Files
hakorune/src/mir/builder/control_flow/mod.rs

904 lines
42 KiB
Rust
Raw Normal View History

//! Control-flow entrypoints (if/loop/try/throw) centralized here.
//!
//! This module is being modularized in phases:
//! - Phase 1: Debug utilities (debug.rs) ✅
//! - Phase 2: Pattern lowerers (joinir/patterns/) ✅
//! - Phase 3: JoinIR routing (joinir/routing.rs) ✅
//! - Phase 4-19: Additional modularization (future)
use super::{Effect, EffectMask, MirInstruction, ValueId};
feat(joinir): Phase 56 ArrayExtBox.filter JoinIR lowering完全実装 ## Summary ArrayExtBox.filter/2 の JoinIR Frontend lowering を完全実装し、 ConditionalMethodCall 命令を導入して filter パターンに対応。 56 JoinIR テスト全て PASS(退行なし)。 ## Technical Changes ### 1. ConditionalMethodCall 命令追加 - **新規命令**: `if pred(v) { acc.push(v) }` パターン用 - **構造**: cond が true なら method 実行、false なら no-op - **MIR 変換**: 4ブロック構造 (cond→then/else→merge) ### 2. AST JSON 拡張 - Break/Continue/FunctionCall に "type" フィールド追加 - ArrayLiteral/MapLiteral に "type" フィールド追加 - JoinIR Frontend 互換性向上 ### 3. Expression Handler 拡張 - Unary 演算子(not, 負号)サポート - Call(変数関数呼び出し)を MethodCall に変換 ### 4. Loop Pattern Binding 修正 - `BoundExpr::Variable("n")` 問題修正 - `MethodCall { receiver: "arr", method: "size" }` に変更 - external_refs (arr, pred) を step 関数に伝播 ### 5. If Statement Handler 拡張 - 条件付き側効果パターン(ケース4)追加 - MethodCall/Method 形式の statement を ConditionalMethodCall に変換 ## Files Modified (10 files, +456/-45 lines) - ast_json.rs: AST JSON "type" フィールド追加 - loop_frontend_binding.rs: n バインディング修正 - control_flow.rs: external_refs params 追加 - loop_patterns.rs: external_refs step 関数伝播 - expr.rs: Unary, Call handler 追加 - stmt_handlers.rs: ConditionalMethodCall パターン追加 - mod.rs: ConditionalMethodCall, UnaryOp 定義 - json.rs: ConditionalMethodCall, UnaryOp シリアライズ - join_ir_runner.rs: ConditionalMethodCall, UnaryOp スタブ - convert.rs: ConditionalMethodCall → MIR 変換 ## Test Results - 56 JoinIR tests: ✅ PASSED - Regression: ❌ None - ArrayExtBox.filter/2: ✅ JoinIR lowering 成功 ## Milestone JoinIR 2ループ完走達成: - ✅ JsonTokenizer.print_tokens/0 - ✅ ArrayExtBox.filter/2 (NEW!) 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com>
2025-11-29 06:51:43 +09:00
use crate::ast::ASTNode;
// Phase 1: Debug utilities
pub(in crate::mir::builder) mod debug;
// Phase 2: JoinIR integration (patterns)
pub(in crate::mir::builder) mod joinir;
impl super::MirBuilder {
/// Control-flow: block
pub(super) fn cf_block(&mut self, statements: Vec<ASTNode>) -> Result<ValueId, String> {
// identical to build_block; kept here for future policy hooks
self.build_block(statements)
}
/// Control-flow: if
Phase 123 proper完了:hako_check JoinIR実装(環境変数選択可能化) ## 実装内容 ### 1. 環境変数フラグ追加 - NYASH_HAKO_CHECK_JOINIR でJoinIR/Legacy経路を切り替え可能 - src/config/env/hako_check.rs で hako_check_joinir_enabled() 実装 - デフォルト: false(レガシー経路)で後方互換性確保 ### 2. MIR Builder JoinIR スイッチ - cf_if() メソッドにフラグチェック追加 - try_cf_if_joinir() プレースホルダー実装(Phase 124で完全実装) - JoinIR → legacy フォールバック機構を構築 ### 3. テストケース作成(4個) - phase123_simple_if.hako - phase123_nested_if.hako - phase123_while_loop.hako - phase123_if_in_loop.hako ### 4. テスト結果 ✅ Legacy path: 4/4 PASS ✅ JoinIR path: 4/4 PASS (JoinIR path は現在フォールバック経由で動作) ### 5. ドキュメント更新 - environment-variables.md: NYASH_HAKO_CHECK_JOINIR 記載 - phase121_hako_check_joinir_design.md: Phase 123実装セクション追加 - hako_check_design.md: 2パス実行フロー図を追加 - CURRENT_TASK.md: Phase 123完了を記録 ## 数値成果 - 新規ファイル: 2個 (config/env/hako_check.rs, test cases × 4, test script) - 修正ファイル: 6個 - 総追加行数: 335行 - ビルド: Zero errors ## 設計・実装の特徴 ✅ Environment variable で簡単に経路切り替え可能 ✅ レガシー経路を完全に保持(後方互換性) ✅ JoinIR基盤を Phase 124 での完全実装に向けて構築 ✅ フォールバック機構でリスク最小化 ## 次のステップ Phase 124: JoinIR 完全実装&デフォルト化 - try_cf_if_joinir() を IfSelectLowerer と統合 - Loop JoinIR 統合追加 - JoinIR をデフォルト経路に変更 - NYASH_LEGACY_PHI=1 で legacy フォールバック可能に 🤖 Generated with Claude Code Co-Authored-By: Claude <noreply@anthropic.com>
2025-12-04 06:17:10 +09:00
///
/// # Phase 124: JoinIR-Only (hako_check専用化完了)
Phase 123 proper完了:hako_check JoinIR実装(環境変数選択可能化) ## 実装内容 ### 1. 環境変数フラグ追加 - NYASH_HAKO_CHECK_JOINIR でJoinIR/Legacy経路を切り替え可能 - src/config/env/hako_check.rs で hako_check_joinir_enabled() 実装 - デフォルト: false(レガシー経路)で後方互換性確保 ### 2. MIR Builder JoinIR スイッチ - cf_if() メソッドにフラグチェック追加 - try_cf_if_joinir() プレースホルダー実装(Phase 124で完全実装) - JoinIR → legacy フォールバック機構を構築 ### 3. テストケース作成(4個) - phase123_simple_if.hako - phase123_nested_if.hako - phase123_while_loop.hako - phase123_if_in_loop.hako ### 4. テスト結果 ✅ Legacy path: 4/4 PASS ✅ JoinIR path: 4/4 PASS (JoinIR path は現在フォールバック経由で動作) ### 5. ドキュメント更新 - environment-variables.md: NYASH_HAKO_CHECK_JOINIR 記載 - phase121_hako_check_joinir_design.md: Phase 123実装セクション追加 - hako_check_design.md: 2パス実行フロー図を追加 - CURRENT_TASK.md: Phase 123完了を記録 ## 数値成果 - 新規ファイル: 2個 (config/env/hako_check.rs, test cases × 4, test script) - 修正ファイル: 6個 - 総追加行数: 335行 - ビルド: Zero errors ## 設計・実装の特徴 ✅ Environment variable で簡単に経路切り替え可能 ✅ レガシー経路を完全に保持(後方互換性) ✅ JoinIR基盤を Phase 124 での完全実装に向けて構築 ✅ フォールバック機構でリスク最小化 ## 次のステップ Phase 124: JoinIR 完全実装&デフォルト化 - try_cf_if_joinir() を IfSelectLowerer と統合 - Loop JoinIR 統合追加 - JoinIR をデフォルト経路に変更 - NYASH_LEGACY_PHI=1 で legacy フォールバック可能に 🤖 Generated with Claude Code Co-Authored-By: Claude <noreply@anthropic.com>
2025-12-04 06:17:10 +09:00
///
/// If statements are now always routed through the canonical lowering path
/// (lower_if_form), which internally uses JoinIR-based PHI generation.
Phase 123 proper完了:hako_check JoinIR実装(環境変数選択可能化) ## 実装内容 ### 1. 環境変数フラグ追加 - NYASH_HAKO_CHECK_JOINIR でJoinIR/Legacy経路を切り替え可能 - src/config/env/hako_check.rs で hako_check_joinir_enabled() 実装 - デフォルト: false(レガシー経路)で後方互換性確保 ### 2. MIR Builder JoinIR スイッチ - cf_if() メソッドにフラグチェック追加 - try_cf_if_joinir() プレースホルダー実装(Phase 124で完全実装) - JoinIR → legacy フォールバック機構を構築 ### 3. テストケース作成(4個) - phase123_simple_if.hako - phase123_nested_if.hako - phase123_while_loop.hako - phase123_if_in_loop.hako ### 4. テスト結果 ✅ Legacy path: 4/4 PASS ✅ JoinIR path: 4/4 PASS (JoinIR path は現在フォールバック経由で動作) ### 5. ドキュメント更新 - environment-variables.md: NYASH_HAKO_CHECK_JOINIR 記載 - phase121_hako_check_joinir_design.md: Phase 123実装セクション追加 - hako_check_design.md: 2パス実行フロー図を追加 - CURRENT_TASK.md: Phase 123完了を記録 ## 数値成果 - 新規ファイル: 2個 (config/env/hako_check.rs, test cases × 4, test script) - 修正ファイル: 6個 - 総追加行数: 335行 - ビルド: Zero errors ## 設計・実装の特徴 ✅ Environment variable で簡単に経路切り替え可能 ✅ レガシー経路を完全に保持(後方互換性) ✅ JoinIR基盤を Phase 124 での完全実装に向けて構築 ✅ フォールバック機構でリスク最小化 ## 次のステップ Phase 124: JoinIR 完全実装&デフォルト化 - try_cf_if_joinir() を IfSelectLowerer と統合 - Loop JoinIR 統合追加 - JoinIR をデフォルト経路に変更 - NYASH_LEGACY_PHI=1 で legacy フォールバック可能に 🤖 Generated with Claude Code Co-Authored-By: Claude <noreply@anthropic.com>
2025-12-04 06:17:10 +09:00
///
/// Phase 123 の環境変数による分岐は削除済み。
pub(super) fn cf_if(
&mut self,
condition: ASTNode,
then_branch: ASTNode,
else_branch: Option<ASTNode>,
) -> Result<ValueId, String> {
// Phase 124: JoinIR-only path (環境変数分岐削除)
// lower_if_form は JoinIR ベースの PHI 生成を使用
self.lower_if_form(condition, then_branch, else_branch)
}
/// Control-flow: loop
///
/// # Phase 49: JoinIR Frontend Mainline Integration
///
/// This is the unified entry point for all loop lowering. Specific functions
/// are routed through JoinIR Frontend instead of the traditional LoopBuilder path
/// when enabled via dev flags (Phase 49) or Core policy (Phase 80):
///
/// - Core ON (`joinir_core_enabled()`): print_tokens / ArrayExt.filter はまず JoinIR Frontend を試す
/// - Dev フラグ(既存):
/// - `HAKO_JOINIR_PRINT_TOKENS_MAIN=1`: JsonTokenizer.print_tokens/0
/// - `HAKO_JOINIR_ARRAY_FILTER_MAIN=1`: ArrayExtBox.filter/2
///
/// Note: Arity does NOT include implicit `me` receiver.
pub(super) fn cf_loop(
&mut self,
condition: ASTNode,
body: Vec<ASTNode>,
) -> Result<ValueId, String> {
// Phase 49/80: Try JoinIR Frontend route for mainline targets
if let Some(result) = self.try_cf_loop_joinir(&condition, &body)? {
return Ok(result);
}
📦 Hotfix 1 & 2: Parameter ValueId Reservation + Exit PHI Validation (Box-First Theory) **箱理論に基づく根治的修正**: ## 🎯 Hotfix 1: Parameter ValueId Reservation (パラメータ ValueId 予約) ### 根本原因 - MirFunction counter が params.len() を考慮していなかった - local variables が parameter ValueIds を上書き ### 箱理論的解決 1. **LoopFormContext Box** - パラメータ予約を明示的に管理 - 境界をはっきりさせる 2. **MirFunction::new() 改善** - `initial_counter = param_count.max(1)` でパラメータ予約 - Parameters are %0, %1, ..., %N-1 3. **ensure_counter_after() 強化** - パラメータ数 + 既存 ValueIds 両方を考慮 - `min_counter = param_count.max(max_id + 1)` 4. **reserve_parameter_value_ids() 追加** - 明示的な予約メソッド(Box-First) ## 🎯 Hotfix 2: Exit PHI Predecessor Validation (Exit PHI 検証) ### 根本原因 - LoopForm builder が存在しないブロックを PHI predecessor に追加 - 「幽霊ブロック」問題 ### 箱理論的解決 1. **LoopFormOps.block_exists() 追加** - CFG 存在確認メソッド - 境界を明確化 2. **build_exit_phis() 検証** - 非存在ブロックをスキップ - デバッグログ付き ### 実装ファイル - `src/mir/function.rs`: Parameter reservation - `src/mir/phi_core/loopform_builder.rs`: Context + validation - `src/mir/loop_builder.rs`: LoopFormOps impl - `src/mir/builder/stmts.rs`: Local variable allocation ### 業界標準準拠 - ✅ LLVM IR: Parameters are %0, %1, ... - ✅ SSA Form: PHI predecessors must exist in CFG - ✅ Cytron et al. (1991): Parameter reservation principle 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com>
2025-11-18 06:39:45 +09:00
if std::env::var("NYASH_LOOPFORM_DEBUG").is_ok() {
eprintln!("[cf_loop] CALLED from somewhere");
eprintln!("[cf_loop] Current stack (simulated): check build_statement vs build_expression_impl");
}
// Phase 186: LoopBuilder Hard Freeze - Legacy path disabled
// Phase 187-2: LoopBuilder module removed - all loops must use JoinIR
return Err(format!(
"[joinir/freeze] Loop lowering failed: JoinIR does not support this pattern, and LoopBuilder has been removed.\n\
Function: {}\n\
Hint: This loop pattern is not supported. All loops must use JoinIR lowering.",
self.current_function.as_ref().map(|f| f.signature.name.as_str()).unwrap_or("<unknown>")
));
}
/// Phase 49: Try JoinIR Frontend for mainline integration
///
/// Returns `Ok(Some(value))` if the current function should use JoinIR Frontend,
/// `Ok(None)` to fall through to the legacy LoopBuilder path.
///
/// # Phase 49-4: Multi-target support
///
/// Targets are enabled via separate dev flags:
/// - `HAKO_JOINIR_PRINT_TOKENS_MAIN=1`: JsonTokenizer.print_tokens/0
/// - `HAKO_JOINIR_ARRAY_FILTER_MAIN=1`: ArrayExtBox.filter/2
///
/// Note: Arity in function names does NOT include implicit `me` receiver.
/// - Instance method `print_tokens()` → `/0` (no explicit params)
/// - Static method `filter(arr, pred)` → `/2` (two params)
fix(joinir): Phase 188-Impl-3 Pattern 3 router integration in control_flow.rs Add Pattern 3 (Loop with If-Else PHI) routing and detection in MIR builder's control flow handler. This enables proper routing of loop_if_phi.hako test case. ## Changes ### Router Integration (lines 179-195) - Add Pattern 3 detection BEFORE Pattern 1 to avoid incorrect routing - Pattern 3 detection: func_name == "main" && variable_map.contains_key("sum") - This distinguishes Pattern 3 (with sum accumulator) from Pattern 1 (without) - Debug logging added for routing decisions ### New cf_loop_pattern3_with_if_phi() Method (lines 665-789) Implements Pattern 3 lowering pipeline: 1. Extract loop variables from condition (i) and variable map (sum) 2. Create minimal LoopScopeShape placeholder 3. Call lower_loop_with_if_phi_pattern() to generate JoinModule 4. Convert JoinModule to MirModule via convert_join_module_to_mir_with_meta() 5. Create JoinInlineBoundary with TWO carriers: i and sum 6. Merge JoinIR-generated MIR blocks into current function 7. Return Void (loops don't produce values) ### Debug Logging Enhancement - NYASH_JOINIR_MAINLINE_DEBUG environment variable for function name routing logs - Phase 188 debug output shows: loop variables extracted, boundary mapping created ## Architecture Notes - Pattern 3 is correctly detected by presence of 'sum' variable in variable_map - Boundary mapping uses sequential ValueIds from JoinIR: ValueId(0) for i, ValueId(1) for sum - Host function's loop_var_id and sum_var_id are mapped via JoinInlineBoundary - Select instruction handling deferred to Phase 189+ (MIR bridge enhancement) ## Status - Pattern 1 test (loop_min_while.hako): ✅ Works - Pattern 2 test (joinir_min_loop.hako): ✅ Works - Pattern 3 test (loop_if_phi.hako): 🔄 Infrastructure complete, Select MIR conversion pending Next: Phase 189 will implement Select instruction conversion (Branch + Then/Else + PHI merge). 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com>
2025-12-05 16:04:15 +09:00
/// Phase 49-3.2: Merge JoinIR-generated MIR blocks into current_function
///
/// # Phase 189: Multi-Function MIR Merge
///
/// This merges JoinIR-generated blocks by:
/// 1. Remapping all block IDs across ALL functions to avoid conflicts
/// 2. Remapping all value IDs across ALL functions to avoid conflicts
/// 3. Adding all blocks from all functions to current_function
/// 4. Jumping from current_block to the entry block
/// 5. Converting Return → Jump to exit block for all functions
///
/// **Multi-Function Support** (Phase 189):
/// - Pattern 1 (Simple While) generates 3 functions: entry + loop_step + k_exit
/// - All functions are flattened into current_function with global ID remapping
/// - Single exit block receives all Return instructions from all functions
///
/// # Phase 188-Impl-3: JoinInlineBoundary Support
///
/// When `boundary` is provided, injects Copy instructions at the entry block
/// to connect host ValueIds to JoinIR local ValueIds:
///
/// ```text
/// entry_block:
/// // Injected by boundary
/// ValueId(100) = Copy ValueId(4) // join_input → host_input
/// // Original JoinIR instructions follow...
/// ```
///
/// This enables clean separation: JoinIR uses local IDs (0,1,2...),
/// host uses its own IDs, and Copy instructions bridge the gap.
///
/// # Returns
///
/// Returns `Ok(Some(exit_phi_id))` if the merged JoinIR functions have return values
refactor(joinir): Phase 189 - Remove hardcoded 'sum' variable, generalize exit PHI connection Key improvements: 1. **Eliminate hardcoded variable name**: Replace hardcoded "sum" with generic ValueId-based variable_map updates. Add new JoinInlineBoundary constructor `new_with_input_and_host_outputs()` for Pattern 3. 2. **Generalize output slot mapping**: Exit PHI result now updates all host_outputs entries in variable_map, not just hardcoded "sum". Prepares for future multi-carrier patterns. 3. **PHI preservation in blocks**: Fix block finalization to preserve existing PHI instructions (from handle_select) instead of overwriting them. 4. **Stable function name→ValueId mapping**: Ensure consistent ValueId assignment for tail call detection across multi-function merges. 5. **Enhanced debugging**: Add detailed logging in block converter and meta analysis for PHI verification. Files modified: - src/mir/builder/control_flow.rs: Remove hardcoded "sum", use boundary outputs - src/mir/join_ir/lowering/inline_boundary.rs: Add new constructor - src/mir/join_ir_vm_bridge/joinir_block_converter.rs: Stable mappings, PHI preservation - src/mir/join_ir_vm_bridge/meta.rs: Debug output for PHI tracking - src/mir/builder/joinir_id_remapper.rs: PHI value remapping - src/mir/builder/joinir_inline_boundary_injector.rs: Span preservation Test status: Pattern 3 (loop_if_phi.hako) still produces correct result (sum=9, RC=9) 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: ChatGPT <noreply@openai.com> Co-Authored-By: Claude <noreply@anthropic.com>
2025-12-05 19:39:54 +09:00
/// that were collected into an exit block PHI. さらに、`boundary` に
/// host_outputs が指定されている場合は、exit PHI の結果をホスト側の
/// SSA スロットへ再接続するvariable_map 内の ValueId を更新する)。
fn merge_joinir_mir_blocks(
&mut self,
mir_module: &crate::mir::MirModule,
boundary: Option<&crate::mir::join_ir::lowering::inline_boundary::JoinInlineBoundary>,
debug: bool,
) -> Result<Option<ValueId>, String> {
use crate::mir::{BasicBlock, BasicBlockId, MirInstruction, ValueId};
use std::collections::HashMap;
// Phase 189: Use new ID remapper Box
use super::joinir_id_remapper::JoinIrIdRemapper;
if debug {
eprintln!(
"[cf_loop/joinir] merge_joinir_mir_blocks called with {} functions",
mir_module.functions.len()
);
}
// Phase 189: Create ID remapper for ValueId/BlockId translation
let mut remapper = JoinIrIdRemapper::new();
// Phase 189: Map function names to their entry blocks (for Call→Jump conversion)
let mut function_entry_map: HashMap<String, BasicBlockId> = HashMap::new();
// 1. Allocate new block IDs for ALL functions (Phase 189)
// DETERMINISM FIX: Sort functions by name to ensure consistent iteration order
if debug {
eprintln!("[cf_loop/joinir] Phase 189: Allocating block IDs for all functions");
}
let mut functions: Vec<_> = mir_module.functions.iter().collect();
functions.sort_by_key(|(name, _)| name.as_str());
for (func_name, func) in functions {
if debug {
eprintln!("[cf_loop/joinir] Function: {}", func_name);
}
// DETERMINISM FIX: Sort blocks by ID to ensure consistent iteration order
let mut blocks: Vec<_> = func.blocks.iter().collect();
blocks.sort_by_key(|(id, _)| id.0);
for (old_block_id, _) in blocks {
let new_block_id = self.block_gen.next();
// Use remapper to store composite key mapping
remapper.set_block(func_name.clone(), *old_block_id, new_block_id);
if debug {
eprintln!(
"[cf_loop/joinir] Block remap: {}:{:?} → {:?}",
func_name, old_block_id, new_block_id
);
}
}
// Map function entry blocks for Call→Jump conversion
let entry_block_new = remapper.get_block(func_name, func.entry_block)
.ok_or_else(|| format!("Entry block not found for {}", func_name))?;
function_entry_map.insert(func_name.clone(), entry_block_new);
if debug {
eprintln!(
"[cf_loop/joinir] Entry map: {} → {:?}",
func_name, entry_block_new
);
}
}
// 2. Create exit block for Return conversion (single for all functions)
let exit_block_id = self.block_gen.next();
if debug {
eprintln!("[cf_loop/joinir] Exit block: {:?}", exit_block_id);
}
// 3. Collect all ValueIds used across ALL functions (Phase 189)
// Also build a map of ValueId → function name for Call→Jump conversion
// Phase 188-Impl-3: Also collect function parameters for tail call conversion
if debug {
eprintln!("[cf_loop/joinir] Phase 189: Collecting value IDs from all functions");
}
refactor(mir): loop_builder.rs モジュール化 - 6ファイルに分割 ## リファクタリング内容 ### ファイル構造変更 - `src/mir/loop_builder.rs` (1515行) 削除 - `src/mir/loop_builder/` ディレクトリ新設(6ファイル、1529行) ### 新規モジュール構成 1. **mod.rs** (6,293行 → 実際は約150行) - モジュール定義とre-export - LoopBuilder構造体定義 2. **loop_form.rs** (25,988行 → 実際は約650行) - メインループlowering pipeline - デバッグ/実験フラグ集約 3. **if_lowering.rs** (15,600行 → 実際は約390行) - In-loop if lowering with JoinIR/PHI bridge - **Phase 61-2コード完全保持**: - JoinIR dry-run検証モード - PhiSpec計算とA/B比較 4. **phi_ops.rs** (12,844行 → 実際は約320行) - PHI emit helpers - LoopFormOps/PhiBuilderOps impls 5. **control.rs** (4,261行 → 実際は約107行) - break/continue capture - predecessor bookkeeping 6. **statements.rs** (1,673行 → 実際は約42行) - loop-body statement lowering entry point 7. **README.md** (752行 → 実際は約19行) - モジュール責務とサブモジュール説明 ### 設計原則 - **責務分離**: CFG構築/PHI生成/制御フロー/文処理を分離 - **Phase 61-2保持**: if_lowering.rsにJoinIR dry-run完全移行 - **phi_core委譲**: PHI構築ロジックは`phi_core`に委譲 ## テスト結果 - Phase 61-2テスト: ✅ 2/2 PASS(dry-runフラグ、PhiSpec) - loopformテスト: ✅ 14/14 PASS(退行なし) - ビルド: ✅ 成功(エラー0件) ## 統計 - **純削減**: -1,521行(25ファイル変更) - **loop_builder**: 1515行 → 1529行(+14行、6ファイル化) - **可読性**: 巨大単一ファイル → 責務別モジュール ## ChatGPT設計・Claude確認 大規模リファクタリングをChatGPTが実施、Claudeが検証完了。 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com>
2025-11-29 12:44:40 +09:00
let mut used_values: std::collections::BTreeSet<ValueId> =
std::collections::BTreeSet::new();
let mut value_to_func_name: HashMap<ValueId, String> = HashMap::new();
let mut function_params: HashMap<String, Vec<ValueId>> = HashMap::new();
for (func_name, func) in &mir_module.functions {
// Phase 188-Impl-3: Collect function parameters for tail call conversion
function_params.insert(func_name.clone(), func.params.clone());
for block in func.blocks.values() {
// Phase 189: Use remapper to collect values
let block_values = remapper.collect_values_in_block(block);
used_values.extend(block_values);
// Phase 189: Track Const String instructions that define function names
for inst in &block.instructions {
if let MirInstruction::Const { dst, value } = inst {
if let crate::mir::types::ConstValue::String(s) = value {
if function_entry_map.contains_key(s) {
value_to_func_name.insert(*dst, s.clone());
// Phase 189 FIX: Also add to used_values so it gets remapped!
// Without this, subsequent instructions referencing dst will fail
used_values.insert(*dst);
if debug {
eprintln!(
"[cf_loop/joinir] Found function name constant: {:?} = '{}'",
dst, s
);
}
}
}
}
}
}
// Also collect parameter ValueIds
for param in &func.params {
used_values.insert(*param);
}
}
// 4. Allocate new ValueIds for all collected values
for old_value in used_values {
let new_value = self.next_value_id();
remapper.set_value(old_value, new_value);
if debug {
eprintln!(
"[cf_loop/joinir] Value remap: {:?} → {:?}",
old_value, new_value
);
}
}
// 5. Merge ALL functions (Phase 189: iterate over all, not just first)
// DETERMINISM FIX: Sort functions by name to ensure consistent iteration order
if debug {
eprintln!("[cf_loop/joinir] Phase 189: Merging {} functions", mir_module.functions.len());
}
// Phase 189: Iterate over both names and functions (need name for composite keys)
let mut functions_merge: Vec<_> = mir_module.functions.iter().collect();
functions_merge.sort_by_key(|(name, _)| name.as_str());
// Phase 189 FIX: Build set of boundary join_inputs to skip their Const initializers
// When BoundaryInjector provides values via Copy, entry function shouldn't also set them via Const
let boundary_input_set: std::collections::HashSet<ValueId> = boundary
.map(|b| b.join_inputs.iter().copied().collect())
.unwrap_or_default();
let entry_func_name = functions_merge.first().map(|(name, _)| name.as_str());
// Phase 189-Fix: Collect return values from JoinIR functions for exit PHI
// Each Return → Jump conversion records (from_block, return_value) here
let mut exit_phi_inputs: Vec<(BasicBlockId, ValueId)> = Vec::new();
for (func_name, func) in functions_merge {
if debug {
eprintln!(
"[cf_loop/joinir] Merging function '{}' with {} blocks, entry={:?}",
func_name,
func.blocks.len(),
func.entry_block
);
}
// Build a local block map for this function (for remap_instruction compatibility)
let mut local_block_map: HashMap<BasicBlockId, BasicBlockId> = HashMap::new();
for old_block_id in func.blocks.keys() {
let new_block_id = remapper.get_block(func_name, *old_block_id)
.ok_or_else(|| format!("Block {:?} not found for {}", old_block_id, func_name))?;
local_block_map.insert(*old_block_id, new_block_id);
}
// Clone and remap all blocks from this function
// DETERMINISM FIX: Sort blocks by ID to ensure consistent iteration order
let mut blocks_merge: Vec<_> = func.blocks.iter().collect();
blocks_merge.sort_by_key(|(id, _)| id.0);
for (old_block_id, old_block) in blocks_merge {
// Use remapper to get correct mapping for this function's block
let new_block_id = remapper.get_block(func_name, *old_block_id)
.ok_or_else(|| format!("Block {:?} not found for {}", old_block_id, func_name))?;
let mut new_block = BasicBlock::new(new_block_id);
// Remap instructions (Phase 189: Convert inter-function Calls to control flow)
let mut found_tail_call = false;
let mut tail_call_target: Option<(BasicBlockId, Vec<ValueId>)> = None;
// DEBUG: Print block being processed
if debug {
eprintln!("[cf_loop/joinir] === Processing block {:?} (from func '{}') ===", old_block_id, func_name);
eprintln!("[cf_loop/joinir] Original block has {} instructions:", old_block.instructions.len());
for (idx, inst) in old_block.instructions.iter().enumerate() {
eprintln!("[cf_loop/joinir] [{}] {:?}", idx, inst);
}
eprintln!("[cf_loop/joinir] Original block terminator: {:?}", old_block.terminator);
}
// Phase 189 FIX: Check if this is entry function's entry block (for boundary input skipping)
let is_entry_func_entry_block = entry_func_name == Some(func_name.as_str())
&& *old_block_id == func.entry_block;
// First pass: Process all instructions, identify tail calls
for inst in &old_block.instructions {
// Phase 189: Skip Const String instructions that define function names
if let MirInstruction::Const { dst, value } = inst {
if let crate::mir::types::ConstValue::String(_) = value {
if value_to_func_name.contains_key(dst) {
if debug {
eprintln!("[cf_loop/joinir] Skipping function name const: {:?}", inst);
}
continue; // Skip this instruction
}
}
// Phase 189 FIX: Skip Const instructions in entry function's entry block
// that initialize boundary inputs. BoundaryInjector provides these values via Copy.
if is_entry_func_entry_block && boundary_input_set.contains(dst) {
if debug {
eprintln!("[cf_loop/joinir] Skipping boundary input const (replaced by BoundaryInjector Copy): {:?}", inst);
}
continue; // Skip - BoundaryInjector will provide the value
}
}
// Phase 189: Detect tail calls and save parameters
if let MirInstruction::Call { func, args, .. } = inst {
if let Some(func_name) = value_to_func_name.get(func) {
if let Some(&target_block) = function_entry_map.get(func_name) {
// This is a tail call - save info and skip the Call instruction itself
let remapped_args: Vec<ValueId> = args
.iter()
.map(|&v| remapper.get_value(v).unwrap_or(v))
.collect();
tail_call_target = Some((target_block, remapped_args));
found_tail_call = true;
if debug {
eprintln!(
"[cf_loop/joinir] Detected tail call to '{}' (args={:?}), will convert to Jump",
func_name, args
);
}
continue; // Skip the Call instruction itself
}
refactor(mir): loop_builder.rs モジュール化 - 6ファイルに分割 ## リファクタリング内容 ### ファイル構造変更 - `src/mir/loop_builder.rs` (1515行) 削除 - `src/mir/loop_builder/` ディレクトリ新設(6ファイル、1529行) ### 新規モジュール構成 1. **mod.rs** (6,293行 → 実際は約150行) - モジュール定義とre-export - LoopBuilder構造体定義 2. **loop_form.rs** (25,988行 → 実際は約650行) - メインループlowering pipeline - デバッグ/実験フラグ集約 3. **if_lowering.rs** (15,600行 → 実際は約390行) - In-loop if lowering with JoinIR/PHI bridge - **Phase 61-2コード完全保持**: - JoinIR dry-run検証モード - PhiSpec計算とA/B比較 4. **phi_ops.rs** (12,844行 → 実際は約320行) - PHI emit helpers - LoopFormOps/PhiBuilderOps impls 5. **control.rs** (4,261行 → 実際は約107行) - break/continue capture - predecessor bookkeeping 6. **statements.rs** (1,673行 → 実際は約42行) - loop-body statement lowering entry point 7. **README.md** (752行 → 実際は約19行) - モジュール責務とサブモジュール説明 ### 設計原則 - **責務分離**: CFG構築/PHI生成/制御フロー/文処理を分離 - **Phase 61-2保持**: if_lowering.rsにJoinIR dry-run完全移行 - **phi_core委譲**: PHI構築ロジックは`phi_core`に委譲 ## テスト結果 - Phase 61-2テスト: ✅ 2/2 PASS(dry-runフラグ、PhiSpec) - loopformテスト: ✅ 14/14 PASS(退行なし) - ビルド: ✅ 成功(エラー0件) ## 統計 - **純削減**: -1,521行(25ファイル変更) - **loop_builder**: 1515行 → 1529行(+14行、6ファイル化) - **可読性**: 巨大単一ファイル → 責務別モジュール ## ChatGPT設計・Claude確認 大規模リファクタリングをChatGPTが実施、Claudeが検証完了。 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com>
2025-11-29 12:44:40 +09:00
}
}
// Process regular instructions - Phase 189: Use remapper.remap_instruction() + manual block remapping
let remapped = remapper.remap_instruction(inst);
// Phase 189 FIX: Manual block remapping for Branch/Phi (JoinIrIdRemapper doesn't know func_name)
let remapped_with_blocks = match remapped {
MirInstruction::Branch { condition, then_bb, else_bb } => {
MirInstruction::Branch {
condition,
then_bb: local_block_map.get(&then_bb).copied().unwrap_or(then_bb),
else_bb: local_block_map.get(&else_bb).copied().unwrap_or(else_bb),
}
}
MirInstruction::Phi { dst, inputs, type_hint: None } => {
MirInstruction::Phi {
dst,
inputs: inputs.iter().map(|(bb, val)| {
(local_block_map.get(bb).copied().unwrap_or(*bb), *val)
}).collect(),
type_hint: None,
}
}
other => other,
};
if debug {
match inst {
MirInstruction::BoxCall { .. } => {
eprintln!("[cf_loop/joinir] Adding BoxCall to block {:?}: {:?}", new_block_id, inst);
}
_ => {}
}
}
new_block.instructions.push(remapped_with_blocks);
}
// DEBUG: Print what was added to the block after first pass
if debug {
eprintln!("[cf_loop/joinir] After first pass, new_block has {} instructions", new_block.instructions.len());
for (idx, inst) in new_block.instructions.iter().enumerate() {
eprintln!("[cf_loop/joinir] [{}] {:?}", idx, inst);
}
}
// Second pass: Insert parameter bindings for tail calls
// Phase 188-Impl-3: Use actual parameter ValueIds from target function
if let Some((target_block, args)) = tail_call_target {
if debug {
eprintln!(
"[cf_loop/joinir] Inserting param bindings for tail call to {:?}",
target_block
);
}
// Find the target function name from the target_block
// We need to reverse-lookup the function name from the entry block
let mut target_func_name: Option<String> = None;
for (fname, &entry_block) in &function_entry_map {
if entry_block == target_block {
target_func_name = Some(fname.clone());
break;
}
}
if let Some(target_func_name) = target_func_name {
if let Some(target_params) = function_params.get(&target_func_name) {
// Insert Copy instructions for parameter binding
for (i, arg_val_remapped) in args.iter().enumerate() {
if i < target_params.len() {
let param_val_original = target_params[i];
if let Some(param_val_remapped) = remapper.get_value(param_val_original) {
new_block.instructions.push(MirInstruction::Copy {
dst: param_val_remapped,
src: *arg_val_remapped,
});
if debug {
eprintln!(
"[cf_loop/joinir] Param binding: arg {:?} → param {:?}",
arg_val_remapped, param_val_remapped
);
}
}
}
}
}
}
// Set terminator to Jump
new_block.terminator = Some(MirInstruction::Jump {
target: target_block,
});
// DEBUG: Print final state after adding parameter bindings
if debug {
eprintln!("[cf_loop/joinir] After adding param bindings, new_block has {} instructions", new_block.instructions.len());
for (idx, inst) in new_block.instructions.iter().enumerate() {
eprintln!("[cf_loop/joinir] [{}] {:?}", idx, inst);
}
}
}
new_block.instruction_spans = old_block.instruction_spans.clone();
if debug {
eprintln!("[cf_loop/joinir] Span sync: new_block.instructions.len()={}, old_block.instruction_spans.len()={}, new_block.instruction_spans.len()={}",
new_block.instructions.len(),
old_block.instruction_spans.len(),
new_block.instruction_spans.len()
);
}
// Remap terminator (convert Return → Jump to exit) if not already set by tail call
if !found_tail_call {
if let Some(ref term) = old_block.terminator {
let remapped_term = match term {
MirInstruction::Return { value } => {
// Convert Return to Jump to exit block
// All functions return to same exit block (Phase 189)
// Phase 189-Fix: Add Copy instruction to pass return value to exit PHI
if let Some(ret_val) = value {
let remapped_val = remapper.get_value(*ret_val).unwrap_or(*ret_val);
if debug {
eprintln!(
"[cf_loop/joinir] Return({:?}) → Jump to exit",
remapped_val
);
}
// Collect (from_block, return_value) for exit PHI generation
exit_phi_inputs.push((new_block_id, remapped_val));
}
MirInstruction::Jump {
target: exit_block_id,
}
}
MirInstruction::Jump { target } => {
// Phase 189 FIX: Remap block ID for Jump
MirInstruction::Jump {
target: local_block_map.get(target).copied().unwrap_or(*target),
}
}
MirInstruction::Branch { condition, then_bb, else_bb } => {
// Phase 189 FIX: Remap block IDs AND condition ValueId for Branch
MirInstruction::Branch {
condition: remapper.remap_value(*condition),
then_bb: local_block_map.get(then_bb).copied().unwrap_or(*then_bb),
else_bb: local_block_map.get(else_bb).copied().unwrap_or(*else_bb),
}
}
_ => remapper.remap_instruction(term),
};
new_block.terminator = Some(remapped_term);
}
}
// Phase 189 FIX: Ensure instruction_spans matches instructions length
// The original spans may not cover all instructions after remapping/adding
// (PHI instructions, tail call parameter bindings, etc.)
let inst_count = new_block.instructions.len();
let span_count = new_block.instruction_spans.len();
if inst_count > span_count {
// Use a default span for the extra instructions
let default_span = new_block.instruction_spans.last()
.copied()
.unwrap_or_else(crate::ast::Span::unknown);
new_block.instruction_spans.resize(inst_count, default_span);
} else if inst_count < span_count {
// Truncate spans to match instructions
new_block.instruction_spans.truncate(inst_count);
}
// Add block to current function
if let Some(ref mut current_func) = self.current_function {
current_func.add_block(new_block);
}
}
}
// Phase 188-Impl-3: Inject Copy instructions for boundary inputs using BoundaryInjector
if let Some(boundary) = boundary {
use super::joinir_inline_boundary_injector::BoundaryInjector;
// Get entry function's entry block (first function by convention)
let (entry_func_name, entry_func) = mir_module
.functions
.iter()
.next()
.ok_or("JoinIR module has no functions")?;
let entry_block_remapped = remapper.get_block(entry_func_name, entry_func.entry_block)
.ok_or_else(|| format!("Entry block not found for {}", entry_func_name))?;
// Create HashMap from remapper for BoundaryInjector (temporary adapter)
let mut value_map_for_injector = HashMap::new();
for join_in in &boundary.join_inputs {
if let Some(remapped) = remapper.get_value(*join_in) {
value_map_for_injector.insert(*join_in, remapped);
}
}
// Use BoundaryInjector to inject Copy instructions
if let Some(ref mut current_func) = self.current_function {
BoundaryInjector::inject_boundary_copies(
current_func,
entry_block_remapped,
boundary,
&value_map_for_injector,
debug,
)?;
}
}
// 6. Create exit block with PHI for return values from JoinIR functions
// Phase 189-Fix: Generate exit PHI if there are multiple return values
let exit_phi_result_id = if let Some(ref mut func) = self.current_function {
let mut exit_block = BasicBlock::new(exit_block_id);
// Phase 189-Fix: If we collected return values, create a PHI in exit block
// This merges all return values from JoinIR functions into a single value
let phi_result = if !exit_phi_inputs.is_empty() {
// Allocate a new ValueId for the PHI result
let phi_dst = self.value_gen.next();
exit_block.instructions.push(MirInstruction::Phi {
dst: phi_dst,
inputs: exit_phi_inputs.clone(),
type_hint: None,
});
exit_block.instruction_spans.push(crate::ast::Span::unknown());
if debug {
eprintln!(
"[cf_loop/joinir] Exit block PHI: {:?} = phi {:?}",
phi_dst, exit_phi_inputs
);
}
Some(phi_dst)
} else {
None
};
func.add_block(exit_block);
if debug {
eprintln!("[cf_loop/joinir] Created exit block: {:?}", exit_block_id);
}
phi_result
} else {
None
};
feat(joinir): Phase 190 - LoopExitBinding boxification Formalize exit PHI → variable_map reconnection with explicit LoopExitBinding structure. Eliminates hardcoded variable names and prepares for Pattern 4+ multi-carrier support. Key changes: 1. **New LoopExitBinding struct**: - carrier_name: String (e.g., "sum", "count") - join_exit_value: ValueId (JoinIR exit value) - host_slot: ValueId (variable_map destination) Makes it explicit: WHICH variable, FROM where, TO where. 2. **Updated JoinInlineBoundary**: - Replaced implicit host_outputs: Vec<ValueId> - With explicit exit_bindings: Vec<LoopExitBinding> - Old APIs marked #[deprecated] for backward compatibility 3. **Pattern 3 now uses explicit bindings**: Before: boundary.host_outputs = vec![sum_var_id] // implicit After: boundary.exit_bindings = vec![LoopExitBinding { carrier_name: "sum".to_string(), join_exit_value: ValueId(18), host_slot: sum_var_id, }] 4. **merge_joinir_mir_blocks() updated**: - Consumes exit_bindings instead of bare ValueIds - Enhanced debug output shows carrier names - Validates carrier name matches variable_map expectations Benefits: - Self-documenting code: bindings explain themselves - Multi-carrier ready: Pattern 4+ just extend the vec![] - Type-safe: No implicit semantics - Debuggable: Explicit carrier name in logs Test status: - Build: ✅ SUCCESS (0 errors, 47 warnings) - Pattern 3: ✅ PASS (no regressions) - Backward compatibility: ✅ Maintained via #[deprecated] Prepare for Phase 191: Pattern Router Table and Phase 192: JoinLoopTrace 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: ChatGPT <noreply@openai.com> Co-Authored-By: Claude <noreply@anthropic.com>
2025-12-05 19:59:40 +09:00
// Phase 190: Use explicit LoopExitBinding to reconnect exit PHI to variable_map
// Each binding explicitly names the carrier variable and maps exit PHI to it.
if let Some(phi_result) = exit_phi_result_id {
if let Some(ref boundary) = boundary {
feat(joinir): Phase 190 - LoopExitBinding boxification Formalize exit PHI → variable_map reconnection with explicit LoopExitBinding structure. Eliminates hardcoded variable names and prepares for Pattern 4+ multi-carrier support. Key changes: 1. **New LoopExitBinding struct**: - carrier_name: String (e.g., "sum", "count") - join_exit_value: ValueId (JoinIR exit value) - host_slot: ValueId (variable_map destination) Makes it explicit: WHICH variable, FROM where, TO where. 2. **Updated JoinInlineBoundary**: - Replaced implicit host_outputs: Vec<ValueId> - With explicit exit_bindings: Vec<LoopExitBinding> - Old APIs marked #[deprecated] for backward compatibility 3. **Pattern 3 now uses explicit bindings**: Before: boundary.host_outputs = vec![sum_var_id] // implicit After: boundary.exit_bindings = vec![LoopExitBinding { carrier_name: "sum".to_string(), join_exit_value: ValueId(18), host_slot: sum_var_id, }] 4. **merge_joinir_mir_blocks() updated**: - Consumes exit_bindings instead of bare ValueIds - Enhanced debug output shows carrier names - Validates carrier name matches variable_map expectations Benefits: - Self-documenting code: bindings explain themselves - Multi-carrier ready: Pattern 4+ just extend the vec![] - Type-safe: No implicit semantics - Debuggable: Explicit carrier name in logs Test status: - Build: ✅ SUCCESS (0 errors, 47 warnings) - Pattern 3: ✅ PASS (no regressions) - Backward compatibility: ✅ Maintained via #[deprecated] Prepare for Phase 191: Pattern Router Table and Phase 192: JoinLoopTrace 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: ChatGPT <noreply@openai.com> Co-Authored-By: Claude <noreply@anthropic.com>
2025-12-05 19:59:40 +09:00
// Phase 190: Use exit_bindings for explicit carrier naming
// This eliminates ambiguity about which variable is being updated
for binding in &boundary.exit_bindings {
// Find the variable in variable_map that matches the binding's host_slot
for (var_name, vid) in self.variable_map.iter_mut() {
if *vid == binding.host_slot {
refactor(joinir): Phase 189 - Remove hardcoded 'sum' variable, generalize exit PHI connection Key improvements: 1. **Eliminate hardcoded variable name**: Replace hardcoded "sum" with generic ValueId-based variable_map updates. Add new JoinInlineBoundary constructor `new_with_input_and_host_outputs()` for Pattern 3. 2. **Generalize output slot mapping**: Exit PHI result now updates all host_outputs entries in variable_map, not just hardcoded "sum". Prepares for future multi-carrier patterns. 3. **PHI preservation in blocks**: Fix block finalization to preserve existing PHI instructions (from handle_select) instead of overwriting them. 4. **Stable function name→ValueId mapping**: Ensure consistent ValueId assignment for tail call detection across multi-function merges. 5. **Enhanced debugging**: Add detailed logging in block converter and meta analysis for PHI verification. Files modified: - src/mir/builder/control_flow.rs: Remove hardcoded "sum", use boundary outputs - src/mir/join_ir/lowering/inline_boundary.rs: Add new constructor - src/mir/join_ir_vm_bridge/joinir_block_converter.rs: Stable mappings, PHI preservation - src/mir/join_ir_vm_bridge/meta.rs: Debug output for PHI tracking - src/mir/builder/joinir_id_remapper.rs: PHI value remapping - src/mir/builder/joinir_inline_boundary_injector.rs: Span preservation Test status: Pattern 3 (loop_if_phi.hako) still produces correct result (sum=9, RC=9) 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: ChatGPT <noreply@openai.com> Co-Authored-By: Claude <noreply@anthropic.com>
2025-12-05 19:39:54 +09:00
*vid = phi_result;
if debug {
eprintln!(
feat(joinir): Phase 190 - LoopExitBinding boxification Formalize exit PHI → variable_map reconnection with explicit LoopExitBinding structure. Eliminates hardcoded variable names and prepares for Pattern 4+ multi-carrier support. Key changes: 1. **New LoopExitBinding struct**: - carrier_name: String (e.g., "sum", "count") - join_exit_value: ValueId (JoinIR exit value) - host_slot: ValueId (variable_map destination) Makes it explicit: WHICH variable, FROM where, TO where. 2. **Updated JoinInlineBoundary**: - Replaced implicit host_outputs: Vec<ValueId> - With explicit exit_bindings: Vec<LoopExitBinding> - Old APIs marked #[deprecated] for backward compatibility 3. **Pattern 3 now uses explicit bindings**: Before: boundary.host_outputs = vec![sum_var_id] // implicit After: boundary.exit_bindings = vec![LoopExitBinding { carrier_name: "sum".to_string(), join_exit_value: ValueId(18), host_slot: sum_var_id, }] 4. **merge_joinir_mir_blocks() updated**: - Consumes exit_bindings instead of bare ValueIds - Enhanced debug output shows carrier names - Validates carrier name matches variable_map expectations Benefits: - Self-documenting code: bindings explain themselves - Multi-carrier ready: Pattern 4+ just extend the vec![] - Type-safe: No implicit semantics - Debuggable: Explicit carrier name in logs Test status: - Build: ✅ SUCCESS (0 errors, 47 warnings) - Pattern 3: ✅ PASS (no regressions) - Backward compatibility: ✅ Maintained via #[deprecated] Prepare for Phase 191: Pattern Router Table and Phase 192: JoinLoopTrace 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: ChatGPT <noreply@openai.com> Co-Authored-By: Claude <noreply@anthropic.com>
2025-12-05 19:59:40 +09:00
"[cf_loop/joinir] Phase 190: Reconnected exit PHI {:?} to variable_map['{}'] (carrier: {})",
phi_result, var_name, binding.carrier_name
);
}
// Validate carrier name matches
if var_name != &binding.carrier_name && debug {
eprintln!(
"[cf_loop/joinir] WARNING: Carrier name mismatch: expected '{}', found '{}'",
binding.carrier_name, var_name
refactor(joinir): Phase 189 - Remove hardcoded 'sum' variable, generalize exit PHI connection Key improvements: 1. **Eliminate hardcoded variable name**: Replace hardcoded "sum" with generic ValueId-based variable_map updates. Add new JoinInlineBoundary constructor `new_with_input_and_host_outputs()` for Pattern 3. 2. **Generalize output slot mapping**: Exit PHI result now updates all host_outputs entries in variable_map, not just hardcoded "sum". Prepares for future multi-carrier patterns. 3. **PHI preservation in blocks**: Fix block finalization to preserve existing PHI instructions (from handle_select) instead of overwriting them. 4. **Stable function name→ValueId mapping**: Ensure consistent ValueId assignment for tail call detection across multi-function merges. 5. **Enhanced debugging**: Add detailed logging in block converter and meta analysis for PHI verification. Files modified: - src/mir/builder/control_flow.rs: Remove hardcoded "sum", use boundary outputs - src/mir/join_ir/lowering/inline_boundary.rs: Add new constructor - src/mir/join_ir_vm_bridge/joinir_block_converter.rs: Stable mappings, PHI preservation - src/mir/join_ir_vm_bridge/meta.rs: Debug output for PHI tracking - src/mir/builder/joinir_id_remapper.rs: PHI value remapping - src/mir/builder/joinir_inline_boundary_injector.rs: Span preservation Test status: Pattern 3 (loop_if_phi.hako) still produces correct result (sum=9, RC=9) 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: ChatGPT <noreply@openai.com> Co-Authored-By: Claude <noreply@anthropic.com>
2025-12-05 19:39:54 +09:00
);
}
}
}
}
feat(joinir): Phase 190 - LoopExitBinding boxification Formalize exit PHI → variable_map reconnection with explicit LoopExitBinding structure. Eliminates hardcoded variable names and prepares for Pattern 4+ multi-carrier support. Key changes: 1. **New LoopExitBinding struct**: - carrier_name: String (e.g., "sum", "count") - join_exit_value: ValueId (JoinIR exit value) - host_slot: ValueId (variable_map destination) Makes it explicit: WHICH variable, FROM where, TO where. 2. **Updated JoinInlineBoundary**: - Replaced implicit host_outputs: Vec<ValueId> - With explicit exit_bindings: Vec<LoopExitBinding> - Old APIs marked #[deprecated] for backward compatibility 3. **Pattern 3 now uses explicit bindings**: Before: boundary.host_outputs = vec![sum_var_id] // implicit After: boundary.exit_bindings = vec![LoopExitBinding { carrier_name: "sum".to_string(), join_exit_value: ValueId(18), host_slot: sum_var_id, }] 4. **merge_joinir_mir_blocks() updated**: - Consumes exit_bindings instead of bare ValueIds - Enhanced debug output shows carrier names - Validates carrier name matches variable_map expectations Benefits: - Self-documenting code: bindings explain themselves - Multi-carrier ready: Pattern 4+ just extend the vec![] - Type-safe: No implicit semantics - Debuggable: Explicit carrier name in logs Test status: - Build: ✅ SUCCESS (0 errors, 47 warnings) - Pattern 3: ✅ PASS (no regressions) - Backward compatibility: ✅ Maintained via #[deprecated] Prepare for Phase 191: Pattern Router Table and Phase 192: JoinLoopTrace 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: ChatGPT <noreply@openai.com> Co-Authored-By: Claude <noreply@anthropic.com>
2025-12-05 19:59:40 +09:00
// Phase 190: Backward compatibility - also check deprecated host_outputs
#[allow(deprecated)]
if !boundary.host_outputs.is_empty() && debug {
eprintln!(
"[cf_loop/joinir] WARNING: Using deprecated host_outputs. Migrate to exit_bindings."
);
}
}
}
// 7. Jump from current block to entry function's entry block
// Entry function is first function by convention
let (entry_func_name, entry_func) = mir_module
.functions
.iter()
.next()
.ok_or("JoinIR module has no functions")?;
// Use remapper to get entry block mapping
let entry_block = remapper.get_block(entry_func_name, entry_func.entry_block)
.ok_or_else(|| format!("Entry block not found for {}", entry_func_name))?;
if debug {
eprintln!("[cf_loop/joinir] Entry function name: {}", entry_func_name);
eprintln!("[cf_loop/joinir] Entry function's entry_block (JoinIR local): {:?}", entry_func.entry_block);
eprintln!("[cf_loop/joinir] Entry block (remapped): {:?}", entry_block);
eprintln!("[cf_loop/joinir] Current block before emit_jump: {:?}", self.current_block);
eprintln!("[cf_loop/joinir] Jumping to entry block: {:?}", entry_block);
}
crate::mir::builder::emission::branch::emit_jump(self, entry_block)?;
if debug {
eprintln!("[cf_loop/joinir] After emit_jump, current_block: {:?}", self.current_block);
}
// 8. Switch to exit block for subsequent code
self.start_new_block(exit_block_id)?;
if debug {
eprintln!(
"[cf_loop/joinir] Phase 189: Merge complete: {} functions merged, continuing from {:?}",
mir_module.functions.len(),
exit_block_id
);
// DEBUG: Check bb0's terminator after merge
if let Some(ref func) = self.current_function {
if let Some(bb0) = func.get_block(BasicBlockId(0)) {
eprintln!("[cf_loop/joinir] bb0 terminator after merge: {:?}", bb0.terminator);
}
// DEBUG: Check bb9's PHI after merge (PHI merge block)
if let Some(bb9) = func.get_block(BasicBlockId(9)) {
let phi_count = bb9.instructions.iter().filter(|i| matches!(i, MirInstruction::Phi { .. })).count();
eprintln!("[cf_loop/joinir] bb9 after merge: {} instructions, {} PHI", bb9.instructions.len(), phi_count);
for (idx, inst) in bb9.instructions.iter().take(3).enumerate() {
eprintln!("[cf_loop/joinir] bb9[{}]: {:?}", idx, inst);
}
}
}
}
Ok(exit_phi_result_id)
}
// Phase 189: collect_values_in_block/collect_values_in_instruction removed
// These functions are now provided by JoinIrIdRemapper::collect_values_in_block()
// Phase 189: remap_joinir_instruction/remap_instruction removed
// These functions are now provided by JoinIrIdRemapper::remap_instruction()
/// Control-flow: try/catch/finally
pub(super) fn cf_try_catch(
&mut self,
try_body: Vec<ASTNode>,
catch_clauses: Vec<crate::ast::CatchClause>,
finally_body: Option<Vec<ASTNode>>,
) -> Result<ValueId, String> {
if std::env::var("NYASH_BUILDER_DISABLE_TRYCATCH")
.ok()
.as_deref()
== Some("1")
{
let try_ast = ASTNode::Program {
statements: try_body,
span: crate::ast::Span::unknown(),
};
let result = self.build_expression(try_ast)?;
return Ok(result);
}
let try_block = self.block_gen.next();
let catch_block = self.block_gen.next();
let finally_block = if finally_body.is_some() {
Some(self.block_gen.next())
} else {
None
};
let exit_block = self.block_gen.next();
// Snapshot deferred-return state
let saved_defer_active = self.return_defer_active;
let saved_defer_slot = self.return_defer_slot;
let saved_defer_target = self.return_defer_target;
let saved_deferred_flag = self.return_deferred_emitted;
let saved_in_cleanup = self.in_cleanup_block;
let saved_allow_ret = self.cleanup_allow_return;
let saved_allow_throw = self.cleanup_allow_throw;
let ret_slot = self.next_value_id();
self.return_defer_active = true;
self.return_defer_slot = Some(ret_slot);
self.return_deferred_emitted = false;
self.return_defer_target = Some(finally_block.unwrap_or(exit_block));
if let Some(catch_clause) = catch_clauses.first() {
if std::env::var("NYASH_DEBUG_TRYCATCH").ok().as_deref() == Some("1") {
eprintln!(
"[BUILDER] Emitting catch handler for {:?}",
catch_clause.exception_type
);
}
let exception_value = self.next_value_id();
self.emit_instruction(MirInstruction::Catch {
exception_type: catch_clause.exception_type.clone(),
exception_value,
handler_bb: catch_block,
})?;
}
// Enter try block
crate::mir::builder::emission::branch::emit_jump(self, try_block)?;
self.start_new_block(try_block)?;
let try_ast = ASTNode::Program {
statements: try_body,
span: crate::ast::Span::unknown(),
};
let _try_result = self.build_expression(try_ast)?;
if !self.is_current_block_terminated() {
let next_target = finally_block.unwrap_or(exit_block);
crate::mir::builder::emission::branch::emit_jump(self, next_target)?;
}
// Catch block
self.start_new_block(catch_block)?;
if std::env::var("NYASH_DEBUG_TRYCATCH").ok().as_deref() == Some("1") {
eprintln!("[BUILDER] Enter catch block {:?}", catch_block);
}
if let Some(catch_clause) = catch_clauses.first() {
let catch_ast = ASTNode::Program {
statements: catch_clause.body.clone(),
span: crate::ast::Span::unknown(),
};
self.build_expression(catch_ast)?;
}
if !self.is_current_block_terminated() {
let next_target = finally_block.unwrap_or(exit_block);
crate::mir::builder::emission::branch::emit_jump(self, next_target)?;
}
// Finally
let mut cleanup_terminated = false;
if let (Some(finally_block_id), Some(finally_statements)) = (finally_block, finally_body) {
self.start_new_block(finally_block_id)?;
self.in_cleanup_block = true;
self.cleanup_allow_return = crate::config::env::cleanup_allow_return();
self.cleanup_allow_throw = crate::config::env::cleanup_allow_throw();
self.return_defer_active = false; // do not defer inside cleanup
let finally_ast = ASTNode::Program {
statements: finally_statements,
span: crate::ast::Span::unknown(),
};
self.build_expression(finally_ast)?;
cleanup_terminated = self.is_current_block_terminated();
if !cleanup_terminated {
crate::mir::builder::emission::branch::emit_jump(self, exit_block)?;
}
self.in_cleanup_block = false;
}
// Exit block
self.start_new_block(exit_block)?;
let result = if self.return_deferred_emitted && !cleanup_terminated {
self.emit_instruction(MirInstruction::Return {
value: Some(ret_slot),
})?;
crate::mir::builder::emission::constant::emit_void(self)
} else {
crate::mir::builder::emission::constant::emit_void(self)
};
// Restore context
self.return_defer_active = saved_defer_active;
self.return_defer_slot = saved_defer_slot;
self.return_defer_target = saved_defer_target;
self.return_deferred_emitted = saved_deferred_flag;
self.in_cleanup_block = saved_in_cleanup;
self.cleanup_allow_return = saved_allow_ret;
self.cleanup_allow_throw = saved_allow_throw;
Ok(result)
}
/// Phase 188-Impl-2: Extract loop variable name from condition
///
/// For `i < 3`, extracts `i`.
/// For `arr.length() > 0`, extracts `arr`.
///
/// This is a minimal implementation that handles simple comparison patterns.
fn extract_loop_variable_from_condition(&self, condition: &ASTNode) -> Result<String, String> {
use crate::ast::BinaryOperator;
match condition {
ASTNode::BinaryOp {
operator, left, ..
} if matches!(
operator,
BinaryOperator::Less
| BinaryOperator::Greater
| BinaryOperator::LessEqual
| BinaryOperator::GreaterEqual
) =>
{
// Binary comparison: extract variable from left side
match &**left {
ASTNode::Variable { name, .. } => Ok(name.clone()),
_ => Err(format!(
"[cf_loop/pattern1] Cannot extract loop variable from condition: {:?}",
condition
)),
}
}
_ => Err(format!(
"[cf_loop/pattern1] Unsupported loop condition pattern: {:?}",
condition
)),
}
}
/// Control-flow: throw
pub(super) fn cf_throw(&mut self, expression: ASTNode) -> Result<ValueId, String> {
if self.in_cleanup_block && !self.cleanup_allow_throw {
return Err("throw is not allowed inside cleanup block (enable NYASH_CLEANUP_ALLOW_THROW=1 to permit)".to_string());
}
if std::env::var("NYASH_BUILDER_DISABLE_THROW").ok().as_deref() == Some("1") {
let v = self.build_expression(expression)?;
self.emit_instruction(MirInstruction::ExternCall {
dst: None,
iface_name: "env.debug".to_string(),
method_name: "trace".to_string(),
args: vec![v],
effects: EffectMask::PURE.add(Effect::Debug),
})?;
return Ok(v);
}
let exception_value = self.build_expression(expression)?;
self.emit_instruction(MirInstruction::Throw {
exception: exception_value,
effects: EffectMask::PANIC,
})?;
Ok(exception_value)
}
}