workers.forEach { it.await() }
```
### Ping-pong between two coroutines
```lyng
val ping = Channel()
val pong = Channel()
launch {
repeat(3) {
val msg = ping.receive()
println("got: $msg → sending pong")
pong.send("pong")
}
}
repeat(3) {
ping.send("ping")
println(pong.receive())
}
```
## Channel vs Flow
| | [Flow] | Channel |
|---|---|---|
| **temperature** | cold (lazy) | hot (eager) |
| **replay** | every collector gets a fresh run | each item is consumed once |
| **consumers** | any number; each gets all items | one receiver per item |
| **back-pressure** | built-in via rendezvous | configurable (rendezvous / buffered / unlimited) |
| **typical use** | transform pipelines, sequences | producer–consumer, fan-out |
## See also
- [parallelism] — `launch`, `Deferred`, `Mutex`, `Flow`, and the full concurrency picture
- [Flow] — cold async sequences
[Flow]: parallelism.md#flow
[parallelism]: parallelism.md

View File

@ -6,13 +6,6 @@ Is a [Iterable] with known `size`, a finite [Iterable]:
val size
}
`Collection` is a read/traversal contract shared by mutable and immutable collections.
Concrete collection classes:
- Mutable: [List], [Set], [Map]
- Immutable: [ImmutableList], [ImmutableSet], [ImmutableMap]
- Observable mutable lists (opt-in module): [ObservableList]
| name | description |
|------------------------|------------------------------------------------------|
@ -23,9 +16,4 @@ See [List], [Set], [Iterable] and [Efficient Iterables in Kotlin Interop](Effici
[Iterable]: Iterable.md
[List]: List.md
[Set]: Set.md
[Map]: Map.md
[ImmutableList]: ImmutableList.md
[ImmutableSet]: ImmutableSet.md
[ImmutableMap]: ImmutableMap.md
[ObservableList]: ObservableList.md
[Set]: Set.md

View File

@ -1,82 +0,0 @@
# Complex Numbers (`lyng.complex`)
`lyng.complex` adds a pure-Lyng `Complex` type backed by `Real` components.
Import it when you want ordinary complex arithmetic:
```lyng
import lyng.complex
```
## Construction
Use any of these:
```lyng
import lyng.complex
val a = Complex(1.0, 2.0)
val b = complex(1.0, 2.0)
val c = 2.i
val d = 3.re
assertEquals(Complex(1.0, 2.0), 1 + 2.i)
```
Convenience extensions:
- `Int.re`, `Real.re`: embed a real value into the complex plane
- `Int.i`, `Real.i`: create a pure imaginary value
- `cis(angle)`: shorthand for `cos(angle) + i sin(angle)`
## Core Operations
`Complex` supports:
- `+`
- `-`
- `*`
- `/`
- unary `-`
- `conjugate`
- `magnitude`
- `phase`
Mixed arithmetic with `Int` and `Real` is enabled through `lyng.operators`, so both sides work naturally:
```lyng
import lyng.complex
assertEquals(Complex(1.0, 2.0), 1 + 2.i)
assertEquals(Complex(1.5, 2.0), 1.5 + 2.i)
assertEquals(Complex(2.0, 2.0), 2.i + 2)
```
Mixed equality with built-in numeric types is intentionally not promised yet. Keep equality checks in the `Complex` domain for now.
## Transcendental Functions
For now, use member-style calls:
```lyng
import lyng.complex
val z = 1 + π.i
val w = z.exp()
val s = z.sin()
val r = z.sqrt()
```
This is deliberate. Lyng already has built-in top-level real-valued functions such as `exp(x)` and `sin(x)`, and imported modules do not currently replace those root bindings. So plain `exp(z)` is not yet the right extension mechanism for complex math.
## Design Scope
This module intentionally uses `Complex` with `Real` parts, not `Complex<T>`.
Reasons:
- the existing math runtime is `Real`-centric
- the operator interop registry works with concrete runtime classes
- transcendental functions (`exp`, `sin`, `ln`, `sqrt`) are defined over the `Real` math backend here
If Lyng later gets a more general numeric-trait or callable-overload registry, a generic algebraic `Complex<T>` can be revisited on firmer ground.

View File

@ -1,325 +0,0 @@
# Decimal (`lyng.decimal`)
`lyng.decimal` adds an arbitrary-precision decimal type to Lyng as a normal library module.
Import it when you need decimal arithmetic that should not inherit `Real`'s binary floating-point behavior:
```lyng
import lyng.decimal
```
## What `Decimal` Is For
Use `Decimal` when values are fundamentally decimal:
- money
- human-entered quantities
- exact decimal text
- predictable decimal rounding
- user-facing formatting and tests
Do not use it just because a number has a fractional part. `Real` is still the right type for ordinary double-precision numeric work.
## Creating Decimal Values
There are three supported conversions:
```lyng
import lyng.decimal
val a = 1.d
val b = 2.2.d
val c = "2.2".d
assertEquals("1", a.toStringExpanded())
assertEquals("2.2", b.toStringExpanded())
assertEquals("2.2", c.toStringExpanded())
```
The three forms mean different things:
- `1.d`: convert `Int -> Decimal`
- `2.2.d`: convert `Real -> Decimal`
- `"2.2".d`: parse exact decimal text
That distinction is intentional.
### `Real.d` vs `"..." .d`
`Real.d` preserves the current `Real` value. It does not pretend the source code was exact decimal text.
```lyng
import lyng.decimal
assertEquals("0.30000000000000004", (0.1 + 0.2).d.toStringExpanded())
assertEquals("0.3", "0.3".d.toStringExpanded())
```
This follows the "minimal confusion" rule:
- if you start from a `Real`, you get a decimal representation of that `Real`
- if you want exact decimal source text, use a `String`
## Factory Functions
The explicit factory methods are:
```lyng
import lyng.decimal
Decimal.fromInt(10)
Decimal.fromReal(2.5)
Decimal.fromString("12.34")
```
These are equivalent to the conversion-property forms, but sometimes clearer in APIs or generated code.
## From Kotlin
If you already have an ionspin `BigDecimal` on the host side, the simplest supported way to create a Lyng `Decimal` is:
```kotlin
import com.ionspin.kotlin.bignum.decimal.BigDecimal
import net.sergeych.lyng.EvalSession
import net.sergeych.lyng.asFacade
import net.sergeych.lyng.newDecimal
val scope = EvalSession().getScope()
val decimal = scope.asFacade().newDecimal(BigDecimal.parseStringWithMode("12.34"))
```
Notes:
- `newDecimal(...)` loads `lyng.decimal` if needed
- it returns a real Lyng `Decimal` object instance
- this is the preferred Kotlin-side construction path when you already hold a host `BigDecimal`
## Core Operations
`Decimal` supports:
- `+`
- `-`
- `*`
- `/`
- `%`
- unary `-`
- comparison operators
- equality operators
Examples:
```lyng
import lyng.decimal
assertEquals("3.75", ("1.5".d + "2.25".d).toStringExpanded())
assertEquals("1.25", ("2.5".d - "1.25".d).toStringExpanded())
assertEquals("3.0", ("1.5".d * 2.d).toStringExpanded())
assertEquals("0.5", (1.d / 2.d).toStringExpanded())
assert("2.0".d > "1.5".d)
assert("2.0".d == 2.d)
```
## Interoperability With `Int` and `Real`
The decimal module registers mixed-operand operator bridges so both sides read naturally:
```lyng
import lyng.decimal
assertEquals(3.d, 1 + 2.d)
assertEquals(3.d, 2.d + 1)
assertEquals(1.5.d, 1.d + 0.5)
assertEquals(1.5.d, 0.5 + 1.d)
assert(2 == 2.d)
assert(3 > 2.d)
```
Without this registration mechanism, only the cases directly implemented on the left-hand class would work. The bridge fills the gap for expressions such as `Int + Decimal` and `Real + Decimal`.
See [OperatorInterop.md](OperatorInterop.md) for the generic mechanism behind that.
## String Representation
Use `toStringExpanded()` when you want plain decimal output without scientific notation:
```lyng
import lyng.decimal
assertEquals("12.34", "12.34".d.toStringExpanded())
```
This is the recommended representation for:
- tests
- user-visible diagnostics
- decimal formatting checks
## Conversions Back To Built-ins
```lyng
import lyng.decimal
assertEquals(2, "2.9".d.toInt())
assertEquals(2.9, "2.9".d.toReal())
```
Use `toReal()` only when you are willing to return to binary floating-point semantics.
## Non-Finite Checks
`Decimal` values are always finite, so these helpers exist for API symmetry with `Real` and always return `false`:
```lyng
import lyng.decimal
assertEquals(false, "2.9".d.isInfinite())
assertEquals(false, "2.9".d.isNaN())
```
## Division Context
Division is the operation where precision and rounding matter most.
By default, decimal division uses:
- precision: `34` significant digits
- rounding: `HalfEven`
Example:
```lyng
import lyng.decimal
assertEquals("0.3333333333333333333333333333333333", (1.d / 3.d).toStringExpanded())
assertEquals("0.6666666666666666666666666666666667", ("2".d / 3.d).toStringExpanded())
```
## `withDecimalContext(...)`
Use `withDecimalContext(...)` to override decimal division rules inside a block:
```lyng
import lyng.decimal
assertEquals(
"0.3333333333",
withDecimalContext(10) { (1.d / 3.d).toStringExpanded() }
)
```
You can also pass an explicit context object:
```lyng
import lyng.decimal
val ctx = DecimalContext(6, DecimalRounding.HalfAwayFromZero)
assertEquals("0.666667", withDecimalContext(ctx) { ("2".d / 3.d).toStringExpanded() })
```
The context is dynamic and local to the block. After the block exits, the previous context is restored.
## Rounding Modes
Available rounding modes:
- `HalfEven`
- `HalfAwayFromZero`
- `HalfTowardsZero`
- `Ceiling`
- `Floor`
- `AwayFromZero`
- `TowardsZero`
Tie example at precision `2`:
```lyng
import lyng.decimal
assertEquals("0.12", withDecimalContext(2, DecimalRounding.HalfEven) { (1.d / 8.d).toStringExpanded() })
assertEquals("0.13", withDecimalContext(2, DecimalRounding.HalfAwayFromZero) { (1.d / 8.d).toStringExpanded() })
assertEquals("0.12", withDecimalContext(2, DecimalRounding.HalfTowardsZero) { (1.d / 8.d).toStringExpanded() })
assertEquals("0.13", withDecimalContext(2, DecimalRounding.Ceiling) { (1.d / 8.d).toStringExpanded() })
assertEquals("0.12", withDecimalContext(2, DecimalRounding.Floor) { (1.d / 8.d).toStringExpanded() })
```
Negative values follow the same named policy in the obvious direction:
```lyng
import lyng.decimal
assertEquals("-0.13", withDecimalContext(2, DecimalRounding.HalfAwayFromZero) { (-1.d / 8.d).toStringExpanded() })
assertEquals("-0.12", withDecimalContext(2, DecimalRounding.HalfTowardsZero) { (-1.d / 8.d).toStringExpanded() })
```
## Recommended Usage Rules
## Decimal With Stdlib Math Functions
Core math helpers such as `abs`, `floor`, `ceil`, `round`, `sin`, `exp`, `ln`, `sqrt`, `log10`, `log2`, and `pow`
now also accept `Decimal`.
Current behavior is intentionally split:
- exact decimal implementation:
- `abs(x)`
- `floor(x)`
- `ceil(x)`
- `round(x)`
- `pow(x, y)` when `x` is `Decimal` and `y` is an integral exponent
- temporary bridge through `Real`:
- `sin`, `cos`, `tan`
- `asin`, `acos`, `atan`
- `sinh`, `cosh`, `tanh`
- `asinh`, `acosh`, `atanh`
- `exp`, `ln`, `log10`, `log2`
- `sqrt`
- `pow` for the remaining non-integral decimal exponent cases
The temporary bridge is:
```lyng
Decimal -> Real -> host math -> Decimal
```
This is a compatibility step, not the long-term design. Native decimal implementations will replace these bridge-based
paths over time.
Examples:
```lyng
import lyng.decimal
assertEquals("2.5", (abs("-2.5".d) as Decimal).toStringExpanded())
assertEquals("2", (floor("2.9".d) as Decimal).toStringExpanded())
// Temporary Real bridge:
assertEquals((exp(1.25) as Real).d.toStringExpanded(), (exp("1.25".d) as Decimal).toStringExpanded())
assertEquals((sqrt(2.0) as Real).d.toStringExpanded(), (sqrt("2".d) as Decimal).toStringExpanded())
```
If you care about exact decimal source text:
```lyng
"12.34".d
```
If you intentionally convert an existing binary floating-point value:
```lyng
someReal.d
```
If you want local control over division:
```lyng
withDecimalContext(precision, rounding) { ... }
```
If you want custom mixed operators for your own type, follow the same pattern as Decimal and use the operator registry:
- define the operators on your own class
- choose a common class
- register mixed operand bridges
See [OperatorInterop.md](OperatorInterop.md).

View File

@ -1,37 +0,0 @@
# ImmutableList built-in class
`ImmutableList` is an immutable, indexable list value.
It implements [Array], therefore [Collection] and [Iterable].
Use it when API contracts require a list that cannot be mutated through aliases.
## Creating
val a = ImmutableList(1,2,3)
val b = [1,2,3].toImmutable()
val c = (1..3).toImmutableList()
>>> void
## Converting
val i = ImmutableList(1,2,3)
val m = i.toMutable()
m += 4
assertEquals( ImmutableList(1,2,3), i )
assertEquals( [1,2,3,4], m )
>>> void
## Members
| name | meaning |
|---------------|-----------------------------------------|
| `size` | number of elements |
| `[index]` | element access by index |
| `[Range]` | immutable slice |
| `+` | append element(s), returns new immutable list |
| `-` | remove element(s), returns new immutable list |
| `toMutable()` | create mutable copy |
[Array]: Array.md
[Collection]: Collection.md
[Iterable]: Iterable.md

View File

@ -1,36 +0,0 @@
# ImmutableMap built-in class
`ImmutableMap` is an immutable map of key-value pairs.
It implements [Collection] and [Iterable] of [MapEntry].
## Creating
val a = ImmutableMap("a" => 1, "b" => 2)
val b = Map("a" => 1, "b" => 2).toImmutable()
val c = ["a" => 1, "b" => 2].toImmutableMap
>>> void
## Converting
val i = ImmutableMap("a" => 1)
val m = i.toMutable()
m["a"] = 2
assertEquals( 1, i["a"] )
assertEquals( 2, m["a"] )
>>> void
## Members
| name | meaning |
|-----------------|------------------------------------------|
| `size` | number of entries |
| `[key]` | get value by key, or `null` if absent |
| `getOrNull(key)`| same as `[key]` |
| `keys` | list of keys |
| `values` | list of values |
| `+` | merge (rightmost wins), returns new immutable map |
| `toMutable()` | create mutable copy |
[Collection]: Collection.md
[Iterable]: Iterable.md
[MapEntry]: Map.md

View File

@ -1,34 +0,0 @@
# ImmutableSet built-in class
`ImmutableSet` is an immutable set of unique elements.
It implements [Collection] and [Iterable].
## Creating
val a = ImmutableSet(1,2,3)
val b = Set(1,2,3).toImmutable()
val c = [1,2,3].toImmutableSet
>>> void
## Converting
val i = ImmutableSet(1,2,3)
val m = i.toMutable()
m += 4
assertEquals( ImmutableSet(1,2,3), i )
assertEquals( Set(1,2,3,4), m )
>>> void
## Members
| name | meaning |
|---------------|-----------------------------------------------------|
| `size` | number of elements |
| `contains(x)` | membership test |
| `+`, `union` | union, returns new immutable set |
| `-`, `subtract` | subtraction, returns new immutable set |
| `*`, `intersect` | intersection, returns new immutable set |
| `toMutable()` | create mutable copy |
[Collection]: Collection.md
[Iterable]: Iterable.md

View File

@ -55,23 +55,6 @@ Here is the sample:
assertEquals( (1..3).joinToString { it * 10 }, "10 20 30")
>>> void
## joinAll
`joinAll()` is an `Iterable<Deferred>` helper that awaits every deferred in iteration order and returns a `List`
with the collected results.
val jobs = (1..4).map { n ->
launch { n * n }
}
assertEquals([1, 4, 9, 16], jobs.joinAll())
>>> void
Notes:
- it does not start any task by itself; it only awaits the deferreds already present in the iterable.
- awaiting happens in iteration order, so the result list keeps the same order as the input iterable.
- if any deferred fails or was cancelled, that `await()` error is propagated from `joinAll()`.
## `sum` and `sumOf`
These, again, does the thing:
@ -125,8 +108,8 @@ You can also use flow variations that return a cold `Flow` instead of a `List`,
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 })
assertEquals(2, source.minOf { it.length })
assertEquals(4, source.maxOf { it.length })
>>> void
## flatten and flatMap
@ -164,15 +147,12 @@ Search for the first element that satisfies the given predicate:
| fun/method | description |
|------------------------|---------------------------------------------------------------------------------|
| toList() | create a list from iterable |
| toImmutableList() | create an immutable list from iterable |
| toSet() | create a set from iterable |
| toImmutableSet | create an immutable set from iterable |
| contains(i) | check that iterable contains `i` |
| `i in iterable` | 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) |
| toImmutableMap | create an immutable map from list of key-value pairs |
| 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 |
@ -201,7 +181,6 @@ Search for the first element that satisfies the given predicate:
| 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) |
| joinAll() | for `Iterable<Deferred>`, await all items in order and collect results to [List] |
| reversed() | create a list containing items from this in reverse order |
| shuffled() | create a list of shuffled elements |
@ -227,20 +206,16 @@ For high-performance Kotlin-side interop and custom iterable implementation deta
## Implemented in classes:
- [List], [ImmutableList], [Range], [Buffer](Buffer.md), [BitBuffer], [Buffer], [Set], [ImmutableSet], [Map], [ImmutableMap], [RingBuffer]
- [List], [Range], [Buffer](Buffer.md), [BitBuffer], [Buffer], [Set], [RingBuffer]
[Collection]: Collection.md
[List]: List.md
[ImmutableList]: ImmutableList.md
[Flow]: parallelism.md#flow
[Range]: Range.md
[Set]: Set.md
[ImmutableSet]: ImmutableSet.md
[Map]: Map.md
[ImmutableMap]: ImmutableMap.md
[RingBuffer]: RingBuffer.md
[RingBuffer]: RingBuffer.md

View File

@ -1,110 +0,0 @@
# LaunchPool
`LaunchPool` is a bounded-concurrency task pool: you submit lambdas with `launch`, and the pool runs them using a fixed number of worker coroutines.
## Constructor
```
LaunchPool(maxWorkers, maxQueueSize = Channel.UNLIMITED)
```
| Parameter | Description |
|-----------|-------------|
| `maxWorkers` | Maximum number of tasks that run in parallel. |
| `maxQueueSize` | Maximum number of tasks that may wait in the queue. When the queue is full, `launch` suspends the caller until space becomes available. Defaults to `Channel.UNLIMITED` (no bound). |
## Methods
### `launch(lambda): Deferred`
Schedules `lambda` for execution and returns a `Deferred` for its result.
- Suspends if the queue is full (`maxQueueSize` reached).
- Throws `IllegalStateException` if the pool is already closed or cancelled.
- Any exception thrown by `lambda` is captured in the returned `Deferred` and **does not escape the pool**.
```lyng
val pool = LaunchPool(4)
val d1 = pool.launch { computeSomething() }
val d2 = pool.launch { computeOther() }
pool.closeAndJoin()
println(d1.await())
println(d2.await())
```
### `closeAndJoin()`
Stops accepting new tasks and suspends until all queued and running tasks complete normally. After this call, any further `launch` throws `IllegalStateException`. Idempotent — safe to call multiple times.
### `cancel()`
Immediately closes the queue and cancels all worker coroutines. Queued but unstarted tasks are discarded. After this call, `launch` throws `IllegalStateException`. Idempotent.
### `cancelAndJoin()`
Like `cancel()`, but also suspends until all worker coroutines have stopped. Useful when you need to be sure no worker code is still running before proceeding. Idempotent.
## Exception handling
Exceptions from submitted lambdas are captured per-task in the returned `Deferred`. The pool itself continues running after a task failure:
```lyng
val pool = LaunchPool(2)
val good = pool.launch { 42 }
val bad = pool.launch { throw IllegalArgumentException("boom") }
pool.closeAndJoin()
assertEquals(42, good.await())
assertThrows(IllegalArgumentException) { bad.await() }
```
## Bounded queue / back-pressure
When `maxQueueSize` is set, the producer suspends if the queue fills up, providing automatic back-pressure:
```lyng
// 1 worker, queue of 2 — producer can be at most 2 tasks ahead of what's running
val pool = LaunchPool(1, 2)
val d1 = pool.launch { delay(10); "a" }
val d2 = pool.launch { delay(10); "b" }
val d3 = pool.launch { delay(10); "c" } // suspends until d1 is picked up by the worker
pool.closeAndJoin()
```
## Collecting all results
`launch` returns a `Deferred`, so you can collect results with `joinAll()`:
```lyng
val pool = LaunchPool(4)
val jobs = (1..10).map { n -> pool.launch { n * n } }
pool.closeAndJoin()
val results = jobs.joinAll()
// results == [1, 4, 9, 16, 25, 36, 49, 64, 81, 100]
```
## Concurrency limit in practice
With `maxWorkers = 2`, at most 2 tasks run simultaneously regardless of how many are queued:
```lyng
val mu = Mutex()
var active = 0
var maxSeen = 0
val pool = LaunchPool(2)
(1..8).map {
pool.launch {
mu.withLock { active++; if (active > maxSeen) maxSeen = active }
delay(5)
mu.withLock { active-- }
}
}
pool.closeAndJoin()
assert(maxSeen <= 2)
```
## See also
- [parallelism.md](parallelism.md) — `launch`, `Deferred`, `Mutex`, `Channel`, and coroutine basics
- [Channel.md](Channel.md) — the underlying channel primitive used by `LaunchPool`

View File

@ -1,66 +0,0 @@
# Legacy Digest Functions (`lyng.legacy_digest`)
> ⚠️ **Security warning:** The functions in this module use cryptographically broken
> algorithms. Do **not** use them for passwords, digital signatures, integrity
> verification against adversarial tampering, or any other security-sensitive
> purpose. They exist solely for compatibility with legacy protocols and file
> formats that require specific hash values.
Import when you need to produce a SHA-1 digest for an existing protocol or format:
```lyng
import lyng.legacy_digest
```
## `LegacyDigest` Object
### `sha1(data): String`
Computes the SHA-1 digest of `data` and returns it as a 40-character lowercase
hex string.
`data` can be:
| Type | Behaviour |
|----------|----------------------------------------|
| `String` | Encoded as UTF-8, then hashed |
| `Buffer` | Raw bytes hashed directly |
| anything | Falls back to `toString()` then UTF-8 |
```lyng
import lyng.legacy_digest
// String input
val h = LegacyDigest.sha1("abc")
assertEquals("a9993e364706816aba3e25717850c26c9cd0d89d", h)
// Empty string
assertEquals("da39a3ee5e6b4b0d3255bfef95601890afd80709", LegacyDigest.sha1(""))
```
```lyng
import lyng.legacy_digest
import lyng.buffer
// Buffer input (raw bytes)
val buf = Buffer.decodeHex("616263") // 0x61 0x62 0x63 = "abc"
assertEquals("a9993e364706816aba3e25717850c26c9cd0d89d", LegacyDigest.sha1(buf))
```
## Implementation Notes
- Pure Kotlin/KMP — no native libraries or extra dependencies.
- Follows FIPS 180-4.
- The output is always lowercase hex, never uppercase or binary.
## When to Use
Use `lyng.legacy_digest` only when an external system you cannot change requires
a SHA-1 value, for example:
- old git-style content addresses
- some OAuth 1.0 / HMAC-SHA1 signature schemes
- legacy file checksums defined in published specs
For any new design choose a current hash function (SHA-256 or better) once
Lyng adds a `lyng.digest` module.

View File

@ -1,8 +1,6 @@
# List built-in class
Mutable list of any objects.
For immutable list values, see [ImmutableList].
For observable mutable lists and change hooks, see [ObservableList].
It's class in Lyng is `List`:
@ -30,13 +28,6 @@ There is a shortcut for the last:
__Important__ negative indexes works wherever indexes are used, e.g. in insertion and removal methods too.
The language also allows multi-selector indexing syntax such as `value[i, j]`, but `List` itself uses a single selector only:
- `list[index]` for one element
- `list[range]` for a slice copy
Multi-selector indexing is intended for custom indexers such as `Matrix`.
## Concatenation
You can concatenate lists or iterable objects:
@ -45,16 +36,6 @@ You can concatenate lists or iterable objects:
assert( [4,5] + (1..3) == [4, 5, 1, 2, 3])
>>> void
## Constructing lists
Besides literals, you can build a list by size using `List.fill`:
val squares = List.fill(5) { i -> i * i }
assertEquals([0, 1, 4, 9, 16], squares)
>>> void
`List.fill(size) { ... }` calls the block once for each index from `0` to `size - 1` and returns a new mutable list.
## Appending
To append to lists, use `+=` with elements, lists and any [Iterable] instances, but beware it will
@ -174,9 +155,6 @@ List could be sorted in place, just like [Collection] provide sorted copies, in
| `[index]` | get or set element at index | Int |
| `[Range]` | get slice of the array (copy) | Range |
| `+=` | append element(s) (2) | List or Obj |
| `List.fill(size, block)` | build a new list from indices `0..<size` | Int, Callable |
| `List.fill(size,capacity,block)` | same, pre-allocating capacity slots | Int, Int, Callable |
| `ensureCapacity(count)` | pre-allocate storage for at least `count` elements without reallocation (5) | Int |
| `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 |
@ -199,57 +177,8 @@ order, e.g. is same as `list.sortWith { a,b -> predicate(a) <=> predicate(b) }`
positive if first is greater, and zero if they are equal. For example, the equvalent comparator
for `sort()` will be `sort { a, b -> a <=> b }
(5)
: if the current capacity is already ≥ `count`, this is a no-op. Otherwise the internal storage
is reallocated to hold at least `count` elements. Use this before a bulk `+=` loop to avoid
repeated reallocations. `List.fill(size, capacity, block)` calls this automatically.
It inherits from [Iterable] too and thus all iterable methods are applicable to any list.
## Observable list hooks
Observable hooks are provided by module `lyng.observable` and are opt-in:
import lyng.observable
val src = [1,2,3]
val xs = src.observable()
assert(xs is ObservableList<Int>)
var before = 0
var after = 0
xs.beforeChange { before++ }
xs.onChange { after++ }
xs += 4
xs[0] = 100
assertEquals([100,2,3,4], xs)
assertEquals(2, before)
assertEquals(2, after)
>>> void
`beforeChange` runs before mutation commit and may reject it by throwing exception (typically `ChangeRejectionException` from the same module):
import lyng.observable
val xs = [1,2].observable()
xs.beforeChange { throw ChangeRejectionException("read only") }
assertThrows(ChangeRejectionException) { xs += 3 }
assertEquals([1,2], xs)
>>> void
`changes()` returns `Flow<ListChange<T>>` of committed events:
import lyng.observable
val xs = [10,20].observable()
val it = xs.changes().iterator()
xs += 30
assert(it.hasNext())
val e = it.next()
assert(e is ListInsert<Int>)
assertEquals([30], (e as ListInsert<Int>).values)
it.cancelIteration()
>>> void
## Member inherited from Array
| name | meaning | type |
@ -267,6 +196,4 @@ Observable hooks are provided by module `lyng.observable` and are opt-in:
[Range]: Range.md
[Iterable]: Iterable.md
[ImmutableList]: ImmutableList.md
[ObservableList]: ObservableList.md
[Iterable]: Iterable.md

View File

@ -3,7 +3,6 @@
Map is a mutable collection of key-value pairs, where keys are unique. You can create maps in two ways:
- with the constructor `Map(...)` or `.toMap()` helpers; and
- with map literals using braces: `{ "key": value, id: expr, id: }`.
For immutable map values, see [ImmutableMap].
When constructing from a list, each list item must be a [Collection] with exactly 2 elements, for example, a [List].
@ -95,8 +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
@ -177,5 +175,4 @@ Notes:
- 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.
[Collection](Collection.md)
[ImmutableMap]: ImmutableMap.md
[Collection](Collection.md)

View File

@ -1,192 +0,0 @@
# Matrix (`lyng.matrix`)
`lyng.matrix` adds dense immutable `Matrix` and `Vector` types for linear algebra.
Import it when you need matrix or vector arithmetic:
```lyng
import lyng.matrix
```
## Construction
Create vectors from a flat list and matrices from nested row lists:
```lyng
import lyng.matrix
val v: Vector = vector([1, 2, 3])
val m: Matrix = matrix([[1, 2, 3], [4, 5, 6]])
assertEquals([1.0, 2.0, 3.0], v.toList())
assertEquals([[1.0, 2.0, 3.0], [4.0, 5.0, 6.0]], m.toList())
```
Factory methods are also available:
```lyng
import lyng.matrix
val z: Vector = Vector.zeros(3)
val i: Matrix = Matrix.identity(3)
val m: Matrix = Matrix.zeros(2, 4)
```
All elements are standard double-precision numeric values internally.
## Shapes
Matrices may have any rectangular geometry:
```lyng
import lyng.matrix
val a: Matrix = matrix([[1, 2, 3], [4, 5, 6]])
assertEquals(2, a.rows)
assertEquals(3, a.cols)
assertEquals([2, 3], a.shape)
assertEquals(false, a.isSquare)
```
Vectors expose:
- `size`
- `length` as an alias of `size`
## Matrix Operations
Supported matrix operations:
- `+` and `-` for element-wise matrix arithmetic
- `*` for matrix-matrix product
- `*` and `/` by a scalar
- `transpose()`
- `trace()`
- `rank()`
- `determinant()`
- `inverse()`
- `solve(rhs)` for `Vector` or `Matrix` right-hand sides
Example:
```lyng
import lyng.matrix
val a: Matrix = matrix([[1, 2, 3], [4, 5, 6]])
val b: Matrix = matrix([[7, 8], [9, 10], [11, 12]])
val product: Matrix = a * b
assertEquals([[58.0, 64.0], [139.0, 154.0]], product.toList())
assertEquals([[1.0, 4.0], [2.0, 5.0], [3.0, 6.0]], a.transpose().toList())
```
Inverse and solve:
```lyng
import lyng.matrix
val a: Matrix = matrix([[4, 7], [2, 6]])
val rhs: Vector = vector([1, 0])
val inv: Matrix = a.inverse()
val x: Vector = a.solve(rhs)
assert(abs(a.determinant() - 10.0) < 1e-9)
assert(abs(inv.get(0, 0) - 0.6) < 1e-9)
assert(abs(x.get(0) - 0.6) < 1e-9)
```
## Vector Operations
Supported vector operations:
- `+` and `-`
- scalar `*` and `/`
- `dot(other)`
- `norm()`
- `normalize()`
- `cross(other)` for 3D vectors
- `outer(other)` producing a matrix
```lyng
import lyng.matrix
val a: Vector = vector([1, 2, 3])
val b: Vector = vector([2, 0, 0])
assertEquals(2.0, a.dot(b))
assertEquals([0.2672612419124244, 0.5345224838248488, 0.8017837257372732], a.normalize().toList())
```
## Indexing and Slicing
`Matrix` supports both method-style indexing and bracket syntax.
Scalar access:
```lyng
import lyng.matrix
val m: Matrix = matrix([[1, 2, 3], [4, 5, 6]])
assertEquals(6.0, m.get(1, 2))
assertEquals(6.0, m[1, 2])
```
Bracket indexing accepts two selectors: `[row, col]`.
Each selector may be either:
- an `Int`
- a `Range`
Examples:
```lyng
import lyng.matrix
val m: Matrix = matrix([[1, 2, 3, 4], [5, 6, 7, 8], [9, 10, 11, 12]])
assertEquals(7.0, m[1, 2])
val columnSlice: Matrix = m[0..2, 2]
val topLeft: Matrix = m[0..1, 0..1]
val tail: Matrix = m[1.., 1..]
assertEquals([[3.0], [7.0], [11.0]], columnSlice.toList())
assertEquals([[1.0, 2.0], [5.0, 6.0]], topLeft.toList())
assertEquals([[6.0, 7.0, 8.0], [10.0, 11.0, 12.0]], tail.toList())
```
Shape rules:
- `m[Int, Int]` returns a `Real`
- `m[Range, Int]` returns an `Nx1` `Matrix`
- `m[Int, Range]` returns a `1xM` `Matrix`
- `m[Range, Range]` returns a submatrix
Open-ended ranges are supported:
- `m[..1, ..1]`
- `m[1.., 1..]`
- `m[.., 2]`
Stepped ranges are not supported in matrix slicing.
Slices currently return new matrices, not views.
## Rows and Columns
If you want plain lists instead of a sliced matrix:
```lyng
import lyng.matrix
val a: Matrix = matrix([[1, 2, 3], [4, 5, 6]])
assertEquals([4.0, 5.0, 6.0], a.row(1))
assertEquals([2.0, 5.0], a.column(1))
```
## Backend Notes
The matrix module uses a platform-specific backend where available and falls back to pure Kotlin where needed.
The public Lyng API stays the same across platforms.

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:
@ -71,99 +71,14 @@ object DefaultLogger : Logger("Default") {
}
```
## 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.
Properties are declared using `val` (read-only) or `var` (read-write) followed by a name and `get()`/`set()` blocks:
The standard syntax uses parentheses:
```lyng
class Person(private var _age: Int) {
// Read-only property
@ -179,17 +94,21 @@ class Person(private var _age: Int) {
if (value >= 0) _age = value
}
}
val p = Person(15)
assertEquals("Minor", p.ageCategory)
p.age = 20
assertEquals("Adult", p.ageCategory)
```
### Laconic Syntax (Optional Parentheses)
### Laconic Expression Shorthand
For even cleaner code, you can omit the parentheses for `get` and `set`. This is especially useful for simple expression shorthand:
For simple accessors and methods, you can use the `=` shorthand for a more elegant and laconic form:
```lyng
class Circle(val radius: Real) {
// Laconic expression shorthand
val area get = π * radius * radius
val circumference get = 2 * π * radius
val area get() = π * radius * radius
val circumference get() = 2 * π * radius
fun diameter() = radius * 2
}
@ -198,16 +117,15 @@ fun median(a, b) = (a + b) / 2
class Counter {
private var _count = 0
var count get = _count set(v) = _count = v
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.
- **`val` properties** must have a `get()` accessor 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.
@ -264,19 +182,6 @@ A delegate is any object that provides the following methods (all optional depen
- `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
@ -376,10 +281,11 @@ 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
@ -454,43 +360,6 @@ Key rules and features:
- For arbitrary receivers, use casts: `(expr as Type).member(...)` or `(expr as? Type)?.member(...)`.
- Qualified access does not relax visibility.
### Receiver-stack lambdas
Qualified `this@Type` is also used outside inheritance when a lambda has multiple visible receivers.
This is common in DSL-style builders.
- `A & B` means one receiver value that implements both types.
- `context(A, B) C.()->R` means a receiver stack:
- primary `this` is `C`
- outer/context receivers are `A`, then `B`
- Unqualified lookup checks the primary receiver first.
- If the primary receiver does not define a member and several outer/context receivers do, Lyng reports a compile-time ambiguity. Use `this@Type` to select one explicitly.
Example:
```lyng
class Html { fun title() = "html" }
class Head { fun title() = "head" }
class Body
val block: context(Html, Head) Body.()->String = {
// title() // compile-time ambiguity: Html vs Head
this@Html.title()
}
```
Context receivers can also constrain extension functions. The extension is visible only when the required receiver is
already in the implicit receiver stack:
```lyng
class Tag { fun addText(text: String) { /* ... */ } }
context(Tag)
fun String.unaryPlus() {
this@Tag.addText(this)
}
```
- Field inheritance (`val`/`var`) and collisions
- Instance storage is kept per declaring class, internally disambiguated; unqualified read/write resolves to the first match in the resolution order (leftmost base).
- Qualified read/write (via `this@Type` or casts) targets the chosen ancestor’s storage.
@ -503,7 +372,7 @@ fun String.unaryPlus() {
- 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
@ -620,103 +489,6 @@ class Critical {
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` | `unaryPlus()` |
| `-a` | `negate()` |
| `!a` | `logicalNot()` |
| `~a` | `bitNot()` |
`unaryPlus()` is useful in DSL-style builders where `+"text"` should append text to
the current receiver. See [samples/html_builder_dsl.lyng](samples/html_builder_dsl.lyng).
### 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.
Compatibility notes:
- Existing single‑inheritance code continues to work unchanged; its resolution order reduces to the single base.
@ -800,23 +572,6 @@ 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:
@ -939,44 +694,23 @@ 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
@ -1019,7 +753,7 @@ 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" )
@ -1030,164 +764,57 @@ 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
# Extension members
# Extending classes
Sometimes an existing type or named singleton object is missing some particular functionality that can be _added to it_ without rewriting its inner logic and without using its private state. In this case, _extension members_ can 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 members_ could be used.
## Extension methods
For example, we want to create an extension method that would test if a value can be interpreted as an integer:
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:
fun Int.isInteger() { true }
fun Real.isInteger() { this.toInt() == this }
fun String.isInteger() { (this.toReal() as Real).isInteger() }
fun Object.isInteger() {
when(this) {
// already Int?
is Int -> true
// Let's test:
// real, but with no declimal part?
is Real -> toInt() == this
// string with int or real reuusig code above
is String -> toReal().isInteger()
// otherwise, no:
else -> false
}
}
// Let's test:
assert( 12.isInteger() == true )
assert( 12.1.isInteger() == false )
assert( "5".isInteger() )
assert( ! "5.2".isInteger() )
>>> void
Extension methods normally act like instance members. If declared as `static`, they are called on the type object itself:
```lyng
static fun List<T>.fill(size: Int, block: (Int)->T): List<T> { ... }
val tens = List.fill(5) { it * 10 }
```
## Extending singleton `object` declarations
Named singleton objects can also be extension receivers. Use the object name as the receiver type:
```lyng
object Config {
fun base() = "cfg"
}
fun Config.describe(value) {
this.base() + ":" + value.toString()
}
val Config.tag get() = this.base() + ":tag"
assertEquals("cfg:42", Config.describe(42))
assertEquals("cfg:tag", Config.tag)
```
This differs from extending a class in one important way:
- `fun Point.foo()` adds a member-like extension for all `Point` instances.
- `fun Config.foo()` adds a member-like extension only for the single named object `Config`.
The same rules still apply:
- Extensions on singleton objects are scope-isolated just like class extensions.
- They cannot access the object's `private` members.
- Inside the extension body, `this` is the singleton object itself.
## Extension indexers
Bracket syntax is just another form of member dispatch. When you write `value[i]` or `value[i] = x`, Lyng lowers it to `getAt(...)` and `putAt(...)`.
That means indexers can be extended in the same way as methods and properties.
### Extending indexers on classes
Use `override fun Type.getAt(...)` and `override fun Type.putAt(...)`:
```lyng
class BoxStore {
val data = {"name": "alice"}
}
override fun BoxStore.getAt(key: String): Object? {
data[key]
}
override fun BoxStore.putAt(key: String, value: Object) {
data[key] = value
}
val store = BoxStore()
assertEquals("alice", store["name"])
store["name"] = "bob"
assertEquals("bob", store["name"])
```
As with other extension members, this does not modify the original class declaration. It adds indexer behavior only in the scope where the extension is visible.
### Extending indexers on singleton `object` declarations
Named singleton objects work the same way:
```lyng
object Storage
var storageData = {}
override fun Storage.getAt(key: String): Object? {
storageData[key]
}
override fun Storage.putAt(key: String, value: Object) {
storageData[key] = value
}
Storage["name"] = "alice"
val name: String? = Storage["name"]
assertEquals("alice", name)
```
This is the indexer equivalent of `fun Config.foo()`: the extension applies to that single named object, not to all instances of some class.
### Selector packing
Index syntax can contain more than one selector:
```lyng
value[i]
value[i, j]
```
The same packing rules still apply for extension indexers:
- `value[i]` calls `getAt(i)` or `putAt(i, value)`
- `value[i, j]` passes `[i, j]` as one list-like index object
- `value[i, j, k]` passes `[i, j, k]`
So if you want multi-selector indexing, define the receiver to accept that packed index object.
### About types and generics
In practice, extension indexers are usually best declared with `Object?` for reads and `Object` for writes:
```lyng
override fun Storage.getAt(key: String): Object? { ... }
override fun Storage.putAt(key: String, value: Object) { ... }
```
Then use the expected type at the call site:
```lyng
val name: String? = Storage["name"]
```
Explicit generic arguments do not fit naturally onto `[]` syntax, so typed assignment on read is usually the clearest approach.
## Extension properties
Just like methods, you can extend existing classes with properties. These can be defined using simple initialization (for `val` only) or with custom accessors.
@ -1283,7 +910,7 @@ 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 ->
@ -1351,24 +978,8 @@ collection's sugar won't work with it:
assertEquals("buzz", x[0])
>>> void
Multiple selectors are packed into one list index object:
val x = dynamic {
get {
if( it == [1, 2] ) "hit"
else null
}
}
assertEquals("hit", x[1, 2])
>>> void
So:
- `x[i]` passes `i`
- `x[i, j]` passes `[i, j]`
- `x[i, j, k]` passes `[i, j, k]`
This is the same rule used by Kotlin-backed `getAt` / `putAt` indexers in embedding.
If you want dynamic to function like an array, create a [feature
request](https://gitea.sergeych.net/SergeychWorks/lyng/issues).
# Theory

View File

@ -1,70 +0,0 @@
# ObservableList module
`ObservableList` lives in explicit module `lyng.observable`.
Import it first:
import lyng.observable
>>> void
Create from a regular mutable list:
import lyng.observable
val xs = [1,2,3].observable()
assert(xs is ObservableList<Int>)
assertEquals([1,2,3], xs)
>>> void
## Hook flow
Event order is:
1. `beforeChange(change)` listeners
2. mutation commit
3. `onChange(change)` listeners
4. `changes()` flow emission
Rejection is done by throwing in `beforeChange`.
import lyng.observable
val xs = [1,2].observable()
xs.beforeChange {
throw ChangeRejectionException("no mutation")
}
assertThrows(ChangeRejectionException) { xs += 3 }
assertEquals([1,2], xs)
>>> void
## Subscriptions
`beforeChange` and `onChange` return `Subscription`.
Call `cancel()` to unsubscribe.
import lyng.observable
val xs = [1].observable()
var hits = 0
val sub = xs.onChange { hits++ }
xs += 2
sub.cancel()
xs += 3
assertEquals(1, hits)
>>> void
## Change events
`changes()` returns `Flow<ListChange<T>>` with concrete event classes:
- `ListInsert`
- `ListSet`
- `ListRemove`
- `ListClear`
- `ListReorder`
import lyng.observable
val xs = [10,20].observable()
val it = xs.changes().iterator()
xs[1] = 200
val ev = it.next()
assert(ev is ListSet<Int>)
assertEquals(20, (ev as ListSet<Int>).oldValue)
assertEquals(200, ev.newValue)
it.cancelIteration()
>>> void

View File

@ -1,309 +0,0 @@
# Operator Interop Registry
`lyng.operators` provides a runtime registry for mixed-class binary operators.
Import it when you want expressions such as:
- `1 + MyType(...)`
- `2 < MyType(...)`
- `3 == MyType(...)`
to work without modifying the built-in `Int` or `Real` classes.
```lyng
import lyng.operators
```
## Why This Exists
If your class defines:
```lyng
class Amount(val value: Int) {
fun plus(other: Amount) = Amount(value + other.value)
}
```
then:
```lyng
Amount(1) + Amount(2)
```
works, because the left operand already knows how to add another `Amount`.
But:
```lyng
1 + Amount(2)
```
does not naturally work, because `Int` has not been rewritten to know about `Amount`.
The operator interop registry solves exactly that problem.
## Mental Model
Registration describes a mixed pair:
- left class `L`
- right class `R`
- common class `C`
When Lyng sees `L op R`, it:
1. converts `L -> C`
2. converts `R -> C`
3. evaluates the operator as `C op C`
So the registry is a bridge, not a separate arithmetic engine.
## API
```lyng
OperatorInterop.register(
leftClass,
rightClass,
commonClass,
operators,
leftToCommon,
rightToCommon
)
```
Parameters:
- `leftClass`: original left operand class
- `rightClass`: original right operand class
- `commonClass`: class that will actually execute the operator methods
- `operators`: list of operators enabled for this pair
- `leftToCommon`: conversion from left operand to common class
- `rightToCommon`: conversion from right operand to common class
## Supported Operators
`BinaryOperator` values:
- `Plus`
- `Minus`
- `Mul`
- `Div`
- `Mod`
- `Compare`
- `Equals`
Meaning:
- `Compare` enables `<`, `<=`, `>`, `>=`, and `<=>`
- `Equals` enables `==` and `!=`
## Minimal Working Example
```lyng
package test.decimalbox
import lyng.operators
class DecimalBox(val value: Int) {
fun plus(other: DecimalBox) = DecimalBox(value + other.value)
fun minus(other: DecimalBox) = DecimalBox(value - other.value)
fun mul(other: DecimalBox) = DecimalBox(value * other.value)
fun div(other: DecimalBox) = DecimalBox(value / other.value)
fun mod(other: DecimalBox) = DecimalBox(value % other.value)
fun compareTo(other: DecimalBox) = value <=> other.value
}
OperatorInterop.register(
Int,
DecimalBox,
DecimalBox,
[
BinaryOperator.Plus,
BinaryOperator.Minus,
BinaryOperator.Mul,
BinaryOperator.Div,
BinaryOperator.Mod,
BinaryOperator.Compare,
BinaryOperator.Equals
],
{ x: Int -> DecimalBox(x) },
{ x: DecimalBox -> x }
)
```
Then:
```lyng
import test.decimalbox
assertEquals(DecimalBox(3), 1 + DecimalBox(2))
assertEquals(DecimalBox(1), 3 - DecimalBox(2))
assertEquals(DecimalBox(8), 4 * DecimalBox(2))
assertEquals(DecimalBox(4), 8 / DecimalBox(2))
assertEquals(DecimalBox(1), 7 % DecimalBox(2))
assert(1 < DecimalBox(2))
assert(2 <= DecimalBox(2))
assert(3 > DecimalBox(2))
assert(2 == DecimalBox(2))
assert(2 != DecimalBox(3))
```
## How Decimal Uses It
`lyng.decimal` uses this same mechanism so that:
```lyng
import lyng.decimal
1 + 2.d
0.5 + 1.d
2 == 2.d
3 > 2.d
```
work naturally even though `Int` and `Real` themselves were not edited to know `Decimal`.
The shape is:
- `leftClass = Int` or `Real`
- `rightClass = Decimal`
- `commonClass = Decimal`
- convert built-ins into `Decimal`
- leave `Decimal` values unchanged
## Step-By-Step Pattern For Your Own Type
### 1. Pick the common class
Choose one class that will be the actual arithmetic domain.
For numeric-like types, that is usually your own class:
```lyng
class Rational(...)
```
### 2. Implement operators on that class
The common class should define the operations you plan to register.
Example:
```lyng
class Rational(val num: Int, val den: Int) {
fun plus(other: Rational) = Rational(num * other.den + other.num * den, den * other.den)
fun minus(other: Rational) = Rational(num * other.den - other.num * den, den * other.den)
fun mul(other: Rational) = Rational(num * other.num, den * other.den)
fun div(other: Rational) = Rational(num * other.den, den * other.num)
fun compareTo(other: Rational) = (num * other.den) <=> (other.num * den)
static fun fromInt(value: Int) = Rational(value, 1)
}
```
### 3. Register the mixed pair
```lyng
import lyng.operators
OperatorInterop.register(
Int,
Rational,
Rational,
[
BinaryOperator.Plus,
BinaryOperator.Minus,
BinaryOperator.Mul,
BinaryOperator.Div,
BinaryOperator.Compare,
BinaryOperator.Equals
],
{ x: Int -> Rational.fromInt(x) },
{ x: Rational -> x }
)
```
### 4. Use it
```lyng
assertEquals(Rational(3, 2), 1 + Rational(1, 2))
assert(Rational(3, 2) == Rational(3, 2))
assert(2 > Rational(3, 2))
```
## Registering More Than One Built-in Type
If you want both `Int + MyType` and `Real + MyType`, register both pairs explicitly:
```lyng
OperatorInterop.register(
Int,
MyType,
MyType,
[BinaryOperator.Plus, BinaryOperator.Compare, BinaryOperator.Equals],
{ x: Int -> MyType.fromInt(x) },
{ x: MyType -> x }
)
OperatorInterop.register(
Real,
MyType,
MyType,
[BinaryOperator.Plus, BinaryOperator.Compare, BinaryOperator.Equals],
{ x: Real -> MyType.fromReal(x) },
{ x: MyType -> x }
)
```
Each mixed pair is independent.
## Pure Lyng Registration
This mechanism is intentionally useful from pure Lyng code, not only from Kotlin-backed modules.
That means you can:
- declare a class in Lyng
- define its operators in Lyng
- register mixed operand bridges in Lyng
without touching compiler internals.
## Where To Register
Register once during module initialization.
Top-level module code is a good place:
```lyng
package my.rational
import lyng.operators
class Rational(...)
OperatorInterop.register(...)
```
That keeps registration close to the type declaration and makes importing the module enough to activate the interop.
## What Registration Does Not Do
The registry does not:
- invent operators your common class does not implement
- change the original `Int`, `Real`, or other built-ins
- automatically cover every class pair
- replace normal method overload resolution when the left-hand class already knows what to do
It only teaches Lyng how to bridge a specific mixed pair into a common class for the listed operators.
## Recommended Design Rules
If you want interop to feel natural:
- choose one obvious common class
- make conversions explicit and unsurprising
- implement `compareTo` if you want ordering operators
- register `Equals` whenever mixed equality should work
- keep the registered operator list minimal and accurate
For decimal-like semantics, also read [Decimal.md](Decimal.md).

View File

@ -25,23 +25,6 @@ Exclusive end ranges are adopted from kotlin either:
assert(4 in r)
>>> void
Descending finite ranges are explicit too:
val r = 5 downTo 1
assert(r.isDescending)
assert(r.toList() == [5,4,3,2,1])
>>> void
Use `downUntil` when the lower bound should be excluded:
val r = 5 downUntil 1
assert(r.toList() == [5,4,3,2])
assert(1 !in r)
>>> void
This is explicit by design: `5..1` is not treated as a reverse range. It is an
ordinary ascending range with no values in it when iterated.
In any case, we can test an object to belong to using `in` and `!in` and
access limits:
@ -62,11 +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
@ -80,8 +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 )
@ -90,57 +70,11 @@ but
>>> 2
>>> void
Descending ranges work in `for` loops exactly the same way:
for( i in 3 downTo 1 )
println(i)
>>> 3
>>> 2
>>> 1
>>> void
And with an exclusive lower bound:
for( i in 3 downUntil 1 )
println(i)
>>> 3
>>> 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( [5,3,1] == (5 downTo 1 step 2).toList() )
assert( ['a','c','e'] == ('a'..'e' step 2).toList() )
>>> void
Descending ranges still use a positive `step`; the direction comes from
`downTo` / `downUntil`:
assert( ['e','c','a'] == ('e' downTo 'a' step 2).toList() )
>>> void
A negative step with `downTo` / `downUntil` is invalid.
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:
val r = 'a'..'c'
val r = 'a' .. 'c'
assert( 'b' in r)
assert( 'e' !in r)
for( ch in r )
@ -162,15 +96,12 @@ Exclusive end char ranges are supported too:
|-----------------|------------------------------|---------------|
| contains(other) | used in `in` | Range, or Any |
| isEndInclusive | true for '..' | Bool |
| isDescending | true for `downTo`/`downUntil`| 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

@ -19,9 +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 |
| `.isInfinite()` | true when the value is `Infinity` or `-Infinity` | Bool |
| `.isNaN()` | true when the value is `NaN` | Bool |
| `.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`.
@ -61,19 +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 )
>>> void
Regex replacement is exposed on `String.replace` and `String.replaceFirst`:
assertEquals( "v#.#.#", "v1.2.3".replace( "\d+".re, "#" ) )
assertEquals( "v[1].[2].[3]", "v1.2.3".replace( "(\d+)".re ) { m -> "[" + m[1] + "]" } )
assertEquals( "year-04-08", "2026-04-08".replaceFirst( "\d+".re, "year" ) )
>>> void
When `replace` takes a plain `String`, it is treated literally, not as a regex pattern:
assertEquals( "a-b-c", "a.b.c".replace( ".", "-" ) )
assert( "cd" == "abcdef"[ "c.".re ].value )
>>> void
@ -102,3 +88,4 @@ When `replace` takes a plain `String`, it is treated literally, not as a regex p
[List]: List.md
[Range]: Range.md

View File

@ -1,8 +1,7 @@
# Set built-in class
# List built-in class
Mutable set of any objects: a group of different objects, no repetitions.
Sets are not ordered, order of appearance does not matter.
For immutable set values, see [ImmutableSet].
val set = Set(1,2,3, "foo")
assert( 1 in set )
@ -27,8 +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 `-`:
@ -92,5 +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
[ImmutableSet]: ImmutableSet.md
[Range]: Range.md

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

@ -1,252 +0,0 @@
# Lyng Language Reference for AI Agents (Current Compiler State)
[//]: # (excludeFromIndex)
Purpose: dense, implementation-first reference for generating valid Lyng code.
Primary sources used: `lynglib/src/commonMain/kotlin/net/sergeych/lyng/{Parser,Token,Compiler,Script,TypeDecl}.kt`, `lynglib/stdlib/lyng/root.lyng`, tests in `lynglib/src/commonTest` and `lynglib/src/jvmTest`.
## 1. Ground Rules
- Resolution is compile-time-first. Avoid runtime name/member lookup assumptions.
- `lyng.stdlib` is auto-seeded for normal scripts (default import manager).
- Use explicit casts when receiver type is unknown (`Object`/`Obj`).
- Prefer modern null-safe operators (`?.`, `?:`/`??`, `?=`, `as?`, `!!`).
- Do not rely on fallback opcodes or dynamic member fallback semantics.
## 2. Lexical Syntax
- Comments: `// line`, `/* block */`.
- Strings: `"..."` or `` `...` `` (supports escapes). Multiline string content is normalized by indentation logic.
- AI generation preference: use `"..."` by default, including multiline strings; `"` strings are also multiline-capable and should be preferred for ordinary code/doc/SQL text. Use backtick strings mainly when the content contains many double quotes and backticks would make the source clearer.
- Shared escapes: `\n`, `\r`, `\t`, `\\`, `\uXXXX` (4 hex digits).
- Delimiter escapes: `\"` inside `"..."`, ``\` `` inside `` `...` ``.
- Unicode escapes use exactly 4 hex digits (for example: `"\u0416"` -> `Ж`).
- Unknown `\x` escapes in strings are preserved literally as two characters (`\` and `x`).
- String interpolation is supported:
- identifier form: `"$name"` or `` `$name` ``
- expression form: `"${expr}"` or `` `${expr}` ``
- escaped dollar: `"\$"`, `"$$"`, `` `\$` ``, and `` `$$` `` all produce literal `$`.
- `\\$x` means backslash + interpolated `x` in either delimiter form.
- Per-file opt-out is supported via leading comment directive:
- `// feature: interpolation: off`
- with this directive, `$...` stays literal text.
- Numbers: `Int` (`123`, `1_000`), `Real` (`1.2`, `1e3`), hex (`0xFF`).
- Char: `'a'`, escaped chars supported.
- Supported escapes: `\n`, `\r`, `\t`, `\'`, `\\`, `\uXXXX` (4 hex digits).
- Backslash character in a char literal must be written as `'\\'` (forms like `'\'` are invalid).
- Labels:
- statement label: `loop@ for (...) { ... }`
- label reference: `break@loop`, `continue@loop`, `return@fnLabel`
- Keywords/tokens include (contextual in many places):
- declarations: `fun`/`fn`, `val`, `var`, `class`, `object`, `interface`, `enum`, `type`, `init`
- modifiers: `private`, `protected`, `static`, `abstract`, `closed`, `override`, `extern`, `open`
- flow: `if`, `else`, `when`, `for`, `while`, `do`, `try`, `catch`, `finally`, `throw`, `return`, `break`, `continue`
## 3. Literals and Core Expressions
- Scalars: `null`, `true`, `false`, `void`.
- List literal: `[a, b, c]`, spreads with `...`.
- Spread positions: beginning, middle, end are all valid: `[...a]`, `[0, ...a, 4]`, `[head, ...mid, tail]`.
- Spread source must be a `List` at runtime (non-list spread raises an error).
- Map literal: `{ key: value, x:, ...otherMap }`.
- `x:` means shorthand `x: x`.
- Map spread source must be a `Map`.
- Range literals:
- inclusive: `a..b`
- exclusive end: `a..<b`
- descending inclusive: `a downTo b`
- descending exclusive end: `a downUntil b`
- open-ended forms are supported (`a..`, `..b`, `..`).
- optional step: `a..b step 2`, `a downTo b step 2`
- Lambda literal:
- with params: `{ x, y -> x + y }`
- implicit `it`: `{ it + 1 }`
- Ternary conditional is supported: `cond ? thenExpr : elseExpr`.
## 3.1 Splats in Calls and Lambdas
- Declaration-side variadic parameters use ellipsis suffix:
- functions: `fun f(head, tail...) { ... }`
- lambdas: `{ x, rest... -> ... }`
- Call-side splats use `...expr` and are expanded by argument kind:
- positional splat: `f(...[1,2,3])`
- named splat: `f(...{ a: 1, b: 2 })` (map-style)
- Runtime acceptance for splats:
- positional splat accepts `List` and general `Iterable` (iterable is converted to list first).
- named splat accepts `Map` with string keys only.
- Ordering/validation rules (enforced):
- positional argument cannot follow named arguments (except trailing-block parsing case).
- positional splat cannot follow named arguments.
- duplicate named arguments are errors (including duplicates introduced via named splat).
- unknown named parameters are errors.
- variadic parameter itself cannot be passed as a named argument (`fun g(args..., tail)` then `g(args: ...)` is invalid).
- Trailing block + named arguments:
- if the last callable parameter is already provided by name in parentheses, adding a trailing block is invalid.
## 4. Operators (implemented)
- Assignment: `=`, `+=`, `-=`, `*=`, `/=`, `%=`, `?=`.
- Logical: `||`, `&&`, unary `!`.
- Unary arithmetic/bitwise: unary `+`, unary `-`, `~`.
- Bitwise: `|`, `^`, `&`, `~`, shifts `<<`, `>>`.
- Equality/comparison: `==`, `!=`, `===`, `!==`, `<`, `<=`, `>`, `>=`, `<=>`, `=~`, `!~`.
- Type/containment: `is`, `!is`, `in`, `!in`, `as`, `as?`.
- Null-safe family:
- member access: `?.`
- safe index: `?[i]`, `?[i, j]`
- safe invoke: `?(...)`
- safe block invoke: `?{ ... }`
- elvis: `?:` and `??`.
- Increment/decrement: prefix and postfix `++`, `--`.
- Indexing syntax:
- single selector: `a[i]`
- multiple selectors: `a[i, j, k]`
- language-level indexing with multiple selectors is passed to `getAt`/`putAt` as one list-like index object, not as multiple method arguments.
- indexers can also be supplied by extension members, including named singleton `object` receivers via `override fun Storage.getAt(...)` / `putAt(...)`.
- example: `m[0..2, 2]`.
## 5. Declarations
- Variables:
- `val` immutable, `var` mutable.
- top-level/local `val` must be initialized.
- class `val` may be late-initialized, but must be assigned in class body/init before class parse ends.
- destructuring declaration: `val [a, b, rest...] = expr`.
- destructuring declaration details:
- allowed in `val` and `var` declarations.
- supports nested patterns: `val [a, [b, c...], d] = rhs`.
- supports at most one splat (`...`) per pattern level.
- RHS must be a `List`.
- without splat: RHS must have at least as many elements as pattern arity.
- with splat: head/tail elements are bound directly, splat receives a `List`.
- Functions:
- `fun` and `fn` are equivalent.
- full body: `fun f(x) { ... }`
- shorthand: `fun f(x) = expr`.
- generics: `fun f<T>(x: T): T`.
- extension functions: `fun Type.name(...) { ... }`.
- context-aware extension functions: `context(Tag) fun String.unaryPlus() { this@Tag.addText(this) }`.
- named singleton `object` declarations can be extension receivers too: `fun Config.describe(...) { ... }`, `val Config.tag get() = ...`.
- static extension functions are callable on the type object: `static fun List<T>.fill(...)` -> `List.fill(...)`.
- delegated callable: `fun f(...) by delegate`.
- Type aliases:
- `type Name = TypeExpr`
- generic: `type Box<T> = List<T>`
- aliases are expanded structurally.
- Classes/objects/enums/interfaces:
- `interface` is parsed as abstract class synonym.
- `object` supports named singleton and anonymous object expression forms.
- enums support lifted entries: `enum E* { A, B }`.
- multiple inheritance is supported; override is enforced when overriding base members.
- Properties/accessors in class body:
- accessor form supports `get`/`set`, including `private set`/`protected set`.
## 6. Control Flow
- `if` is expression-like.
- `compile if (cond) { ... } else { ... }` is a compile-time-only conditional.
- current condition grammar is restricted to `defined(NameOr.Package)`, `!`, `&&`, `||`, and parentheses.
- the untaken branch is skipped by the compiler and is not name-resolved or type-checked.
- `when(value) { ... }` supported.
- branch conditions support equality, `in`, `!in`, `is`, `!is`, and `nullable` predicate.
- `when { ... }` (subject-less) is currently not implemented.
- Loops: `for`, `while`, `do ... while`.
- loop `else` blocks are supported.
- `break value` can return a loop result.
- Exceptions: `try/catch/finally`, `throw`.
## 6.1 Destructuring Assignment (implemented)
- Reassignment form is supported (not only declaration):
- `[x, y] = [y, x]`
- Semantics match destructuring declaration:
- nested patterns allowed.
- at most one splat per pattern level.
- RHS must be a `List`.
- too few RHS elements raises runtime error.
- Targets in pattern are variables parsed from identifier patterns.
## 7. Type System (current behavior)
- Non-null by default (`T`), nullable with `T?`.
- `as` (checked cast), `as?` (safe cast returning `null`), `!!` non-null assertion.
- Type expressions support:
- unions `A | B`
- intersections `A & B`
- function types `(A, B)->R` and receiver form `Receiver.(A)->R`
- receiver-stack function types via `context(A, B) Receiver.(P)->R`
- variadics in function type via ellipsis (`T...`)
- `A & B` means one value implementing both types.
- `context(A, B) Receiver.(P)->R` is different: it declares an ordered implicit-receiver stack where `Receiver` is primary `this`, then `A`, then `B`.
- Nested receiver lambdas keep outer receivers in scope; unqualified lookup prefers the innermost receiver, and `this@Type` can select an outer/context receiver explicitly.
- If the primary receiver does not provide a member and multiple outer/context receivers do, the lookup is a compile-time ambiguity and must be disambiguated with `this@Type`.
- Generics:
- type params on classes/functions/type aliases
- bounds via `:` with union/intersection expressions
- declaration-site variance via `in` / `out`
- Generic function/class/type syntax examples:
- function: `fun choose<T>(a: T, b: T): T = a`
- class: `class Box<T>(val value: T)`
- alias: `type PairList<T> = List<List<T>>`
- Untyped params default to `Object` (`x`) or `Object?` (`x?` shorthand).
- Untyped `var x` starts as `Unset`; first assignment fixes type tracking in compiler.
## 7.1 Generics Runtime Model and Bounds (AI-critical)
- Lyng generic type information is operational in script execution contexts; do not assume JVM-style full erasure.
- Generic call type arguments can be:
- explicit at call site (`f<Int>(1)` style),
- inferred from runtime values/declared arg types,
- defaulted from type parameter defaults (or `Any` fallback).
- At function execution, generic type parameters are runtime-bound as constants in scope:
- simple non-null class-like types are bound as `ObjClass`,
- complex/nullable/union/intersection forms are bound as `ObjTypeExpr`.
- Practical implication for generated code:
- inside generic code, treat type params as usable type objects in `is`/`in`/type-expression logic (not as purely compile-time placeholders).
- example pattern: `if (value is T) { ... }`.
- Bound syntax (implemented):
- intersection bound: `fun f<T: A & B>(x: T) { ... }`
- union bound: `fun g<T: A | B>(x: T) { ... }`
- Bound checks happen at two points:
- compile-time call checking for resolvable generic calls,
- runtime re-check while binding type params for actual invocation.
- Bound satisfaction is currently class-hierarchy based for class-resolvable parts (including union/intersection combination rules).
- Keep expectations realistic:
- extern-generic runtime ABI for full instance-level generic metadata is still proposal-level (`proposals/extern_generic_runtime_abi.md`), so avoid assuming fully materialized generic-instance metadata everywhere.
## 7.2 Differences vs Java / Kotlin / Scala
- Java:
- Java generics are erased at runtime (except reflection metadata and raw `Class` tokens).
- Lyng generic params in script execution are runtime-bound type objects, so generated code can reason about `T` directly.
- Kotlin:
- Kotlin on JVM is mostly erased; full runtime type access usually needs `inline reified`.
- Lyng generic function execution binds `T` without requiring an inline/reified escape hatch.
- Scala:
- Scala has richer static typing but still runs on JVM erasure model unless carrying explicit runtime evidence (`TypeTag`, etc.).
- Lyng exposes runtime-bound type expressions/classes directly in generic execution scope.
- AI generation rule:
- do not port JVM-language assumptions like “`T` unavailable at runtime unless reified/tagged”.
- in Lyng, prefer direct type-expression-driven branching when useful, but avoid assuming extern object generic args are always introspectable today.
## 8. OOP, Members, and Dispatch
- Multiple inheritance with C3-style linearization behavior is implemented in class machinery.
- Disambiguation helpers are supported:
- qualified this: `this@Base.member()`
- cast view: `(obj as Base).member()`
- In nested receiver lambdas, `this@Type` can target any receiver visible through the receiver stack, not just inheritance ancestors.
- On unknown receiver types, compiler allows only Object-safe members:
- `toString`, `toInspectString`, `let`, `also`, `apply`, `run`
- Other members require known receiver type or explicit cast.
## 9. Delegation (`by`)
- Works for `val`, `var`, and `fun`.
- Expected delegate hooks in practice:
- `getValue(thisRef, name)`
- `setValue(thisRef, name, newValue)`
- `invoke(thisRef, name, args...)` for delegated callables
- optional `bind(name, access, thisRef)`
- `@Transient` is recognized for declarations/params and affects serialization/equality behavior.
## 10. Modules and Imports
- `package` and `import module.name` are supported.
- Import form is module-only (no aliasing/selective import syntax in parser).
- Default module ecosystem includes:
- auto-seeded: `lyng.stdlib`
- available by import: `lyng.observable`, `lyng.buffer`, `lyng.serialization`, `lyng.time`
- extra module (when installed): `lyng.io.fs`, `lyng.io.process`
## 11. Current Limitations / Avoid
- No subject-less `when { ... }` yet.
- No regex literal tokenization (`/.../`); use `Regex("...")` or `"...".re`.
- Do not generate runtime name fallback patterns from legacy docs.

View File

@ -1,14 +0,0 @@
# AI notes: publish JVM CLI updates with `bin/local_jrelease`
[//]: # (excludeFromIndex)
When a change affects the JVM CLI launcher used as `jlyng`, refresh the installed local distribution with:
```bash
bin/local_jrelease
```
Why:
- `jlyng` in this repo is installed from `~/bin/jlyng-jvm/lyng-jvm`, not directly from `lyng/build/install`.
- Manual copying from Gradle build output can leave the actual launcher on `PATH` stale.
- `bin/local_jrelease` rebuilds `lyng/build/distributions/lyng-jvm.zip`, reinstalls it under `~/bin/jlyng-jvm`, and recreates the `~/bin/jlyng` symlink.

View File

@ -1,10 +0,0 @@
# AI notes: heading levels must be consecutive
[//]: # (excludeFromIndex)
When editing repository documentation:
- Use heading levels in order: `#`, then `##`, then `###`, and so on.
- Do not skip levels, for example `#` directly to `###`.
- Keep the heading tree balanced inside each document; sibling sections should use the same level.
- If you add a subsection and the parent is `##`, the child must be `###`.

View File

@ -1,7 +1,5 @@
# AI notes: avoid Kotlin/Wasm invalid IR with suspend lambdas
[//]: # (excludeFromIndex)
## 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.

View File

@ -1,105 +0,0 @@
# Lyng Stdlib Reference for AI Agents (Compact)
[//]: # (excludeFromIndex)
Purpose: fast overview of what is available by default and what must be imported.
Sources: `lynglib/src/commonMain/kotlin/net/sergeych/lyng/Script.kt`, `lynglib/stdlib/lyng/root.lyng`, `lynglib/src/commonMain/kotlin/net/sergeych/lyng/stdlib_included/observable_lyng.kt`.
## 1. Default Availability
- Normal scripts are auto-seeded with `lyng.stdlib` (default import manager path).
- Root runtime scope also exposes global constants/functions directly.
## 2. Core Global Functions (Root Scope)
- IO/debug: `print`, `println`, `traceScope`.
- Invocation/util: `call`, `run`, `dynamic`, `cached`, `lazy`.
- Assertions/tests: `assert`, `assertEquals`/`assertEqual`, `assertNotEquals`, `assertThrows`.
- Preconditions: `require`, `check`.
- Async/concurrency: `launch`, `yield`, `flow`, `delay`.
- `Deferred.cancel()` cancels an active task.
- `Deferred.await()` throws `CancellationException` if that task was cancelled.
- `Iterable<Deferred>.joinAll()` awaits every deferred in iteration order and returns a `List` of results.
- Math: `floor`, `ceil`, `round`, `sin`, `cos`, `tan`, `asin`, `acos`, `atan`, `sinh`, `cosh`, `tanh`, `asinh`, `acosh`, `atanh`, `exp`, `ln`, `log10`, `log2`, `pow`, `sqrt`, `abs`, `clamp`.
- These helpers also accept `lyng.decimal.Decimal`.
- Exact Decimal path today: `abs`, `floor`, `ceil`, `round`, and `pow` with integral exponent.
- Temporary Decimal path for the rest: convert `Decimal -> Real`, compute, then convert back to `Decimal`.
- Treat that bridge as temporary; prefer native Decimal implementations when they become available.
## 3. Core Global Constants/Types
- Values: `Unset`, `π`.
- Primitive/class symbols: `Object`, `Int`, `Real`, `Bool`, `Char`, `String`, `Class`, `Callable`.
- Collections/types: `Iterable`, `Iterator`, `Collection`, `Array`, `List`, `ImmutableList`, `Set`, `ImmutableSet`, `Map`, `ImmutableMap`, `MapEntry`, `Range`, `RingBuffer`.
- Random: singleton `Random` and class `SeededRandom`.
- Async types: `Deferred`, `CompletableDeferred`, `Mutex`, `Flow`, `FlowBuilder`.
- Async exception: `CancellationException`.
- Delegation types: `Delegate`, `DelegateContext`.
- Regex types: `Regex`, `RegexMatch`.
- Also present: `Math.PI` namespace constant.
## 4. `lyng.stdlib` Module Surface (from `root.lyng`)
### 4.1 Extern class declarations
- Exceptions/delegation base: `Exception`, `CancellationException`, `IllegalArgumentException`, `NotImplementedException`, `Delegate`.
- Collections and iterables: `Iterable<T>`, `Iterator<T>`, `Collection<T>`, `Array<T>`, `List<T>`, `ImmutableList<T>`, `Set<T>`, `ImmutableSet<T>`, `Map<K,V>`, `ImmutableMap<K,V>`, `MapEntry<K,V>`, `RingBuffer<T>`.
- Host iterator bridge: `KotlinIterator<T>`.
- Random APIs: `extern object Random`, `extern class SeededRandom`.
### 4.2 High-use extension APIs
- Iteration/filtering: `forEach`, `filter`, `filterFlow`, `filterNotNull`, `filterFlowNotNull`, `drop`, `dropLast`, `takeLast`.
- Search/predicates: `findFirst`, `findFirstOrNull`, `any`, `all`, `count`, `first`, `last`.
- Mapping/aggregation: `map`, `flatMap`, `flatten`, `sum`, `sumOf`, `minOf`, `maxOf`.
- Ordering and list building: `sorted`, `sortedBy`, `shuffled`, `List.sort`, `List.sortBy`, `List.fill`.
- `List.fill(size) { index -> ... }` constructs a new `List<T>` by evaluating the block once per index from `0` to `size - 1`.
- String helper: `joinToString`, `String.re`.
### 4.3 Delegation helpers
- `enum DelegateAccess { Val, Var, Callable }`
- `interface Delegate<T,ThisRefType=void>` with `getValue`, `setValue`, `invoke`, `bind`.
- `class lazy<T,...>` delegate implementation.
- `fun with(self, block)` helper.
### 4.4 Other module-level symbols
- `$~` (last regex match object).
- `TODO(message?)` utility.
- `StackTraceEntry` class.
- `Random.nextInt()`, `Random.nextFloat()`, `Random.next(range)`, `Random.seeded(seed)`.
- `SeededRandom.nextInt()`, `SeededRandom.nextFloat()`, `SeededRandom.next(range)`.
## 5. Additional Built-in Modules (import explicitly)
- `import lyng.observable`
- `Observable`, `Subscription`, `ObservableList`, `ListChange` and change subtypes, `ChangeRejectionException`.
- `import lyng.decimal`
- `Decimal`, `DecimalContext`, `DecimalRounding`, `withDecimalContext(...)`.
- Kotlin host helper: `ScopeFacade.newDecimal(BigDecimal)` wraps an ionspin host decimal as a Lyng `Decimal`.
- `import lyng.complex`
- `Complex`, `complex(re, im)`, `cis(angle)`, and numeric embedding extensions such as `2.i` / `3.re`.
- `import lyng.matrix`
- `Matrix`, `Vector`, `matrix(rows)`, `vector(values)`, dense linear algebra, inversion, solving, and matrix slicing with `m[row, col]`.
- `import lyng.buffer`
- `Buffer`, `MutableBuffer`.
- `import lyng.legacy_digest`
- `LegacyDigest.sha1(data): String` — SHA-1 hex digest; `data` may be `String` (UTF-8) or `Buffer` (raw bytes).
- ⚠️ Cryptographically broken. Use only for legacy protocol / file-format compatibility.
- `import lyng.serialization`
- `Lynon` serialization utilities.
- `import lyng.time`
- `Instant`, `Date`, `DateTime`, `Duration`, and module `delay`.
## 6. Optional (lyngio) Modules
Requires installing `lyngio` into the import manager from host code.
- `import lyng.io.fs` (filesystem `Path` API)
- `import lyng.io.process` (process execution API)
- `import lyng.io.console` (console capabilities, geometry, ANSI/output, events)
- `import lyng.io.http` (HTTP/HTTPS client API)
- `import lyng.io.http.server` (minimal HTTP/1.1 and WebSocket server API)
- `import lyng.io.ws` (WebSocket client API; currently supported on JVM, capability-gated elsewhere)
- `import lyng.io.net` (TCP/UDP transport API; currently supported on JVM, capability-gated elsewhere)
- `import lyng.io.html` (pure Lyng HTML builder DSL: `html { body { h3 { +"text" } } }`)
- Shared network value-type packages are also available when installed by host code:
- `import lyng.io.http.types` (`HttpHeaders`)
- `import lyng.io.ws.types` (`WsMessage`)
- `import lyng.io.net.types` (`IpVersion`, `SocketAddress`, `Datagram`)
## 7. AI Generation Tips
- Assume `lyng.stdlib` APIs exist in regular script contexts.
- For platform-sensitive code (`fs`, `process`, `console`, `http`, `ws`, `net`), gate assumptions and mention required module install.
- Prefer extension-method style (`items.filter { ... }`) and standard scope helpers (`let`/`also`/`apply`/`run`).

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")

View File

@ -57,8 +57,7 @@ class lazy(val creator) : Delegate {
override fun getValue(thisRef, name) {
if (value == Unset) {
// calculate value using thisRef as this:
value = with(thisRef) creator()
value = creator()
}
value
}
@ -151,24 +150,6 @@ fun test() {
}
```
### 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:

View File

@ -1,12 +0,0 @@
# Some resources to download
## Lync CLI tool
- [lyng-linuxX64.zip](/distributables/lyng-linuxX64.zip) CLI tool for linuxX64: nodependencies, small monolith executable binary.
- [lyng-jvm.zip](/distributables/lyng-jvm.zip) JVM CLI distribution: download, unpack, and run `lyng-jvm/bin/lyng`.
## IDE plugins
- [lyng-textmate.zip](../../lyng/distributables/lyng-textmate.zip) Texmate-compatible bundle with syntax coloring (could be outdated)
- [lyng-idea-0.0.5-SNAPSHOT.zip](/distributables/lyng-idea-0.0.5-SNAPSHOT.zip) - plugin for IntelliJ-compatible IDE

View File

@ -1,12 +1,10 @@
# Embedding Lyng in your Kotlin project
[//]: # (topMenu)
Lyng is a tiny, embeddable, Kotlin‑first scripting language. This page shows, step by step, how to:
- add Lyng to your build
- create a runtime and execute scripts
- declare extern globals in Lyng and bind them from Kotlin
- define functions and variables from Kotlin
- read variable values back in Kotlin
- call Lyng functions from Kotlin
- create your own packages and import them in Lyng
@ -38,60 +36,21 @@ dependencies {
If you use Kotlin Multiplatform, add the dependency in the `commonMain` source set (and platform‑specific sets if you need platform APIs).
### 2) Preferred runtime: `EvalSession`
### 2) Create a runtime (Scope) and execute scripts
For host applications, prefer `EvalSession` as the main way to run scripts.
It owns one reusable Lyng scope, serializes `eval(...)` calls, and governs coroutines started from Lyng `launch { ... }`.
Main entrypoints:
- `session.eval(code)` / `session.eval(source)`
- `session.getScope()` when you need low-level binding APIs
- `session.cancel()` to cancel active session-owned coroutines
- `session.join()` to wait for active session-owned coroutines
```kotlin
fun main() = kotlinx.coroutines.runBlocking {
val session = EvalSession()
// Evaluate a one‑liner
val result = session.eval("1 + 2 * 3")
println("Lyng result: $result") // ObjReal/ObjInt etc.
// Optional lifecycle management
session.join()
}
```
The session creates its underlying scope lazily. If you need raw low-level APIs, get the scope explicitly:
```kotlin
val session = EvalSession()
val scope = session.getScope()
```
Use `cancel()` / `join()` to govern async work started by scripts:
```kotlin
val session = EvalSession()
session.eval("""launch { delay(1000); println("done") }""")
session.cancel()
session.join()
```
### 2.1) Low-level runtime: `Scope`
Use `Scope` directly when you intentionally want lower-level control.
The easiest way to get a ready‑to‑use scope with standard packages is via `Script.newScope()`.
```kotlin
fun main() = kotlinx.coroutines.runBlocking {
val scope = Script.newScope() // suspends on first init
// Evaluate a one‑liner
val result = scope.eval("1 + 2 * 3")
println("Lyng result: $result")
println("Lyng result: $result") // ObjReal/ObjInt etc.
}
```
You can also pre‑compile a script and execute it multiple times on the same scope:
You can also pre‑compile a script and execute it multiple times:
```kotlin
val script = Compiler.compile("""
@ -104,98 +63,34 @@ val run1 = script.execute(scope)
val run2 = script.execute(scope)
```
`Scope.eval("...")` is the low-level shortcut that compiles and executes on the given scope.
For most embedding use cases, prefer `session.eval("...")`.
`Scope.eval("...")` is a shortcut that compiles and executes on the given scope.
### 3) Preferred: bind extern globals from Kotlin
### 3) Define variables from Kotlin
For module-level APIs, the default workflow is:
1. declare globals in Lyng using `extern fun` / `extern val` / `extern var`;
2. bind Kotlin implementation via `ModuleScope.globalBinder()`.
This is also the recommended way to expose a Kotlin-backed value that should behave like a true
Lyng global variable/property. If you need `x` to read/write through Kotlin on every access, use
`extern var` / `extern val` plus `bindGlobalVar(...)`.
Do not use `addConst(...)` for this case: `addConst(...)` installs a value, not a Kotlin-backed
property accessor. It is appropriate for fixed values and objects, but not for a global that should
delegate reads/writes back into Kotlin state.
To expose data to Lyng, add constants (read‑only) or mutable variables to the scope. All values in Lyng are `Obj` instances; the core types live in `net.sergeych.lyng.obj`.
```kotlin
import net.sergeych.lyng.bridge.*
import net.sergeych.lyng.obj.ObjInt
import net.sergeych.lyng.obj.ObjString
// Read‑only constant
scope.addConst("pi", ObjReal(3.14159))
val session = EvalSession()
val scope = session.getScope()
val im = Script.defaultImportManager.copy()
im.addPackage("my.api") { module ->
module.eval("""
extern fun globalFun(v: Int): Int
extern var globalProp: String
extern val globalVersion: String
""".trimIndent())
// Mutable variable: create or update
scope.addOrUpdateItem("counter", ObjInt(0))
val binder = module.globalBinder()
binder.bindGlobalFun1<Int>("globalFun") { v ->
ObjInt.of((v + 1).toLong())
}
var prop = "initial"
binder.bindGlobalVar(
name = "globalProp",
get = { prop },
set = { prop = it }
)
binder.bindGlobalVar(
name = "globalVersion",
get = { "1.0.0" } // readonly: setter omitted
)
}
// Use it from Lyng
scope.eval("counter = counter + 1")
```
Usage from Lyng:
```lyng
import my.api
assertEquals(42, globalFun(41))
assertEquals("initial", globalProp)
globalProp = "changed"
assertEquals("changed", globalProp)
assertEquals("1.0.0", globalVersion)
```
Minimal rule of thumb:
- use `bindGlobalFun(...)` for global functions
- use `bindGlobalVar(...)` for Kotlin-backed global variables/properties
- use `addConst(...)` only for fixed values/objects that do not need getter/setter behavior
For custom argument handling and full runtime access:
Tip: Lyng values can be converted back to Kotlin with `toKotlin(scope)`:
```kotlin
binder.bindGlobalFun("sum3") {
requireExactCount(3)
ObjInt.of((int(0) + int(1) + int(2)).toLong())
}
binder.bindGlobalFunRaw("echoRaw") { _, args ->
args.firstAndOnly()
}
val current = (scope.eval("counter")).toKotlin(scope) // Any? (e.g., Int/Double/String/List)
```
### 4) Low-level: direct functions/variables from Kotlin
### 4) Add Kotlin‑backed functions
Use this when you intentionally want raw `Scope` APIs. For most module APIs, prefer section 3.
Use `Scope.addFn`/`addVoidFn` to register functions implemented in Kotlin. Inside the lambda, use `this.args` to access arguments and return an `Obj`.
```kotlin
val session = EvalSession()
val scope = session.getScope()
// A function returning value
scope.addFn<ObjInt>("inc") {
val x = args.firstAndOnly() as ObjInt
@ -208,101 +103,17 @@ 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
session.eval("val y = inc(41); log('Answer:', y)")
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.
### 4.5) Indexers from Kotlin: `getAt` and `putAt`
Lyng bracket syntax is dispatched through `getAt` and `putAt`.
That means:
- `x[i]` calls `getAt(index)`
- `x[i] = value` calls `putAt(index, value)` or `setAt(index, value)`
- field-like `x["name"]` also uses the same index path unless you expose a real field/property
For Kotlin-backed classes, bind indexers as ordinary methods named `getAt` and `putAt`:
```kotlin
moduleScope.eval("""
extern class Grid {
override fun getAt(index: List<Int>): Int
override fun putAt(index: List<Int>, value: Int): void
}
""".trimIndent())
moduleScope.bind("Grid") {
init { _ -> data = IntArray(4) }
addFun("getAt") {
val index = args.requiredArg<ObjList>(0)
val row = (index.list[0] as ObjInt).value.toInt()
val col = (index.list[1] as ObjInt).value.toInt()
val data = (thisObj as ObjInstance).data as IntArray
ObjInt.of(data[row * 2 + col].toLong())
}
addFun("putAt") {
val index = args.requiredArg<ObjList>(0)
val value = args.requiredArg<ObjInt>(1).value.toInt()
val row = (index.list[0] as ObjInt).value.toInt()
val col = (index.list[1] as ObjInt).value.toInt()
val data = (thisObj as ObjInstance).data as IntArray
data[row * 2 + col] = value
ObjVoid
}
}
```
Usage from Lyng:
```lyng
val g = Grid()
g[0, 1] = 42
assertEquals(42, g[0, 1])
```
Important rule: multiple selectors inside brackets are packed into one index object.
So:
- `x[i]` passes `i`
- `x[i, j]` passes a `List` containing `[i, j]`
- `x[i, j, k]` passes `[i, j, k]`
This applies equally to:
- Kotlin-backed classes
- Lyng classes overriding `getAt`
- `dynamic { get { ... } set { ... } }`
If you want multi-axis slicing semantics, decode that list yourself in `getAt`.
### 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 session = EvalSession()
val scope = session.getScope()
val myClass = ObjClass("MyClass")
// Add a read-only field (constant)
@ -311,9 +122,6 @@ 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)
```
@ -330,8 +138,6 @@ println(instance.count) // -> 5
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 session = EvalSession()
val scope = session.getScope()
val myClass = ObjClass("MyClass")
var internalValue: Long = 10
@ -347,16 +153,6 @@ myClass.addProperty(
}
)
// 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)
```
@ -368,172 +164,17 @@ 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.
Pure extern declarations use the simplified rule set:
- `extern class` / `extern object` are declaration-only ABI surfaces.
- Every member in their body is implicitly extern (you may still write `extern`, but it is redundant).
- Plain Lyng member implementations inside `extern class` / `extern object` are not allowed.
- Put Lyng behavior into regular classes or extension methods.
```lyng
// Lyng side (in a module)
class Counter {
extern var value: Int
extern fun inc(by: Int): Int
}
```
Note: members of `extern class` / `extern object` are treated as extern by default, so the compiler emits ABI slots that Kotlin bindings attach to. This applies to functions and properties bound via `addFun` / `addVal` / `addVar`.
Example of pure extern class declaration:
```lyng
extern class HostCounter {
var value: Int
fun inc(by: Int): Int
}
```
If you need Lyng-side convenience behavior, add it as an extension:
```lyng
fun HostCounter.bump() = inc(1)
```
```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 = { thisObj.readField(this, "value").value },
set = { v -> thisObj.writeField(this, "value", v) }
)
addFun("inc") {
val by = args.requiredArg<ObjInt>(0).value
val current = thisObj.readField(this, "value").value as ObjInt
val next = ObjInt(current.value + by)
thisObj.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.
As with class binding, you must first add/evaluate the Lyng declaration into that module scope, then bind Kotlin handlers.
```kotlin
// Kotlin side (binding)
val moduleScope = importManager.createModuleScope(Pos.builtIn, "bridge.obj")
// 1) Seed the module with the Lyng declaration first
moduleScope.eval("""
extern object HostObject {
extern fun add(a: Int, b: Int): Int
extern val status: String
extern var count: Int
}
""".trimIndent())
// 2) Then bind Kotlin implementations to that declared object
moduleScope.bindObject("HostObject") {
classData = "OK"
init { _ -> data = 0L }
addFun("add") {
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 = { ObjInt.of((thisObj as ObjInstance).data as Long) },
set = { value -> (thisObj as ObjInstance).data = (value as ObjInt).value }
)
}
```
Notes:
- Required order: declare/eval Lyng object in the module first, then call `bindObject(...)`.
This is the pattern covered by `BridgeBindingTest.testExternObjectBinding`.
- Members must be extern (explicitly, or implicitly via `extern object`) so the compiler emits ABI slots for Kotlin bindings.
- You can also bind by name/module via `LyngObjectBridge.bind(...)`.
Minimal `extern fun` example:
```kotlin
val moduleScope = importManager.createModuleScope(Pos.builtIn, "bridge.ping")
moduleScope.eval("""
extern object HostObject {
extern fun ping(): Int
}
""".trimIndent())
moduleScope.bindObject("HostObject") {
addFun("ping") { ObjInt.of(7) }
}
```
### 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 session = EvalSession()
val scope = session.getScope()
session.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 = session.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
The simplest approach: evaluate an expression that yields the value and convert it.
```kotlin
val session = EvalSession()
val scope = session.getScope()
val kotlinAnswer = session.eval("(1 + 2) * 3").toKotlin(scope) // -> 9 (Int)
val kotlinAnswer = scope.eval("(1 + 2) * 3").toKotlin(scope) // -> 9 (Int)
// After scripts manipulate your vars:
scope.addOrUpdateItem("name", ObjString("Lyng"))
session.eval("name = name + ' rocks!'")
val kotlinName = session.eval("name").toKotlin(scope) // -> "Lyng rocks!"
scope.eval("name = name + ' rocks!'")
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.
@ -546,20 +187,16 @@ There are two convenient patterns.
```kotlin
// Suppose Lyng defines: fun add(a, b) = a + b
val session = EvalSession()
val scope = session.getScope()
session.eval("fun add(a, b) = a + b")
scope.eval("fun add(a, b) = a + b")
val sum = session.eval("add(20, 22)").toKotlin(scope) // -> 42
val sum = scope.eval("add(20, 22)").toKotlin(scope) // -> 42
```
2) Call a Lyng function by name via a prepared call scope:
```kotlin
// Ensure the function exists in the scope
val session = EvalSession()
val scope = session.getScope()
session.eval("fun add(a, b) = a + b")
scope.eval("fun add(a, b) = a + b")
// Look up the function object
val addFn = scope.get("add")!!.value as Statement
@ -587,47 +224,27 @@ Key concepts:
Register a Kotlin‑built package:
```kotlin
import net.sergeych.lyng.bridge.*
import net.sergeych.lyng.obj.ObjInt
val session = EvalSession()
val scope = session.getScope()
val scope = Script.newScope()
// Access the import manager behind this scope
val im: ImportManager = scope.importManager
// Register a package "my.tools"
im.addPackage("my.tools") { module: ModuleScope ->
module.eval(
"""
extern val version: String
extern var status: String
extern fun triple(x: Int): Int
""".trimIndent()
)
val binder = module.globalBinder()
var status = "ready"
binder.bindGlobalVar(
name = "version",
get = { "1.0" }
)
binder.bindGlobalVar(
name = "status",
get = { status },
set = { status = it }
)
binder.bindGlobalFun1<Int>("triple") { x ->
ObjInt.of((x * 3).toLong())
// Expose symbols inside the module scope
module.addConst("version", ObjString("1.0"))
module.addFn<ObjInt>("triple") {
val x = args.firstAndOnly() as ObjInt
ObjInt(x.value * 3)
}
}
// Use it from Lyng
session.eval("""
scope.eval("""
import my.tools.*
val v = triple(14)
status = "busy"
""")
val v = session.eval("v").toKotlin(scope) // -> 42
val v = scope.eval("v").toKotlin(scope) // -> 42
```
Register a package from Lyng source text:
@ -641,27 +258,24 @@ val pkgText = """
scope.importManager.addTextPackages(pkgText)
session.eval("""
scope.eval("""
import math.extra.*
val s = sqr(12)
""")
val s = session.eval("s").toKotlin(scope) // -> 144
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
- To run code from a file, read it and pass to `session.eval(text)` or compile with `Compiler.compile(Source(fileName, text))`.
- 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.
- For isolation, prefer a fresh `EvalSession()` per request. Use `Scope.new()` / `Script.newScope()` when you specifically need low-level raw scopes or modules.
- For isolation, create fresh modules/scopes via `Scope.new()` or `Script.newScope()` when you need a clean environment per request.
```kotlin
// Preferred per-request runtime:
val isolatedSession = EvalSession()
// Low-level fresh module based on the default manager, without the standard prelude:
val isolatedScope = net.sergeych.lyng.Scope.new()
// Fresh module based on the default manager, without the standard prelude
val isolated = net.sergeych.lyng.Scope.new()
```
### 11) Tips and troubleshooting
@ -672,49 +286,6 @@ val isolatedScope = 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
val session = EvalSession()
val scope = session.getScope()
try {
session.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

@ -128,11 +128,9 @@ 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
@ -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 | |
| UnknownException | unexpected kotlin exception caught |
| | |
### Symbol resolution errors

View File

@ -1,146 +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
# Nullability checks for types
Use `is nullable` to check whether a type expression accepts `null`:
T is nullable
T !is nullable
This works with concrete and generic types:
fun describe<T>(x: T): String = when (T) {
nullable -> "nullable"
else -> "non-null"
}
Equivalent legacy form:
null is T
# 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
@ -27,6 +24,6 @@ See `docs/lyng_d_files.md` for `.lyng.d` syntax and examples.
- Alternatively, if/when the plugin is published to a marketplace, you will be able to install it
directly from the “Marketplace” tab (not yet available).
### [Download plugin v0.0.5-SNAPSHOT](https://lynglang.com/distributables/lyng-idea-0.0.5-SNAPSHOT.zip)
### [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

@ -1,32 +1,9 @@
# Json support
Lyng now has two distinct JSON-facing layers:
Since 1.0.5 we start adding JSON support. Versions 1,0,6* support serialization of the basic types, including lists and
maps, and simple classes. Multiple inheritance may produce incorrect results, it is work in progress.
- plain JSON projection:
- `Obj.toJson()`
- `Obj.toJsonString()`
- canonical JSON round-trip format:
- `Json.encode(value)`
- `Json.decode(text)`
- typed canonical JSON round-trip format:
- `Json.encodeAs(Type, value)`
- `Json.decodeAs(Type, text)`
Use the first when you need ordinary JSON for interop.
Use the second when you need Lyng value round-trip semantics through JSON text with no schema.
Use the third when both sides already know the Lyng type and you want the same round-trip semantics with fewer type
tags in the JSON.
This distinction is intentional:
- plain JSON projection is optimized for compatibility with ordinary JSON tooling
- canonical `Json.encode()` is optimized for semantic fidelity to Lyng and Lynon and stays self-describing
- typed canonical `Json.encodeAs()` is optimized for the same fidelity when the schema is provided externally
- these goals conflict for values such as sets, exceptions, singleton objects, buffers, and maps with non-string keys
## Plain JSON projection in Lyng
## Serialization in Lyng
// in lyng
assertEquals("{\"a\":1}", {a: 1}.toJsonString())
@ -43,19 +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 plain JSON serialization:
Note that mutable members are serialized:
import lyng.serialization
@ -70,8 +35,8 @@ Note that if you override plain JSON serialization:
assertEquals( "{\"custom\":true}", Point2(1,2).toJsonString() )
>>> void
Custom serialization of user classes is possible by overriding `toJsonObject`. It must return an object which is
serializable to JSON. Most often it is a map, but any object is accepted:
Custom serialization of user classes is possible by overriding `toJsonObject` method. It must return an object which is
serializable to Json. Most often it is a map, but any object is accepted, that makes it very flexible:
import lyng.serialization
@ -94,87 +59,12 @@ serializable to JSON. Most often it is a map, but any object is accepted:
Please note that `toJsonString` should be used to get serialized string representation of the object. Don't call
`toJsonObject` directly, it is not intended to be used outside the serialization library.
## Canonical Json round-trip format
`Json.encode()` and `Json.decode()` are now the JSON equivalents of `Lynon.encode()` and `Lynon.decode()`.
They still use JSON text, but they add Lyng-specific type tags where plain JSON would otherwise lose information.
When a map already fits ordinary JSON object rules, canonical JSON keeps that traditional object shape. In particular,
maps with string keys are still serialized as JSON objects, not as tagged entry lists.
Example:
```lyng
import lyng.serialization
import lyng.time
enum Color { Red, Green }
class Point(x,y) { var z = 42 }
val p = Point(1,2)
p.z = 99
val value = List(
p,
Map([1, "one"], ["two", 2]),
Set(1,2,3),
"hello".encodeUtf8(),
Date(2026,4,15),
Color.Green
)
assertEquals(value, Json.decode(Json.encode(value)))
```
The canonical `Json` format is intended for Lyng-to-Lyng transfer through JSON text.
The plain `toJson()` projection is intended for ordinary JSON interop.
Canonical `Json.encode()` should be read as the JSON analogue of `Lynon.encode()`: when Lynon already preserves a
Lyng distinction, canonical JSON tries to preserve it too, using tags only where ordinary JSON is insufficient.
## Typed canonical Json round-trip format
`Json.encodeAs(Type, value)` and `Json.decodeAs(Type, text)` use the same canonical rules, but with a declared target
type available during the whole traversal.
This changes one thing only: type tags may be omitted when the declared type is already exact enough to restore the
value unambiguously.
The same map rule still applies here: `Map<String, T>` stays a normal JSON object, while non-string-key maps fall back
to canonical entry encoding.
Example:
```lyng
import lyng.serialization
closed class Point(x: Int, y: Int)
closed class Segment(a: Point, b: Point)
val value = Segment(Point(0, 1), Point(2, 3))
val encoded = Json.encodeAs(Segment, value)
assertEquals("{\"a\":{\"x\":0,\"y\":1},\"b\":{\"x\":2,\"y\":3}}", encoded)
assertEquals(value, Json.decodeAs(Segment, encoded))
```
Subtype information is still preserved when the declared type is wider than the runtime one. For example, if a field is
declared as `Base` but contains `Derived`, canonical subtype tags remain in that field.
This is why the APIs are split:
- `toJson()` stays plain and interop-friendly
- `Json.encode()` stays fully self-describing and safe to decode without a schema
- `Json.encodeAs()` uses the supplied schema to reduce noise, but only where that schema is sufficient
## Kotlin side interfaces
The "Batteries included" principle is also applied to serialization.
- `Obj.toJson()` provides Kotlin `JsonElement` for the plain JSON projection
- `Obj.toJsonString()` provides plain JSON string representation
- `Obj.toJson()` provides Kotlin `JsonElement`
- `Obj.toJsonString()` provides Json string representation
- `Obj.decodeSerializableWith()` and `Obj.decodeSerializable()` allows to decode Lyng classes as Kotlin objects using
`kotlinx.serialization`:
@ -203,9 +93,10 @@ suspend inline fun <reified T> Obj.decodeSerializable(scope: Scope = Scope()) =
decodeSerializableWith<T>(serializer<T>(), scope)
```
Note that Lyng-to-Kotlin deserialization with `kotlinx.serialization` is based on the plain JSON projection,
not the canonical `Json.encode()` format. It uses `JsonElement` as the information carrier without formatting and
parsing actual JSON strings. This is why we use `Json.decodeFromJsonElement` instead of `Json.decodeFromString`.
Note that lyng-2-kotlin deserialization with `kotlinx.serialization` uses JsonElement as information carrier without
formatting and parsing actual Json strings. This is why we use `Json.decodeFromJsonElement` instead of
`Json.decodeFromString`. Such an approach gives satisfactory performance without writing and supporting custom
`kotlinx.serialization` codecs.
### Pitfall: JSON objects and Map<String, Any?>
@ -220,8 +111,7 @@ data class TestJson2(
@Test
fun deserializeMapWithJsonTest() = runTest {
val session = EvalSession()
val x = session.eval("""
val x = eval("""
import lyng.serialization
{ value: 1, inner: { "foo": 1, "bar": 2 }}
""".trimIndent()).decodeSerializable<TestJson2>()
@ -232,8 +122,7 @@ fun deserializeMapWithJsonTest() = runTest {
But what if your map has objects of different types? The approach of using polymorphism is partially applicable, but what to do with `{ one: 1, two: "two" }`?
The answer is simple: use `JsonObject` in your deserializable object. This class is capable of holding any JSON types
and structures:
The answer is pretty simple: use `JsonObject` in your deserializable object. This class is capable of holding any JSON types and structures and is sort of a silver bullet for such cases:
~~~kotlin
@Serializable
@ -243,8 +132,7 @@ data class TestJson3(
)
@Test
fun deserializeAnyMapWithJsonTest() = runTest {
val session = EvalSession()
val x = session.eval("""
val x = eval("""
import lyng.serialization
{ value: 12, inner: { "foo": 1, "bar": "two" }}
""".trimIndent()).decodeSerializable<TestJson3>()
@ -253,71 +141,27 @@ fun deserializeAnyMapWithJsonTest() = runTest {
~~~
## Supported shapes
### Plain JSON projection
# List of supported types
| Lyng type | JSON type | notes |
|-----------|-----------|-------------|
| `Int` | number | |
| `Real` | number | finite values only as plain numbers |
| `Real` | number | |
| `String` | string | |
| `Bool` | boolean | |
| `null` | null | |
| `Instant` | string | ISO8601 (1) |
| `List` | array | (2) |
| `Map` | object | string keys only |
| simple class instance | object | constructor fields + mutable vars |
| enum | string | entry name |
| `Map` | object | (2) |
### Canonical `Json.encode`
This format can also round-trip:
- maps with non-string keys
- sets
- immutable collections
- buffers and bit buffers
- class instances
- singleton objects
- enums
- exceptions
- `Date`, `Instant`, `DateTime`
- non-finite reals
- `void`
### Typed canonical `Json.encodeAs`
This format round-trips the same value space as canonical `Json.encode`, but it can emit simpler JSON for:
- closed classes and other exactly-known class fields
- enums when the enum type is known
- typed collections whose element types are known
- nested object graphs where declared field types are precise
It still falls back to canonical tagged encoding when exact runtime type information would otherwise be lost.
It does so by adding Lyng-specific type tags only when necessary.
## Kotlin-side extension point for more formats
Additional formats can be exported from Kotlin modules by subclassing `ObjSerializationFormatClass` and registering the
format in module scope with `bindSerializationFormat(...)`.
```kotlin
module.bindSerializationFormat(
object : ObjSerializationFormatClass("MyFormat") {
override suspend fun encodeValue(scope: Scope, value: Obj): Obj = ...
override suspend fun decodeValue(scope: Scope, encoded: Obj): Obj = ...
}
)
```
This makes `MyFormat.encode(...)` and `MyFormat.decode(...)` available from Lyng after importing the module.
(1)
: ISO8601 flavor `1970-05-06T06:00:00.000Z` is used; number of fractional digits depends on truncation on
`Instant`, see `Instant.truncateTo...` functions.
: ISO8601 flavor 1970-05-06T06:00:00.000Z in used; number of fractional digits depends on the truncation
on [Instant](time.md), see `Instant.truncateTo...` functions.
(2)
: Lists may contain any values serializable by the selected JSON layer.
: List may contain any objects serializable to Json.
(3)
: Map keys must be strings, map values may be any objects serializable to Json.

View File

@ -1,113 +0,0 @@
### lyng.io.console
`lyng.io.console` provides optional rich console support for terminal applications.
> **Note:** this module is part of `lyngio`. It must be explicitly installed into the import manager by host code.
>
> **CLI note:** the `lyng` CLI now installs `lyng.io.console` in its base scope by default, so scripts can simply `import lyng.io.console`.
#### Install in host
```kotlin
import net.sergeych.lyng.EvalSession
import net.sergeych.lyng.io.console.createConsoleModule
import net.sergeych.lyngio.console.security.PermitAllConsoleAccessPolicy
suspend fun initScope() {
val session = EvalSession()
val scope = session.getScope()
createConsoleModule(PermitAllConsoleAccessPolicy, scope)
}
```
#### Use in Lyng script
```lyng
import lyng.io.console
println("supported = " + Console.isSupported())
println("tty = " + Console.isTty())
println("ansi = " + Console.ansiLevel())
println("geometry = " + Console.geometry())
Console.write("hello\n")
Console.home()
Console.clear()
Console.moveTo(1, 1)
Console.clearLine()
Console.enterAltScreen()
Console.leaveAltScreen()
Console.setCursorVisible(true)
Console.flush()
```
#### Tetris sample
The repository includes a full interactive Tetris sample that demonstrates:
- alternate screen rendering
- raw keyboard input
- resize handling
- typed console events
![Lyng Tetris sample](/tetris.png)
Run it from the project root in a real TTY:
```bash
lyng examples/tetris_console.lyng
```
#### API
- `Console.isSupported(): Bool` — whether console control is available on this platform/runtime.
- `Console.isTty(): Bool` — whether output is attached to a TTY.
- `Console.ansiLevel(): ConsoleAnsiLevel``NONE`, `BASIC16`, `ANSI256`, `TRUECOLOR`.
- `Console.geometry(): ConsoleGeometry?``{columns, rows}` as typed object or `null`.
- `Console.details(): ConsoleDetails` — consolidated capability object.
- `Console.write(text: String)` — writes to console output.
- `Console.flush()` — flushes buffered output.
- `Console.home()` — moves cursor to top-left.
- `Console.clear()` — clears visible screen.
- `Console.moveTo(row: Int, column: Int)` — moves cursor to 1-based row/column.
- `Console.clearLine()` — clears current line.
- `Console.enterAltScreen()` — switch to alternate screen buffer.
- `Console.leaveAltScreen()` — return to normal screen buffer.
- `Console.setCursorVisible(visible: Bool)` — shows/hides cursor.
- `Console.events(): ConsoleEventStream` — endless iterable source of typed events: `ConsoleResizeEvent`, `ConsoleKeyEvent`.
- `Console.setRawMode(enabled: Bool): Bool` — requests raw input mode, returns `true` if changed.
#### Event Iteration
Use events from a loop, typically in a separate coroutine:
```lyng
launch {
for (ev in Console.events()) {
if (ev is ConsoleKeyEvent) {
// handle key
}
}
}
```
#### Event format
`Console.events()` emits `ConsoleEvent` with:
- `type: ConsoleEventType``UNKNOWN`, `RESIZE`, `KEY_DOWN`, `KEY_UP`
Additional fields:
- `ConsoleResizeEvent`: `columns`, `rows`
- `ConsoleKeyEvent`: `key`, `code`, `ctrl`, `alt`, `shift`, `meta`
#### Security policy
The module uses `ConsoleAccessPolicy` with operations:
- `WriteText(length)`
- `ReadEvents`
- `SetRawMode(enabled)`
For permissive mode, use `PermitAllConsoleAccessPolicy`.

View File

@ -1,565 +0,0 @@
# lyng.io.db — SQL database access for Lyng scripts
This module provides the portable SQL database contract for Lyng. The current shipped providers are SQLite via `lyng.io.db.sqlite` and a JVM-only JDBC bridge via `lyng.io.db.jdbc`.
> **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.
---
## Install the module into a Lyng session
For SQLite-backed database access, install both the generic DB module and the SQLite provider:
```kotlin
import net.sergeych.lyng.EvalSession
import net.sergeych.lyng.Scope
import net.sergeych.lyng.io.db.createDbModule
import net.sergeych.lyng.io.db.sqlite.createSqliteModule
suspend fun bootstrapDb() {
val session = EvalSession()
val scope: Scope = session.getScope()
createDbModule(scope)
createSqliteModule(scope)
session.eval("""
import lyng.io.db
import lyng.io.db.sqlite
""".trimIndent())
}
```
`createSqliteModule(...)` also registers the `sqlite:` scheme for generic `openDatabase(...)`.
For JVM JDBC-backed access, install the JDBC provider as well:
```kotlin
import net.sergeych.lyng.EvalSession
import net.sergeych.lyng.Scope
import net.sergeych.lyng.io.db.createDbModule
import net.sergeych.lyng.io.db.jdbc.createJdbcModule
suspend fun bootstrapJdbc() {
val session = EvalSession()
val scope: Scope = session.getScope()
createDbModule(scope)
createJdbcModule(scope)
session.eval("""
import lyng.io.db
import lyng.io.db.jdbc
""".trimIndent())
}
```
`createJdbcModule(...)` registers `jdbc:`, `h2:`, `postgres:`, and `postgresql:` for `openDatabase(...)`.
---
## Using from Lyng scripts
Typed SQLite open helper:
```lyng
import lyng.io.db.sqlite
val db = openSqlite(":memory:")
val userCount = db.transaction { tx ->
tx.execute("create table user(id integer primary key autoincrement, name text not null)")
tx.execute("insert into user(name) values(?)", "Ada")
tx.execute("insert into user(name) values(?)", "Linus")
tx.select("select count(*) as count from user").toList()[0]["count"]
}
assertEquals(2, userCount)
```
Generic provider-based open:
```lyng
import lyng.io.db
import lyng.io.db.sqlite
val db = openDatabase(
"sqlite:./app.db",
Map(
"foreignKeys" => true,
"busyTimeoutMillis" => 5000
)
)
```
JVM JDBC open with H2:
```lyng
import lyng.io.db.jdbc
val db = openH2("mem:demo;DB_CLOSE_DELAY=-1")
val names = db.transaction { tx ->
tx.execute("create table person(id bigint auto_increment primary key, name varchar(120) not null)")
tx.execute("insert into person(name) values(?)", "Ada")
tx.execute("insert into person(name) values(?)", "Linus")
tx.select("select name from person order by id").toList()
}
assertEquals("Ada", names[0]["name"])
assertEquals("Linus", names[1]["name"])
```
Generic JDBC open through `openDatabase(...)`:
```lyng
import lyng.io.db
import lyng.io.db.jdbc
val db = openDatabase(
"jdbc:h2:mem:demo2;DB_CLOSE_DELAY=-1",
Map()
)
val answer = db.transaction { tx ->
tx.select("select 42 as answer").toList()[0]["answer"]
}
assertEquals(42, answer)
```
PostgreSQL typed open:
```lyng
import lyng.io.db.jdbc
val db = openPostgres(
"jdbc:postgresql://127.0.0.1/appdb",
"appuser",
"secret"
)
val titles = db.transaction { tx ->
tx.execute("create table if not exists task(id bigserial primary key, title text not null)")
tx.execute("insert into task(title) values(?)", "Ship JDBC provider")
tx.execute("insert into task(title) values(?)", "Test PostgreSQL path")
tx.select("select title from task order by id").toList()
}
assertEquals("Ship JDBC provider", titles[0]["title"])
```
Nested transactions use real savepoint semantics:
```lyng
import lyng.io.db
import lyng.io.db.sqlite
val db = openSqlite(":memory:")
db.transaction { tx ->
tx.execute("create table item(id integer primary key autoincrement, name text not null)")
tx.execute("insert into item(name) values(?)", "outer")
try {
tx.transaction { inner ->
inner.execute("insert into item(name) values(?)", "inner")
throw IllegalStateException("rollback nested")
}
} catch (_: IllegalStateException) {
}
assertEquals(1, tx.select("select count(*) as count from item").toList()[0]["count"])
}
```
Intentional rollback without treating it as a backend failure:
```lyng
import lyng.io.db
import lyng.io.db.sqlite
val db = openSqlite(":memory:")
assertThrows(RollbackException) {
db.transaction { tx ->
tx.execute("create table item(id integer primary key autoincrement, name text not null)")
tx.execute("insert into item(name) values(?)", "temporary")
throw RollbackException("stop here")
}
}
```
---
## Runnable serialization sample
A complete runnable example is in [examples/sqlite_serialization.lyng](../examples/sqlite_serialization.lyng).
It uses:
- `@DbJson`
- `@DbLynon`
- `@DbExcept`
- `@cols(...)`, `@vals(...)`, `@set(...)`
- `decodeAs<T>()`
The current direct read form that works under `jlyng` is:
```lyng
tx.select("select * from item where id = ?", 1).decodeAs<Item>().first
```
If we want a shorter form such as:
```lyng
tx.selectAllAs<Item>("item where id = ?", 1).first
```
it should be added as a built-in `SqlTransaction` API. A pure Lyng generic wrapper around `decodeAs<T>()` does not currently preserve `T` reliably enough under `jlyng`.
---
## Portable API
### `Database`
- `transaction(block)` — opens a transaction, commits on normal exit, rolls back on uncaught failure.
### `SqlTransaction`
- `select(clause, params...)` — execute a statement whose primary result is a row set.
- `execute(clause, params...)` — execute a side-effect statement and return `ExecutionResult`.
- `transaction(block)` — nested transaction with real savepoint semantics.
`select(...)` and `execute(...)` also support SQL object-expansion macros for declaration-driven writes:
- `@cols(?1)` — expand object argument `?1` to a comma-separated column list
- `@vals(?1)` — expand object argument `?1` to matching placeholders and bind values
- `@set(?1)` — expand object argument `?1` to `column = ?` pairs and bind values
Each macro also supports an optional clause-local exclusion list:
```lyng
tx.execute("update item set @set(?1 except: \"id\", \"createdAt\") where id = ?2", item, item.id)
```
Example:
```lyng
tx.execute("insert into item(@cols(?1)) values(@vals(?1))", item)
tx.execute("update item set @set(?1) where id = ?2", item, item.id)
```
When a clause uses any of these macros, non-expanded scalar parameters in the same SQL string must use explicit indexed placeholders such as `?2`, `?3`, and so on.
### `ResultSet`
- `columns` — positional `SqlColumn` metadata, available before iteration.
- `size()` — result row count.
- `isEmpty()` — fast emptiness check where possible.
- `iterator()` — normal row iteration while the transaction is active.
- `toList()` — materialize detached `SqlRow` snapshots that may be used after the transaction ends.
- `decodeAs<T>()` — transaction-scoped iterable view that decodes each row into `T`.
### `SqlRow`
- `row[index]` — zero-based positional access.
- `row["columnName"]` — case-insensitive lookup by output column label.
- `row.decodeAs<T>()` — decode one row into a typed Lyng value.
Name-based access fails with `SqlUsageException` if the name is missing or ambiguous.
### `DbFieldAdapter`
Custom DB field projection hook used by `@DbDecodeWith(...)` and `@DbSerializeWith(...)`.
- `decode(rawValue, column, row, targetType)` — adapt one raw DB field value to a Lyng value for the requested target type.
- `encode(value, targetType)` — adapt one Lyng value to a direct DB-bindable value for SQL object expansion.
Use `@DbDecodeWith(adapter)` on class constructor parameters and class-body fields/properties that participate in `decodeAs<T>()`.
Use `@DbSerializeWith(adapter)` on constructor parameters and class-body fields/properties that participate in `@cols(...)`, `@vals(...)`, and `@set(...)` object expansion.
Annotation arguments are evaluated once when the declaration is created, and the resulting adapter instance is retained in declaration metadata.
### `ExecutionResult`
- `affectedRowsCount`
- `getGeneratedKeys()`
Statements that return rows directly, such as `... returning ...`, should use `select(...)`, not `execute(...)`.
---
## Value mapping
Portable bind values:
- `null`
- `Bool`
- `Int`, `Double`, `Decimal`
- `String`
- `Buffer`
- `Date`, `DateTime`, `Instant`
Unsupported parameter values fail with `SqlUsageException`.
SQL object-expansion write rules:
- constructor parameters participate in projection by declaration order
- matching serializable class-body fields/properties also participate
- `@Transient` fields are excluded automatically
- `@DbExcept` fields are excluded automatically
- `except:` excludes additional fields for one specific macro use
- direct DB-bindable values are written as-is
- `@DbJson` fields are encoded as canonical JSON text
- `@DbLynon` fields are encoded as Lynon binary
- `@DbSerializeWith(adapter)` fields are encoded through the adapter
- unannotated non-bindable object fields fail with `SqlUsageException`
Write-side encoding is intentionally explicit. The runtime does not try to infer target DB column types from SQL text or backend metadata during statement preparation.
Example:
```lyng
import lyng.io.db
import lyng.io.db.sqlite
class Payload(name: String, count: Int)
object TrimAdapter: DbFieldAdapter {
override fun encode(value, targetType) =
when(value) {
null -> null
else -> value.toString().trim()
}
}
class Item(
id: Int,
@DbSerializeWith(TrimAdapter) title: String,
@DbJson meta: Payload,
@DbLynon state: Payload
) {
var note: String = ""
@DbExcept var cache: String = ""
}
val db = openSqlite(":memory:")
val restored = db.transaction { tx ->
tx.execute(
"create table item(id integer not null, title text not null, meta text not null, state blob not null, note text not null)"
)
val item = Item(1, " first ", Payload("json", 10), Payload("bin", 20))
item.note = "created"
item.cache = "not stored"
tx.execute("insert into item(@cols(?1)) values(@vals(?1))", item)
item.title = " second "
item.meta = Payload("json2", 11)
item.state = Payload("bin2", 21)
item.note = "updated"
tx.execute(
"update item set @set(?1 except: \"id\") where id = ?2",
item,
item.id
)
tx.select("select id, title, meta, state, note from item").decodeAs<Item>().first
}
assertEquals("second", restored.title)
assertEquals("json2", restored.meta.name)
assertEquals(21, restored.state.count)
assertEquals("updated", restored.note)
```
This example shows:
- `@DbSerializeWith(...)` trimming a string before write
- `@DbJson` storing structured data in a text column
- `@DbLynon` storing structured data in a binary column
- `@DbExcept` excluding a field from automatic projection
- `@set(... except: "id")` skipping one field for an update clause
- `decodeAs<Item>()` reconstructing the object on read
Portable result metadata categories:
- `Binary`
- `String`
- `Int`
- `Double`
- `Decimal`
- `Bool`
- `Date`
- `DateTime`
- `Instant`
Typed row decode rules:
- object/class targets map constructor parameters by column label, case-insensitively
- remaining matching serializable mutable fields are assigned after constructor call
- `@DbDecodeWith(adapter)` on a constructor parameter or class-body field/property takes precedence over built-in JSON/Lynon decoding
- `@DbDecodeWith(adapter)` must receive exactly one adapter instance implementing `DbFieldAdapter`
- adapter output must match the target member type or decoding fails with `SqlUsageException`
- missing required non-null constructor fields fail
- defaulted or nullable constructor fields may be omitted from the result
- extra result columns currently fail in strict mode
- if a row has exactly one column, that value may be decoded directly as the requested target type
- JSON-like native column types (`json`, `jsonb`) are decoded through typed canonical `Json` when the target type is not `String`
- binary columns are decoded through `Lynon` when the target type is not `Buffer`
- `Buffer` targets keep the raw binary payload without Lynon decoding
- plain text columns are not implicitly treated as JSON
For temporal types, see [time functions](time.md).
---
## SQLite provider
`lyng.io.db.sqlite` currently provides the first concrete backend.
Typed helper:
```lyng
openSqlite(
path: String,
readOnly: Bool = false,
createIfMissing: Bool = true,
foreignKeys: Bool = true,
busyTimeoutMillis: Int = 5000
): Database
```
Accepted generic URL forms:
- `sqlite::memory:`
- `sqlite:relative/path.db`
- `sqlite:/absolute/path.db`
Supported `openDatabase(..., extraParams)` keys for SQLite:
- `readOnly: Bool`
- `createIfMissing: Bool`
- `foreignKeys: Bool`
- `busyTimeoutMillis: Int`
SQLite write/read policy in v1:
- `Bool` writes as `0` / `1`
- `Decimal` writes as canonical text
- `Date` writes as `YYYY-MM-DD`
- `DateTime` writes as ISO local timestamp text without timezone
- `Instant` writes as ISO UTC timestamp text with explicit timezone marker
- `TIME*` values stay `String`
- `TIMESTAMP` / `DATETIME` reject timezone-bearing stored text
Open-time validation failures:
- malformed URL or bad option shape -> `IllegalArgumentException`
- runtime open failure -> `DatabaseException`
## JDBC provider
`lyng.io.db.jdbc` is currently implemented on the JVM target only. The `lyngio-jvm` artifact bundles and explicitly loads these JDBC drivers:
- SQLite
- H2
- PostgreSQL
Typed helpers:
```lyng
openJdbc(
connectionUrl: String,
user: String? = null,
password: String? = null,
driverClass: String? = null,
properties: Map<String, Object?>? = null
): Database
openH2(
connectionUrl: String,
user: String? = null,
password: String? = null,
properties: Map<String, Object?>? = null
): Database
openPostgres(
connectionUrl: String,
user: String? = null,
password: String? = null,
properties: Map<String, Object?>? = null
): Database
```
Accepted generic URL forms:
- `jdbc:h2:mem:test;DB_CLOSE_DELAY=-1`
- `h2:mem:test;DB_CLOSE_DELAY=-1`
- `jdbc:postgresql://localhost/app`
- `postgres://localhost/app`
- `postgresql://localhost/app`
Supported `openDatabase(..., extraParams)` keys for JDBC:
- `driverClass: String`
- `user: String`
- `password: String`
- `properties: Map<String, Object?>`
Behavior notes for the JDBC bridge:
- the portable `Database` / `SqlTransaction` API stays the same as for SQLite
- nested transactions use JDBC savepoints
- JDBC connection properties are built from `user`, `password`, and `properties`
- `properties` values are stringified before being passed to JDBC
- statements with row-returning clauses still must use `select(...)`, not `execute(...)`
Platform support for this provider:
- `lyng.io.db.jdbc` — JVM only
- `openH2(...)` — works out of the box with `lyngio-jvm`
- `openPostgres(...)` — driver included, but an actual PostgreSQL server is still required
PostgreSQL-specific notes:
- `openPostgres(...)` accepts either a full JDBC URL or shorthand forms such as `//localhost/app`
- local peer/trust setups may use an empty password string
- generated keys work with PostgreSQL `bigserial` / identity columns through `ExecutionResult.getGeneratedKeys()`
- for reproducible automated tests, prefer a disposable PostgreSQL instance such as Docker/Testcontainers instead of a long-lived shared server
---
## Lifetime rules
`ResultSet` is valid only while its owning transaction is active.
`SqlRow` values are detached snapshots once materialized, so this pattern is valid:
```lyng
val rows = db.transaction { tx ->
tx.select("select name from person order by id").toList()
}
assertEquals("Ada", rows[0]["name"])
```
This means:
- do not keep `ResultSet` objects after the transaction block returns
- materialize rows with `toList()` inside the transaction when they must outlive it
- the iterable returned by `decodeAs<T>()` is also transaction-scoped
- decoded objects produced while iterating `decodeAs<T>()` are detached ordinary Lyng values
The same rule applies to generated keys from `ExecutionResult.getGeneratedKeys()`: the `ResultSet` is transaction-scoped, but rows returned by `toList()` are detached.
---
## Platform support
- `lyng.io.db` — generic contract, available when host code installs it
- `lyng.io.db.sqlite` — implemented on JVM and Linux Native in the current release tree
- `lyng.io.db.jdbc` — implemented on JVM in the current release tree
For the broader I/O overview, see [lyngio overview](lyngio.md).

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 {
@ -39,27 +39,19 @@ This brings in:
---
#### Install the module into a Lyng session
#### Install the module into a Lyng Scope
The filesystem module is not installed automatically. The preferred host runtime is `EvalSession`: create the session, get its underlying scope, install the module there, and execute scripts through the session. You can customize access control via `FsAccessPolicy`.
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.EvalSession
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
suspend fun bootstrapFs() {
val session = EvalSession()
val scope: Scope = session.getScope()
val installed: Boolean = createFs(PermitAllAccessPolicy, scope)
// installed == true on first registration in this ImportManager, false on repeats
// In scripts (or via session.eval), import the module to use its symbols:
session.eval("import lyng.io.fs")
}
// In scripts (or via scope.eval), import the module to use its symbols:
scope.eval("import lyng.io.fs")
```
You can install with a custom policy too (see Access policy below).
@ -189,7 +181,7 @@ val denyWrites = object : FsAccessPolicy {
}
createFs(denyWrites, scope)
session.eval("import lyng.io.fs")
scope.eval("import lyng.io.fs")
```
Composite operations like `copy` and `move` are checked as a set of primitives (e.g., `OpenRead(src)` + `Delete(dst)` if overwriting + `CreateFile(dst)` + `OpenWrite(dst)`).

View File

@ -1,164 +0,0 @@
# lyng.io.html
`lyng.io.html` provides a pure Lyng HTML builder DSL. It uses Lyng context
receiver extensions, so text can be appended with `+"text"` inside tag blocks
without global builder state.
Host code installs the package from `lyngio` with `createHtmlModule(...)`:
```kotlin
val scope = Script.newScope()
createHtmlModule(scope.importManager)
```
Lyng code can then import it:
```lyng
import lyng.io.html
val page = html {
head {
title { +"Demo" }
}
body {
nav {
a(href: "/") { +"Home" }
}
h3 { +"Heading 3" }
p {
attr("data-id", 123)
+"Text is escaped: <safe>"
}
img(src: "/logo.png", alt: "Logo")
}
}
```
`html { ... }` returns a `String` beginning with `<!doctype html>`.
## Escaping
Text appended with unary `+` is HTML-escaped:
```lyng
html {
body {
p { +"Text & <more>" }
}
}
```
produces:
```html
<!doctype html><html><body><p>Text &amp; &lt;more&gt;</p></body></html>
```
Attribute values are escaped with HTML attribute rules:
```lyng
p {
attr("data-x", "\"quoted\" & <tag>")
+"content"
}
```
Use `raw(...)` only for trusted markup:
```lyng
div {
raw("<span>already escaped or trusted</span>")
}
```
## Tag Helpers
Current tag helpers cover common structural tags (`head`, `body`, `main`,
`section`, `article`, `header`, `footer`, `nav`, `div`, `span`, `p`), headings
(`h1` through `h6`), lists (`ul`, `ol`, `li`), and text/code tags (`strong`,
`em`, `code`, `pre`, `script`, `style`).
```lyng
body {
main {
section {
h2 { +"News" }
p { +"First item" }
}
}
}
```
Common void tags are also available: `meta`, `link`, `img`, `br`, and `input`.
```lyng
head {
meta { attr("charset", "utf-8") }
link {
attr("rel", "stylesheet")
attr("href", "/site.css")
}
}
```
## Attributes
Use `attr(name, value)` inside a tag block to set an escaped attribute value.
`id(...)` and `classes(...)` are small aliases:
```lyng
div {
id("root")
classes("app shell")
}
```
Use `flag(name)` for boolean attributes:
```lyng
input {
attr("type", "checkbox")
flag("checked")
}
```
## Convenience Helpers
Convenience helpers include `metaCharset()`, `stylesheet(href)`,
`a(href) { ... }`, `img(src, alt)`, and `input(type, name, value)`.
```lyng
head {
metaCharset()
stylesheet("/site.css")
}
body {
nav {
a(href: "/home") { +"Home" }
}
img(src: "/logo.png", alt: "Logo & mark")
input(type: "hidden", name: "token", value: "abc")
}
```
## Generic Elements
Use `tag(name) { ... }` and `voidTag(name) { ... }` for elements that do not
have dedicated helpers yet:
```lyng
body {
tag("custom-element") {
flag("hidden")
+"Secret"
}
voidTag("source") {
attr("srcset", "/image.webp")
attr("type", "image/webp")
}
}
```
These helpers are intentionally simple escape hatches. Prefer a dedicated helper
when one exists because it can encode safer defaults and clearer parameter names.

View File

@ -1,181 +0,0 @@
# lyng.io.http — HTTP/HTTPS client for Lyng scripts
This module provides a compact HTTP client API for Lyng scripts. It is implemented in `lyngio` and backed by Ktor on supported runtimes.
> **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.
>
> **Shared type note:** `HttpHeaders` is also available from `lyng.io.http.types` when host code wants the reusable value type without relying on the HTTP client module itself.
---
## 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 also use the Lyng Maven repository described in `lyng.io.fs`.
---
## Install the module into a Lyng session
The HTTP module is not installed automatically. Install it into the session scope and provide a policy.
Kotlin (host) bootstrap example:
```kotlin
import net.sergeych.lyng.EvalSession
import net.sergeych.lyng.Scope
import net.sergeych.lyng.io.http.createHttpModule
import net.sergeych.lyngio.http.security.PermitAllHttpAccessPolicy
suspend fun bootstrapHttp() {
val session = EvalSession()
val scope: Scope = session.getScope()
createHttpModule(PermitAllHttpAccessPolicy, scope)
session.eval("import lyng.io.http")
}
```
---
## Using from Lyng scripts
Simple GET:
import lyng.io.http
val r = Http.get(HTTP_TEST_URL + "/hello")
[r.status, r.text()]
>>> [200,hello from test]
Headers and response header access:
import lyng.io.http
val r = Http.get(HTTP_TEST_URL + "/headers")
[r.headers["X-Reply"], r.headers.getAll("X-Reply").size, r.text()]
>>> [one,2,header demo]
Programmatic request object:
import lyng.io.http
val q = HttpRequest()
q.method = "POST"
q.url = HTTP_TEST_URL + "/echo"
q.headers = Map("Content-Type" => "text/plain")
q.bodyText = "ping"
val r = Http.request(q)
r.text()
>>> "POST:ping"
HTTPS GET:
import lyng.io.http
val r = Http.get(HTTPS_TEST_URL + "/hello")
[r.status, r.text()]
>>> [200,hello from test]
---
## API reference
### `Http` (static methods)
- `isSupported(): Bool` — Whether HTTP client support is available on the current runtime.
- `request(req: HttpRequest): HttpResponse` — Execute a request described by a mutable request object.
- `get(url: String, headers...): HttpResponse` — Convenience GET request.
- `post(url: String, bodyText: String = "", contentType: String? = null, headers...): HttpResponse` — Convenience text POST request.
- `postBytes(url: String, body: Buffer, contentType: String? = null, headers...): HttpResponse` — Convenience binary POST request.
For convenience methods, `headers...` accepts:
- `MapEntry`, e.g. `"Accept" => "text/plain"`
- 2-item lists, e.g. `["Accept", "text/plain"]`
### `HttpRequest`
- `method: String`
- `url: String`
- `headers: Map<String, String>`
- `bodyText: String?`
- `bodyBytes: Buffer?`
- `timeoutMillis: Int?`
Only one of `bodyText` and `bodyBytes` should be set.
### `HttpResponse`
- `status: Int`
- `statusText: String`
- `headers: HttpHeaders`
- `text(): String`
- `bytes(): Buffer`
Response body decoding is cached inside the response object.
### `HttpHeaders`
`HttpHeaders` behaves like `Map<String, String>` for the first value of each header name and additionally exposes:
- `get(name: String): String?`
- `getAll(name: String): List<String>`
- `names(): List<String>`
Header lookup is case-insensitive.
---
## Security policy
The module uses `HttpAccessPolicy` to authorize requests before they are sent.
- `HttpAccessPolicy` — interface for custom policies
- `PermitAllHttpAccessPolicy` — allows all requests
- `HttpAccessOp.Request(method, url)` — operation checked by the policy
Example restricted policy in Kotlin:
```kotlin
import net.sergeych.lyngio.fs.security.AccessContext
import net.sergeych.lyngio.fs.security.AccessDecision
import net.sergeych.lyngio.fs.security.Decision
import net.sergeych.lyngio.http.security.HttpAccessOp
import net.sergeych.lyngio.http.security.HttpAccessPolicy
val allowLocalOnly = object : HttpAccessPolicy {
override suspend fun check(op: HttpAccessOp, ctx: AccessContext): AccessDecision =
when (op) {
is HttpAccessOp.Request ->
if (
op.url.startsWith("http://127.0.0.1:") ||
op.url.startsWith("https://127.0.0.1:") ||
op.url.startsWith("http://localhost:") ||
op.url.startsWith("https://localhost:")
)
AccessDecision(Decision.Allow)
else
AccessDecision(Decision.Deny, "only local HTTP/HTTPS requests are allowed")
}
}
```
---
## Platform support
- **JVM:** supported
- **Android:** supported via the Ktor CIO client backend
- **JS:** supported via the Ktor JS client backend
- **Linux native:** supported via the Ktor Curl client backend
- **Windows native:** supported via the Ktor WinHttp client backend
- **Apple native:** supported via the Ktor Darwin client backend
- **Other targets:** may report unsupported; use `Http.isSupported()` before relying on it

View File

@ -1,446 +0,0 @@
# `lyng.io.http.server` - Minimal HTTP/1.1 And WebSocket Server
This module provides a small server-side HTTP API for Lyng scripts. It is implemented in `lyngio` on top of the existing TCP layer and is intended for embedded tools, local services, test fixtures, and lightweight app backends.
It supports:
- HTTP/1.1 request parsing
- keep-alive
- exact-path routing
- regex routing
- path-template routing with named parameters
- websocket upgrade and server-side sessions
It does not aim to replace a full reverse proxy. Typical deployment is behind nginx, Caddy, or another frontend that handles TLS and public-facing edge concerns.
> **Security note:** this module uses the same `NetAccessPolicy` capability model as raw TCP sockets. If scripts are allowed to listen on TCP, they can host an HTTP server.
## Install The Module Into A Lyng Session
Kotlin bootstrap example:
```kotlin
import net.sergeych.lyng.EvalSession
import net.sergeych.lyng.Scope
import net.sergeych.lyng.io.http.server.createHttpServerModule
import net.sergeych.lyngio.net.security.PermitAllNetAccessPolicy
suspend fun bootstrapHttpServer() {
val session = EvalSession()
val scope: Scope = session.getScope()
createHttpServerModule(PermitAllNetAccessPolicy, scope)
session.eval("import lyng.io.http.server")
}
```
## RequestContext Sugar
Route handlers use `RequestContext` as the receiver, so inside handlers you normally write direct calls such as:
- `jsonBody<T>()`
- `respondJson(...)`
- `respondHtml { ... }`
- `respondText(...)`
- `setHeader(...)`
- `request.path`
- `routeParams["id"]`
This keeps ordinary HTTP endpoints compact and avoids passing an explicit request or exchange parameter through every route lambda.
## HTML Response Sugar
Use `respondHtml { ... }` to render an HTML document with the `lyng.io.html` DSL and send it as `text/html; charset=utf-8`.
```lyng
import lyng.io.http.server
import lyng.io.html
val server = HttpServer()
server.get("/") {
respondHtml {
head {
title { +"Lyng status" }
}
body {
h3 { +"Service is running" }
p { +"Path: ${request.path}" }
}
}
}
server.listen(8080, "127.0.0.1")
```
Pass `code:` when the route should return a non-200 status:
```lyng
server.get("/accepted") {
respondHtml(code: 202) {
body { h3 { +"Accepted" } }
}
}
```
## JSON API Sugar
For ordinary JSON APIs, `RequestContext` includes two primary helpers:
- `jsonBody<T>()` decodes the request body with typed `Json.decodeAs(...)`
- `respondJson(body, status = 200)` sets JSON content type and responds with plain `toJsonString()`
These helpers intentionally use ordinary JSON projection for HTTP interop, not canonical `Json.encode(...)`.
### Typed JSON POST With Route Params
```lyng
import lyng.io.http.server
closed class CreateResultRequest(title: String, score: Int)
closed class CreateResultResponse(id: String, userId: String, title: String, score: Int)
val server = HttpServer()
server.postPath("/api/users/{userId}/results") {
val req = jsonBody<CreateResultRequest>()
if (req.title.isBlank()) {
respondJson({ error: "title must not be empty" }, 400)
return
}
respondJson(
CreateResultResponse("r-101", routeParams["userId"], req.title, req.score),
201
)
}
server.listen(8080, "127.0.0.1")
```
### JSON Response With Route Params
```lyng
import lyng.io.http.server
val server = HttpServer()
server.getPath("/api/users/{id}") {
respondJson({
id: routeParams["id"],
path: request.path,
ok: true
})
}
server.listen(8080, "127.0.0.1")
```
## Request And Route Data
`ServerRequest` exposes parsed HTTP request data:
- `method: String`
- `target: String`
- `path: String`
- `pathParts: List<String>`
- `queryString: String?`
- `query: Map<String, String>`
- `headers: HttpHeaders`
- `body: Buffer`
`RequestContext` exposes routing context and response controls:
- `request: ServerRequest`
- `routeMatch: RegexMatch?`
- `routeParams: Map<String, String>`
- `jsonBody<T>()`
- `respond(...)`
- `respondText(...)`
- `respondJson(body, status = 200)`
- `respondHtml(code: 200) { ... }`
- `setHeader(...)`
- `addHeader(...)`
- `acceptWebSocket(...)`
For exact routes, `routeMatch` is `null` and `routeParams` is empty.
For regex routes, `routeMatch` is set and `routeParams` is empty.
For path-template routes, both `routeMatch` and `routeParams` are set.
## Reusable Routers
`Router` collects the same route kinds as `HttpServer`, but does not listen on sockets by itself.
Mount it into `HttpServer` or another `Router`.
```lyng
import lyng.io.http.server
val api = Router()
api.get("/health") {
respondText(200, "ok")
}
val users = Router()
users.getPath("/users/{id}") {
respondJson({ id: routeParams["id"] })
}
api.mount(users)
val server = HttpServer()
server.mount(api)
server.listen(8080, "127.0.0.1")
```
Mounted routers reuse the built-in server router. They are configuration-time composition, not an extra per-request Lyng dispatch layer.
## WebSocket Routes
You can route websocket upgrades by exact path, regex, or path template.
```lyng
server.ws("/chat") { ws ->
ws.sendText("hello")
ws.close()
}
server.wsPath("/ws/{room}") { ws ->
ws.sendText("room=" + routeParams["room"])
ws.close()
}
```
A websocket handler runs only for requests that actually ask for websocket upgrade. Ordinary HTTP requests to the same path are not treated as websocket sessions.
### Choosing Between `ws(...)` And `acceptWebSocket(...)`
Use `server.ws(...)` or `server.wsPath(...)` when the route is always a websocket endpoint.
Use `acceptWebSocket(...)` inside a normal HTTP handler when the same route may inspect the request first and then decide whether to upgrade.
```lyng
server.get("/maybe-upgrade") {
if (!request.isWebSocketUpgrade()) {
respondText(400, "websocket upgrade required")
return
}
acceptWebSocket { ws ->
ws.sendText("connected")
ws.close()
}
}
```
### Reading Incoming Messages
Inside a websocket handler, call `ws.receive()` to wait for the next application message.
What `receive()` returns:
- `WsMessage` for the next text or binary message.
- `null` after the client sends a close frame.
- `null` after the socket is already closed and no more frames can arrive.
What reaches Lyng code:
- Text frames become `WsMessage(isText = true, text = ...)`.
- Binary frames become `WsMessage(isText = false, data = ...)`.
- Fragmented websocket messages are reassembled before they are returned.
- Ping and pong control frames are handled internally and do not appear in Lyng.
- A client close frame is answered by the server close handshake, then `receive()` returns `null`.
Typical server receive loop:
```lyng
import lyng.buffer
server.ws("/echo") { ws ->
while (true) {
val msg = ws.receive() ?: break
if (msg.isText) {
ws.sendText("echo:" + msg.text)
} else {
ws.sendBytes(msg.data as Buffer)
}
}
}
```
### Sending Outgoing Messages
Use:
- `ws.sendText(text)` for text messages.
- `ws.sendBytes(data)` for binary messages.
Example:
```lyng
import lyng.buffer
server.ws("/push") { ws ->
ws.sendText("ready")
ws.sendBytes(Buffer(1, 2, 3))
ws.close()
}
```
Send behavior:
- Each call sends one websocket message.
- The server API does not expose frame-by-frame streaming.
- Once the session is closed, send calls fail with a websocket error.
### What Happens When The Connection Closes
There are three practical cases:
1. The client closes first.
The runtime replies with a close frame, releases the socket, and `receive()` returns `null`.
2. Your handler closes first with `ws.close(...)`.
The runtime sends a close frame and releases the socket locally.
3. The transport disappears unexpectedly.
The session is released and no more messages can be received; subsequent sends fail.
What Lyng code should do:
- Treat `receive() == null` as end-of-session.
- Exit the handler or break the receive loop at that point.
- Do not keep sending after close has been observed.
The current server-side API does not expose the peer close code or close reason to Lyng.
### Closing The Connection Yourself
Call `ws.close()` when you want to terminate the websocket session.
```lyng
server.ws("/chat") { ws ->
ws.sendText("server shutting down")
ws.close(1000, "done")
}
```
Close semantics:
- `close()` sends a websocket close frame with the given code and reason.
- Defaults are `code = 1000` and `reason = ""`.
- `close()` is idempotent; calling it again after close does nothing.
- After local close, the session should be treated as unusable.
- After close, `isOpen()` becomes false and further sends fail.
### WebSocket Handler Pattern
```lyng
import lyng.io.http.server
val server = HttpServer()
server.wsPath("/rooms/{room}") { ws ->
val room = routeParams["room"] ?: "<unknown>"
ws.sendText("joined:" + room)
while (true) {
val msg = ws.receive() ?: break
if (msg.isText) {
ws.sendText(room + ":" + msg.text)
}
}
ws.close()
}
server.listen(8080, "127.0.0.1")
```
## Path-Template Routes
Path templates are sugar on top of regex routes. Template parameters are exposed as decoded `routeParams`.
```lyng
server.getPath("/users/{userId}/posts/{postId}") {
respondText(
200,
routeParams["userId"] + ":" + routeParams["postId"]
)
}
```
Template rules:
- template must start with `/`
- a segment is either literal text or `{name}`
- parameter names must be valid identifiers
- parameter values match one path segment only
- parameter values use path decoding rules:
- valid percent-encoding is decoded
- `+` stays `+`
- malformed `%` stays literal
## Regex Routes
Regex routes match the whole request path, not a substring.
```lyng
server.get("^/users/([0-9]+)/posts/([0-9]+)$".re) {
val m = routeMatch!!
respondText(200, "user=" + m[1] + ", post=" + m[2])
}
```
## Basic Exact Route
```lyng
import lyng.io.http.server
val server = HttpServer()
server.get("/hello") {
setHeader("Content-Type", "text/plain")
respondText(200, "hello")
}
server.listen(8080, "127.0.0.1")
```
## Route Precedence
Dispatch order is:
1. exact method route
2. exact `any` route
3. regex method route, registration order
4. regex `any` route, registration order
5. fallback
This means exact routes stay fast and always win over template or regex routes for the same path.
## API Surface
### `Router` Route Registration Methods
- `get(path: String|Regex, handler)`
- `getPath(pathTemplate: String, handler)`
- `post(path: String|Regex, handler)`
- `postPath(pathTemplate: String, handler)`
- `put(path: String|Regex, handler)`
- `putPath(pathTemplate: String, handler)`
- `delete(path: String|Regex, handler)`
- `deletePath(pathTemplate: String, handler)`
- `any(path: String|Regex, handler)`
- `anyPath(pathTemplate: String, handler)`
- `ws(path: String|Regex, handler)`
- `wsPath(pathTemplate: String, handler)`
- `fallback(handler)`
- `mount(router)`
### `HttpServer` Route Registration Methods
- `get(path: String|Regex, handler)`
- `getPath(pathTemplate: String, handler)`
- `post(path: String|Regex, handler)`
- `postPath(pathTemplate: String, handler)`
- `put(path: String|Regex, handler)`
- `putPath(pathTemplate: String, handler)`
- `delete(path: String|Regex, handler)`
- `deletePath(pathTemplate: String, handler)`
- `any(path: String|Regex, handler)`
- `anyPath(pathTemplate: String, handler)`
- `ws(path: String|Regex, handler)`
- `wsPath(pathTemplate: String, handler)`
- `fallback(handler)`
- `mount(router)`
- `listen(port, host = null, backlog = 128)`

View File

@ -1,177 +0,0 @@
### lyng.io.net — TCP and UDP sockets for Lyng scripts
This module provides minimal raw transport networking for Lyng scripts. It is implemented in `lyngio` and backed by Ktor sockets on the JVM and Linux Native, and by Node networking APIs on JS/Node runtimes.
> **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.
>
> **Shared type note:** `IpVersion`, `SocketAddress`, and `Datagram` are also available from `lyng.io.net.types` when host code wants reusable transport value types without depending on the `Net` capability object itself.
>
> **Important native platform limit:** current native TCP/UDP support is backed by a selector with a per-process file descriptor ceiling. On Linux/macOS native targets this makes high-connection-count servers and same-process load tests unsuitable once the process approaches that limit.
>
> **Recommendation:** for serious HTTP/TCP servers, prefer the JVM target today. On native targets, keep concurrency bounded, batch local load tests in waves, and use multiple worker processes behind a reverse proxy if you need more throughput before the backend is reworked.
>
> **Need this fixed?** Please open or upvote an issue at <https://github.com/sergeych/lyng/issues> so native high-concurrency networking can be prioritized.
---
#### Install the module into a Lyng session
Kotlin (host) bootstrap example:
```kotlin
import net.sergeych.lyng.EvalSession
import net.sergeych.lyng.Scope
import net.sergeych.lyng.io.net.createNetModule
import net.sergeych.lyngio.net.security.PermitAllNetAccessPolicy
suspend fun bootstrapNet() {
val session = EvalSession()
val scope: Scope = session.getScope()
createNetModule(PermitAllNetAccessPolicy, scope)
session.eval("import lyng.io.net")
}
```
---
#### Using from Lyng scripts
Capability checks and address resolution:
import lyng.io.net
val a: SocketAddress = Net.resolve("127.0.0.1", 4040)[0]
[Net.isSupported(), a.toString(), a.resolved, a.ipVersion == IpVersion.IPV4]
>>> [true,127.0.0.1:4040,true,true]
TCP client connect, write, read, and close:
import lyng.buffer
import lyng.io.net
val socket = Net.tcpConnect("127.0.0.1", NET_TEST_TCP_PORT)
socket.writeUtf8("ping")
socket.flush()
val reply = (socket.read(16) as Buffer).decodeUtf8()
socket.close()
reply
>>> "reply:ping"
Lyng TCP server socket operations with `tcpListen()` and `accept()`:
import lyng.buffer
import lyng.io.net
val server = Net.tcpListen(0, "127.0.0.1")
val port = server.localAddress().port
val accepted = launch {
val client = server.accept()
val line = (client.read(4) as Buffer).decodeUtf8()
client.writeUtf8("echo:" + line)
client.flush()
client.close()
server.close()
line
}
val socket = Net.tcpConnect("127.0.0.1", port)
socket.writeUtf8("ping")
socket.flush()
val reply = (socket.read(16) as Buffer).decodeUtf8()
socket.close()
[accepted.await(), reply]
>>> [ping,echo:ping]
UDP bind, send, receive, and inspect sender address:
import lyng.buffer
import lyng.io.net
val server = Net.udpBind(0, "127.0.0.1")
val client = Net.udpBind(0, "127.0.0.1")
client.send(Buffer("ping"), "127.0.0.1", server.localAddress().port)
val d = server.receive()
client.close()
server.close()
[d.data.decodeUtf8(), d.address.port > 0]
>>> [ping,true]
---
#### API reference
##### `Net` (static methods)
- `isSupported(): Bool` — Whether any raw networking support is available.
- `isTcpAvailable(): Bool` — Whether outbound TCP sockets are available.
- `isTcpServerAvailable(): Bool` — Whether listening TCP server sockets are available.
- `isUdpAvailable(): Bool` — Whether UDP datagram sockets are available.
- `resolve(host: String, port: Int): List<SocketAddress>` — Resolve a host and port into concrete addresses.
- `tcpConnect(host: String, port: Int, timeoutMillis: Int? = null, noDelay: Bool = true): TcpSocket` — Open an outbound TCP socket.
- `tcpListen(port: Int, host: String? = null, backlog: Int = 128, reuseAddress: Bool = true): TcpServer` — Start a listening TCP server socket.
- `udpBind(port: Int = 0, host: String? = null, reuseAddress: Bool = true): UdpSocket` — Bind a UDP socket.
##### `SocketAddress`
- `host: String`
- `port: Int`
- `ipVersion: IpVersion`
- `resolved: Bool`
- `toString(): String`
##### `TcpSocket`
- `isOpen(): Bool`
- `localAddress(): SocketAddress`
- `remoteAddress(): SocketAddress`
- `read(maxBytes: Int = 65536): Buffer?`
- `readLine(): String?`
- `write(data: Buffer): void`
- `writeUtf8(text: String): void`
- `flush(): void`
- `close(): void`
##### `TcpServer`
- `isOpen(): Bool`
- `localAddress(): SocketAddress`
- `accept(): TcpSocket`
- `close(): void`
##### `UdpSocket`
- `isOpen(): Bool`
- `localAddress(): SocketAddress`
- `receive(maxBytes: Int = 65536): Datagram?`
- `send(data: Buffer, host: String, port: Int): void`
- `close(): void`
##### `Datagram`
- `data: Buffer`
- `address: SocketAddress`
---
#### Security policy
The module uses `NetAccessPolicy` to authorize network operations before they are executed.
- `NetAccessPolicy` — interface for custom policies
- `PermitAllNetAccessPolicy` — allows all network operations
- `NetAccessOp.Resolve(host, port)`
- `NetAccessOp.TcpConnect(host, port)`
- `NetAccessOp.TcpListen(host, port, backlog)`
- `NetAccessOp.UdpBind(host, port)`
---
#### Platform support
- **JVM:** supported
- **Android:** supported via the Ktor CIO and Ktor sockets backends
- **JS/Node:** supported for `resolve`, TCP client/server, and UDP
- **JS/browser:** unsupported; capability checks report unavailable
- **Linux Native:** supported via Ktor sockets
- **Apple Native:** enabled via the shared native Ktor sockets backend; compile-verified, runtime not yet host-verified
- **Other native targets:** currently report unsupported; use capability checks before relying on raw sockets

View File

@ -1,138 +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 session
The process module is not installed automatically. The preferred host runtime is `EvalSession`: create the session, get its underlying scope, install the module there, and execute scripts through the session. You can customize access control via `ProcessAccessPolicy`.
Kotlin (host) bootstrap example:
```kotlin
import net.sergeych.lyng.Scope
import net.sergeych.lyng.EvalSession
import net.sergeych.lyng.io.process.createProcessModule
import net.sergeych.lyngio.process.security.PermitAllProcessAccessPolicy
suspend fun bootstrapProcess() {
val session = EvalSession()
val scope: Scope = session.getScope()
createProcessModule(PermitAllProcessAccessPolicy, scope)
// In scripts (or via session.eval), import the module:
session.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

@ -1,279 +0,0 @@
# `lyng.io.ws` - WebSocket client for Lyng scripts
This module provides a compact WebSocket client API for Lyng scripts. It is implemented in `lyngio` and currently backed by Ktor WebSockets on the JVM.
> **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.
>
> **Shared type note:** `WsMessage` is also available from `lyng.io.ws.types` when host code wants the reusable message type without depending on the WebSocket client module itself.
## Install The Module Into A Lyng Session
Kotlin host bootstrap example:
```kotlin
import net.sergeych.lyng.EvalSession
import net.sergeych.lyng.Scope
import net.sergeych.lyng.io.ws.createWsModule
import net.sergeych.lyngio.ws.security.PermitAllWsAccessPolicy
suspend fun bootstrapWs() {
val session = EvalSession()
val scope: Scope = session.getScope()
createWsModule(PermitAllWsAccessPolicy, scope)
session.eval("import lyng.io.ws")
}
```
## Using From Lyng Scripts
### Text Exchange
```lyng
import lyng.io.ws
val ws = Ws.connect(WS_TEST_URL)
ws.sendText("ping")
val m: WsMessage = ws.receive()
ws.close()
[ws.url() == WS_TEST_URL, m.isText, m.text]
>>> [true,true,echo:ping]
```
### Binary Exchange
```lyng
import lyng.buffer
import lyng.io.ws
val ws = Ws.connect(WS_TEST_BINARY_URL)
ws.sendBytes(Buffer(9, 8, 7))
val m: WsMessage = ws.receive()
ws.close()
[m.isText, (m.data as Buffer).hex]
>>> [false,010203090807]
```
### Secure `wss` Exchange
```lyng
import lyng.io.ws
val ws = Ws.connect(WSS_TEST_URL)
ws.sendText("ping")
val m: WsMessage = ws.receive()
ws.close()
[ws.url() == WSS_TEST_URL, m.text]
>>> [true,secure:ping]
```
## Message Flow And Session Lifecycle
### Reading Incoming Messages
Call `ws.receive()` to wait for the next application message.
What `receive()` returns:
- `WsMessage` for the next text or binary message.
- `null` after the peer closes the connection cleanly.
- `null` after the transport has already been closed and no more messages can arrive.
What reaches Lyng code:
- Text frames are exposed as `WsMessage(isText = true, text = ...)`.
- Binary frames are exposed as `WsMessage(isText = false, data = ...)`.
- Fragmented websocket messages are reassembled before they are returned.
- Ping and pong control frames are handled internally and are not returned by `receive()`.
- Incoming close frames are handled internally; after that `receive()` returns `null`.
Typical receive loop:
```lyng
import lyng.buffer
import lyng.io.ws
val ws = Ws.connect(WS_URL)
while (true) {
val msg = ws.receive() ?: break
if (msg.isText) {
println("text=" + msg.text)
} else {
println("bytes=" + ((msg.data as Buffer).size))
}
}
println("peer closed the websocket")
```
### Sending Outgoing Messages
Use:
- `ws.sendText(text)` for UTF-8 text messages.
- `ws.sendBytes(data)` for binary messages.
Example:
```lyng
import lyng.buffer
import lyng.io.ws
val ws = Ws.connect(WS_URL)
ws.sendText("hello")
ws.sendBytes(Buffer(1, 2, 3, 4))
```
Send behavior:
- Each call sends one websocket message.
- The API does not expose partial-frame streaming; send the whole message in one call.
- If the session is already closed, `sendText(...)` and `sendBytes(...)` fail with a websocket error.
- If the transport breaks during send, the session is released and the send call fails.
### Detecting Closed Connections
Use both signals together:
- `ws.isOpen()` tells you whether the session is still considered open right now.
- `ws.receive() == null` tells you the receive side has reached the end of the websocket session.
Practical rule:
- If `receive()` returns `null`, stop reading and treat the session as closed.
- After close has been observed, do not attempt further sends.
The API does not currently expose the peer close code or close reason to Lyng code.
### Closing The Connection Yourself
Call `ws.close()` when you are done.
```lyng
import lyng.io.ws
val ws = Ws.connect(WS_URL)
ws.sendText("bye")
ws.close(1000, "done")
```
Close semantics:
- `close()` sends a websocket close frame with the given code and reason.
- Defaults are `code = 1000` and `reason = ""`.
- After `close()`, the session is released locally and should be treated as closed immediately.
- Calling `close()` on an already closed session is a no-op.
- After local close, `receive()` returns `null` and further sends fail.
### Recommended Usage Pattern
For request-response style exchanges:
```lyng
import lyng.io.ws
val ws = Ws.connect(WS_URL)
try {
ws.sendText("ping")
val reply = ws.receive() ?: error("socket closed before reply")
println(reply.text)
} finally {
ws.close()
}
```
For long-lived consumers:
```lyng
import lyng.io.ws
val ws = Ws.connect(WS_URL)
try {
while (true) {
val msg = ws.receive() ?: break
if (msg.isText) {
println(msg.text)
}
}
} finally {
ws.close()
}
```
## API Reference
### `Ws`
- `isSupported(): Bool` - whether WebSocket client support is available on the current runtime.
- `connect(url: String, headers...): WsSession` - open a client websocket session.
`headers...` accepts:
- `MapEntry`, for example `"Authorization" => "Bearer x"`
- 2-item lists, for example `["Authorization", "Bearer x"]`
### `WsSession`
- `isOpen(): Bool`
- `url(): String`
- `sendText(text: String): void`
- `sendBytes(data: Buffer): void`
- `receive(): WsMessage?`
- `close(code: Int = 1000, reason: String = ""): void`
Behavior summary:
- `receive()` returns `null` after close.
- `close()` is safe to call more than once.
- send operations require an open session.
### `WsMessage`
- `isText: Bool`
- `text: String?`
- `data: Buffer?`
Payload rules:
- Text messages populate `text` and leave `data == null`.
- Binary messages populate `data` and leave `text == null`.
## Security Policy
The module uses `WsAccessPolicy` to authorize websocket operations.
- `WsAccessPolicy` - interface for custom policies.
- `PermitAllWsAccessPolicy` - allows all websocket operations.
- `WsAccessOp.Connect(url)`
- `WsAccessOp.Send(url, bytes, isText)`
- `WsAccessOp.Receive(url)`
Example restricted policy in Kotlin:
```kotlin
import net.sergeych.lyngio.fs.security.AccessContext
import net.sergeych.lyngio.fs.security.AccessDecision
import net.sergeych.lyngio.fs.security.Decision
import net.sergeych.lyngio.ws.security.WsAccessOp
import net.sergeych.lyngio.ws.security.WsAccessPolicy
val allowLocalOnly = object : WsAccessPolicy {
override suspend fun check(op: WsAccessOp, ctx: AccessContext): AccessDecision =
when (op) {
is WsAccessOp.Connect ->
if (
op.url.startsWith("ws://127.0.0.1:") ||
op.url.startsWith("wss://127.0.0.1:") ||
op.url.startsWith("ws://localhost:") ||
op.url.startsWith("wss://localhost:")
)
AccessDecision(Decision.Allow)
else
AccessDecision(Decision.Deny, "only local ws/wss connections are allowed")
else -> AccessDecision(Decision.Allow)
}
}
```
## Platform Support
- **JVM:** supported.
- **Android:** supported via the Ktor CIO websocket client backend.
- **JS:** supported via the Ktor JS websocket client backend.
- **Linux native:** supported via the Ktor Curl websocket client backend.
- **Windows native:** supported via the Ktor WinHttp websocket client backend.
- **Apple native:** supported via the Ktor Darwin websocket client backend.
- **Other targets:** may report unsupported; use `Ws.isSupported()` before relying on websocket client access.

View File

@ -1,15 +1,13 @@
# Lyng CLI (`lyng`)
### Lyng CLI (`lyng`)
The Lyng CLI is the reference command-line tool for the Lyng language. It lets you:
- Run Lyng scripts from files or inline strings (shebangs accepted)
- Use standard argument passing (`ARGV`) to your scripts.
- Resolve local file imports from the executed script's directory tree.
- Format Lyng source files via the built-in `fmt` subcommand.
- Register synchronous process-exit handlers with `atExit(...)`.
## Building on Linux
#### Building on Linux
Requirements:
- JDK 17+ (for Gradle and the JVM distribution)
@ -21,7 +19,7 @@ The repository provides convenience scripts in `bin/` for local builds and insta
Note: In this repository the scripts are named `bin/local_release` and `bin/local_jrelease`. In some environments these may be aliased as `bin/release` and `bin/jrelease`. The steps below use the actual file names present here.
### Option A: Native linuxX64 executable (`lyng`)
##### Option A: Native linuxX64 executable (`lyng`)
1) Build the native binary:
@ -40,27 +38,26 @@ What this does:
- Produces `distributables/lyng-linuxX64.zip` containing the `lyng` executable.
### Option B: JVM distribution (`jlyng` launcher)
##### Option B: JVM distribution (`jlyng` launcher)
This creates a JVM distribution with a launcher script, packages it as a downloadable zip, and links it to `~/bin/jlyng`.
This creates a JVM distribution with a launcher script and links it to `~/bin/jlyng`.
```
bin/local_jrelease
```
What this does:
- Runs `./gradlew :lyng:jvmDistZip` to build the JVM app distribution archive at `lyng/build/distributions/lyng-jvm.zip`.
- Copies the archive to `distributables/lyng-jvm.zip`.
- Unpacks that distribution under `~/bin/jlyng-jvm`.
- Runs `./gradlew :lyng:installJvmDist` to build the JVM app distribution to `lyng/build/install/lyng-jvm`.
- Copies the distribution under `~/bin/jlyng-jvm`.
- Creates a symlink `~/bin/jlyng` pointing to the launcher script.
## Usage
#### Usage
Once installed, ensure `~/bin` is on your `PATH`. You can then use either the native `lyng` or the JVM `jlyng` launcher (both have the same CLI surface).
### Running scripts
##### Running scripts
- Run a script by file name and pass arguments to `ARGV`:
@ -75,7 +72,6 @@ lyng -- -my-script.lyng arg1 arg2
```
- Execute inline code with `-x/--execute` and pass positional args to `ARGV`:
- Inline execution does not scan the filesystem for local modules; only file-based execution does.
```
lyng -x "println(\"Hello\")" more args
@ -88,101 +84,7 @@ lyng --version
lyng --help
```
### Exit handlers: `atExit(...)`
The CLI exposes a CLI-only builtin:
```lyng
extern fun atExit(append: Bool=true, handler: ()->Void)
```
Use it to register synchronous cleanup handlers that should run when the CLI process is leaving.
Semantics:
- `append=true` appends the handler to the end of the queue.
- `append=false` inserts the handler at the front of the queue.
- Handlers run one by one.
- Exceptions thrown by a handler are ignored, and the next handler still runs.
- Handlers are best-effort and run on:
- normal script completion
- script failure
- script `exit(code)`
- process shutdown such as `SIGTERM`
Non-goals:
- `SIGKILL`, hard crashes, and power loss cannot be intercepted.
- `atExit` is currently a CLI feature only; it is not part of the general embedding/runtime surface.
Examples:
```lyng
atExit {
println("closing resources")
}
atExit(false) {
println("runs first")
}
```
### Local imports for file execution
When you execute a script file, the CLI builds a temporary local import manager rooted at the directory that contains the entry script.
Formal structure:
- Root directory: the parent directory of the script passed to `lyng`.
- Scan scope: every `.lyng` file under that root directory, recursively.
- Entry script: the executed file itself is not registered as an importable module.
- Module name mapping: `relative/path/to/file.lyng` maps to import name `relative.path.to.file`.
- Package declaration: if a scanned file starts with `package ...` as its first non-blank line, that package name must exactly match the relative path mapping.
- Package omission: if there is no leading `package` declaration, the CLI uses the relative path mapping as the module name.
- Duplicates: if two files resolve to the same module name, CLI execution fails before script execution starts.
- Import visibility: only files inside the entry root subtree are considered. Parent directories and sibling projects are not searched.
Examples:
```
project/
main.lyng
util/answer.lyng
math/add.lyng
```
`util/answer.lyng` is imported as `import util.answer`.
`math/add.lyng` is imported as `import math.add`.
Example contents:
```lyng
// util/answer.lyng
package util.answer
import math.add
fun answer() = plus(40, 2)
```
```lyng
// math/add.lyng
fun plus(a, b) = a + b
```
```lyng
// main.lyng
import util.answer
println(answer())
```
Rationale:
- The module name is deterministic from the filesystem layout.
- Explicit `package` remains available as a consistency check instead of a second, conflicting naming system.
- The import search space stays local to the executed script, which avoids accidental cross-project resolution.
## Use in shell scripts
### Use in shell scripts
Standard unix shebangs (`#!`) are supported, so you can make Lyng scripts directly executable on Unix-like systems. For example:
@ -190,7 +92,7 @@ Standard unix shebangs (`#!`) are supported, so you can make Lyng scripts direct
println("Hello, world!")
### Formatting source: `fmt` subcommand
##### Formatting source: `fmt` subcommand
Format Lyng files with the built-in formatter.
@ -232,7 +134,7 @@ lyng fmt --spacing --wrap src/file.lyng
```
## Notes
#### Notes
- Both native and JVM distributions expose the same CLI interface. Use whichever best fits your environment.
- When executing scripts, all positional arguments after the script name are available in Lyng as `ARGV`.

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

@ -1,146 +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.
> **Important native networking limit:** `lyng.io.net` on current native targets is suitable for modest workloads, local tools, and test servers, but not yet for high-connection-count production servers. For serious HTTP/TCP serving, prefer the JVM target for now. If native high-concurrency networking matters for your use case, please open or upvote an issue at <https://github.com/sergeych/lyng/issues>.
#### 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`, `ConsoleAccessPolicy`) that allow you to control exactly what a script can do.
#### Included Modules
- **[lyng.io.db](lyng.io.db.md):** Portable SQL database access. Provides `Database`, `SqlTransaction`, `ResultSet`, SQLite support through `lyng.io.db.sqlite`, and JVM JDBC support through `lyng.io.db.jdbc`.
- **[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.
- **[lyng.io.console](lyng.io.console.md):** Rich console/TTY access. Provides `Console` capability detection, geometry, output, and iterable events.
- **[lyng.io.http](lyng.io.http.md):** HTTP/HTTPS client access. Provides `Http`, `HttpRequest`, `HttpResponse`, and `HttpHeaders`.
- **[lyng.io.http.server](lyng.io.http.server.md):** Minimal HTTP/1.1 and WebSocket server. Provides `HttpServer`, `Router`, `ServerRequest`, `RequestContext`, and `ServerWebSocket`.
- **[lyng.io.ws](lyng.io.ws.md):** WebSocket client access. Provides `Ws`, `WsSession`, and `WsMessage`.
- **[lyng.io.net](lyng.io.net.md):** Transport networking. Provides `Net`, `TcpSocket`, `TcpServer`, `UdpSocket`, and `SocketAddress`.
- **Shared networking type packages:** `lyng.io.http.types`, `lyng.io.ws.types`, and `lyng.io.net.types` expose reusable value types such as `HttpHeaders`, `WsMessage`, `IpVersion`, `SocketAddress`, and `Datagram` when host code wants type-only imports without installing the corresponding capability object module.
---
#### 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.EvalSession
import net.sergeych.lyng.io.db.createDbModule
import net.sergeych.lyng.io.db.jdbc.createJdbcModule
import net.sergeych.lyng.io.db.sqlite.createSqliteModule
import net.sergeych.lyng.io.fs.createFs
import net.sergeych.lyng.io.process.createProcessModule
import net.sergeych.lyng.io.console.createConsoleModule
import net.sergeych.lyng.io.http.createHttpModule
import net.sergeych.lyng.io.net.createNetModule
import net.sergeych.lyng.io.ws.createWsModule
import net.sergeych.lyngio.fs.security.PermitAllAccessPolicy
import net.sergeych.lyngio.process.security.PermitAllProcessAccessPolicy
import net.sergeych.lyngio.console.security.PermitAllConsoleAccessPolicy
import net.sergeych.lyngio.http.security.PermitAllHttpAccessPolicy
import net.sergeych.lyngio.net.security.PermitAllNetAccessPolicy
import net.sergeych.lyngio.ws.security.PermitAllWsAccessPolicy
suspend fun runMyScript() {
val session = EvalSession()
val scope = session.getScope()
// Install modules with policies
createDbModule(scope)
createJdbcModule(scope)
createSqliteModule(scope)
createFs(PermitAllAccessPolicy, scope)
createProcessModule(PermitAllProcessAccessPolicy, scope)
createConsoleModule(PermitAllConsoleAccessPolicy, scope)
createHttpModule(PermitAllHttpAccessPolicy, scope)
createNetModule(PermitAllNetAccessPolicy, scope)
createWsModule(PermitAllWsAccessPolicy, scope)
// Now scripts can import them
session.eval("""
import lyng.io.db
import lyng.io.db.jdbc
import lyng.io.db.sqlite
import lyng.io.fs
import lyng.io.process
import lyng.io.console
import lyng.io.http
import lyng.io.net
import lyng.io.ws
println("H2 JDBC available: " + (openH2("mem:demo;DB_CLOSE_DELAY=-1") != null))
println("SQLite available: " + (openSqlite(":memory:") != null))
println("Working dir: " + Path(".").readUtf8())
println("OS: " + Platform.details().name)
println("TTY: " + Console.isTty())
println("HTTP available: " + Http.isSupported())
println("TCP available: " + Net.isTcpAvailable())
println("WS available: " + Ws.isSupported())
""")
}
```
---
#### 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).
- **Database Installation:** Database access is still explicit-capability style. The host must install `lyng.io.db` and at least one provider such as `lyng.io.db.sqlite` or `lyng.io.db.jdbc`; otherwise scripts cannot open databases.
- **Process Security:** Implement `ProcessAccessPolicy` to restrict which executables can be run or to disable shell execution entirely.
- **Console Security:** Implement `ConsoleAccessPolicy` to control output writes, event reads, and raw mode switching.
- **HTTP Security:** Implement `HttpAccessPolicy` to restrict which requests scripts may send.
- **Transport Security:** Implement `NetAccessPolicy` to restrict DNS resolution and TCP/UDP socket operations.
- **WebSocket Security:** Implement `WsAccessPolicy` to restrict websocket connects and message flow.
For more details, see the specific module documentation:
- [Database Module Details](lyng.io.db.md)
- [Filesystem Security Details](lyng.io.fs.md#access-policy-security)
- [Process Security Details](lyng.io.process.md#security-policy)
- [Console Module Details](lyng.io.console.md)
- [HTTP Module Details](lyng.io.http.md)
- [HTTP Server Module Details](lyng.io.http.server.md)
- [Transport Networking Details](lyng.io.net.md)
- [WebSocket Module Details](lyng.io.ws.md)
---
#### Platform Support Overview
| Platform | lyng.io.db/sqlite | lyng.io.db/jdbc | lyng.io.fs | lyng.io.process | lyng.io.console | lyng.io.http | lyng.io.ws | lyng.io.net |
| :--- | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: |
| **JVM** | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ |
| **Linux Native** | ✅ | ❌ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ |
| **Apple Native** | ❌ | ❌ | ✅ | ✅ | ✅ | ✅ | ✅ | ⚠️ |
| **Windows Native** | ❌ | ❌ | ✅ | ❌ | ✅ | ✅ | ✅ | ❌ |
| **Android** | ⚠️ | ❌ | ✅ | ❌ | ⚠️ | ✅ | ✅ | ✅ |
| **JS / Node** | ❌ | ❌ | ✅ | ❌ | ⚠️ | ✅ | ✅ | ✅ |
| **JS / Browser** | ❌ | ❌ | ✅ (in-memory) | ❌ | ⚠️ | ✅ | ✅ | ❌ |
| **Wasm** | ❌ | ❌ | ✅ (in-memory) | ❌ | ⚠️ | ❌ | ❌ | ❌ |
Legend:
- `✅` supported
- `⚠️` available but environment-dependent or not fully host-verified yet
- `❌` unsupported

View File

@ -60,13 +60,8 @@ but:
## Round and range
The following functions return the argument unchanged if it is `Int`.
For `Decimal`:
- `floor(x)`, `ceil(x)`, and `round(x)` currently use exact decimal operations
- the result stays `Decimal`
For `Real`, the result is a transformed `Real`.
The following functions return its argument if it is `Int`,
or transformed `Real` otherwise.
| name | description |
|----------------|--------------------------------------------------------|
@ -77,14 +72,6 @@ For `Real`, the result is a transformed `Real`.
## Lyng math functions
Decimal note:
- all scalar math helpers accept `Decimal`
- `abs(x)` stays exact for `Decimal`
- `pow(x, y)` is exact for `Decimal` when `y` is an integral exponent
- the remaining `Decimal` cases currently use a temporary bridge:
`Decimal -> Real -> host math -> Decimal`
- this is temporary; native decimal implementations are planned
| name | meaning |
|-----------|------------------------------------------------------|
| sin(x) | sine |
@ -104,8 +91,7 @@ Decimal note:
| log10(x) | $log_{10}(x)$ |
| pow(x, y) | ${x^y}$ |
| sqrt(x) | $ \sqrt {x}$ |
| abs(x) | absolute value of x. Int if x is Int, Decimal if x is Decimal, Real otherwise |
| clamp(x, range) | limit x to be inside range boundaries |
| abs(x) | absolute value of x. Int if x is Int, Real otherwise |
For example:
@ -116,91 +102,6 @@ For example:
// abs() keeps the argument type:
assert( abs(-1) is Int)
assert( abs(-2.21) == 2.21 )
import lyng.decimal
// Decimal-aware math works too. Some functions are exact, some still bridge through Real temporarily:
assert( (abs("-2.5".d) as Decimal).toStringExpanded() == "2.5" )
assert( (floor("2.9".d) as Decimal).toStringExpanded() == "2" )
assert( sin("0.5".d) is Decimal )
// 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
## Linear algebra: `lyng.matrix`
For vectors and dense matrices, import `lyng.matrix`:
```lyng
import lyng.matrix
```
It provides:
- `Vector`
- `Matrix`
- `vector(values)`
- `matrix(rows)`
Core operations include:
- matrix addition and subtraction
- matrix-matrix multiplication
- matrix-vector multiplication
- transpose
- determinant
- inverse
- linear solve
- vector dot, norm, normalize, cross, outer product
Example:
```lyng
import lyng.matrix
val a: Matrix = matrix([[1, 2, 3], [4, 5, 6]])
val b: Matrix = matrix([[7, 8], [9, 10], [11, 12]])
val product: Matrix = a * b
assertEquals([[58.0, 64.0], [139.0, 154.0]], product.toList())
```
Matrices also support two-axis bracket indexing and slicing:
```lyng
import lyng.matrix
val m: Matrix = matrix([[1, 2, 3], [4, 5, 6], [7, 8, 9]])
assertEquals(6.0, m[1, 2])
val sub: Matrix = m[0..1, 1..2]
assertEquals([[2.0, 3.0], [5.0, 6.0]], sub.toList())
```
See [Matrix](Matrix.md) for the full API.
## Random values
Lyng stdlib provides a global random singleton and deterministic seeded generators:
| name | meaning |
|--------------------------|---------|
| Random.nextInt() | random `Int` from full platform range |
| Random.nextFloat() | random `Real` in `[0,1)` |
| Random.next(range) | random value from the given finite range |
| Random.seeded(seed) | creates deterministic generator |
| SeededRandom.nextInt() | deterministic random `Int` |
| SeededRandom.nextFloat() | deterministic random `Real` in `[0,1)` |
| SeededRandom.next(range) | deterministic random value from range |
Examples:
val rng = Random.seeded(1234)
assert( rng.next(1..10) in 1..10 )
assert( rng.next('a'..<'f') in 'a'..<'f' )
assert( rng.next(0.0..<1.0) >= 0.0 )
assert( rng.next(0.0..<1.0) < 1.0 )
>>> void
## Scientific constant

View File

@ -4,7 +4,7 @@
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`.
When kotlin 2.0 was released, or soon after, JetBrains made an exptic 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).
@ -12,14 +12,14 @@ Later JetBrains added serializers for their new `Instant` and `Clock` types, but
## 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`.)
We hereby publish a new version of Lyng, 1.0.8-SNAPSHOT, which uses `ktlin.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. Still you might need to migrate your code to use `kotlinx.datetime` types.
So, if you are getting errors with new version, please do:
So, if you are getting errors with new version, plase 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`.
- 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.
This should solve the problem and hopefully we'll see no more suh a brillant ideas from IDEA ideologspersons.
Sorry for inconvenience and send a ray of hate to JetBrains ;)
Sorry for inconvenicence and send a ray of hate to JetBrains ;)

View File

@ -32,36 +32,10 @@ Depending on the platform, these coroutines may be executed on different CPU and
assert(xIsCalled)
>>> void
This example shows how to launch a coroutine with `launch` which returns [Deferred] instance, the latter have ways to await for the coroutine completion, cancel it if it is no longer needed, and retrieve possible result.
This example shows how to launch a coroutine with `launch` which returns [Deferred] instance, the latter have ways to await for the coroutine completion and retrieve possible result.
Launch has the only argument which should be a callable (lambda usually) that is run in parallel (or cooperatively in parallel), and return anything as the result.
When you have an iterable of deferreds, use `joinAll()` to await all of them and collect results in input order:
val jobs = (1..4).map { n ->
launch {
delay(1)
n * 10
}
}
assertEquals([10, 20, 30, 40], jobs.joinAll())
>>> void
If you no longer need the result, cancel the deferred. Awaiting a cancelled deferred throws `CancellationException`:
var reached = false
val work = launch {
delay(100)
reached = true
"ok"
}
work.cancel()
assertThrows(CancellationException) { work.await() }
assert(work.isCancelled)
assert(!work.isActive)
assert(!reached)
>>> void
## Synchronization: Mutex
Suppose we have a resource, that could be used concurrently, a counter in our case. If we won't protect it, concurrent usage cause RC, Race Condition, providing wrong result:
@ -75,7 +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
@ -90,12 +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.
@ -230,73 +205,6 @@ Flows allow easy transforming of any [Iterable]. See how the standard Lyng libra
}
}
## Channel
A [Channel] is a **hot pipe** between coroutines: values are pushed in by a producer and pulled out by a consumer, with each value consumed exactly once.
Unlike a `Flow` (which is cold and re-runs its generator on every collection), a `Channel` is stateful — the right tool for classic _producer / consumer_ work.
val ch = Channel() // rendezvous: sender waits for receiver
val producer = launch {
for (i in 1..5) ch.send(i)
ch.close() // signal: no more values
}
var item = ch.receive() // suspends until a value is ready
while (item != null) {
println(item)
item = ch.receive()
}
// prints 1 2 3 4 5
`receive()` returns `null` when the channel is both closed and fully drained — that is the idiomatic loop termination condition.
Channels can also be buffered so the producer can run ahead:
val ch = Channel(4) // buffer up to 4 items without blocking
ch.send(10)
ch.send(20)
ch.send(30)
ch.close()
assertEquals(10, ch.receive())
assertEquals(20, ch.receive())
assertEquals(30, ch.receive())
assertEquals(null, ch.receive()) // drained
For the full API — including `tryReceive`, `Channel.UNLIMITED`, and the fan-out / ping-pong patterns — see the [Channel] reference page.
## LaunchPool
When you need **bounded concurrency** — run at most *N* tasks at the same time without spawning a new coroutine per task — use [LaunchPool]:
```lyng
val pool = LaunchPool(4) // at most 4 tasks run in parallel
val jobs = (1..20).map { n ->
pool.launch { expensiveCompute(n) }
}
pool.closeAndJoin() // wait for all tasks to complete
val results = jobs.joinAll()
```
Exceptions thrown inside a submitted lambda are captured in the returned `Deferred` and do not crash the pool, so other tasks continue running normally.
See [LaunchPool] for the full API including bounded queues and cancellation.
[LaunchPool]: LaunchPool.md
| | Flow | Channel |
|---|---|---|
| **temperature** | cold (lazy) | hot (eager) |
| **replay** | every collector gets a fresh run | each item consumed once |
| **consumers** | any number, each gets all items | one receiver per item |
| **typical use** | transform pipelines, sequences | producer–consumer, fan-out |
[Channel]: Channel.md
[Iterable]: Iterable.md
@ -316,14 +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.
@ -114,12 +114,10 @@ When running end‑to‑end “book” workloads or heavier benches, you can ena
Flags are mutable at runtime, e.g.:
```kotlin
runTest {
PerfFlags.ARG_BUILDER = false
val r1 = (EvalSession(Scope()).eval(script) as ObjInt).value
PerfFlags.ARG_BUILDER = true
val r2 = (EvalSession(Scope()).eval(script) as ObjInt).value
}
PerfFlags.ARG_BUILDER = false
val r1 = (Scope().eval(script) as ObjInt).value
PerfFlags.ARG_BUILDER = true
val r2 = (Scope().eval(script) as ObjInt).value
```
Reset flags at the end of a test to avoid impacting other tests.
@ -621,3 +619,4 @@ Reproduce
Notes
- Negative caches are installed only after a real miss throws (cache‑after‑miss), preserving error semantics and invalidation on `layoutVersion` changes.
- IndexRef PIC augments the existing direct path and uses move‑to‑front promotion; it is keyed on `(classId, layoutVersion)` like other PICs.

View File

@ -1,62 +0,0 @@
## Pi Spigot JVM Baseline
Saved on April 4, 2026 before the `List<Int>` indexed-access follow-up fix.
Benchmark target:
- [examples/pi-bench.py](/home/sergeych/dev/lyng/examples/pi-bench.py)
- [examples/pi-bench.lyng](../examples/pi-bench.lyng)
Execution path:
- Python: `python3 examples/pi-bench.py`
- Lyng JVM: `./gradlew :lyng:runJvm --args='/home/sergeych/dev/lyng/examples/pi-bench.lyng'`
- Constraint: do not use Kotlin/Native `lyng` CLI for perf comparisons
Baseline measurements:
- Python full script: `167 ms`
- Lyng JVM full script: `1.287097604 s`
- Python warm function average over 5 runs: `126.126 ms`
- Lyng JVM warm function average over 5 runs: about `1071.6 ms`
Baseline ratio:
- Full script: about `7.7x` slower on Lyng JVM
- Warm function only: about `8.5x` slower on Lyng JVM
Primary finding at baseline:
- The hot `reminders[j]` accesses in `piSpigot` were still lowered through boxed object index ops and boxed arithmetic.
- Newly added `GET_INDEX_INT` and `SET_INDEX_INT` only reached `pi`, not `reminders`.
- Root cause: initializer element inference handled list literals, but not `List.fill(boxes) { 2 }`, so `reminders` did not become known `List<Int>` at compile time.
## After Optimizations 1-4
Follow-up change:
- propagate inferred lambda return class into bytecode compilation
- infer `List.fill(...)` element type from the fill lambda
- lower `reminders[j]` reads and writes to `GET_INDEX_INT` and `SET_INDEX_INT`
- add primitive-backed `ObjList` storage for all-int lists
- lower `List.fill(Int) { Int }` to `LIST_FILL_INT`
- stop boxing the integer index inside `GET_INDEX_INT` / `SET_INDEX_INT`
Verification:
- `piSpigot` disassembly now contains typed ops for `reminders`, for example:
- `GET_INDEX_INT s5(reminders), s10(j), ...`
- `SET_INDEX_INT s5(reminders), s10(j), ...`
Post-change measurements using `jlyng`:
- Full script: `655.819559 ms`
- Warm 5-run total: `1.430945810 s`
- Warm average per run: about `286.2 ms`
Observed improvement vs baseline:
- Full script: about `1.96x` faster (`1.287 s -> 0.656 s`)
- Warm function: about `3.74x` faster (`1071.6 ms -> 286.2 ms`)
Residual gap vs Python baseline:
- Full script: Lyng JVM is still about `3.9x` slower than Python (`655.8 ms` vs `167 ms`)
- Warm function: Lyng JVM is still about `2.3x` slower than Python (`286.2 ms` vs `126.126 ms`)
Current benchmark-test snapshot (`n=200`, JVM test harness):
- `optimized-int-division-rval-off`: `135 ms`
- `optimized-int-division-rval-on`: `125 ms`
- `piSpigot` bytecode now contains:
- `LIST_FILL_INT` for both `pi` and `reminders`
- `GET_INDEX_INT` / `SET_INDEX_INT` for the hot indexed loop

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

@ -1,50 +0,0 @@
class Tag(name: String) {
val name = name
var inner = ""
fun child(tagName: String, block: Tag.()->void) {
val child = Tag(tagName)
with(child) { block(this) }
inner += child.render()
}
fun head(block: Tag.()->void) { child("head", block) }
fun body(block: Tag.()->void) { child("body", block) }
fun title(block: Tag.()->void) { child("title", block) }
fun h1(block: Tag.()->void) { child("h1", block) }
fun addText(text: String) {
inner += text
}
fun render() {
"<" + name + ">" + inner + "</" + name + ">"
}
}
context(Tag)
fun String.unaryPlus() {
this@Tag.addText(this)
}
fun html(block: Tag.()->void) {
val root = Tag("html")
with(root) { block(this) }
root.render()
}
val page = html {
head {
title {
+"Demo"
}
}
body {
h1 {
+"Heading 1"
}
}
}
println(page)
assertEquals("<html><head><title>Demo</title></head><body><h1>Heading 1</h1></body></html>", page)

View File

@ -1,71 +0,0 @@
// Sample: Operator Overloading in Lyng
class Vector<T>(val x: T, val y: T) {
// Overload unary +
fun unaryPlus() = this
// 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 unary +
val v0 = +v1
println("+v1 = " + v0)
assertEquals(Vector(10, 20), v0)
// 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

@ -1,10 +0,0 @@
fun describe<T>(x: T): String = when (T) {
nullable -> "nullable"
else -> "non-null"
}
type MaybeInt = Int?
assert(MaybeInt is nullable)
assert(!(Int is nullable))
assertEquals("nullable", describe<Int?>(null))
assertEquals("non-null", describe<Int>(1))

View File

@ -1,19 +1,95 @@
# 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.
- **Exception: `compile if` can skip dead branches**: inside an untaken `compile if (...)` branch, names are not resolved or type-checked at all. This is the supported way to guard optional classes or packages such as `defined(Udp)` or `defined(lyng.io.net)`.
- **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.
## Practical Example: `cached`
The `cached` function (defined in `lyng.stdlib`) is a classic example of using closures to maintain state. It wraps a builder into a zero-argument function that computes once and remembers the result:
```lyng
fun cached(builder) {
var calculated = false
var value = null
{ // This lambda captures `calculated`, `value`, and `builder`
if( !calculated ) {
value = builder()
calculated = true
}
value
}
}
```
Because Lyng now correctly isolates closures for each evaluation of a lambda literal, using `cached` inside a class instance works as expected: each instance maintains its own private `calculated` and `value` state, even if they share the same property declaration.
## 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

@ -1,40 +1,6 @@
# Lyng serialization
Lyng has a built-in serialization module, `lyng.serialization`.
There are now two built-in formats with different goals:
- `Lynon`: the canonical binary format for Lyng values.
- `Json`: the canonical JSON-based round-trip format for Lyng values.
In addition, `Obj.toJson()` / `toJsonString()` remain available as a plain JSON projection for interoperability with
regular JSON tools and Kotlin `kotlinx.serialization`.
## Canonical formats
`Lynon` and `Json` are both exposed as format objects with the same surface:
- `Format.encode(value)`
- `Format.decode(encodedValue)`
For the built-in formats:
- `Lynon.encode(x)` returns `BitBuffer`
- `Lynon.decode(bitBuffer)` returns the original Lyng value
- `Json.encode(x)` returns `String`
- `Json.decode(jsonString)` returns the original Lyng value
`Json` also provides a typed canonical mode:
- `Json.encodeAs(Type, value)` returns `String`
- `Json.decodeAs(Type, jsonString)` returns the original Lyng value of the specified type
This is still canonical JSON, but it is schema-driven instead of fully self-describing.
## Lynon
Lynon is LYng Object Notation. It is typed, binary, bit-effective, implements caching, automatic compression,
variable-length integers, one-bit booleans, and preserves Lyng runtime structure well.
Lyng has builting binary bit-effective serialization format, called Lynon for LYng Object Notation. It is typed, binary, implements caching, automatic compression, variable-length ints, one-bit Booleans an many nice features.
It is as simple as:
@ -51,164 +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)
}
}
```
Transient fields:
- Are **omitted** from Lynon binary streams.
- Are **omitted** from JSON output (`toJson`) and canonical `Json.encode(...)`.
- 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`).
- **Exceptions**: canonical formats preserve exception class, message, extra data, and captured stack trace.
## Plain JSON projection vs canonical Json format
There are two JSON-related APIs and they serve different purposes:
- `Obj.toJson()` / `toJsonString()`
- produce ordinary JSON values
- best for interop with external JSON systems
- best for `Obj.decodeSerializable()` / `decodeSerializableWith()`
- may be lossy for Lyng-specific structures
- `Json.encode()` / `Json.decode()`
- produce JSON text too
- use Lyng-specific type tags so the payload is self-describing
- intended for round-tripping Lyng values
- intended to match Lynon semantics where JSON can carry them
- still keep ordinary string-key maps in traditional JSON object form
- can preserve values that plain JSON cannot represent directly, such as:
- maps with non-string keys
- sets
- buffers and bit buffers
- class instances
- singleton objects
- enums
- exceptions
- date/time objects
- non-finite reals
- `void`
- `Json.encodeAs(Type, value)` / `Json.decodeAs(Type, text)`
- also round-trip Lyng values through JSON text
- use the declared or requested type as decoding schema
- recursively omit type tags when the declared type is already exact enough
- keep canonical tags when the runtime value is more specific than the declared type
- produce less noisy JSON for closed and otherwise precisely-typed object graphs
- still keep ordinary `Map<String, ...>` values in traditional JSON object form
Why this split exists:
- plain `toJson()` must remain ordinary JSON so it stays convenient for external JSON systems and Kotlin
`kotlinx.serialization`
- canonical `Json.encode()` is for Lyng-to-Lyng transport through JSON text without any external schema, so it must
remain self-describing and preserve Lyng runtime
distinctions whenever possible
- `Json.encodeAs()` exists for the cases where a schema is known on both sides and we want canonical round-trip
behavior with fewer tags
- one API cannot optimize for both goals at once: either you get too many Lyng tags for ordinary JSON interop, or you
get lossy round-trips
Example:
import lyng.serialization
import lyng.time
enum Color { Red, Green }
class Point(x,y) { var z = 42 }
val p = Point(1,2)
p.z = 99
val x = List(
p,
Map([1, "one"], ["two", 2]),
Set(1,2,3),
"hello".encodeUtf8(),
Date(2026,4,15),
Color.Green
)
assertEquals(x, Json.decode(Json.encode(x)))
assertEquals( 5, p.x )
assertEquals( 6, p.y )
>>> void
Typed canonical example:
import lyng.serialization
closed class Point(x: Int, y: Int)
closed class Segment(a: Point, b: Point)
val value = Segment(Point(0,1), Point(2,3))
val json = Json.encodeAs(Segment, value)
assertEquals("{\"a\":{\"x\":0,\"y\":1},\"b\":{\"x\":2,\"y\":3}}", json)
assertEquals(value, Json.decodeAs(Segment, json))
>>> void
## Adding more formats from Kotlin modules
External modules can add new formats on the Kotlin side.
The common base class is:
```kotlin
abstract class ObjSerializationFormatClass(className: String) : ObjClass(className) {
abstract suspend fun encodeValue(scope: Scope, value: Obj): Obj
abstract suspend fun decodeValue(scope: Scope, encoded: Obj): Obj
}
```
To export a new format from a module:
```kotlin
im.addPackage("test.formats") { module ->
module.bindSerializationFormat(
object : ObjSerializationFormatClass("Reverse") {
override suspend fun encodeValue(scope: Scope, value: Obj): Obj =
ObjString(value.toString(scope).value.reversed())
override suspend fun decodeValue(scope: Scope, encoded: Obj): Obj =
ObjString((encoded as ObjString).value.reversed())
}
)
}
```
Then from Lyng, after importing the Kotlin module above, usage looks like:
```lyng
import test.formats
assertEquals("cba", Reverse.encode("abc"))
assertEquals("abc", Reverse.decode("cba"))
```
## Notes
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:
@ -217,3 +42,5 @@ Important is to understand that normally `Lynon.decode` wants [BitBuffer], as `L
this possibly creates extra zero bits at the end, as bit content could be shorter than byte-grained but for the Lynon format it does not make sense. Note that when you serialize [BitBuffer], exact number of bits is written. To convert bit buffer to bytes:
Lynon.encode("hello").toBuffer()
(topic is incomplete and under construction)

View File

@ -1,185 +1,184 @@
# Lyng time functions
Lyng date and time support requires importing `lyng.time`. The module provides four related types:
Lyng date and time support requires importing `lyng.time` packages. Lyng uses simple yet modern time object models:
- `Instant` for absolute timestamps.
- `Date` for calendar dates without time-of-day or timezone.
- `DateTime` for calendar-aware points in time in a specific timezone.
- `Duration` for absolute elapsed time.
- `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`
`Instant` represents some moment of time independently of the calendar. It is similar to SQL `TIMESTAMP`
or Kotlin `Instant`.
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()
val t2 = Instant(1704110400)
val t3 = Instant("2024-01-01T12:00:00.123456Z")
val t4 = t3.truncateToMinute()
assertEquals("2024-01-01T12:00:00Z", t4.toRFC3339())
val dt = t3.toDateTime("+02:00")
assertEquals(14, dt.hour)
val d = t3.toDate("Z")
assertEquals(Date(2024, 1, 1), d)
### Instant members
| member | description |
|--------------------------------|------------------------------------------------------|
| epochSeconds: Real | offset in seconds since Unix epoch |
| epochWholeSeconds: Int | whole seconds since Unix epoch |
| nanosecondsOfSecond: Int | nanoseconds within the current second |
| isDistantFuture: Bool | true if it is `Instant.distantFuture` |
| isDistantPast: Bool | true if it is `Instant.distantPast` |
| truncateToMinute(): Instant | truncate to minute precision |
| truncateToSecond(): Instant | truncate to second precision |
| truncateToMillisecond(): Instant | truncate to millisecond precision |
| truncateToMicrosecond(): Instant | truncate to microsecond precision |
| toRFC3339(): String | format as RFC3339 string in UTC |
| toDateTime(tz?): DateTime | localize to a timezone |
| toDate(tz?): Date | convert to a calendar date in a timezone |
## Calendar date: `Date`
`Date` represents a pure calendar date. It has no time-of-day and no attached timezone. Use it for values
like birthdays, due dates, invoice dates, and SQL `DATE` columns.
### Constructing
import lyng.time
val today = Date()
val d1 = Date(2026, 4, 15)
val d2 = Date("2024-02-29")
val d3 = Date.parseIso("2024-02-29")
val d4 = Date(DateTime(2024, 5, 20, 15, 30, 45, "+02:00"))
val d5 = Date(Instant("2024-01-01T23:30:00Z"), "+02:00")
### Date members
| member | description |
|--------------------------------|------------------------------------------------------------|
| year: Int | year component |
| month: Int | month component (1..12) |
| day: Int | day of month (alias `dayOfMonth`) |
| dayOfWeek: Int | day of week (1=Monday, 7=Sunday) |
| dayOfYear: Int | day of year (1..365/366) |
| isLeapYear: Bool | whether this date is in a leap year |
| lengthOfMonth: Int | number of days in this month |
| lengthOfYear: Int | 365 or 366 |
| toIsoString(): String | ISO `YYYY-MM-DD` string |
| toSortableString(): String | alias to `toIsoString()` |
| toDateTime(tz="Z"): DateTime | start-of-day `DateTime` in the specified timezone |
| atStartOfDay(tz="Z"): DateTime | alias to `toDateTime()` |
| addDays(n): Date | add or subtract calendar days |
| addMonths(n): Date | add or subtract months, normalizing end-of-month |
| addYears(n): Date | add or subtract years |
| daysUntil(other): Int | calendar days until `other` |
| daysSince(other): Int | calendar days since `other` |
| static today(tz?): Date | today in the specified timezone |
| static parseIso(s): Date | parse ISO `YYYY-MM-DD` |
### Date arithmetic
`Date` supports only whole-day arithmetic. This is deliberate: calendar dates should not silently accept
sub-day durations.
import lyng.time
val d1 = Date(2026, 4, 15)
val d2 = d1.addDays(10)
assertEquals(Date(2026, 4, 25), d2)
assertEquals(Date(2026, 4, 18), d1 + 3.days)
assertEquals(Date(2026, 4, 12), d1 - 3.days)
assertEquals(10, d1.daysUntil(d2))
assertEquals(10, d2.daysSince(d1))
assertEquals(10, d2 - d1)
### Date conversions
import lyng.time
val i = Instant("2024-01-01T23:30:00Z")
assertEquals(Date(2024, 1, 1), i.toDate("Z"))
assertEquals(Date(2024, 1, 2), i.toDate("+02:00"))
val dt = DateTime(2024, 5, 20, 15, 30, 45, "+02:00")
assertEquals(Date(2024, 5, 20), dt.date)
assertEquals(Date(2024, 5, 20), dt.toDate())
assertEquals(DateTime(2024, 5, 20, 0, 0, 0, "Z"), Date(2024, 5, 20).toDateTime("Z"))
assertEquals(DateTime(2024, 5, 20, 0, 0, 0, "+02:00"), Date(2024, 5, 20).atStartOfDay("+02:00"))
## Calendar time: `DateTime`
`DateTime` represents a point in time in a specific timezone. It provides access to calendar components
such as year, month, day, and hour.
### Constructing
import lyng.time
val now = DateTime.now()
val offsetTime = DateTime.now("+02:00")
val dt = Instant().toDateTime("Z")
val dt2 = DateTime(2024, 1, 1, 12, 0, 0, "Z")
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 |
| date: Date | calendar date component |
| toInstant(): Instant | convert back to absolute Instant |
| toDate(): Date | extract the calendar date in this timezone |
| 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:
import lyng.time
val leapDay = Instant("2024-02-29T12:00:00Z").toDateTime("Z")
val nextYear = leapDay.addYears(1)
assertEquals(28, nextYear.day)
# `Duration` class
`Duration` represents absolute elapsed time between two instants.
import lyng.time
val t1 = Instant()
delay(1.millisecond)
val t2 = Instant()
assert(t2 - t1 >= 1.millisecond)
assert(t2 - t1 < 100.millisecond)
assert( t2 - t1 < 1.millisecond )
assert( t2.epochSeconds - t1.epochSeconds < 0.001 )
>>> void
Duration values can be created from numbers using extensions on `Int` and `Real`:
## 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 (1) |
| isDistantFuture: Bool | true if it `Instant.distantFuture` |
| isDistantPast: Bool | true if it `Instant.distantPast` |
| truncateToSecond: Intant | create new instnce truncated to second |
| truncateToMillisecond: Instant | truncate new instance with to millisecond |
| truncateToMicrosecond: Instant | truncate new instance to microsecond |
(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)
## Class members
| 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 |
# `Duraion` class
Represent absolute time distance between two `Instant`.
import lyng.time
val t1 = Instant()
// yes we can delay to period, and it is not blocking. is suspends!
delay(1.millisecond)
val t2 = Instant()
// be suspend, so actual time may vary:
assert( t2 - t1 >= 1.millisecond)
assert( t2 - t1 < 100.millisecond)
>>> void
Duration can be converted from numbers, like `5.minutes` and so on. Extensions are created for
`Int` and `Real`, so for n as Real or Int it is possible to create durations::
- `n.millisecond`, `n.milliseconds`
- `n.second`, `n.seconds`
@ -187,9 +186,10 @@ Duration values can be created from numbers using extensions on `Int` and `Real`
- `n.hour`, `n.hours`
- `n.day`, `n.days`
Larger units like months or years are calendar-dependent and are intentionally not part of `Duration`.
The bigger time units like months or years are calendar-dependent and can't be used with `Duration`.
Each duration instance can be converted to numbers in these units:
Each duration instance can be converted to number of any of these time units, as `Real` number, if `d` is a `Duration`
instance:
- `d.microseconds`
- `d.milliseconds`
@ -198,16 +198,18 @@ Each duration instance can be converted to numbers in these units:
- `d.hours`
- `d.days`
Example:
for example
import lyng.time
assertEquals( 60, 1.minute.seconds )
assertEquals( 10.milliseconds, 0.01.seconds )
assertEquals(60, 1.minute.seconds)
assertEquals(10.milliseconds, 0.01.seconds)
>>> void
# Utility functions
## `delay(duration: Duration)`
## delay(duration: Duration)
Suspends current coroutine for at least the specified duration.
Suspends the current coroutine for at least the specified duration.

View File

@ -8,13 +8,12 @@ __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)
- [math in Lyng](math.md), [the `when` statement](when.md)
- [Testing and Assertions](Testing.md)
- [Generics and type expressions](generics.md)
- [time](time.md) and [parallelism](parallelism.md)
- [parallelism] - multithreaded code, coroutines, etc.
- Some class
references: [List], [ImmutableList], [Set], [ImmutableSet], [Map], [ImmutableMap], [Real], [Range], [Iterable], [Iterator], [time manipulation](time.md), [Array], [RingBuffer], [Buffer].
references: [List], [Set], [Map], [Real], [Range], [Iterable], [Iterator], [time manipulation](time.md), [Array], [RingBuffer], [Buffer].
- Some samples: [combinatorics](samples/combinatorics.lyng.md), national vars and
loops: [сумма ряда](samples/сумма_ряда.lyng.md). More at [samples folder](samples)
@ -33,15 +32,6 @@ 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() {
@ -107,23 +97,6 @@ Singleton objects are declared using the `object` keyword. They define a class a
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.
@ -206,13 +179,14 @@ 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 +203,9 @@ 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 +216,29 @@ 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 +283,8 @@ 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
@ -336,56 +292,12 @@ 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
Receiver lambdas can also keep outer receivers in scope. The primary receiver wins for unqualified lookup, and `this@Type`
selects an outer receiver explicitly:
class Html { fun lang() = "en" }
class Body { fun lang() = "body" }
fun html(block: Html.()->String) = with(Html()) { block(this) }
fun body(block: Body.()->String) = with(Body()) { block(this) }
val result = html {
body {
lang() + ":" + this@Html.lang()
}
}
assertEquals("body:en", result)
>>> void
You can declare the same requirement in a function type:
val block: context(Html) Body.()->String = {
lang() + ":" + this@Html.lang()
}
If the primary receiver does not define a member and multiple outer/context receivers do, Lyng reports an ambiguity instead of picking one silently:
class A { fun title() = "a" }
class B { fun title() = "b" }
class C
val block: context(A, B) C.()->String = {
// title() // compile-time ambiguity
this@A.title()
}
## run
Executes a block after it returning the value passed by the block. for example, can be used with elvis operator:
@ -409,18 +321,6 @@ It is rather simple, like everywhere else:
See [math](math.md) for more on it. Notice using Greek as identifier, all languages are allowed.
For linear algebra, import `lyng.matrix`:
import lyng.matrix
val a: Matrix = matrix([[1, 2], [3, 4]])
val i: Matrix = Matrix.identity(2)
val sum: Matrix = a + i
assertEquals([[2.0, 2.0], [3.0, 5.0]], sum.toList())
>>> void
See [Matrix](Matrix.md) for vectors, matrix multiplication, inversion, and slicing such as `m[0..2, 1]`.
Logical operation could be used the same
var x = 10
@ -500,6 +400,8 @@ Almost the same, using `val`:
val foo = 1
foo += 1 // this will throw exception
# Constants
Same as in kotlin:
val HalfPi = π / 2
@ -507,163 +409,6 @@ 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.
For type expressions, you can check nullability directly:
T is nullable
T !is nullable
This is especially useful in generic code and in `when` over a type parameter:
fun describe<T>(x: T): String = when (T) {
nullable -> "nullable"
else -> "non-null"
}
## 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) {
@ -705,9 +450,8 @@ 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
@ -800,11 +544,6 @@ 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)
@ -815,7 +554,6 @@ See also: [Testing and Assertions](Testing.md)
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
@ -843,13 +581,6 @@ Lyng has built-in mutable array class `List` with simple literals:
[List] is an implementation of the type `Array`, and through it `Collection` and [Iterable]. Please read [Iterable],
many collection based methods are implemented there.
For immutable list values, use `list.toImmutable()` and [ImmutableList].
To construct a list programmatically, use the static helper `List.fill`:
val tens = List.fill(5) { index -> index * 10 }
assertEquals([0, 10, 20, 30, 40], tens)
>>> void
Lists can contain any type of objects, lists too:
@ -858,19 +589,11 @@ 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).
In general, bracket indexing may contain more than one selector:
value[i]
value[i, j]
For built-in lists, strings, maps, and buffers, the selector is usually a single value such as an `Int`, `Range`, or `Regex`.
For types with custom indexers, multiple selectors are packed into one list-like index object and passed to `getAt` / `putAt`.
When you want to "flatten" it to single array, you can use splat syntax:
[1, ...[2,3], 4]
@ -1040,7 +763,6 @@ Set are unordered collection of unique elements, see [Set]. Sets are [Iterable]
>>> void
Please see [Set] for detailed description.
For immutable set values, use `set.toImmutable()` and [ImmutableSet].
# Maps
@ -1101,7 +823,6 @@ Notes:
- When you need computed (expression) keys or non-string keys, use `Map(...)` constructor with entries, e.g. `Map( ("a" + "b") => 1 )`, then merge with a literal if needed: `{ base: } + (computedKey => 42)`.
Please see the [Map] reference for a deeper guide.
For immutable map values, use `map.toImmutable()` and [ImmutableMap].
# Flow control operators
@ -1128,37 +849,6 @@ Or, more neat:
>>> just 3
>>> void
## compile if
`compile if` is a compile-time conditional. Unlike normal `if`, the compiler evaluates its condition while compiling
the file and completely skips the untaken branch. This is useful when some class or package may or may not be
available:
compile if (defined(Udp)) {
val socket = Udp()
println("udp is available")
} else {
println("udp is not available")
}
`compile if` also supports single-statement branches:
compile if (defined(lyng.io.net) && !defined(Udp))
println("network module exists, but Udp is not visible here")
else
println("either Udp exists or the module is unavailable")
Current condition syntax is intentionally limited to compile-time symbol checks:
- `defined(Name)`
- `defined(package.name)`
- `!`, `&&`, `||`
- parentheses
Examples:
compile if (defined(Udp) && defined(Tcp))
println("both transports are available")
## When
See also: [Comprehensive guide to `when`](when.md)
@ -1347,8 +1037,8 @@ 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
@ -1423,48 +1113,12 @@ 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"
When you need a counting loop that goes backwards, use an explicit descending
range:
var sum = 0
for( i in 5 downTo 1 ) {
sum += i
}
sum
>>> 15
If the lower bound should be excluded, use `downUntil`:
val xs = []
for( i in 5 downUntil 1 ) {
xs.add(i)
}
xs
>>> [5,4,3,2]
This is intentionally explicit: `5..1` is an empty ascending range, not an
implicit reverse loop.
Descending loops also support `step`:
val xs = []
for( i in 10 downTo 1 step 3 ) {
xs.add(i)
}
xs
>>> [10,7,4,1]
For descending ranges, `step` stays positive. The direction comes from
`downTo` / `downUntil`, so `10 downTo 1 step 3` is valid, while
`10 downTo 1 step -3` is an error.
For loop support breaks the same as while loops above:
fun search(haystack, needle) {
@ -1573,7 +1227,7 @@ 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
@ -1594,12 +1248,6 @@ It could be open and closed:
assert( 5 !in (1..<5) )
>>> void
Descending ranges are explicit too:
assertEquals([5,4,3,2,1], (5 downTo 1).toList())
assertEquals([5,4,3,2], (5 downUntil 1).toList())
>>> void
Ranges could be inside other ranges:
assert( (2..3) in (1..10) )
@ -1612,19 +1260,11 @@ There are character ranges too:
and you can use ranges in for-loops:
for( x in 'a'..<'c' ) println(x)
for( x in 'a' ..< 'c' ) println(x)
>>> a
>>> b
>>> void
Descending character ranges work the same way:
for( ch in 'e' downTo 'a' step 2 ) println(ch)
>>> e
>>> c
>>> a
>>> void
See [Ranges](Range.md) for detailed documentation on it.
# Time routines
@ -1665,7 +1305,7 @@ 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 |
|--------|---------------------------------|---------------------|
@ -1675,7 +1315,6 @@ 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 | |
@ -1688,94 +1327,28 @@ The type for the character objects is `Char`.
### String literal escapes
Lyng string literals can use either double quotes or backticks:
val a = "hello"
val b = `hello`
assert(a == b)
| escape | ASCII value |
|--------|-----------------------|
| \n | 0x10, newline |
| \r | 0x13, carriage return |
| \t | 0x07, tabulation |
| \\ | \ slash character |
| \uXXXX | unicode code point |
Delimiter-specific escapes:
| form | escape | value |
|--------|--------|------------------|
| `"..."` | \" | " double quote |
| `` `...` `` | \` | ` backtick |
Unicode escape form is exactly 4 hex digits, e.g. `"\u263A"` -> `☺`.
| \" | " double quote |
Other `\c` combinations, where c is any char except mentioned above, are left intact, e.g.:
val s = "\a"
assert(s[0] == '\\')
assert(s[0] == '\')
assert(s[1] == 'a')
>>> void
same as:
val s = "\\a"
assert(s[0] == '\\')
assert(s[0] == '\')
assert(s[1] == 'a')
>>> void
### String interpolation
Supported forms:
- `$name`
- `${expr}`
Literal dollar forms:
- `\$` -> `$`
- `$$` -> `$`
Example:
val name = "Lyng"
assertEquals("hello, Lyng!", "hello, $name!")
assertEquals("hello, Lyng!", `hello, $name!`)
assertEquals("sum=3", "sum=${1+2}")
assertEquals("sum=3", `sum=${1+2}`)
assertEquals("\$name", "\$name")
assertEquals("\$name", "$$name")
assertEquals("\$name", `\$name`)
assertEquals("\$name", `$$name`)
assertEquals("\\Lyng", "\\$name")
assertEquals("\\Lyng", `\\$name`)
>>> void
Interpolation and `printf`-style formatting can be combined when needed:
val method = "transfer"
val argc = 2
val compact = "%s:%d"(method, argc)
assertEquals("call=transfer:2", "call=$compact")
assertEquals("[transfer:2] ok", "[${"%s:%d"(method, argc)}] ok")
>>> void
Interpolation also works well with regex patterns. To keep a literal `$` in the
regex, escape it in the resulting pattern:
val currency = "USD"
val amount = 15
val escapedDollar = "\\$"
val re = Regex("^${currency}${escapedDollar}${amount}$")
assert("USD$15" =~ re)
assert("USD15" !~ re)
>>> void
If you need old literal behavior in a file, add a leading directive comment:
// feature: interpolation: off
### Char literal escapes
Are the same as in string literals with little difference:
@ -1787,9 +1360,6 @@ Are the same as in string literals with little difference:
| \t | 0x07, tabulation |
| \\ | \ slash character |
| \' | ' apostrophe |
| \uXXXX | unicode code point |
For char literals, use `'\\'` to represent a single backslash character; `'\'` is invalid.
### Char instance members
@ -1856,14 +1426,6 @@ Open-ended ranges could be used to get start and end too:
assertEquals( "pult", "catapult"[ 4.. ])
>>> void
The same bracket syntax is also used by imported numeric modules such as `lyng.matrix`, where indexing can be multi-axis:
import lyng.matrix
val m: Matrix = matrix([[1, 2, 3], [4, 5, 6]])
assertEquals(6.0, m[1, 2])
>>> void
### String operations
Concatenation is a `+`: `"hello " + name` works as expected. No confusion. There is also
@ -1875,39 +1437,22 @@ Concatenation is a `+`: `"hello " + name` works as expected. No confusion. There
Extraction:
("abcd42def"[ "\d+".re ] as RegexMatch).value
"abcd42def"[ "\d+".re ].value
>>> "42"
Part match:
assert( "abc foo def" =~ "f[oO]+".re )
assert( "foo" == ($~ as RegexMatch).value )
assert( "foo" == $~.value )
>>> void
Replacing text:
assertEquals("bonono", "banana".replace('a', 'o'))
assertEquals("a-b-c", "a.b.c".replace(".", "-")) // string patterns are literal
assertEquals("v#.#.#", "v1.2.3".replace("\d+".re, "#"))
assertEquals("v[1].[2].[3]", "v1.2.3".replace("(\d+)".re) { m -> "[" + m[1] + "]" })
>>> void
Repeating the fragment:
assertEquals("hellohello", "hello"*2)
assertEquals("", "hello"*0)
>>> void
A typical set of String functions includes:
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 |
@ -1924,11 +1469,9 @@ A typical set of String functions includes:
| 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) |
| characters() | create [List] of characters (1) |
| encodeUtf8() | returns [Buffer] with characters encoded to utf8 |
| matches(re) | matches the regular expression (2) |
| replace(old, new) | replace all literal or regex matches; regex needs [Regex] |
| replaceFirst(old,new)| replace the first literal or regex match |
| | |
(1)
@ -1976,7 +1519,7 @@ See [math functions](math.md). Other general purpose functions are:
| 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 |
| let, also, apply, run | see above, flow controls |
(1)
: `fn` is optional lambda returning string message to add to exception string.
@ -1992,7 +1535,6 @@ Lambda avoid unnecessary execution if assertion is not failed. for example:
| π | See [math](math.md) |
[List]: List.md
[ImmutableList]: ImmutableList.md
[Testing]: Testing.md
@ -2004,15 +1546,13 @@ 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
[Set]: Set.md
[ImmutableSet]: ImmutableSet.md
[Map]: Map.md
[ImmutableMap]: ImmutableMap.md
[Buffer]: Buffer.md
@ -2130,7 +1670,7 @@ 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
@ -2145,7 +1685,7 @@ You can add new methods and properties to existing classes without modifying the
### Extension properties
val Int.isEven get() = this % 2 == 0
val Int.isEven = this % 2 == 0
4.isEven
>>> true
@ -2155,30 +1695,6 @@ Example with custom accessors:
"abc".firstChar
>>> 'a'
### Extension indexers
Indexers can also be extended by overriding `getAt` and `putAt` on the receiver:
```lyng
object Storage
var storageData = {}
override fun Storage.getAt(key: String): Object? {
storageData[key]
}
override fun Storage.putAt(key: String, value: Object) {
storageData[key] = value
}
Storage["answer"] = 42
val answer: Int? = Storage["answer"]
assertEquals(42, answer)
```
This works for classes and named singleton `object` declarations. Bracket syntax is lowered to `getAt` / `putAt`, and multiple selectors are packed into one list-like index object the same way as other custom indexers.
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

@ -1,235 +1,9 @@
# What's New in Lyng
This document highlights the current Lyng release, **1.5.5**, and the broader additions from the 1.5 cycle.
It is intentionally user-facing: new language features, new modules, new tools, and the practical things you can build with them.
For a programmer-focused migration summary across 1.5.x, see `docs/whats_new_1_5.md`.
## Release 1.5.5 Highlights
- `1.5.5` extends the 1.5 line with practical database APIs, first-class calendar dates, and better coroutine building blocks.
- The 1.5 line now brings together richer ranges and loops, interpolation, math modules, immutable and observable collections, richer `lyngio`, and much better CLI/IDE support.
- `1.5.5` adds `Channel`, `LaunchPool`, and `joinAll()` so coroutine-heavy scripts can coordinate work more directly.
- `1.5.5` adds `Date`, the portable `lyng.io.db` layer, SQLite/JDBC providers, and a compatibility `lyng.legacy_digest` module.
- `1.5.5` also continues runtime/compiler hardening with better import dispatch, faster exact lambda calls, and correct `val +=`/`-=` behavior for mutating types versus real reassignment.
- The docs, homepage samples, and release metadata now point at the current stable version.
## User Highlights Across 1.5.x
- Descending ranges and loops with `downTo` / `downUntil`
- String interpolation with `$name` and `${expr}`
- Backtick string literals for raw-ish string text
- Decimal arithmetic, matrices/vectors, and complex numbers
- Calendar `Date` support in `lyng.time`
- `Channel`, `LaunchPool`, and `joinAll()` for coroutine workflows
- Immutable collections and opt-in `ObservableList`
- Rich `lyngio` modules for SQLite/JDBC databases, console, HTTP, WebSocket, TCP, and UDP
- Legacy SHA-1 compatibility helpers in `lyng.legacy_digest`
- CLI improvements including the built-in formatter `lyng fmt`
- Better IDE support and stronger docs around the released feature set
This document highlights the latest additions and improvements to the Lyng language and its ecosystem.
## Language Features
### Descending Ranges and Loops
Lyng ranges are no longer just ascending. You can now write explicit descending ranges with inclusive or exclusive lower bounds.
```lyng
assertEquals([5,4,3,2,1], (5 downTo 1).toList())
assertEquals([5,4,3,2], (5 downUntil 1).toList())
for (i in 10 downTo 1 step 3) {
println(i)
}
```
This also works for characters:
```lyng
assertEquals(['e','c','a'], ('e' downTo 'a' step 2).toList())
```
See [Range](Range.md).
### String Interpolation
Lyng 1.5.1 added built-in string interpolation:
- `$name`
- `${expr}`
Literal dollar forms are explicit too:
- `\$` -> `$`
- `$$` -> `$`
```lyng
val name = "Lyng"
assertEquals("hello, Lyng!", "hello, $name!")
assertEquals("sum=3", "sum=${1+2}")
assertEquals("\$name", "\$name")
assertEquals("\$name", "$$name")
```
If you need legacy literal-dollar behavior in a file, add:
```lyng
// feature: interpolation: off
```
See [Tutorial](tutorial.md).
### Matrix and Vector Module (`lyng.matrix`)
Lyng now ships a dense linear algebra module with immutable double-precision `Matrix` and `Vector` types.
It provides:
- `matrix([[...]])` and `vector([...])`
- matrix multiplication
- matrix inversion
- determinant, trace, rank
- solving `A * x = b`
- vector operations such as `dot`, `normalize`, `cross`, and `outer`
```lyng
import lyng.matrix
val a: Matrix = matrix([[4, 7], [2, 6]])
val inv: Matrix = a.inverse()
assert(abs(inv.get(0, 0) - 0.6) < 1e-9)
```
Matrices also support Lyng-style slicing:
```lyng
import lyng.matrix
val m: Matrix = matrix([[1, 2, 3], [4, 5, 6], [7, 8, 9]])
assertEquals(6.0, m[1, 2])
val column: Matrix = m[0..2, 2]
val tail: Matrix = m[1.., 1..]
assertEquals([[3.0], [6.0], [9.0]], column.toList())
assertEquals([[5.0, 6.0], [8.0, 9.0]], tail.toList())
```
See [Matrix](Matrix.md).
### Multiple Selectors in Bracket Indexing
Bracket indexing now accepts more than one selector:
```lyng
value[i]
value[i, j]
value[i, j, k]
```
For custom indexers, multiple selectors are packed into one list-like index object and dispatched through `getAt` / `putAt`.
This is the rule used by `lyng.matrix` and by embedding APIs for Kotlin-backed indexers.
### Decimal Arithmetic Module (`lyng.decimal`)
Lyng now ships a first-class decimal module built as a regular extension library rather than a deep core special case.
It provides:
- `Decimal`
- convenient `.d` conversions from `Int`, `Real`, and `String`
- mixed arithmetic with `Int` and `Real`
- local division precision and rounding control via `withDecimalContext(...)`
```lyng
import lyng.decimal
assertEquals("3", (1 + 2.d).toStringExpanded())
assertEquals("0.30000000000000004", (0.1 + 0.2).d.toStringExpanded())
assertEquals("0.3", "0.3".d.toStringExpanded())
assertEquals(
"0.3333333333",
withDecimalContext(10) { (1.d / 3.d).toStringExpanded() }
)
```
The distinction between `Real -> Decimal` and exact decimal parsing is explicit by design:
- `2.2.d` converts the current `Real` value
- `"2.2".d` parses exact decimal text
See [Decimal](Decimal.md).
### Complex Numbers (`lyng.complex`)
Lyng also ships a complex-number module for ordinary arithmetic in the complex plane.
```lyng
import lyng.complex
assertEquals(Complex(1.0, 2.0), 1 + 2.i)
assertEquals(Complex(2.0, 2.0), 2.i + 2)
val z = 1 + π.i
println(z.exp())
```
See [Complex](Complex.md).
### Legacy Digest Module (`lyng.legacy_digest`)
For situations where an external protocol or file format requires a SHA-1 value,
Lyng now ships a `lyng.legacy_digest` module backed by a pure Kotlin/KMP
implementation with no extra dependencies.
> ⚠️ SHA-1 is cryptographically broken. Use only for legacy-compatibility work.
```lyng
import lyng.legacy_digest
val hex = LegacyDigest.sha1("abc")
// → "a9993e364706816aba3e25717850c26c9cd0d89d"
// Also accepts raw bytes:
import lyng.buffer
val buf = Buffer.decodeHex("616263")
assertEquals(hex, LegacyDigest.sha1(buf))
```
The name `LegacyDigest` is intentional: it signals that these algorithms belong
to a compatibility layer, not to a current security toolkit.
See [LegacyDigest](LegacyDigest.md).
### Binary Operator Interop Registry
Lyng now provides a general mechanism for mixed binary operators through `lyng.operators`.
This solves cases like:
- `Int + MyType`
- `Real < MyType`
- `Int == MyType`
without requiring changes to built-in classes.
```lyng
import lyng.operators
class DecimalBox(val value: Int) {
fun plus(other: DecimalBox) = DecimalBox(value + other.value)
fun compareTo(other: DecimalBox) = value <=> other.value
}
OperatorInterop.register(
Int,
DecimalBox,
DecimalBox,
[BinaryOperator.Plus, BinaryOperator.Compare, BinaryOperator.Equals],
{ x: Int -> DecimalBox(x) },
{ x: DecimalBox -> x }
)
assertEquals(DecimalBox(3), 1 + DecimalBox(2))
assert(1 < DecimalBox(2))
assert(2 == DecimalBox(2))
```
`lyng.decimal` uses this same mechanism internally to interoperate with `Int` and `Real`.
See [Operator Interop Registry](OperatorInterop.md).
### 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.
@ -250,9 +24,6 @@ class Person(private var _age: Int) {
### 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
@ -327,64 +98,13 @@ Singleton objects are declared using the `object` keyword. They provide a conven
```lyng
object Config {
val version = "1.5.6-SNAPSHOT"
val version = "1.2.3"
fun show() = println("Config version: " + version)
}
Config.show()
```
Named singleton objects can also be used as extension receivers:
```lyng
object X {
fun base() = "base"
}
fun X.decorate(value): String {
this.base() + ":" + value.toString()
}
val X.tag get() = this.base() + ":tag"
assertEquals("base:42", X.decorate(42))
assertEquals("base:tag", X.tag)
```
### 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.
@ -403,188 +123,8 @@ var name by Observable("initial") { n, old, new ->
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.
### Immutable Collections
Lyng 1.5 adds immutable collection types for APIs that should not expose mutable state through aliases:
- `ImmutableList`
- `ImmutableSet`
- `ImmutableMap`
```lyng
val a = ImmutableList(1,2,3)
val b = a + 4
assertEquals(ImmutableList(1,2,3), a)
assertEquals(ImmutableList(1,2,3,4), b)
```
See [ImmutableList](ImmutableList.md), [ImmutableSet](ImmutableSet.md), and [ImmutableMap](ImmutableMap.md).
### Observable Mutable Lists
For reactive-style code, `lyng.observable` provides `ObservableList` with hooks and change streams.
```lyng
import lyng.observable
val xs = [1,2].observable()
xs.onChange { println("changed") }
xs += 3
```
You can validate or reject mutations in `beforeChange`, listen in `onChange`, and consume structured change events from `changes()`.
See [ObservableList](ObservableList.md).
### Random API
The standard library now includes a built-in random API plus deterministic seeded generators.
```lyng
val rng = Random.seeded(1234)
assert(rng.next(1..10) in 1..10)
assert(rng.next('a'..<'f') in 'a'..<'f')
```
Use:
- `Random.nextInt()`
- `Random.nextFloat()`
- `Random.next(range)`
- `Random.seeded(seed)`
## Tooling and Infrastructure
### Rich Console Apps with `lyng.io.console`
`lyngio` now includes a real console module for terminal applications:
- TTY detection
- screen clearing and cursor movement
- alternate screen buffer
- raw input mode
- typed key and resize events
```lyng
import lyng.io.console
Console.enterAltScreen()
Console.clear()
Console.moveTo(1, 1)
Console.write("Hello from Lyng console app")
Console.flush()
Console.leaveAltScreen()
```
The repository includes a full interactive Tetris sample built on this API.
See [lyng.io.console](lyng.io.console.md).
### HTTP, WebSocket, TCP, and UDP in `lyngio`
`lyngio` grew from filesystem/process support into a broader application-facing I/O library. In 1.5.x it includes:
- `lyng.io.http` for HTTP/HTTPS client calls
- `lyng.io.ws` for WebSocket clients
- `lyng.io.net` for raw TCP/UDP transport
HTTP example:
```lyng
import lyng.io.http
val r = Http.get("https://example.com")
println(r.status)
println(r.text())
```
TCP example:
```lyng
import lyng.io.net
val socket = Net.tcpConnect("127.0.0.1", 4040)
socket.writeUtf8("ping")
socket.flush()
println(socket.readLine())
socket.close()
```
WebSocket example:
```lyng
import lyng.io.ws
val ws = Ws.connect("wss://example.com/socket")
ws.sendText("hello")
println(ws.receive())
ws.close()
```
These modules are capability-gated and host-installed, keeping Lyng safe by default while making networked scripts practical when enabled.
See [lyngio overview](lyngio.md), [lyng.io.db](lyng.io.db.md), [lyng.io.http](lyng.io.http.md), [lyng.io.ws](lyng.io.ws.md), and [lyng.io.net](lyng.io.net.md).
### CLI: Formatting Command
A new `fmt` subcommand has been added to the Lyng CLI.
@ -594,28 +134,7 @@ lyng fmt --in-place MyFile.lyng # Format file in-place
lyng fmt --check MyFile.lyng # Check if file needs formatting
```
### CLI: Better Terminal Workflows
The CLI is no longer just a script launcher. In the 1.5 line it also gained:
- built-in formatter support
- integrated `lyng.io.console` support for terminal programs
- downloadable packaged distributions for easier local use
This makes CLI-first scripting and console applications much more practical than in earlier releases.
### 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 extern surfaces in Lyng (`extern` members, or members inside `extern class/object`) and bind the implementations in Kotlin before the first instance is created.
- **Extern declaration rule**: `extern class` / `extern object` are declaration-only; all members in their bodies are implicitly extern.
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

@ -1,145 +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).
Extern declaration clarification:
- `extern class` / `extern object` are pure extern surfaces.
- Members inside them are implicitly extern (`extern` on a member is optional/redundant).
- Lyng method/property bodies for these declarations should be implemented as extensions instead.
### 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

@ -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}_]*)\\.)?([\\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)\\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)\\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

@ -1,325 +0,0 @@
#!/usr/bin/env lyng
import lyng.io.db
import lyng.io.db.sqlite
import lyng.io.fs
val DB_FILE_NAME = "contents.db"
val ANSI_ESC = "\u001b["
val NEWLINE = "\n"
val WINDOWS_SEPARATOR = "\\"
val SQLITE_JOURNAL_SUFFIXES = ["-wal", "-shm", "-journal"]
val USAGE_TEXT = "
Lyng content index
Scan a directory tree, diff it against a SQLite snapshot, and optionally refresh the snapshot.
usage:
lyng examples/content_index_db.lyng <root> [-u|--update]
options:
-u, --update write the current scan back to $DB_FILE_NAME
notes:
- the database lives inside <root>/$DB_FILE_NAME
- on first run the snapshot is created automatically
- the script ignores its own SQLite sidecar files
"
val CREATE_FILE_INDEX_SQL = "
create table if not exists file_index(
path text primary key not null,
size integer not null,
mtime integer not null
)
"
val CREATE_CURRENT_SCAN_SQL = "
create temp table current_scan(
path text primary key not null,
size integer not null,
mtime integer not null
)
"
val SELECT_ADDED_SQL = "
select
c.path,
c.size,
c.mtime
from current_scan c
left join file_index f on f.path = c.path
where f.path is null
order by c.path
"
val SELECT_REMOVED_SQL = "
select f.path, f.size, f.mtime
from file_index f
left join current_scan c on c.path = f.path
where c.path is null
order by f.path
"
val SELECT_CHANGED_SQL = "
select c.path, f.size as old_size, c.size as new_size, f.mtime as old_mtime,
c.mtime as new_mtime
from current_scan c
join file_index f on f.path = c.path
where c.size != f.size or c.mtime != f.mtime
order by c.path
"
val DELETE_MISSING_SQL = "
delete from file_index
where not exists (
select 1
from current_scan c
where c.path = file_index.path
)
"
val UPSERT_SCAN_SQL = "
insert or replace into file_index(path, size, mtime)
select path, size, mtime
from current_scan
"
val INSERT_SCAN_ROW_SQL = "
insert into current_scan(path, size, mtime)
values(?, ?, ?)
"
val USE_COLOR = true
class CliOptions(val rootText: String, val updateSnapshot: Bool) {}
fun out(text: String? = null): Void {
if (text == null) {
print(NEWLINE)
return
}
print(text + NEWLINE)
}
fun paint(code: String, text: String): String {
if (!USE_COLOR) return text
ANSI_ESC + code + "m" + text + ANSI_ESC + "0m"
}
fun bold(text: String): String = paint("1", text)
fun dim(text: String): String = paint("2", text)
fun cyan(text: String): String = paint("36", text)
fun green(text: String): String = paint("32", text)
fun yellow(text: String): String = paint("33", text)
fun red(text: String): String = paint("31", text)
fun signed(value: Int): String = if (value > 0) "+" + value else value.toString()
fun plural(count: Int, one: String, many: String): String {
if (count == 1) return one
many
}
fun childPath(parent: Path, name: String): Path {
val base = parent.toString()
if (base.endsWith("/") || base.endsWith(WINDOWS_SEPARATOR)) {
return Path(base + name)
}
Path(base + "/" + name)
}
fun relativePath(root: Path, file: Path): String {
val parts: List<String> = []
for (i in root.segments.size..<file.segments.size) {
parts.add(file.segments[i] as String)
}
parts.joinToString("/")
}
fun isDatabaseArtifact(relative: String): Bool {
relative == DB_FILE_NAME || SQLITE_JOURNAL_SUFFIXES.any { relative == DB_FILE_NAME + (it as String) }
}
fun printUsage(message: String? = null): Void {
if (message != null && message.trim().isNotEmpty()) {
out(red("error: ") + message)
out()
}
out(bold(USAGE_TEXT))
}
fun parseArgs(argv: List<String>): CliOptions? {
var rootText: String? = null
var updateSnapshot = false
for (arg in argv) {
when (arg) {
"-u", "--update" -> updateSnapshot = true
"-h", "--help" -> {
printUsage()
return null
}
else -> {
if (arg.startsWith("-")) {
printUsage("unknown option: " + arg)
return null
}
if (rootText != null) {
printUsage("only one root path is allowed")
return null
}
rootText = arg
}
}
}
if (rootText == null) {
printUsage("missing required <root> argument")
return null
}
CliOptions(rootText as String, updateSnapshot)
}
fun printBanner(root: Path, dbFile: Path, dbWasCreated: Bool, updateSnapshot: Bool): Void {
val mode =
if (dbWasCreated) "bootstrap snapshot"
else if (updateSnapshot) "scan + refresh snapshot"
else "scan only"
out(cyan("== Lyng content index =="))
out(dim("root: " + root))
out(dim("db: " + dbFile))
out(dim("mode: " + mode))
out()
}
fun printSection(title: String, accent: (String)->String, rows: List<SqlRow>, render: (SqlRow)->String): Void {
out(accent(title + " (" + rows.size + ")"))
if (rows.isEmpty()) {
out(dim(" none"))
out()
return
}
for (row in rows) {
out(render(row))
}
out()
}
fun renderAdded(row: SqlRow): String {
val path = row["path"] as String
val size = row["size"] as Int
val mtime = row["mtime"] as Int
" " + green("+") + " " + bold(path) + dim(" %12d B mtime %d"(size, mtime))
}
fun renderRemoved(row: SqlRow): String {
val path = row["path"] as String
val size = row["size"] as Int
val mtime = row["mtime"] as Int
" " + red("-") + " " + bold(path) + dim(" %12d B mtime %d"(size, mtime))
}
fun renderChanged(row: SqlRow): String {
val path = row["path"] as String
val oldSize = row["old_size"] as Int
val newSize = row["new_size"] as Int
val oldMtime = row["old_mtime"] as Int
val newMtime = row["new_mtime"] as Int
val sizeDelta = newSize - oldSize
val mtimeDelta = newMtime - oldMtime
" " + yellow("~") + " " + bold(path) +
dim(
" size %d -> %d (%s B), mtime %d -> %d (%s ms)"(
oldSize,
newSize,
signed(sizeDelta),
oldMtime,
newMtime,
signed(mtimeDelta)
)
)
}
fun loadRows(tx: SqlTransaction, query: String): List<SqlRow> = tx.select(query).toList()
fun main() {
val argv: List<String> = []
for (raw in ARGV as List) {
argv.add(raw as String)
}
val options = parseArgs(argv)
if (options == null) {
return
}
val root = Path(options.rootText)
if (!root.exists()) {
printUsage("root does not exist: " + root)
return
}
if (!root.isDirectory()) {
printUsage("root is not a directory: " + root)
return
}
val dbFile = childPath(root, DB_FILE_NAME)
val dbWasCreated = !dbFile.exists()
val shouldUpdateSnapshot = dbWasCreated || options.updateSnapshot
printBanner(root, dbFile, dbWasCreated, shouldUpdateSnapshot)
val db = openSqlite(dbFile.toString())
db.transaction { tx ->
tx.execute(CREATE_FILE_INDEX_SQL)
tx.execute("drop table if exists temp.current_scan")
tx.execute(CREATE_CURRENT_SCAN_SQL)
var scannedFiles = 0
for (rawEntry in root.glob("**")) {
val entry = rawEntry as Path
if (!entry.isFile()) continue
val relative = relativePath(root, entry)
if (isDatabaseArtifact(relative)) continue
val size = entry.size() ?: 0
val mtime = entry.modifiedAtMillis() ?: 0
tx.execute(INSERT_SCAN_ROW_SQL, relative, size, mtime)
scannedFiles++
}
val added = loadRows(tx, SELECT_ADDED_SQL)
val removed = loadRows(tx, SELECT_REMOVED_SQL)
val changed = loadRows(tx, SELECT_CHANGED_SQL)
val totalChanges = added.size + removed.size + changed.size
out(dim("scanned %d %s under %s"(scannedFiles, plural(scannedFiles, "file", "files"), root.toString())))
out(dim("detected %d %s"(totalChanges, plural(totalChanges, "change", "changes"))))
out()
printSection("Added", { green(it) }, added) { renderAdded(it) }
printSection("Removed", { red(it) }, removed) { renderRemoved(it) }
printSection("Changed", { yellow(it) }, changed) { renderChanged(it) }
if (shouldUpdateSnapshot) {
tx.execute(DELETE_MISSING_SQL)
tx.execute(UPSERT_SCAN_SQL)
val action = if (dbWasCreated) "created" else "updated"
out(cyan("snapshot " + action + " in " + dbFile.name))
} else {
out(dim("snapshot unchanged; re-run with -u or --update to persist the scan"))
}
}
}
main()

View File

@ -1,23 +0,0 @@
#!/env/bin lyng
import lyng.io.http
// Step 1: download the main lynglang.com page.
val home = Http.get("https://lynglang.com").text()
// Step 2: find the version-script reference in the page HTML.
val jsRef = "src=\"([^\"]*lyng-version\\.js)\"".re.find(home)
require(jsRef != null, "lyng-version.js reference not found on the homepage")
// Step 3: extract the referenced script path from the first regex capture.
val versionJsPath = jsRef[1]
// Step 4: download the script that exposes `window.LYNG_VERSION`.
val versionJs = Http.get("https://lynglang.com/" + versionJsPath).text()
// Step 5: pull the actual version string from the JavaScript source.
val versionMatch = "LYNG_VERSION\\s*=\\s*\"([^\"]+)\"".re.find(versionJs)
require(versionMatch != null, "LYNG_VERSION assignment not found")
// Step 6: print the discovered version for the user.
println("Lynglang.com version: " + ((versionMatch as RegexMatch)[1]))

View File

@ -1,15 +0,0 @@
import lyng.time
val n = 700_000
fun tm<T>(block: ()->T): T {
val t = Instant()
block().also {
println("tm: ${Instant() - t}")
}
}
val x = tm { List.fill(n) { it * 10 + 1 } }
val y = tm { List.fill(n, n + 10) { it * 10 + 1 } }
tm { x.add(-1) }
tm { y.add(-2) }

View File

@ -1,76 +0,0 @@
fun calculateDepth(
T: Real,
m: Real,
d: Real,
rho: Real = 1.2,
c: Real = 340.0,
g: Real = 9.81,
Cd: Real = 0.5,
eps: Real = 1e-3,
maxIter: Int = 100
): Real? {
// Площадь миделя
val r = d / 2.0
val A = π * r * r
// Коэффициент сопротивления
val k = 0.5 * Cd * rho * A
// Предельная скорость
val vTerm = sqrt(m * g / k)
// Функция времени падения с высоты h
fun tFall(h: Real): Real {
// Для численной стабильности при больших h используем логарифмическую форму
val arg = exp(g * h / (vTerm * vTerm))
// arcosh(x) = ln(x + sqrt(x^2 - 1))
val arcosh = ln(arg + sqrt(arg * arg - 1.0))
return vTerm / g * arcosh
}
// Полное расчётное время
fun Tcalc(h: Real): Real = tFall(h) + h / c
// Находим интервал, содержащий корень
// Нижняя граница: глубина не может быть отрицательной
var lo = 0.0
// Верхняя граница: сначала попробуем оценку по свободному падению (без звука)
var hi = 0.5 * g * T * T // максимальная глубина, если бы не было сопротивления и звука
// Уточним hi, чтобы Tcalc(hi) было заведомо больше T
while (Tcalc(hi) < T && hi < 1e4) {
hi *= 2.0
}
// Проверка, что hi достаточно велико
if (Tcalc(hi) < T) return null // слишком большая глубина, не укладываемся в разумное
// Бисекция
var iter = 0
var h = (lo + hi) / 2.0
while (iter < maxIter && (hi - lo) > eps) {
val f = Tcalc(h) - T
if (abs(f) < eps) break
if (f > 0) {
hi = h
} else {
lo = h
}
h = (lo + hi) / 2.0
iter++
}
return h
}
// Пример: T=12 секунд
val T = 26.0
val m = 1.0 // кг
val d = 0.1 // м
val depth = calculateDepth(T, m, d)
if (depth != null) {
println("Глубина: %.2f м"(depth))
// Для проверки выведем теоретическое время при найденной глубине
// (можно добавить функцию для самопроверки)
} else {
println("Расчёт не сошёлся")
}

View File

@ -1,43 +0,0 @@
import lyng.io.db
import lyng.io.db.jdbc
println("H2 JDBC demo: typed open, generic open, generated keys")
val db = openH2("mem:lyng_h2_demo;DB_CLOSE_DELAY=-1")
db.transaction { tx ->
tx.execute("create table if not exists person(id bigint auto_increment primary key, name varchar(120) not null, active boolean not null)")
tx.execute("delete from person")
val firstInsert = tx.execute(
"insert into person(name, active) values(?, ?)",
"Ada",
true
)
val firstId = firstInsert.getGeneratedKeys().toList()[0][0]
assertEquals(1, firstId)
tx.execute(
"insert into person(name, active) values(?, ?)",
"Linus",
false
)
val rows = tx.select("select id, name, active from person order by id").toList()
assertEquals(2, rows.size)
println("#" + rows[0]["id"] + " " + rows[0]["name"] + " active=" + rows[0]["active"])
println("#" + rows[1]["id"] + " " + rows[1]["name"] + " active=" + rows[1]["active"])
}
val genericDb = openDatabase(
"jdbc:h2:mem:lyng_h2_generic;DB_CLOSE_DELAY=-1",
Map()
)
val answer = genericDb.transaction { tx ->
tx.select("select 42 as answer").toList()[0]["answer"]
}
assertEquals(42, answer)
println("Generic JDBC openDatabase(...) also works: answer=$answer")
println("OK")

View File

@ -1,19 +0,0 @@
import lyng.io.http.server
closed class CreateUserRequest(name: String, age: Int)
closed class CreateUserResponse(id: Int, name: String, age: Int)
val server = HttpServer()
server.postPath("/api/users") {
val req = jsonBody<CreateUserRequest>()
if (req.name.isBlank()) {
respondJson({ error: "name must not be empty" }, 400)
return
}
respondJson(CreateUserResponse(101, req.name, req.age), 201)
}
server.listen(8080, "127.0.0.1")

View File

@ -1,67 +0,0 @@
import lyng.time
val WORK_SIZE = 500
val THREADS = 1
fn piSpigot(iThread: Int, n: Int) {
var piIter = 0
var pi = List.fill(n) { 0 }
val boxes = n * 10 / 3
var reminders = List.fill(boxes) { 2 }
var heldDigits = 0
for (i in 0..<n) {
var carriedOver = 0
var sum = 0
for (j in (boxes - 1) downTo 0) {
val denom = j * 2 + 1
reminders[j] *= 10
sum = reminders[j] + carriedOver
val quotient = sum / denom
reminders[j] = sum % denom
carriedOver = quotient * j
}
reminders[0] = sum % 10
var q = sum / 10
if (q == 9) {
++heldDigits
} else if (q == 10) {
q = 0
for (k in 1..heldDigits) {
var replaced = pi[i - k]
if (replaced == 9) {
replaced = 0
} else {
++replaced
}
pi[i - k] = replaced
}
heldDigits = 1
} else {
heldDigits = 1
}
pi[piIter] = q
++piIter
}
var res = ""
for (i in (n - 8)..<n) {
res += pi[i]
}
println(iThread.toString() + ": " + res)
res
}
for( r in 0..100 ) {
val t0 = Instant()
println("piBench (lyng): THREADS = " + THREADS + ", WORK_SIZE = " + WORK_SIZE)
for (i in 0..<THREADS) {
piSpigot(i, WORK_SIZE)
}
val dt = Instant() - t0
println("all done, dt = ", dt)
delay(800)
}

View File

@ -1,83 +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.
#
import time
from multiprocessing import Process
def piSpigot(iThread, nx):
piIter = 0
pi = [None] * nx
boxes = nx * 10 // 3
reminders = [None]*boxes
i = 0
while i < boxes:
reminders[i] = 2
i += 1
heldDigits = 0
i = 0
while i < nx:
carriedOver = 0
sum = 0
j = boxes - 1
while j >= 0:
reminders[j] *= 10
sum = reminders[j] + carriedOver
quotient = sum // (j * 2 + 1)
reminders[j] = sum % (j * 2 + 1)
carriedOver = quotient * j
j -= 1
reminders[0] = sum % 10
q = sum // 10
if q == 9:
heldDigits += 1
elif q == 10:
q = 0
k = 1
while k <= heldDigits:
replaced = pi[i - k]
if replaced == 9:
replaced = 0
else:
replaced += 1
pi[i - k] = replaced
k += 1
heldDigits = 1
else:
heldDigits = 1
pi[piIter] = q
piIter += 1
i += 1
res = ""
for i in range(len(pi)-8, len(pi), 1):
res += str(pi[i])
print(str(iThread) + ": " + res)
def createProcesses():
THREADS = 1
WORK_SIZE = 500
print("piBench (python3): THREADS = " + str(THREADS) + ", WORK_SIZE = " + str(WORK_SIZE))
pa = []
for i in range(THREADS):
p = Process(target=piSpigot, args=(i, WORK_SIZE))
p.start()
pa.append(p)
for p in pa:
p.join()
if __name__ == "__main__":
t1 = time.time()
createProcesses()
dt = time.time() - t1
print("total time: %i ms" % (dt*1000))

View File

@ -1,71 +0,0 @@
import lyng.io.db.jdbc
/*
PostgreSQL JDBC demo.
Usage:
lyng examples/postgres_basic.lyng [jdbc-url] [user] [password]
Typical local URL:
jdbc:postgresql://127.0.0.1/postgres
*/
fun cliArgs(): List<String> {
val result: List<String> = []
for (raw in ARGV as List) {
result.add(raw as String)
}
return result
}
val argv = cliArgs()
val URL = if (argv.size > 0) argv[0] else "jdbc:postgresql://127.0.0.1/postgres"
val USER = if (argv.size > 1) argv[1] else ""
val PASSWORD = if (argv.size > 2) argv[2] else ""
println("PostgreSQL JDBC demo: typed open, generated keys, nested transaction")
val db = openPostgres(URL, USER, PASSWORD)
db.transaction { tx ->
tx.execute("create table if not exists lyng_pg_demo(id bigserial primary key, title text not null, done boolean not null)")
tx.execute("delete from lyng_pg_demo")
val firstInsert = tx.execute(
"insert into lyng_pg_demo(title, done) values(?, ?)",
"Verify PostgreSQL JDBC support",
false
)
val firstId = firstInsert.getGeneratedKeys().toList()[0][0]
println("First generated id=" + firstId)
tx.execute(
"insert into lyng_pg_demo(title, done) values(?, ?)",
"Review documentation",
true
)
try {
tx.transaction { inner ->
inner.execute(
"insert into lyng_pg_demo(title, done) values(?, ?)",
"This row is rolled back",
false
)
throw IllegalStateException("rollback nested")
}
} catch (_: IllegalStateException) {
println("Nested transaction rolled back as expected")
}
val rows = tx.select("select id, title, done from lyng_pg_demo order by id").toList()
for (row in rows) {
println("#" + row["id"] + " " + row["title"] + " done=" + row["done"])
}
val count = tx.select("select count(*) as count from lyng_pg_demo").toList()[0]["count"]
assertEquals(2, count)
println("Visible rows after nested rollback: " + count)
}
println("OK")

View File

@ -1,89 +0,0 @@
import lyng.io.db
import lyng.io.db.sqlite
import lyng.time
println("SQLite demo: typed open, generic open, result sets, generated keys, nested rollback")
// The typed helper is the simplest entry point when you know you want SQLite.
val db = openSqlite(":memory:")
db.transaction { tx ->
// Keep schema creation and data changes inside one transaction block.
tx.execute("create table task(id integer primary key autoincrement, title text not null, done integer not null, due_date date not null)")
// execute(...) is for side-effect statements. Generated keys are read from
// ExecutionResult rather than from a synthetic row-returning INSERT.
val firstInsert = tx.execute(
"insert into task(title, done, due_date) values(?, ?, ?)",
"Write a SQLite example",
false,
Date(2026, 4, 15)
)
val firstGeneratedKeys = firstInsert.getGeneratedKeys()
val firstId = firstGeneratedKeys.toList()[0][0]
assertEquals(1, firstId)
tx.execute(
"insert into task(title, done, due_date) values(?, ?, ?)",
"Review the DB API",
true,
Date(2026, 4, 16)
)
// Nested transactions are real savepoints. If the inner block fails,
// only the nested work is rolled back.
try {
tx.transaction { inner ->
inner.execute(
"insert into task(title, done, due_date) values(?, ?, ?)",
"This row is rolled back",
false,
Date(2026, 4, 17)
)
throw IllegalStateException("demonstrate nested rollback")
}
} catch (_: IllegalStateException) {
println("Nested transaction rolled back as expected")
}
// select(...) is for row-producing statements. ResultSet exposes metadata,
// cheap emptiness checks, iteration, and conversion to a plain list.
val tasks = tx.select("select id, title, done, due_date from task order by id")
assertEquals(false, tasks.isEmpty())
assertEquals(2, tasks.size())
println("Columns:")
for (column in tasks.columns) {
println(" " + column.name + " -> " + column.sqlType + " (native " + column.nativeType + ")")
}
val taskRows = tasks.toList()
println("Rows:")
for (row in taskRows) {
// Name lookups are case-insensitive and values are already converted.
println(" #" + row["ID"] + " " + row["title"] + " done=" + row["done"] + " due=" + row["due_date"])
}
// toList() materializes detached rows that stay usable after transaction close.
val snapshot = tx.select("select title, due_date from task order by id").toList()
assertEquals("Write a SQLite example", snapshot[0]["title"])
assertEquals(Date(2026, 4, 16), snapshot[1]["due_date"])
val count = tx.select("select count(*) as count from task").toList()[0]["count"]
assertEquals(2, count)
println("Visible rows after nested rollback: $count")
}
// The generic entry point stays useful for config-driven code.
val genericDb = openDatabase(
"sqlite::memory:",{ foreignKeys: true, busyTimeoutMillis: 1000 }
)
val answer = genericDb.transaction { tx ->
tx.select("select 42 as answer").toList()[0]["answer"]
}
assertEquals(42, answer)
println("Generic openDatabase(...) also works: answer=$answer")
println("OK")

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