Files
hakorune/src/mir/builder.rs
nyash-codex d605611a16 feat(canonicalizer): Phase 143-P0 - parse_number pattern support
Add parse_number pattern recognition to canonicalizer, expanding adaptation
range for digit collection loops with break in THEN clause.

## Changes

### New Recognizer (ast_feature_extractor.rs)
- `detect_parse_number_pattern()`: Detects `if invalid { break }` pattern
- `ParseNumberInfo`: Struct for extracted pattern info
- ~150 lines added

### Canonicalizer Integration (canonicalizer.rs)
- Parse_number pattern detection before skip_whitespace
- LoopSkeleton construction with 4 steps (Header + Body x2 + Update)
- Routes to Pattern2Break (has_break=true)
- ~60 lines modified

### Export Chain (6 files)
- patterns/mod.rs → joinir/mod.rs → control_flow/mod.rs
- builder.rs → mir/mod.rs
- 8 lines total

### Tests
- `test_parse_number_pattern_recognized()`: Unit test for recognition
- Strict parity verification: GREEN (canonical and router agree)
- ~130 lines added

## Pattern Comparison

| Aspect | Skip Whitespace | Parse Number |
|--------|----------------|--------------|
| Break location | ELSE clause | THEN clause |
| Pattern | `if cond { update } else { break }` | `if invalid { break } rest... update` |
| Body after if | None | Required (result append) |

## Results

-  Skeleton creation successful
-  RoutingDecision matches router (Pattern2Break)
-  Strict parity OK (canonicalizer ↔ router agreement)
-  Unit test PASS
-  Manual test: test_pattern2_parse_number.hako executes correctly

## Statistics

- New patterns: 1 (parse_number)
- Total patterns: 3 (skip_whitespace, parse_number, continue)
- Lines added: ~280
- Files modified: 8
- Parity status: Green 

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
2025-12-16 09:08:37 +09:00

1156 lines
49 KiB
Rust
Raw Blame History

This file contains ambiguous Unicode characters

This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

/*!
* MIR Builder - Converts AST to MIR/SSA form
*
* Implements AST → MIR conversion with SSA construction
*/
use super::slot_registry::resolve_slot_by_type_name;
use super::{
BasicBlock, BasicBlockId, BasicBlockIdGenerator, CompareOp, ConstValue, Effect, EffectMask,
FunctionSignature, MirFunction, MirInstruction, MirModule, MirType, ValueId, ValueIdGenerator,
};
use crate::ast::{ASTNode, LiteralValue, Span};
use crate::mir::builder::builder_calls::CallTarget;
use crate::mir::region::function_slot_registry::FunctionSlotRegistry;
use crate::mir::region::RegionId;
use std::collections::HashSet;
use std::collections::{BTreeMap, HashMap};
mod binding_context; // Phase 136 follow-up (Step 4/7): BindingContext extraction
mod builder_calls;
mod call_resolution; // ChatGPT5 Pro: Type-safe call resolution utilities
mod calls; // Call system modules (refactored from builder_calls)
mod compilation_context; // Phase 136 follow-up (Step 7/7): CompilationContext extraction
mod context; // BoxCompilationContext - 箱理論による静的Boxコンパイル時のコンテキスト分離
mod core_context; // Phase 136 follow-up (Step 2/7): CoreContext extraction
mod decls; // declarations lowering split
mod exprs; // expression lowering split
mod exprs_call;
mod metadata_context; // Phase 136 follow-up (Step 6/7): MetadataContext extraction
mod method_call_handlers;
mod variable_context; // Phase 136 follow-up (Step 5/7): VariableContext extraction // Method call handler separation (Phase 3) // call(expr)
// include lowering removed (using is handled in runner)
mod control_flow; // thin wrappers to centralize control-flow entrypoints
// Phase 140-P4-A: Re-export for loop_canonicalizer SSOT (crate-wide visibility)
pub(crate) use control_flow::{detect_skip_whitespace_pattern, SkipWhitespaceInfo};
// Phase 142-P1: Re-export continue pattern detection for loop_canonicalizer
pub(crate) use control_flow::{detect_continue_pattern, ContinuePatternInfo};
// Phase 143-P0: Re-export parse_number pattern detection for loop_canonicalizer
pub(crate) use control_flow::{detect_parse_number_pattern, ParseNumberInfo};
mod exprs_lambda; // lambda lowering
mod exprs_peek; // peek expression
mod exprs_qmark; // ?-propagate
mod fields; // field access/assignment lowering split
mod if_form;
mod joinir_id_remapper; // Phase 189: JoinIR ID remapping (ValueId/BlockId translation)
mod joinir_inline_boundary_injector; // Phase 189: JoinInlineBoundary Copy instruction injector
mod lifecycle;
mod loop_frontend_binding; // Phase 50: Loop Frontend Binding (JoinIR variable mapping)
pub(crate) mod loops;
mod ops;
mod phi;
mod phi_merge; // Phase 25.1q: Unified PHI merge helper // prepare/lower_root/finalize split
// legacy large-match remains inline for now (planned extraction)
mod emission; // emission::*Const/Compare/Branch の薄い発行箱)
mod emit_guard; // EmitGuardBoxemit直前の最終関所
mod metadata; // MetadataPropagationBoxtype/originの伝播
mod name_const; // NameConstBox関数名Const生成
mod observe; // P0: dev-only observability helpersssa/resolve
mod origin; // P0: origin inferenceme/Knownと PHI 伝播(軽量)
mod plugin_sigs; // plugin signature loader
mod receiver; // ReceiverMaterializationBoxMethod recv の pin+LocalSSA 集約)
mod rewrite; // P1: Known rewrite & special consolidation
mod router; // RouterPolicyBoxUnified vs BoxCall
mod schedule; // BlockScheduleBox物理順序: PHI→materialize→body
mod scope_context; // Phase 136 follow-up (Step 3/7): ScopeContext extraction
mod ssa; // LocalSSA helpers (in-block materialization)
mod stmts;
mod type_context; // Phase 136 follow-up: TypeContext extraction
mod type_facts; // Phase 136 follow-up: Type inference facts box
pub(crate) mod type_registry;
mod types; // types::annotation / inference型注釈/推論の箱: 推論は後段)
mod utils;
mod vars; // variables/scope helpers // small loop helpers (header/exit context) // TypeRegistryBox型情報管理の一元化
// Unified member property kinds for computed/once/birth_once
#[derive(Debug, Clone, PartialEq, Eq, Hash)]
pub(crate) enum PropertyKind {
Computed,
Once,
BirthOnce,
}
/// MIR builder for converting AST to SSA form
pub struct MirBuilder {
/// Current module being built
pub(super) current_module: Option<MirModule>,
/// Current basic block being built
pub(super) current_block: Option<BasicBlockId>,
/// Phase 136 follow-up (Step 2/7): Core ID generation context
/// Consolidates value_gen, block_gen, next_binding_id, temp_slot_counter, debug_join_counter.
/// Direct field access for backward compatibility (migration in progress).
pub(super) core_ctx: core_context::CoreContext,
/// Phase 136 follow-up: Type information context
/// Consolidates value_types, value_kinds, value_origin_newbox for better organization.
/// Direct field access for backward compatibility (migration in progress).
pub(super) type_ctx: type_context::TypeContext,
/// Phase 136 follow-up (Step 3/7): Scope and control flow context
/// Consolidates lexical_scope_stack, loop stacks, if_merge_stack, current_function,
/// function_param_names, debug_scope_stack for better organization.
/// Direct field access for backward compatibility (migration in progress).
pub(super) scope_ctx: scope_context::ScopeContext,
/// Phase 136 follow-up (Step 4/7): Binding context
/// Consolidates binding_map (String -> BindingId mapping).
/// Direct field access for backward compatibility (migration in progress).
pub(super) binding_ctx: binding_context::BindingContext,
/// Phase 136 follow-up (Step 5/7): Variable context
/// Consolidates variable_map (String -> ValueId mapping for SSA conversion).
/// Direct field access for backward compatibility (migration in progress).
pub(super) variable_ctx: variable_context::VariableContext,
/// Phase 136 follow-up (Step 6/7): Metadata context
/// Consolidates current_span, source_file, hint_sink, current_region_stack.
/// Direct field access for backward compatibility (migration in progress).
pub(super) metadata_ctx: metadata_context::MetadataContext,
/// Phase 136 follow-up (Step 7/7): Compilation context
/// Consolidates compilation_context, current_static_box, user_defined_boxes, reserved_value_ids,
/// fn_body_ast, weak_fields_by_box, property_getters_by_box, field_origin_class, field_origin_by_box,
/// static_method_index, method_tail_index, type_registry, current_slot_registry, plugin_method_sigs.
/// Direct field access for backward compatibility (migration in progress).
pub(super) comp_ctx: compilation_context::CompilationContext,
/// Pending phi functions to be inserted
#[allow(dead_code)]
pub(super) pending_phis: Vec<(BasicBlockId, ValueId, String)>,
// Phase 2-5: binding_map removed - use binding_ctx.binding_map instead
// include guards removed
// フェーズM: no_phi_modeフィールド削除常にPHI使用
// ---- Try/Catch/Cleanup lowering context ----
/// When true, `return` statements are deferred: they assign to `return_defer_slot`
/// and jump to `return_defer_target` (typically the cleanup/exit block).
pub(super) return_defer_active: bool,
/// Slot value to receive deferred return values (edge-copy mode friendly).
pub(super) return_defer_slot: Option<ValueId>,
/// Target block to jump to on deferred return.
pub(super) return_defer_target: Option<BasicBlockId>,
/// Set to true when a deferred return has been emitted in the current context.
pub(super) return_deferred_emitted: bool,
/// True while lowering the cleanup block.
pub(super) in_cleanup_block: bool,
/// Policy flags (snapshotted at entry of try/catch lowering)
pub(super) cleanup_allow_return: bool,
pub(super) cleanup_allow_throw: bool,
/// If true, skip entry materialization of pinned slots on the next start_new_block call.
suppress_pin_entry_copy_next: bool,
// ----------------------
// Debug scope context (dev only; zero-cost when unused)
// ----------------------
/// Local SSA cache: ensure per-block materialization for critical operands (e.g., recv)
/// Key: (bb, original ValueId, kind) -> local ValueId
/// kind: 0=recv, 1+ reserved for future (args etc.)
pub(super) local_ssa_map: HashMap<(BasicBlockId, ValueId, u8), ValueId>,
/// BlockSchedule cache: deduplicate materialize copies per (bb, src)
pub(super) schedule_mat_map: HashMap<(BasicBlockId, ValueId), ValueId>,
/// Mapping from ValueId to its pin slot name (e.g., "__pin$3$@recv")
/// Used by LocalSSA to redirect old pinned values to the latest slot value.
pub(super) pin_slot_names: HashMap<ValueId, String>,
/// Guard flag to prevent re-entering emit_unified_call from BoxCall fallback.
/// Used when RouterPolicyBox in emit_unified_call has already decided to
/// route a given Method call to BoxCall; emit_box_or_plugin_call must not
/// bounce back into the unified path for the same call, otherwise an
/// infinite recursion (emit_unified_call → emit_box_or_plugin_call →
/// emit_unified_call …) can occur when routing decisions disagree.
pub(super) in_unified_boxcall_fallback: bool,
/// Recursion depth counter for debugging stack overflow
/// Tracks the depth of build_expression calls to detect infinite loops
pub(super) recursion_depth: usize,
/// Root lowering mode: how to treat top-level Program
/// - None: not decided yet (lower_root not called)
/// - Some(true): App mode (static box Main.main is entry)
/// - Some(false): Script/Test mode (top-level Program runs sequentially)
pub(super) root_is_app_mode: Option<bool>,
/// 🎯 Phase 21.7: Static box singleton instances for methodization
/// Maps BoxName → ValueId of singleton instance (created on demand)
/// Used when HAKO_MIR_BUILDER_METHODIZE=1 to convert Global("BoxName.method/arity")
/// to Method{receiver=singleton} calls
pub(super) static_box_singletons: HashMap<String, ValueId>,
}
impl MirBuilder {
/// Create a new MIR builder
pub fn new() -> Self {
let plugin_method_sigs = plugin_sigs::load_plugin_method_sigs();
let core_ctx = core_context::CoreContext::new();
// Phase 136 Step 7/7: Compilation context (new SSOT)
let comp_ctx =
compilation_context::CompilationContext::with_plugin_sigs(plugin_method_sigs.clone());
// フェーズM: no_phi_mode初期化削除
#[allow(deprecated)]
Self {
current_module: None,
current_block: None,
// Phase 136 Step 2/7: Core context (new SSOT)
core_ctx,
type_ctx: type_context::TypeContext::new(), // Phase 136: Type context
scope_ctx: scope_context::ScopeContext::new(), // Phase 136 Step 3/7: Scope context
binding_ctx: binding_context::BindingContext::new(), // Phase 136 Step 4/7: Binding context
variable_ctx: variable_context::VariableContext::new(), // Phase 136 Step 5/7: Variable context
metadata_ctx: metadata_context::MetadataContext::new(), // Phase 136 Step 6/7: Metadata context
comp_ctx, // Phase 136 Step 7/7: Compilation context
pending_phis: Vec::new(),
// Phase 2-5: binding_map initialization removed
// フェーズM: no_phi_modeフィールド削除
return_defer_active: false,
return_defer_slot: None,
return_defer_target: None,
return_deferred_emitted: false,
in_cleanup_block: false,
cleanup_allow_return: false,
cleanup_allow_throw: false,
suppress_pin_entry_copy_next: false,
local_ssa_map: HashMap::new(),
schedule_mat_map: HashMap::new(),
pin_slot_names: HashMap::new(),
in_unified_boxcall_fallback: false,
recursion_depth: 0,
root_is_app_mode: None,
static_box_singletons: HashMap::new(), // Phase 21.7: methodization support
}
}
// Phase 2-5: BindingContext sync helpers removed - binding_ctx is now SSOT
// Phase 2-6: VariableContext sync helpers removed - variable_ctx is now SSOT
/// Push/pop helpers for If merge context (best-effort; optional usage)
pub(super) fn push_if_merge(&mut self, bb: BasicBlockId) {
// Phase 2-4: Use scope_ctx only (legacy field removed)
self.scope_ctx.push_if_merge(bb);
}
pub(super) fn pop_if_merge(&mut self) {
// Phase 2-4: Use scope_ctx only (legacy field removed)
let _ = self.scope_ctx.pop_if_merge();
}
/// Suppress entry pin copy for the next start_new_block (used for merge blocks).
pub(super) fn suppress_next_entry_pin_copy(&mut self) {
self.suppress_pin_entry_copy_next = true;
}
// ---- Phase 74: BindingId allocation ----
/// Allocate a new BindingId (parallel to ValueId allocation)
///
/// ## Parallel ValueId/BindingId Allocation
///
/// BindingId allocation is completely independent from ValueId allocation:
/// - `next_value_id()` increments `value_gen` counter
/// - `allocate_binding_id()` increments `next_binding_id` counter
///
/// This parallelism enables:
/// 1. **Stable binding identity** across SSA transformations
/// 2. **Independent shadowing tracking** separate from SSA renaming
/// 3. **Future ScopeManager migration** (Phase 75+) without breaking SSA
///
/// Example:
/// ```ignore
/// // local x = 1; <- allocate_binding_id() -> BindingId(0)
/// // next_value_id() -> ValueId(10)
/// // {
/// // local x = 2; <- allocate_binding_id() -> BindingId(1)
/// // next_value_id() -> ValueId(20)
/// // }
/// ```
pub fn allocate_binding_id(&mut self) -> super::BindingId {
// Phase 136 Step 2/7 + Phase 2-2: Use core_ctx as SSOT (no sync needed)
self.core_ctx.next_binding()
}
// ---- Hint helpers (no-op by default) ----
// Phase 136 Step 6/7: Delegate to metadata_ctx with legacy sync
#[inline]
pub(crate) fn hint_scope_enter(&mut self, id: u32) {
self.metadata_ctx.hint_scope_enter(id);
}
#[inline]
pub(crate) fn hint_scope_leave(&mut self, id: u32) {
self.metadata_ctx.hint_scope_leave(id);
}
#[inline]
pub(crate) fn hint_join_result<S: Into<String>>(&mut self, var: S) {
self.metadata_ctx.hint_join_result(var);
}
// ----------------------
// Debug scope helpers (region_id for DebugHub events)
// ----------------------
#[inline]
pub(crate) fn debug_next_join_id(&mut self) -> u32 {
// Phase 136 Step 2/7 + Phase 2-2: Use core_ctx as SSOT (no sync needed)
self.core_ctx.next_debug_join()
}
#[inline]
pub(crate) fn debug_push_region<S: Into<String>>(&mut self, region: S) {
// Phase 2-4: Use scope_ctx only (legacy field removed)
let region = region.into();
self.scope_ctx.debug_push_region(region);
}
#[inline]
pub(crate) fn debug_pop_region(&mut self) {
// Phase 2-4: Use scope_ctx only (legacy field removed)
self.scope_ctx.debug_pop_region();
}
#[inline]
#[allow(deprecated)]
pub(crate) fn debug_current_region_id(&self) -> Option<String> {
// Phase 136 Step 3/7: Read from scope_ctx (SSOT)
self.scope_ctx.debug_current_region_id()
}
/// Hint for downstream metadata: set the logical source file name/path for the next build.
/// Phase 136 Step 6/7: Delegate to metadata_ctx
pub fn set_source_file_hint<S: Into<String>>(&mut self, source: S) {
self.metadata_ctx.set_source_file(source);
}
/// Clear the source file hint (used when reusing the builder across modules).
/// Phase 136 Step 6/7: Delegate to metadata_ctx
pub fn clear_source_file_hint(&mut self) {
self.metadata_ctx.clear_source_file();
}
/// Resolve current source file hint (builder field or env fallback).
/// Phase 136 Step 6/7: Delegate to metadata_ctx
fn current_source_file(&self) -> Option<String> {
self.metadata_ctx
.current_source_file()
.or_else(|| std::env::var("NYASH_SOURCE_FILE_HINT").ok())
}
/// Create a new MirFunction with source metadata applied.
fn new_function_with_metadata(
&self,
signature: FunctionSignature,
entry_block: BasicBlockId,
) -> MirFunction {
let mut f = MirFunction::new(signature, entry_block);
f.metadata.source_file = self.current_source_file();
f
}
// ----------------------
// Compile trace helpers (dev only; env-gated)
// ----------------------
#[inline]
pub(super) fn compile_trace_enabled() -> bool {
std::env::var("NYASH_MIR_COMPILE_TRACE").ok().as_deref() == Some("1")
}
#[inline]
pub(super) fn trace_compile<S: AsRef<str>>(&self, msg: S) {
if Self::compile_trace_enabled() {
eprintln!("[mir-compile] {}", msg.as_ref());
}
}
// ----------------------
// Method tail index (performance helper)
// ----------------------
fn rebuild_method_tail_index(&mut self) {
self.comp_ctx.method_tail_index.clear();
if let Some(ref module) = self.current_module {
for name in module.functions.keys() {
if let (Some(dot), Some(slash)) = (name.rfind('.'), name.rfind('/')) {
if slash > dot {
let tail = &name[dot..];
self.comp_ctx
.method_tail_index
.entry(tail.to_string())
.or_insert_with(Vec::new)
.push(name.clone());
}
}
}
self.comp_ctx.method_tail_index_source_len = module.functions.len();
} else {
self.comp_ctx.method_tail_index_source_len = 0;
}
}
fn ensure_method_tail_index(&mut self) {
let need_rebuild = match self.current_module {
Some(ref refmod) => {
self.comp_ctx.method_tail_index_source_len != refmod.functions.len()
}
None => self.comp_ctx.method_tail_index_source_len != 0,
};
if need_rebuild {
self.rebuild_method_tail_index();
}
}
pub(super) fn method_candidates(&mut self, method: &str, arity: usize) -> Vec<String> {
self.ensure_method_tail_index();
let tail = format!(".{}{}", method, format!("/{}", arity));
self.comp_ctx
.method_tail_index
.get(&tail)
.cloned()
.unwrap_or_default()
}
pub(super) fn method_candidates_tail<S: AsRef<str>>(&mut self, tail: S) -> Vec<String> {
self.ensure_method_tail_index();
self.comp_ctx
.method_tail_index
.get(tail.as_ref())
.cloned()
.unwrap_or_default()
}
/// Build a complete MIR module from AST
pub fn build_module(&mut self, ast: ASTNode) -> Result<MirModule, String> {
self.prepare_module()?;
let result_value = self.lower_root(ast)?;
self.finalize_module(result_value)
}
/// Build an expression and return its value ID
pub(super) fn build_expression(&mut self, ast: ASTNode) -> Result<ValueId, String> {
// Delegated to exprs.rs to keep this file lean
// Debug: Track recursion depth to detect infinite loops
const MAX_RECURSION_DEPTH: usize = 200;
self.recursion_depth += 1;
if self.recursion_depth > MAX_RECURSION_DEPTH {
eprintln!("\n[FATAL] ============================================");
eprintln!(
"[FATAL] Recursion depth exceeded {} in build_expression",
MAX_RECURSION_DEPTH
);
eprintln!("[FATAL] Current depth: {}", self.recursion_depth);
eprintln!("[FATAL] AST node type: {:?}", std::mem::discriminant(&ast));
eprintln!("[FATAL] ============================================\n");
return Err(format!(
"Recursion depth exceeded: {} (possible infinite loop)",
self.recursion_depth
));
}
let result = self.build_expression_impl(ast);
self.recursion_depth -= 1;
result
}
/// Build a literal value
pub(super) fn build_literal(&mut self, literal: LiteralValue) -> Result<ValueId, String> {
// Determine type without moving literal
let ty_for_dst = match &literal {
LiteralValue::Integer(_) => Some(super::MirType::Integer),
LiteralValue::Float(_) => Some(super::MirType::Float),
LiteralValue::Bool(_) => Some(super::MirType::Bool),
LiteralValue::String(_) => Some(super::MirType::String),
_ => None,
};
// Emit via ConstantEmissionBox仕様不変の統一ルート
let dst = match literal {
LiteralValue::Integer(n) => {
crate::mir::builder::emission::constant::emit_integer(self, n)
}
LiteralValue::Float(f) => crate::mir::builder::emission::constant::emit_float(self, f),
LiteralValue::String(s) => {
crate::mir::builder::emission::constant::emit_string(self, s)
}
LiteralValue::Bool(b) => crate::mir::builder::emission::constant::emit_bool(self, b),
LiteralValue::Null => crate::mir::builder::emission::constant::emit_null(self),
LiteralValue::Void => crate::mir::builder::emission::constant::emit_void(self),
};
// Annotate type
if let Some(ty) = ty_for_dst {
self.type_ctx.value_types.insert(dst, ty);
}
Ok(dst)
}
/// Build variable access
pub(super) fn build_variable_access(&mut self, name: String) -> Result<ValueId, String> {
// Step 5-5-G: __pin$ variables should NEVER be accessed from variable_map
// They are transient temporaries created during expression building and
// should not persist across blocks. If we see one here, it's a compiler bug.
if name.starts_with("__pin$") {
return Err(format!(
"COMPILER BUG: Attempt to access __pin$ temporary '{}' from variable_map. \
__pin$ variables should only exist as direct SSA values, not as named variables.",
name
));
}
if let Some(&value_id) = self.variable_ctx.variable_map.get(&name) {
Ok(value_id)
} else {
Err(self.undefined_variable_message(&name))
}
}
pub(in crate::mir::builder) fn undefined_variable_message(&self, name: &str) -> String {
// Enhance diagnostics using Using simple registry (Phase 1)
let mut msg = format!("Undefined variable: {}", name);
// Stage-3 keyword diagnostic (local/flow/try/catch/throw)
if name == "local" && !crate::config::env::parser_stage3_enabled() {
msg.push_str("\nHint: 'local' is a Stage-3 keyword. Prefer NYASH_FEATURES=stage3 (legacy: NYASH_PARSER_STAGE3=1 / HAKO_PARSER_STAGE3=1 for Stage-B).");
msg.push_str("\nFor AotPrep verification, use tools/hakorune_emit_mir.sh which sets these automatically.");
} else if (name == "flow" || name == "try" || name == "catch" || name == "throw")
&& !crate::config::env::parser_stage3_enabled()
{
msg.push_str(&format!("\nHint: '{}' is a Stage-3 keyword. Prefer NYASH_FEATURES=stage3 (legacy: NYASH_PARSER_STAGE3=1 / HAKO_PARSER_STAGE3=1 for Stage-B).", name));
}
let suggest = crate::using::simple_registry::suggest_using_for_symbol(name);
if !suggest.is_empty() {
msg.push_str("\nHint: symbol appears in using module(s): ");
msg.push_str(&suggest.join(", "));
msg.push_str(
"\nConsider adding 'using <module> [as Alias]' or check nyash.toml [using].",
);
}
msg
}
/// Build assignment
pub(super) fn build_assignment(
&mut self,
var_name: String,
value: ASTNode,
) -> Result<ValueId, String> {
// SSOT (LANGUAGE_REFERENCE_2025 / syntax-cheatsheet):
// - Assignment to an undeclared name is an error.
// - Use `local name = ...` (or `local name; name = ...`) to declare.
vars::assignment_resolver::AssignmentResolverBox::ensure_declared(self, &var_name)?;
let value_id = self.build_expression(value)?;
// Step 5-5-E: FIX variable map corruption bug
// REMOVED pin_to_slot() call - it was causing __pin$ temporaries to overwrite
// real variable names in the variable map.
//
// Root cause: pin_to_slot(raw_value_id, "@assign") would sometimes return
// a ValueId from a previous __pin$ temporary (e.g., __pin$767$@binop_lhs),
// causing variable_map["m"] to point to the wrong ValueId.
//
// SSA + PHI merges work correctly without explicit pinning here.
// The expression building already creates necessary temporaries.
// Step 5-5-F: NEVER insert __pin$ temporaries into variable_map
// __pin$ variables are transient compiler-generated temporaries that should
// never be tracked as real variables. They are used only within expression
// building and should not persist across blocks or loops.
//
// BUG FIX: Previously, __pin$ variables would be inserted into variable_map,
// causing stale references after LoopForm transformation renumbers blocks.
// Result: VM would try to read undefined ValueIds (e.g., ValueId(270) at bb303).
if !var_name.starts_with("__pin$") {
// In SSA form, each assignment creates a new value
self.variable_ctx
.variable_map
.insert(var_name.clone(), value_id);
}
Ok(value_id)
}
/// Emit an instruction to the current basic block
pub(super) fn emit_instruction(&mut self, instruction: MirInstruction) -> Result<(), String> {
let block_id = self.current_block.ok_or("No current basic block")?;
// Make instruction mutable for potential receiver materialization
let mut instruction = instruction;
// Precompute debug metadata to avoid borrow conflicts later
let _dbg_fn_name = self
.scope_ctx
.current_function
.as_ref()
.map(|f| f.signature.name.clone());
let _dbg_region_id = self.debug_current_region_id();
// P0: PHI の軽量補強と観測は、関数ブロック取得前に実施して借用競合を避ける
if let MirInstruction::Phi { dst, inputs, .. } = &instruction {
origin::phi::propagate_phi_meta(self, *dst, inputs);
observe::ssa::emit_phi(self, *dst, inputs);
}
// CRITICAL: Final receiver materialization for MethodCall
// This ensures the receiver has an in-block definition in the same block as the Call.
// Must happen BEFORE function mutable borrow to avoid borrowck conflicts.
if let MirInstruction::Call {
callee: Some(callee),
dst,
args,
effects,
..
} = &instruction
{
use crate::mir::definitions::call_unified::Callee;
if let Callee::Method {
box_name,
method,
receiver: Some(r),
certainty,
box_kind,
} = callee.clone()
{
// LocalSSA: ensure receiver has a Copy in current_block
let r_local = crate::mir::builder::ssa::local::recv(self, r);
// Update instruction with materialized receiver
let new_callee = Callee::Method {
box_name: box_name.clone(),
method: method.clone(),
receiver: Some(r_local),
certainty,
box_kind,
};
instruction = MirInstruction::Call {
dst: *dst,
func: crate::mir::ValueId::INVALID, // Legacy dummy (not a real SSA use)
callee: Some(new_callee),
args: args.clone(),
effects: *effects,
};
}
}
if let Some(ref mut function) = self.scope_ctx.current_function {
// Pre-capture branch/jump targets for predecessor update after we finish
// mutably borrowing the current block.
let (then_t, else_t, jump_t) = match &instruction {
MirInstruction::Branch {
then_bb, else_bb, ..
} => (Some(*then_bb), Some(*else_bb), None),
MirInstruction::Jump { target } => (None, None, Some(*target)),
_ => (None, None, None),
};
// Extract function name before mutable borrow to avoid borrowck error
let current_fn_name = function.signature.name.clone();
if let Some(block) = function.get_block_mut(block_id) {
// CRITICAL: Copy専用トレースLocalSSA調査用
if let MirInstruction::Copy { dst, src } = &instruction {
if std::env::var("NYASH_LOCAL_SSA_TRACE").ok().as_deref() == Some("1") {
eprintln!(
"[emit-inst] fn={} bb={:?} COPY %{} <- %{}",
current_fn_name,
self.current_block.map(|b| b.0).unwrap_or(0),
dst.0,
src.0
);
}
}
// Invariant: Call must always carry a Callee (unified path).
if let MirInstruction::Call { callee, .. } = &instruction {
if callee.is_none() {
return Err("builder invariant violated: MirInstruction::Call.callee must be Some (unified call)".into());
} else if std::env::var("NYASH_LOCAL_SSA_TRACE").ok().as_deref() == Some("1") {
use crate::mir::definitions::call_unified::Callee;
if let Some(Callee::Method {
box_name,
method,
receiver: Some(r),
..
}) = callee
{
eprintln!(
"[emit-inst] fn={} bb={:?} Call {}.{} recv=%{}",
current_fn_name,
self.current_block.map(|b| b.0).unwrap_or(0),
box_name,
method,
r.0
);
}
} else if std::env::var("NYASH_BUILDER_TRACE_RECV").ok().as_deref() == Some("1")
{
use crate::mir::definitions::call_unified::Callee;
if let Some(Callee::Method {
box_name,
method,
receiver: Some(r),
..
}) = callee
{
let names: Vec<String> = self
.variable_ctx
.variable_map
.iter()
.filter(|(_, &vid)| vid == *r)
.map(|(k, _)| k.clone())
.collect();
eprintln!(
"[builder/recv-trace] fn={} bb={:?} method={}.{} recv=%{} aliases={:?}",
current_fn_name,
self.current_block,
box_name,
method,
r.0,
names
);
}
}
}
if utils::builder_debug_enabled() {
eprintln!(
"[BUILDER] emit @bb{} -> {}",
block_id,
match &instruction {
MirInstruction::TypeOp { dst, op, value, ty } =>
format!("typeop {:?} {} {:?} -> {}", op, value, ty, dst),
MirInstruction::Print { value, .. } => format!("print {}", value),
MirInstruction::BoxCall {
box_val,
method,
method_id,
args,
dst,
..
} => {
if let Some(mid) = method_id {
format!(
"boxcall {}.{}[#{}]({:?}) -> {:?}",
box_val, method, mid, args, dst
)
} else {
format!(
"boxcall {}.{}({:?}) -> {:?}",
box_val, method, args, dst
)
}
}
MirInstruction::Call {
func, args, dst, ..
} => format!("call {}({:?}) -> {:?}", func, args, dst),
MirInstruction::NewBox {
dst,
box_type,
args,
} => format!("new {}({:?}) -> {}", box_type, args, dst),
MirInstruction::Const { dst, value } =>
format!("const {:?} -> {}", value, dst),
MirInstruction::Branch {
condition,
then_bb,
else_bb,
} => format!("br {}, {}, {}", condition, then_bb, else_bb),
MirInstruction::Jump { target } => format!("br {}", target),
_ => format!("{:?}", instruction),
}
);
}
// Phase 136 Step 6/7: Use metadata_ctx for span
block.add_instruction_with_span(
instruction.clone(),
self.metadata_ctx.current_span(),
);
// Drop the mutable borrow of `block` before updating other blocks
}
// Update predecessor sets for branch/jump immediately so that
// debug_verify_phi_inputs can observe a consistent CFG without
// requiring a full function.update_cfg() pass.
if let Some(t) = then_t {
if let Some(succ) = function.get_block_mut(t) {
succ.add_predecessor(block_id);
}
}
if let Some(t) = else_t {
if let Some(succ) = function.get_block_mut(t) {
succ.add_predecessor(block_id);
}
}
if let Some(t) = jump_t {
if let Some(succ) = function.get_block_mut(t) {
succ.add_predecessor(block_id);
}
}
Ok(())
} else {
Err(format!("Basic block {} does not exist", block_id))
}
}
/// Update an existing PHI instruction's inputs (for loop sealing)
/// Used by LoopFormBuilder to complete incomplete PHI nodes
#[allow(dead_code)]
pub(super) fn update_phi_instruction(
&mut self,
block: BasicBlockId,
phi_id: ValueId,
new_inputs: Vec<(BasicBlockId, ValueId)>,
) -> Result<(), String> {
if let Some(ref mut function) = self.scope_ctx.current_function {
if let Some(block_data) = function.get_block_mut(block) {
// Find PHI instruction with matching dst
for inst in &mut block_data.instructions {
if let MirInstruction::Phi { dst, inputs, .. } = inst {
if *dst == phi_id {
*inputs = new_inputs;
return Ok(());
}
}
}
Err(format!(
"PHI instruction {} not found in block {}",
phi_id, block
))
} else {
Err(format!("Block {} not found", block))
}
} else {
Err("No current function".to_string())
}
}
// フェーズM: is_no_phi_mode()メソッド削除
// フェーズM: insert_edge_copy()メソッド削除no_phi_mode撤廃により不要
/// Build new expression: new ClassName(arguments)
pub(super) fn build_new_expression(
&mut self,
class: String,
arguments: Vec<ASTNode>,
) -> Result<ValueId, String> {
// Phase 9.78a: Unified Box creation using NewBox instruction
// Core-13 pure mode: emit ExternCall(env.box.new) with type name const only
if crate::config::env::mir_core13_pure() {
// Emit Const String for type nameConstantEmissionBox
let ty_id = crate::mir::builder::emission::constant::emit_string(self, class.clone());
// Evaluate arguments (pass through to env.box.new shim)
let mut arg_vals: Vec<ValueId> = Vec::with_capacity(arguments.len());
for a in arguments {
arg_vals.push(self.build_expression(a)?);
}
// Build arg list: [type, a1, a2, ...]
let mut args: Vec<ValueId> = Vec::with_capacity(1 + arg_vals.len());
args.push(ty_id);
args.extend(arg_vals);
// Call env.box.new
// 📦 Hotfix 3: Use next_value_id() to respect function parameter reservation
let dst = self.next_value_id();
self.emit_instruction(MirInstruction::ExternCall {
dst: Some(dst),
iface_name: "env.box".to_string(),
method_name: "new".to_string(),
args,
effects: EffectMask::PURE,
})?;
// 型注釈(最小)
self.type_ctx
.value_types
.insert(dst, super::MirType::Box(class.clone()));
return Ok(dst);
}
// Optimization: Primitive wrappers → emit Const directly when possible
if class == "IntegerBox" && arguments.len() == 1 {
if let ASTNode::Literal {
value: LiteralValue::Integer(n),
..
} = arguments[0].clone()
{
// 📦 Hotfix 3: Use next_value_id() to respect function parameter reservation
let dst = self.next_value_id();
self.emit_instruction(MirInstruction::Const {
dst,
value: ConstValue::Integer(n),
})?;
self.type_ctx
.value_types
.insert(dst, super::MirType::Integer);
return Ok(dst);
}
}
// First, evaluate all arguments to get their ValueIds
let mut arg_values = Vec::new();
for arg in arguments {
let arg_value = self.build_expression(arg)?;
arg_values.push(arg_value);
}
// Generate the destination ValueId
// 📦 Hotfix 3: Use next_value_id() to respect function parameter reservation
let dst = self.next_value_id();
// Emit NewBox instruction for all Box types
// VM will handle optimization for basic types internally
self.emit_instruction(MirInstruction::NewBox {
dst,
box_type: class.clone(),
args: arg_values.clone(),
})?;
// Phase 15.5: Unified box type handling
// All boxes (including former core boxes) are treated uniformly as Box types
self.type_ctx
.value_types
.insert(dst, super::MirType::Box(class.clone()));
// Record origin for optimization: dst was created by NewBox of class
self.type_ctx.value_origin_newbox.insert(dst, class.clone());
// birth 呼び出しBuilder 正規化)
// 優先: 低下済みグローバル関数 `<Class>.birth/Arity`Arity は me を含まない)
// 代替: 既存互換として BoxCall("birth")(プラグイン/ビルトインの初期化に対応)
if class != "StringBox" {
let arity = arg_values.len();
let lowered =
crate::mir::builder::calls::function_lowering::generate_method_function_name(
&class, "birth", arity,
);
let use_lowered = if let Some(ref module) = self.current_module {
module.functions.contains_key(&lowered)
} else {
false
};
if use_lowered {
// Call Global("Class.birth/Arity") with argv = [me, args...]
let mut argv: Vec<ValueId> = Vec::with_capacity(1 + arity);
argv.push(dst);
argv.extend(arg_values.iter().copied());
self.emit_legacy_call(None, CallTarget::Global(lowered), argv)?;
} else {
// Fallback policy:
// - For user-defined boxes (no explicit constructor), do NOT emit BoxCall("birth").
// VM will treat plain NewBox as constructed; dev verify warns if needed.
// - For builtins/plugins, keep BoxCall("birth") fallback to preserve legacy init.
let is_user_box = self.comp_ctx.user_defined_boxes.contains(&class);
// Dev safety: allow disabling birth() injection for builtins to avoid
// unified-call method dispatch issues while migrating. Off by default unless explicitly enabled.
let allow_builtin_birth = std::env::var("NYASH_DEV_BIRTH_INJECT_BUILTINS")
.ok()
.as_deref()
== Some("1");
if !is_user_box && allow_builtin_birth {
let birt_mid = resolve_slot_by_type_name(&class, "birth");
self.emit_box_or_plugin_call(
None,
dst,
"birth".to_string(),
birt_mid,
arg_values,
EffectMask::READ.add(Effect::ReadHeap),
)?;
}
}
}
Ok(dst)
}
/// Check if the current basic block is terminated
fn is_current_block_terminated(&self) -> bool {
if let (Some(block_id), Some(ref function)) =
(self.current_block, &self.scope_ctx.current_function)
{
if let Some(block) = function.get_block(block_id) {
return block.is_terminated();
}
}
false
}
// ============================================================================
// Phase 26-A: ValueId型安全化メソッド
// ============================================================================
/// 型付きValueIdを発行新API
/// Phase 136 P0: Use SSOT allocator (next_value_id) to respect function context
pub fn new_typed_value(&mut self, kind: super::MirValueKind) -> super::TypedValueId {
let id = self.next_value_id();
self.type_ctx.value_kinds.insert(id, kind);
super::TypedValueId::new(id, kind)
}
/// 既存ValueIdの型情報を取得
pub fn get_value_kind(&self, id: ValueId) -> Option<super::MirValueKind> {
self.type_ctx.value_kinds.get(&id).copied()
}
/// 既存ValueIdに型情報を後付けレガシー互換用
pub fn register_value_kind(&mut self, id: ValueId, kind: super::MirValueKind) {
self.type_ctx.value_kinds.insert(id, kind);
}
/// 型安全なパラメータ判定ValueIdベース - GUARD Bug Prevention
pub fn is_value_parameter(&self, id: ValueId) -> bool {
self.get_value_kind(id)
.map(|kind| kind.is_parameter())
.unwrap_or(false)
}
/// 型安全なローカル変数判定ValueIdベース
pub fn is_value_local(&self, id: ValueId) -> bool {
self.get_value_kind(id)
.map(|kind| kind.is_local())
.unwrap_or(false)
}
/// 型安全なLoopCarrier判定ValueIdベース
pub fn is_value_loop_carrier(&self, id: ValueId) -> bool {
self.get_value_kind(id)
.map(|kind| kind.is_loop_carrier())
.unwrap_or(false)
}
}
impl Default for MirBuilder {
fn default() -> Self {
Self::new()
}
}
// Phase 79: BindingMapProvider implementation
// Centralizes feature-gated binding_map access for promoters
use crate::mir::loop_pattern_detection::BindingMapProvider;
impl BindingMapProvider for MirBuilder {
#[cfg(feature = "normalized_dev")]
fn get_binding_map(
&self,
) -> Option<&std::collections::BTreeMap<String, crate::mir::BindingId>> {
// Phase 136 Step 4/7: Use binding_ctx (SSOT)
Some(self.binding_ctx.binding_map())
}
#[cfg(not(feature = "normalized_dev"))]
fn get_binding_map(
&self,
) -> Option<&std::collections::BTreeMap<String, crate::mir::BindingId>> {
None
}
}
#[cfg(test)]
mod binding_id_tests {
use super::*;
#[test]
fn test_binding_map_initialization() {
let builder = MirBuilder::new();
assert_eq!(builder.core_ctx.next_binding_id, 0);
// Phase 2-6: binding_ctx is now SSOT (legacy field removed)
assert!(builder.binding_ctx.is_empty());
}
#[test]
fn test_binding_allocation_sequential() {
let mut builder = MirBuilder::new();
let bid0 = builder.allocate_binding_id();
let bid1 = builder.allocate_binding_id();
let bid2 = builder.allocate_binding_id();
assert_eq!(bid0.raw(), 0);
assert_eq!(bid1.raw(), 1);
assert_eq!(bid2.raw(), 2);
assert_eq!(builder.core_ctx.next_binding_id, 3);
}
#[test]
fn test_shadowing_binding_restore() {
let mut builder = MirBuilder::new();
// Simulate function entry scope
builder.push_lexical_scope();
// Declare outer x
// Phase 136 P0: Use SSOT allocator for function scope simulation
let outer_vid = builder.next_value_id();
builder
.declare_local_in_current_scope("x", outer_vid)
.unwrap();
// Phase 2-6: Check binding_ctx (SSOT)
let outer_bid = builder.binding_ctx.lookup("x").unwrap();
assert_eq!(outer_bid.raw(), 0);
// Enter inner scope and shadow x
builder.push_lexical_scope();
// Phase 136 P0: Use SSOT allocator for function scope simulation
let inner_vid = builder.next_value_id();
builder
.declare_local_in_current_scope("x", inner_vid)
.unwrap();
// Phase 2-6: Check binding_ctx (SSOT)
let inner_bid = builder.binding_ctx.lookup("x").unwrap();
assert_eq!(inner_bid.raw(), 1);
// Exit inner scope - should restore outer binding
builder.pop_lexical_scope();
// Phase 2-6: Check binding_ctx (SSOT)
let restored_bid = builder.binding_ctx.lookup("x").unwrap();
assert_eq!(restored_bid, outer_bid);
assert_eq!(restored_bid.raw(), 0);
// Cleanup
builder.pop_lexical_scope();
}
#[test]
fn test_valueid_binding_parallel_allocation() {
let mut builder = MirBuilder::new();
// Phase 136 P0: Use SSOT allocator (next_value_id)
// Note: Without current_function, next_value_id() falls back to value_gen.next()
// so this test still validates ValueId/BindingId independence
// Allocate ValueIds and BindingIds in parallel
let vid0 = builder.next_value_id();
let bid0 = builder.allocate_binding_id();
let vid1 = builder.next_value_id();
let bid1 = builder.allocate_binding_id();
// ValueId and BindingId should be independent
assert_eq!(vid0.0, 0);
assert_eq!(bid0.raw(), 0);
assert_eq!(vid1.0, 1);
assert_eq!(bid1.raw(), 1);
// Allocating more ValueIds should not affect BindingId counter
let _ = builder.next_value_id();
let _ = builder.next_value_id();
let bid2 = builder.allocate_binding_id();
assert_eq!(bid2.raw(), 2); // Still sequential
// Allocating more BindingIds should not affect ValueId counter
let _ = builder.allocate_binding_id();
let _ = builder.allocate_binding_id();
let vid2 = builder.next_value_id();
assert_eq!(vid2.0, 4); // Continues from where we left off
}
}