pyvm: split op handlers into ops_core/ops_box/ops_ctrl; add ops_flow + intrinsic; delegate vm.py without behavior change
net-plugin: modularize constants (consts.rs) and sockets (sockets.rs); remove legacy commented socket code; fix unused imports mir: move instruction unit tests to tests/mir_instruction_unit.rs (file lean-up); no semantic changes runner/pyvm: ensure using pre-strip; misc docs updates Build: cargo build ok; legacy cfg warnings remain as before
This commit is contained in:
Selfhosting Dev
@ -17,7 +17,7 @@ logic := compare (('&&' | '||') compare)*
|
||||
compare := sum (( '==' | '!=' | '<' | '>' | '<=' | '>=' ) sum)?
|
||||
sum := term (('+' | '-') term)*
|
||||
term := unary (('*' | '/') unary)*
|
||||
unary := '-' unary | factor
|
||||
unary := ('-' | '!' | 'not') unary | factor
|
||||
|
||||
factor := INT
|
||||
| STRING
|
||||
@ -50,6 +50,8 @@ args := expr (',' expr)*
|
||||
|
||||
Notes
|
||||
- ASI: Newline is the primary statement separator. Do not insert a semicolon between a closed block and a following 'else'.
|
||||
- Semicolon (optional): When `NYASH_PARSER_ALLOW_SEMICOLON=1` is set, `;` is accepted as an additional statement separator (equivalent to newline). It is not allowed between `}` and a following `else`.
|
||||
- Do‑while: not supported by design. Prefer a single‑entry, pre‑condition loop normalized via sugar (e.g., `repeat N {}` / `until cond {}`) to a `loop` with clear break conditions.
|
||||
- Short-circuit: '&&' and '||' must not evaluate the RHS when not needed.
|
||||
- Unary minus has higher precedence than '*' and '/'.
|
||||
- IDENT names consist of [A-Za-z_][A-Za-z0-9_]*
|
||||
@ -92,6 +94,11 @@ block_as_role := block 'as' ( 'once' | 'birth_once' )? IDENT ':' TYPE
|
||||
handler_tail := ( catch_block )? ( cleanup_block )?
|
||||
catch_block := 'catch' ( '(' ( IDENT IDENT | IDENT )? ')' )? block
|
||||
cleanup_block := 'cleanup' block
|
||||
|
||||
; Stage‑3 (Phase 1 via normalization gate NYASH_CATCH_NEW=1)
|
||||
; Postfix handlers for expressions and calls
|
||||
postfix_catch := primary_expr 'catch' ( '(' ( IDENT IDENT | IDENT )? ')' )? block
|
||||
postfix_cleanup := primary_expr 'cleanup' block
|
||||
```
|
||||
|
||||
Semantics (summary)
|
||||
|
||||
@ -27,7 +27,7 @@ Rust製インタープリターによる高性能実行と、直感的な構文
|
||||
| `loop` | ループ（唯一の形式） | `loop(condition) { }` |
|
||||
| `continue` | ループ継続 | `continue` |
|
||||
| `match` | パターンマッチング（構造/型/ガード） | `match value { "A" => 1, _ => 0 }` |
|
||||
| `try` | 例外捕獲開始 | `try { }` |
|
||||
| `try` | 例外捕獲開始 | `try { }`（非推奨。postfix `catch/cleanup` を使用） |
|
||||
| `interface` | インターフェース定義 | `interface Comparable { }` |
|
||||
| `once` | **NEW** 遅延評価プロパティ | `once cache: CacheBox { build() }` |
|
||||
| `birth_once` | **NEW** 即座評価プロパティ | `birth_once config: ConfigBox { load() }` |
|
||||
@ -37,8 +37,8 @@ Rust製インタープリターによる高性能実行と、直感的な構文
|
||||
|-------|------|---|
|
||||
| `override` | 明示的オーバーライド | `override speak() { }` |
|
||||
| `break` | ループ脱出 | `break` |
|
||||
| `catch` | 例外処理 | `catch (e) { }` |
|
||||
| `cleanup` | 最終処理（finally の後継） | `cleanup { }` |
|
||||
| `catch` | 例外処理 | `catch (e) { }`（式/呼び出しの後置も可・Stage‑3） |
|
||||
| `cleanup` | 最終処理（finally の後継） | `cleanup { }`（式/呼び出しの後置も可・Stage‑3） |
|
||||
| `throw` | 例外発生 | `throw error` |
|
||||
| `nowait` | 非同期実行 | `nowait future = task()` |
|
||||
| `await` | 待機・結果取得 | `result = await future` |
|
||||
@ -55,7 +55,7 @@ Rust製インタープリターによる高性能実行と、直感的な構文
|
||||
### **演算子・論理**
|
||||
| 演算子/キーワード | 用途 | 例 |
|
||||
|-------|------|---|
|
||||
| `not` | 論理否定 | `not condition` |
|
||||
| `not` / `!` | 論理否定 | `not condition` / `!condition` |
|
||||
| `and` | 論理積 | `a and b` |
|
||||
| `or` | 論理和 | `a or b` |
|
||||
| `true`/`false` | 真偽値 | `flag = true` |
|
||||
@ -188,9 +188,12 @@ loop(condition) {
|
||||
}
|
||||
}
|
||||
|
||||
# ❌ 削除済み - 使用不可
|
||||
while condition { } # パーサーエラー
|
||||
loop() { } # パーサーエラー
|
||||
# ❌ 採用しない構文（設計方針）
|
||||
while condition { } # 先頭条件は `loop(condition){}` へ統一
|
||||
do { body } while(cond) # do‑while は不採用。`repeat/ until` 糖衣で表現し、先頭条件に正規化
|
||||
loop() { } # 無条件ループは `loop(true){}` を意図明確に書く
|
||||
|
||||
> 設計メモ: Nyashは「単一入口・先頭条件」の制御フロー規律を重視するため、do‑whileは採用しません。必ず実行の表現は `loop(1)` ラッパーや `repeat/until` 糖衣からゼロコストで正規化します。
|
||||
```
|
||||
|
||||
#### **Peek式（Phase 12.7で追加）**
|
||||
@ -716,3 +719,20 @@ let [first, second, ...rest] = array
|
||||
**🎉 Nyash 2025は、AI協働設計による最先端言語システムとして、シンプルさと強力さを完全に両立しました。**
|
||||
|
||||
*最終更新: 2025年9月4日 - Phase 12.7実装済み機能の正確な反映*
|
||||
### 2.x 例外・エラーハンドリング（postfix / cleanup）
|
||||
|
||||
方針
|
||||
- try は非推奨。postfix `catch` と `cleanup` を用いる。
|
||||
- `catch` は直前の式/呼び出しで発生した例外を処理。
|
||||
- `cleanup` は常に実行（finally の後継）。
|
||||
|
||||
例（式レベルの postfix）
|
||||
```
|
||||
do_work() catch(Error e) { env.console.log(e) }
|
||||
open(path) cleanup { env.console.log("close") }
|
||||
connect(url)
|
||||
catch(NetworkError e) { env.console.warn(e) }
|
||||
cleanup { env.console.log("done") }
|
||||
```
|
||||
|
||||
注: Phase 1 は正規化（ゲート `NYASH_CATCH_NEW=1`）で legacy TryCatch へ展開。Phase 2 でパーサが直接受理。
|
||||
|
||||
41
docs/reference/language/match-guards.md
Normal file
41
docs/reference/language/match-guards.md
Normal file
@ -0,0 +1,41 @@
|
||||
# Match Guards — Syntax and Lowering (MVP + Design Notes)
|
||||
|
||||
Status: reference + design additions during freeze (no implementation changes)
|
||||
|
||||
Scope
|
||||
- Guarded branches as a readable form of first-match selection.
|
||||
- Canonical lowering target: if/else chain + PHI merges.
|
||||
|
||||
Syntax (MVP)
|
||||
- Guard chain (first-match wins):
|
||||
```nyash
|
||||
guard <cond> -> { /* then */ }
|
||||
guard <cond> -> { /* then */ }
|
||||
else -> { /* else */ }
|
||||
```
|
||||
- Conditions may combine comparisons, `is/as` type checks, and literals with `&&` / `||`.
|
||||
|
||||
Lowering
|
||||
- Always lowers to a linear if/else chain with early exit on first true guard.
|
||||
- Merge points use normal PHI formation invariants (see `reference/mir/phi_invariants.md`).
|
||||
|
||||
Design additions (frozen; docs only)
|
||||
- Range Pattern (sugar):
|
||||
- `guard x in '0'..'9' -> { ... }`
|
||||
- Lowers to: `('0' <= x && x <= '9')`.
|
||||
- Multiple ranges: `in A..B || C..D` → OR of each bound check.
|
||||
- CharClass (predefined sets):
|
||||
- `Digit ≡ '0'..'9'`, `AZ ≡ 'A'..'Z'`, `az ≡ 'a'..'z'`, `Alnum ≡ Digit || AZ || az`, `Space ≡ ' '\t\r\n` (MVP set; expandable later).
|
||||
- `guard ch in Digit -> { ... }` expands to range checks.
|
||||
|
||||
Errors & Rules (MVP)
|
||||
- Default `_` branch does not accept guards.
|
||||
- Type guard succeeds inside the then-branch; bindings (e.g., `StringBox(s)`) are introduced at branch head.
|
||||
- Short-circuit semantics follow standard branch evaluation (right side is evaluated only if needed).
|
||||
|
||||
Observability (design)
|
||||
- `NYASH_FLOW_TRACE=1` may trace how guard chains desugar into if/else.
|
||||
|
||||
Notes
|
||||
- This page describes existing guard semantics and adds range/charclass as documentation-only sugar during freeze.
|
||||
|
||||
61
docs/reference/language/strings.md
Normal file
61
docs/reference/language/strings.md
Normal file
@ -0,0 +1,61 @@
|
||||
# Nyash Strings: UTF‑8 First, Bytes Separate
|
||||
|
||||
Status: Design committed. This document defines how Nyash treats text vs bytes and the minimal APIs we expose in each layer.
|
||||
|
||||
## Principles
|
||||
- UTF‑8 is the only in‑memory encoding for `StringBox`.
|
||||
- Text operations are defined in terms of Unicode code points (CP). Grapheme cluster (GC) helpers may be added on top.
|
||||
- Bytes are not text. Byte operations live in a separate `ByteCursorBox` and byte‑level instructions.
|
||||
- Conversions are explicit.
|
||||
|
||||
## Model
|
||||
- `StringBox`: immutable UTF‑8 string value. Public text APIs are CP‑indexed.
|
||||
- `Utf8CursorBox`: delegated implementation for scanning and slicing `StringBox` as CPs.
|
||||
- `ByteCursorBox`: independent binary view/holder for byte sequences.
|
||||
|
||||
## Invariants
|
||||
- Indices are zero‑based. Slices use half‑open intervals `[i, j)`.
|
||||
- CP APIs never intermix with byte APIs. GC APIs are explicitly suffixed (e.g., `*_gc`).
|
||||
- Conversions must be explicit. No implicit transcoding.
|
||||
|
||||
## Core APIs (MVP)
|
||||
|
||||
Text (UTF‑8/CP): implemented by `StringBox` delegating to `Utf8CursorBox`.
|
||||
- `length() -> i64` — number of code points.
|
||||
- `substring(i,j) -> StringBox` — CP slice.
|
||||
- `indexOf(substr, from=0) -> i64` — CP index or `-1`.
|
||||
- Optional helpers: `startsWith/endsWith/replace/split/trim` as sugar.
|
||||
|
||||
Bytes: handled by `ByteCursorBox`.
|
||||
- `len_bytes() -> i64`
|
||||
- `slice_bytes(i,j) -> ByteCursorBox`
|
||||
- `find_bytes(pattern, from=0) -> i64`
|
||||
- `to_string_utf8(strict=true) -> StringBox | Error` — strict throws on invalid UTF‑8 (MVP may replace with U+FFFD when `strict=false`).
|
||||
|
||||
## Errors
|
||||
- CP APIs clamp out‑of‑range indices (dev builds may enable strict). Byte APIs mirror the same behavior for byte indices.
|
||||
- `to_string_utf8(strict=true)` fails on invalid input; `strict=false` replaces invalid sequences by U+FFFD.
|
||||
|
||||
## Interop
|
||||
- FFI/ABI boundaries use UTF‑8. Non‑UTF‑8 sources must enter via `ByteCursorBox` + explicit transcoding.
|
||||
- Other encodings (e.g., UTF‑16) are future work via separate cursor boxes; `StringBox` remains UTF‑8.
|
||||
|
||||
## Roadmap
|
||||
1) Provide Nyash‑level MVP boxes: `Utf8CursorBox`, `ByteCursorBox`.
|
||||
2) Route `StringBox` public methods through `Utf8CursorBox`.
|
||||
3) Migrate Mini‑VM and macro scanners to use `Utf8CursorBox` helpers.
|
||||
4) Add CP/byte parity smokes; later add GC helpers and normalizers.
|
||||
|
||||
## Proposed Convenience (design only)
|
||||
|
||||
Parsing helpers (sugar; freeze-era design, not implemented):
|
||||
- `toDigitOrNull(base=10) -> i64 | null`
|
||||
- Returns 0..9 when the code point is a decimal digit (or base subset), otherwise `null`.
|
||||
- CP based; delegates to `Utf8CursorBox` to read the leading code point.
|
||||
- `toIntOrNull() -> i64 | null`
|
||||
- Parses the leading consecutive decimal digits into an integer; returns `null` when no digit at head.
|
||||
- Pure function; does not move any external cursor (callers decide how to advance).
|
||||
|
||||
Notes
|
||||
- Zero new runtime opcodes; compiled as comparisons and simple arithmetic.
|
||||
- `Option/Maybe` may replace `null` in a future revision; documenting `null` keeps MVP simple.
|
||||
Reference in New Issue
Block a user