Files
hakorune/src/mir/builder.rs

702 lines
30 KiB
Rust
Raw Normal View History

/*!
* 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};
use crate::mir::builder::builder_calls::CallTarget;
use std::collections::HashMap;
docs: Add field visibility analysis and MIR BoxCall documentation ## Field Visibility Analysis Results - Confirmed init{} fields are **public** in current Nyash implementation - No access modifier (private/public) system currently implemented - All fields accessible via me.fieldname syntax - Documented findings for future reference ## MIR Documentation Enhancements - Created comprehensive mir-dumper-guide.md for reading MIR dumps - Enhanced mir-26-specification.md with BoxCall vs regular Call examples - Added clear identification patterns: * BoxCall: `call %value.method(args)` (plugins/builtins) * Regular Call: `call %func(%me, args)` (user-defined boxes) ## VM Backend BoxRef Handling Improvements - Fixed BoxRef method dispatch using share_box() instead of clone_box() - Prevents unintended constructor calls during method resolution - Maintains proper instance identity throughout VM execution ## MIR Builder User-Defined Box Tracking - Added user_defined_boxes HashSet to track declared user boxes - Improved method lowering decisions for user-defined vs builtin boxes - Enhanced AST→MIR conversion accuracy for method calls ## Plugin Tester Lifecycle Enhancements - Added comprehensive FileBox lifecycle testing (open/write/close) - Enhanced cloneSelf() and copyFrom() testing with proper Handle parsing - Added TLV encoding helpers for strings and bytes - Improved error reporting and step-by-step validation 🤖 Generated with [Claude Code](https://claude.ai/code) Co-Authored-By: Claude <noreply@anthropic.com>
2025-08-21 01:18:25 +09:00
use std::collections::HashSet;
mod calls; // Call system modules (refactored from builder_calls)
mod builder_calls;
mod call_resolution; // ChatGPT5 Pro: Type-safe call resolution utilities
mod method_call_handlers; // Method call handler separation (Phase 3)
mod decls; // declarations lowering split
mod exprs; // expression lowering split
mod exprs_call; // call(expr)
feat: using構文完全実装&json_native大幅進化 ## 🎉 using構文の完全実装(ChatGPT作業) - ✅ **include → using移行完了**: 全ファイルでusing構文に統一 - `local X = include` → `using "path" as X` - 約70ファイルを一括変換 - ✅ **AST/パーサー/MIR完全対応**: using専用処理実装 - ASTNode::Using追加 - MIRビルダーでの解決処理 - include互換性も維持 ## 🚀 json_native実装進化(ChatGPT追加実装) - ✅ **浮動小数点対応追加**: is_float/parse_float実装 - ✅ **配列/オブジェクトパーサー実装**: parse_array/parse_object完成 - ✅ **エスケープ処理強化**: Unicode対応、全制御文字サポート - ✅ **StringUtils大幅拡張**: 文字列操作メソッド多数追加 - contains, index_of_string, split, join等 - 大文字小文字変換(全アルファベット対応) ## 💡 MIR SIMD & ハイブリッド戦略考察 - **MIR15 SIMD命令案**: SimdLoad/SimdScan等の新命令セット - **C ABIハイブリッド**: ホットパスのみC委託で10倍速化可能 - **並行処理でyyjson超え**: 100KB以上で2-10倍速の可能性 - **3層アーキテクチャ**: Nyash層/MIR層/C ABI層の美しい分離 ## 📊 技術的成果 - using構文により名前空間管理が明確化 - json_nativeが実用レベルに接近(完成度25%→40%) - 将来的にyyjsonの70%速度達成可能と判明 ChatGPT爆速実装×Claude深い考察の完璧な協働! 🤖 Generated with [Claude Code](https://claude.ai/code) Co-Authored-By: Claude <noreply@anthropic.com>
2025-09-25 00:41:56 +09:00
// include lowering removed (using is handled in runner)
mod exprs_lambda; // lambda lowering
mod exprs_peek; // peek expression
mod exprs_qmark; // ?-propagate
mod fields; // field access/assignment lowering split
pub(crate) mod loops;
mod ops;
mod phi;
mod if_form;
mod control_flow; // thin wrappers to centralize control-flow entrypoints
mod lifecycle; // prepare/lower_root/finalize split
// legacy large-match remains inline for now (planned extraction)
2025-09-11 04:20:28 +09:00
mod plugin_sigs; // plugin signature loader
mod stmts;
mod utils;
mod vars; // variables/scope helpers // small loop helpers (header/exit context)
mod origin; // P0: origin inferenceme/Knownと PHI 伝播(軽量)
mod observe; // P0: dev-only observability helpersssa/resolve
mod rewrite; // P1: Known rewrite & special consolidation
mod ssa; // LocalSSA helpers (in-block materialization)
mod schedule; // BlockScheduleBox物理順序: PHI→materialize→body
mod metadata; // MetadataPropagationBoxtype/originの伝播
mod emission; // emission::*Const/Compare/Branch の薄い発行箱)
mod types; // types::annotation / inference型注釈/推論の箱: 推論は後段)
mod router; // RouterPolicyBoxUnified vs BoxCall
mod emit_guard; // EmitGuardBoxemit直前の最終関所
mod name_const; // NameConstBox関数名Const生成
// 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,
/// Variable name to ValueId mapping (for SSA conversion)
pub(super) variable_map: HashMap<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
pub(super) value_origin_newbox: HashMap<ValueId, String>,
docs: Add field visibility analysis and MIR BoxCall documentation ## Field Visibility Analysis Results - Confirmed init{} fields are **public** in current Nyash implementation - No access modifier (private/public) system currently implemented - All fields accessible via me.fieldname syntax - Documented findings for future reference ## MIR Documentation Enhancements - Created comprehensive mir-dumper-guide.md for reading MIR dumps - Enhanced mir-26-specification.md with BoxCall vs regular Call examples - Added clear identification patterns: * BoxCall: `call %value.method(args)` (plugins/builtins) * Regular Call: `call %func(%me, args)` (user-defined boxes) ## VM Backend BoxRef Handling Improvements - Fixed BoxRef method dispatch using share_box() instead of clone_box() - Prevents unintended constructor calls during method resolution - Maintains proper instance identity throughout VM execution ## MIR Builder User-Defined Box Tracking - Added user_defined_boxes HashSet to track declared user boxes - Improved method lowering decisions for user-defined vs builtin boxes - Enhanced AST→MIR conversion accuracy for method calls ## Plugin Tester Lifecycle Enhancements - Added comprehensive FileBox lifecycle testing (open/write/close) - Enhanced cloneSelf() and copyFrom() testing with proper Handle parsing - Added TLV encoding helpers for strings and bytes - Improved error reporting and step-by-step validation 🤖 Generated with [Claude Code](https://claude.ai/code) Co-Authored-By: Claude <noreply@anthropic.com>
2025-08-21 01:18:25 +09:00
/// 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
pub(super) value_types: HashMap<ValueId, super::MirType>,
/// 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)>>,
/// 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,
feat: using構文完全実装&json_native大幅進化 ## 🎉 using構文の完全実装(ChatGPT作業) - ✅ **include → using移行完了**: 全ファイルでusing構文に統一 - `local X = include` → `using "path" as X` - 約70ファイルを一括変換 - ✅ **AST/パーサー/MIR完全対応**: using専用処理実装 - ASTNode::Using追加 - MIRビルダーでの解決処理 - include互換性も維持 ## 🚀 json_native実装進化(ChatGPT追加実装) - ✅ **浮動小数点対応追加**: is_float/parse_float実装 - ✅ **配列/オブジェクトパーサー実装**: parse_array/parse_object完成 - ✅ **エスケープ処理強化**: Unicode対応、全制御文字サポート - ✅ **StringUtils大幅拡張**: 文字列操作メソッド多数追加 - contains, index_of_string, split, join等 - 大文字小文字変換(全アルファベット対応) ## 💡 MIR SIMD & ハイブリッド戦略考察 - **MIR15 SIMD命令案**: SimdLoad/SimdScan等の新命令セット - **C ABIハイブリッド**: ホットパスのみC委託で10倍速化可能 - **並行処理でyyjson超え**: 100KB以上で2-10倍速の可能性 - **3層アーキテクチャ**: Nyash層/MIR層/C ABI層の美しい分離 ## 📊 技術的成果 - using構文により名前空間管理が明確化 - json_nativeが実用レベルに接近(完成度25%→40%) - 将来的にyyjsonの70%速度達成可能と判明 ChatGPT爆速実装×Claude深い考察の完璧な協働! 🤖 Generated with [Claude Code](https://claude.ai/code) Co-Authored-By: Claude <noreply@anthropic.com>
2025-09-25 00:41:56 +09:00
// include guards removed
2025-09-15 22:14:42 +09:00
/// 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 counters for region IDs (deterministic across a run).
debug_loop_counter: u32,
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>,
}
impl MirBuilder {
/// Create a new MIR builder
pub fn new() -> Self {
2025-09-11 04:20:28 +09:00
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(),
variable_map: HashMap::new(),
pending_phis: Vec::new(),
value_origin_newbox: HashMap::new(),
docs: Add field visibility analysis and MIR BoxCall documentation ## Field Visibility Analysis Results - Confirmed init{} fields are **public** in current Nyash implementation - No access modifier (private/public) system currently implemented - All fields accessible via me.fieldname syntax - Documented findings for future reference ## MIR Documentation Enhancements - Created comprehensive mir-dumper-guide.md for reading MIR dumps - Enhanced mir-26-specification.md with BoxCall vs regular Call examples - Added clear identification patterns: * BoxCall: `call %value.method(args)` (plugins/builtins) * Regular Call: `call %func(%me, args)` (user-defined boxes) ## VM Backend BoxRef Handling Improvements - Fixed BoxRef method dispatch using share_box() instead of clone_box() - Prevents unintended constructor calls during method resolution - Maintains proper instance identity throughout VM execution ## MIR Builder User-Defined Box Tracking - Added user_defined_boxes HashSet to track declared user boxes - Improved method lowering decisions for user-defined vs builtin boxes - Enhanced AST→MIR conversion accuracy for method calls ## Plugin Tester Lifecycle Enhancements - Added comprehensive FileBox lifecycle testing (open/write/close) - Enhanced cloneSelf() and copyFrom() testing with proper Handle parsing - Added TLV encoding helpers for strings and bytes - Improved error reporting and step-by-step validation 🤖 Generated with [Claude Code](https://claude.ai/code) Co-Authored-By: Claude <noreply@anthropic.com>
2025-08-21 01:18:25 +09:00
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: HashMap::new(),
plugin_method_sigs,
current_static_box: None,
static_method_index: std::collections::HashMap::new(),
method_tail_index: std::collections::HashMap::new(),
method_tail_index_source_len: 0,
feat: using構文完全実装&json_native大幅進化 ## 🎉 using構文の完全実装(ChatGPT作業) - ✅ **include → using移行完了**: 全ファイルでusing構文に統一 - `local X = include` → `using "path" as X` - 約70ファイルを一括変換 - ✅ **AST/パーサー/MIR完全対応**: using専用処理実装 - ASTNode::Using追加 - MIRビルダーでの解決処理 - include互換性も維持 ## 🚀 json_native実装進化(ChatGPT追加実装) - ✅ **浮動小数点対応追加**: is_float/parse_float実装 - ✅ **配列/オブジェクトパーサー実装**: parse_array/parse_object完成 - ✅ **エスケープ処理強化**: Unicode対応、全制御文字サポート - ✅ **StringUtils大幅拡張**: 文字列操作メソッド多数追加 - contains, index_of_string, split, join等 - 大文字小文字変換(全アルファベット対応) ## 💡 MIR SIMD & ハイブリッド戦略考察 - **MIR15 SIMD命令案**: SimdLoad/SimdScan等の新命令セット - **C ABIハイブリッド**: ホットパスのみC委託で10倍速化可能 - **並行処理でyyjson超え**: 100KB以上で2-10倍速の可能性 - **3層アーキテクチャ**: Nyash層/MIR層/C ABI層の美しい分離 ## 📊 技術的成果 - using構文により名前空間管理が明確化 - json_nativeが実用レベルに接近(完成度25%→40%) - 将来的にyyjsonの70%速度達成可能と判明 ChatGPT爆速実装×Claude深い考察の完璧な協働! 🤖 Generated with [Claude Code](https://claude.ai/code) Co-Authored-By: Claude <noreply@anthropic.com>
2025-09-25 00:41:56 +09:00
2025-09-15 22:14:42 +09:00
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_loop_counter: 0,
debug_join_counter: 0,
local_ssa_map: HashMap::new(),
schedule_mat_map: HashMap::new(),
}
}
/// 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_loop_header(&mut self) { self.hint_sink.loop_header(); }
#[inline]
pub(crate) fn hint_loop_latch(&mut self) { self.hint_sink.loop_latch(); }
#[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()); }
#[inline]
pub(crate) fn hint_loop_carrier<S: Into<String>>(&mut self, vars: impl IntoIterator<Item = S>) {
self.hint_sink.loop_carrier(vars.into_iter().map(|s| s.into()).collect::<Vec<_>>());
}
// ----------------------
// Debug scope helpers (region_id for DebugHub events)
// ----------------------
#[inline]
pub(crate) fn debug_next_loop_id(&mut self) -> u32 {
let id = self.debug_loop_counter;
self.debug_loop_counter = self.debug_loop_counter.saturating_add(1);
id
}
#[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_replace_region<S: Into<String>>(&mut self, region: S) {
if let Some(top) = self.debug_scope_stack.last_mut() {
*top = region.into();
} else {
self.debug_scope_stack.push(region.into());
}
}
#[inline]
pub(crate) fn debug_current_region_id(&self) -> Option<String> {
self.debug_scope_stack.last().cloned()
}
// ----------------------
// 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> {
2025-09-17 07:29:28 +09:00
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
self.build_expression_impl(ast)
}
/// 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> {
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() {
msg.push_str("\nHint: 'local' is a Stage-3 keyword. Enable NYASH_PARSER_STAGE3=1 (and 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() {
msg.push_str(&format!("\nHint: '{}' is a Stage-3 keyword. Enable NYASH_PARSER_STAGE3=1 (and 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 raw_value_id = self.build_expression(value)?;
// Correctness-first: assignment results may be used across control-flow joins.
// Pin to a slot so the value has a block-local def and participates in PHI merges.
let value_id = self
.pin_to_slot(raw_value_id, "@assign")
.unwrap_or(raw_value_id);
// 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")?;
// 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);
}
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) {
// 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_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(instruction);
// 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))
}
}
// 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
let dst = self.value_gen.next();
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()
{
let dst = self.value_gen.next();
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
let dst = self.value_gen.next();
// 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
}
}
impl Default for MirBuilder {
fn default() -> Self {
Self::new()
}
}