- Fix: Call with Callee::Method now includes receiver in used_values() - Prevents DCE from eliminating Copy instructions that define receivers - Pattern 3 (loop_if_phi.hako) now works correctly (sum=9) - Add: NYASH_DCE_TRACE=1 for debugging eliminated instructions - Shows which pure instructions DCE removes and from which block - Cleanup: Consolidate Call used_values to single source of truth - Early return in methods.rs handles all Call variants - Removed duplicate match arm (now unreachable!()) - ChatGPT's suggestion for cleaner architecture - Docs: Phase 166 analysis of inst_meta layer architecture - Identified CSE pass callee bug (to be fixed next) - Improvement proposals for CallLikeInst 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com>
1030 lines
44 KiB
Rust
1030 lines
44 KiB
Rust
/*!
|
||
* 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 builder_calls;
|
||
mod call_resolution; // ChatGPT5 Pro: Type-safe call resolution utilities
|
||
mod calls; // Call system modules (refactored from builder_calls)
|
||
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 decls; // declarations lowering split
|
||
mod exprs; // expression lowering split
|
||
mod exprs_call;
|
||
mod method_call_handlers; // 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
|
||
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 lifecycle;
|
||
mod loop_frontend_binding; // Phase 50: Loop Frontend Binding (JoinIR variable mapping)
|
||
mod joinir_id_remapper; // Phase 189: JoinIR ID remapping (ValueId/BlockId translation)
|
||
mod joinir_inline_boundary_injector; // Phase 189: JoinInlineBoundary Copy instruction injector
|
||
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; // EmitGuardBox(emit直前の最終関所)
|
||
mod metadata; // MetadataPropagationBox(type/originの伝播)
|
||
mod name_const; // NameConstBox(関数名Const生成)
|
||
mod observe; // P0: dev-only observability helpers(ssa/resolve)
|
||
mod origin; // P0: origin inference(me/Known)と PHI 伝播(軽量)
|
||
mod plugin_sigs; // plugin signature loader
|
||
mod receiver; // ReceiverMaterializationBox(Method recv の pin+LocalSSA 集約)
|
||
mod rewrite; // P1: Known rewrite & special consolidation
|
||
mod router; // RouterPolicyBox(Unified vs BoxCall)
|
||
mod schedule; // BlockScheduleBox(物理順序: PHI→materialize→body)
|
||
mod ssa; // LocalSSA helpers (in-block materialization)
|
||
mod stmts;
|
||
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 function being built
|
||
pub(super) current_function: Option<MirFunction>,
|
||
|
||
/// Current basic block being built
|
||
pub(super) current_block: Option<BasicBlockId>,
|
||
|
||
/// Value ID generator
|
||
pub(super) value_gen: ValueIdGenerator,
|
||
|
||
/// Basic block ID generator
|
||
pub(super) block_gen: BasicBlockIdGenerator,
|
||
|
||
/// 箱理論: Static boxコンパイル時のコンテキスト分離
|
||
/// Some(ctx)の場合、variable_map/value_origin_newbox/value_typesはctxから取得
|
||
/// Noneの場合、従来のフィールドを使用(後方互換性)
|
||
pub(super) compilation_context: Option<context::BoxCompilationContext>,
|
||
|
||
/// Variable name to ValueId mapping (for SSA conversion)
|
||
/// 注意: compilation_contextがSomeの場合は使用されません
|
||
/// Phase 25.1: HashMap → BTreeMap(PHI生成の決定性確保)
|
||
pub(super) variable_map: BTreeMap<String, ValueId>,
|
||
|
||
/// Pending phi functions to be inserted
|
||
#[allow(dead_code)]
|
||
pub(super) pending_phis: Vec<(BasicBlockId, ValueId, String)>,
|
||
|
||
/// Origin tracking for simple optimizations (e.g., object.method after new)
|
||
/// Maps a ValueId to the class name if it was produced by NewBox of that class
|
||
/// 注意: compilation_contextがSomeの場合は使用されません
|
||
// Phase 25.1: HashMap → BTreeMap(決定性確保)
|
||
pub(super) value_origin_newbox: BTreeMap<ValueId, String>,
|
||
|
||
/// Names of user-defined boxes declared in the current module
|
||
pub(super) user_defined_boxes: HashSet<String>,
|
||
|
||
/// Weak field registry: BoxName -> {weak field names}
|
||
pub(super) weak_fields_by_box: HashMap<String, HashSet<String>>,
|
||
|
||
/// Unified members: BoxName -> {propName -> Kind}
|
||
pub(super) property_getters_by_box: HashMap<String, HashMap<String, PropertyKind>>,
|
||
|
||
/// Remember class of object fields after assignments: (base_id, field) -> class_name
|
||
pub(super) field_origin_class: HashMap<(ValueId, String), String>,
|
||
/// Class-level field origin (cross-function heuristic): (BaseBoxName, field) -> FieldBoxName
|
||
pub(super) field_origin_by_box: HashMap<(String, String), String>,
|
||
|
||
/// Optional per-value type annotations (MIR-level): ValueId -> MirType
|
||
/// 注意: compilation_contextがSomeの場合は使用されません
|
||
// Phase 25.1: HashMap → BTreeMap(決定性確保)
|
||
pub(super) value_types: BTreeMap<ValueId, super::MirType>,
|
||
|
||
/// Phase 26-A: ValueId型情報マップ(型安全性強化)
|
||
/// ValueId -> MirValueKind のマッピング
|
||
/// - GUARDバグ予防: ValueId(0)がParameterかLocalか区別可能
|
||
/// - デフォルト: 未登録のValueIdはTemporary扱い
|
||
pub(super) value_kinds: HashMap<ValueId, super::MirValueKind>,
|
||
|
||
/// 関数スコープの SlotRegistry(観測専用)
|
||
/// - current_function と同じライフサイクルを持つよ。
|
||
/// - 既存の variable_map/SSA には影響しない(メタデータのみ)。
|
||
pub(super) current_slot_registry: Option<FunctionSlotRegistry>,
|
||
|
||
/// 🎯 箱理論: 型情報管理の一元化(TypeRegistryBox)
|
||
/// NYASH_USE_TYPE_REGISTRY=1 で有効化(段階的移行用)
|
||
pub(super) type_registry: type_registry::TypeRegistry,
|
||
|
||
/// Plugin method return type signatures loaded from nyash_box.toml
|
||
plugin_method_sigs: HashMap<(String, String), super::MirType>,
|
||
/// Current static box name when lowering a static box body (e.g., "Main")
|
||
current_static_box: Option<String>,
|
||
/// Index of static methods seen during lowering: name -> [(BoxName, arity)]
|
||
pub(super) static_method_index: std::collections::HashMap<String, Vec<(String, usize)>>,
|
||
|
||
/// Function parameter names (for LoopForm PHI construction)
|
||
/// Tracks the original parameter names at function entry
|
||
pub(super) function_param_names: HashSet<String>,
|
||
|
||
/// Fast lookup: method+arity tail → candidate function names (e.g., ".str/0" → ["JsonNode.str/0", ...])
|
||
pub(super) method_tail_index: std::collections::HashMap<String, Vec<String>>,
|
||
/// Source size snapshot to detect when to rebuild the tail index
|
||
pub(super) method_tail_index_source_len: usize,
|
||
|
||
/// Region 観測用のスタックだよ(FunctionRegion がルート)。
|
||
/// - NYASH_REGION_TRACE=1 のときだけ使われる開発用メタデータだよ。
|
||
pub(super) current_region_stack: Vec<RegionId>,
|
||
|
||
// include guards removed
|
||
/// Loop context stacks for lowering break/continue inside nested control flow
|
||
/// Top of stack corresponds to the innermost active loop
|
||
pub(super) loop_header_stack: Vec<BasicBlockId>,
|
||
pub(super) loop_exit_stack: Vec<BasicBlockId>,
|
||
|
||
/// If/merge context stack (innermost first). Used to make merge targets explicit
|
||
/// when lowering nested conditionals and to simplify jump generation.
|
||
pub(super) if_merge_stack: Vec<BasicBlockId>,
|
||
|
||
// 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,
|
||
|
||
/// Hint sink (zero-cost guidance; currently no-op)
|
||
pub(super) hint_sink: crate::mir::hints::HintSink,
|
||
|
||
/// Internal counter for temporary pin slots (block-crossing ephemeral values)
|
||
temp_slot_counter: u32,
|
||
/// 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)
|
||
// ----------------------
|
||
/// Stack of region identifiers like "loop#1/header" or "join#3/join".
|
||
debug_scope_stack: Vec<String>,
|
||
/// Monotonic counter for region IDs (deterministic across a run).
|
||
debug_join_counter: u32,
|
||
|
||
/// 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,
|
||
|
||
/// Current AST span being lowered (used to annotate MIR instructions)
|
||
pub(super) current_span: Span,
|
||
|
||
/// Optional source file hint for metadata/spans
|
||
pub(super) source_file: Option<String>,
|
||
|
||
/// 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();
|
||
// フェーズM: no_phi_mode初期化削除
|
||
Self {
|
||
current_module: None,
|
||
current_function: None,
|
||
current_block: None,
|
||
value_gen: ValueIdGenerator::new(),
|
||
block_gen: BasicBlockIdGenerator::new(),
|
||
compilation_context: None, // 箱理論: デフォルトは従来モード
|
||
variable_map: BTreeMap::new(), // Phase 25.1: 決定性確保
|
||
pending_phis: Vec::new(),
|
||
value_origin_newbox: BTreeMap::new(), // Phase 25.1: 決定性確保
|
||
user_defined_boxes: HashSet::new(),
|
||
weak_fields_by_box: HashMap::new(),
|
||
property_getters_by_box: HashMap::new(),
|
||
field_origin_class: HashMap::new(),
|
||
field_origin_by_box: HashMap::new(),
|
||
value_types: BTreeMap::new(), // Phase 25.1: 決定性確保
|
||
value_kinds: HashMap::new(), // Phase 26-A: ValueId型安全化
|
||
current_slot_registry: None,
|
||
type_registry: type_registry::TypeRegistry::new(),
|
||
plugin_method_sigs,
|
||
current_static_box: None,
|
||
static_method_index: std::collections::HashMap::new(),
|
||
function_param_names: HashSet::new(),
|
||
method_tail_index: std::collections::HashMap::new(),
|
||
method_tail_index_source_len: 0,
|
||
|
||
current_region_stack: Vec::new(),
|
||
|
||
loop_header_stack: Vec::new(),
|
||
loop_exit_stack: Vec::new(),
|
||
if_merge_stack: Vec::new(),
|
||
// 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,
|
||
hint_sink: crate::mir::hints::HintSink::new(),
|
||
temp_slot_counter: 0,
|
||
suppress_pin_entry_copy_next: false,
|
||
|
||
// Debug scope context
|
||
debug_scope_stack: Vec::new(),
|
||
debug_join_counter: 0,
|
||
|
||
local_ssa_map: HashMap::new(),
|
||
schedule_mat_map: HashMap::new(),
|
||
pin_slot_names: HashMap::new(),
|
||
|
||
in_unified_boxcall_fallback: false,
|
||
recursion_depth: 0,
|
||
current_span: Span::unknown(),
|
||
source_file: None,
|
||
root_is_app_mode: None,
|
||
static_box_singletons: HashMap::new(), // Phase 21.7: methodization support
|
||
}
|
||
}
|
||
|
||
/// Push/pop helpers for If merge context (best-effort; optional usage)
|
||
pub(super) fn push_if_merge(&mut self, bb: BasicBlockId) {
|
||
self.if_merge_stack.push(bb);
|
||
}
|
||
pub(super) fn pop_if_merge(&mut self) {
|
||
let _ = self.if_merge_stack.pop();
|
||
}
|
||
|
||
/// 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;
|
||
}
|
||
|
||
// ---- Hint helpers (no-op by default) ----
|
||
#[inline]
|
||
pub(crate) fn hint_scope_enter(&mut self, id: u32) {
|
||
self.hint_sink.scope_enter(id);
|
||
}
|
||
#[inline]
|
||
pub(crate) fn hint_scope_leave(&mut self, id: u32) {
|
||
self.hint_sink.scope_leave(id);
|
||
}
|
||
#[inline]
|
||
pub(crate) fn hint_join_result<S: Into<String>>(&mut self, var: S) {
|
||
self.hint_sink.join_result(var.into());
|
||
}
|
||
|
||
// ----------------------
|
||
// Debug scope helpers (region_id for DebugHub events)
|
||
// ----------------------
|
||
#[inline]
|
||
pub(crate) fn debug_next_join_id(&mut self) -> u32 {
|
||
let id = self.debug_join_counter;
|
||
self.debug_join_counter = self.debug_join_counter.saturating_add(1);
|
||
id
|
||
}
|
||
|
||
#[inline]
|
||
pub(crate) fn debug_push_region<S: Into<String>>(&mut self, region: S) {
|
||
self.debug_scope_stack.push(region.into());
|
||
}
|
||
|
||
#[inline]
|
||
pub(crate) fn debug_pop_region(&mut self) {
|
||
let _ = self.debug_scope_stack.pop();
|
||
}
|
||
|
||
#[inline]
|
||
pub(crate) fn debug_current_region_id(&self) -> Option<String> {
|
||
self.debug_scope_stack.last().cloned()
|
||
}
|
||
|
||
/// Hint for downstream metadata: set the logical source file name/path for the next build.
|
||
pub fn set_source_file_hint<S: Into<String>>(&mut self, source: S) {
|
||
self.source_file = Some(source.into());
|
||
}
|
||
|
||
/// Clear the source file hint (used when reusing the builder across modules).
|
||
pub fn clear_source_file_hint(&mut self) {
|
||
self.source_file = None;
|
||
}
|
||
|
||
/// Resolve current source file hint (builder field or env fallback).
|
||
fn current_source_file(&self) -> Option<String> {
|
||
self.source_file
|
||
.clone()
|
||
.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.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.method_tail_index
|
||
.entry(tail.to_string())
|
||
.or_insert_with(Vec::new)
|
||
.push(name.clone());
|
||
}
|
||
}
|
||
}
|
||
self.method_tail_index_source_len = module.functions.len();
|
||
} else {
|
||
self.method_tail_index_source_len = 0;
|
||
}
|
||
}
|
||
|
||
fn ensure_method_tail_index(&mut self) {
|
||
let need_rebuild = match self.current_module {
|
||
Some(ref refmod) => self.method_tail_index_source_len != refmod.functions.len(),
|
||
None => self.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.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.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.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_map.get(&name) {
|
||
Ok(value_id)
|
||
} else {
|
||
// 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].",
|
||
);
|
||
}
|
||
Err(msg)
|
||
}
|
||
}
|
||
|
||
/// Build assignment
|
||
pub(super) fn build_assignment(
|
||
&mut self,
|
||
var_name: String,
|
||
value: ASTNode,
|
||
) -> Result<ValueId, String> {
|
||
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_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
|
||
.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.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_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),
|
||
}
|
||
);
|
||
}
|
||
block.add_instruction_with_span(instruction.clone(), self.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
|
||
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.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 name(ConstantEmissionBox)
|
||
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.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.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.value_types
|
||
.insert(dst, super::MirType::Box(class.clone()));
|
||
|
||
// Record origin for optimization: dst was created by NewBox of class
|
||
self.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.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.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)
|
||
pub fn new_typed_value(&mut self, kind: super::MirValueKind) -> super::TypedValueId {
|
||
let id = self.value_gen.next();
|
||
self.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.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.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()
|
||
}
|
||
}
|