feat(joinir): Phase 92 P0-3 - ConditionalStep → JoinIR Select generation

Complete implementation of P5b escape sequence handling pattern lowering:

Core changes:
- skeleton_types.rs: Add `cond` field to ConditionalStep for escape condition
- escape_pattern_recognizer.rs: Extract and return escape_cond AST
- pattern_recognizer.rs: Pass escape_cond to ConditionalStep
- canonicalizer.rs: Store escape_cond in ConditionalStep variant

JoinIR lowering:
- carrier_update_emitter.rs: Add emit_conditional_step_update() function
  - Lower condition AST via lower_condition_to_joinir
  - Compute both then/else branches (carrier + delta)
  - Emit JoinInst::Select for conditional carrier update
- loop_with_break_minimal.rs: Add skeleton parameter, detect ConditionalStep
- pattern2_with_break.rs: Pass skeleton to lowering function

Backward compatibility:
- skeleton=None preserves existing Pattern1-4 behavior
- fixtures.rs, tests.rs updated for new signature

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

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
This commit is contained in:
nyash-codex
2025-12-16 15:36:36 +09:00
parent 51017a0f9a
commit a718af3213
9 changed files with 178 additions and 14 deletions

View File

@ -20,6 +20,9 @@ pub struct EscapeSkipPatternInfo {
pub quote_char: char, pub quote_char: char,
pub escape_char: char, pub escape_char: char,
pub body_stmts: Vec<ASTNode>, pub body_stmts: Vec<ASTNode>,
/// Phase 92 P0-3: The condition expression for conditional increment
/// e.g., `ch == '\\'` for escape sequence handling
pub escape_cond: Box<ASTNode>,
} }
/// Detect escape sequence handling pattern in loop body /// Detect escape sequence handling pattern in loop body
@ -57,8 +60,9 @@ pub fn detect_escape_skip_pattern(body: &[ASTNode]) -> Option<EscapeSkipPatternI
// Find escape check after break - scan for second "if" with increment // Find escape check after break - scan for second "if" with increment
let escape_idx = find_escape_in_if(body, break_idx)?; let escape_idx = find_escape_in_if(body, break_idx)?;
// Extract counter_name, escape_delta, and normal_delta from the escape if statement // Extract counter_name, escape_delta, normal_delta, and condition from the escape if statement
let (counter_name, escape_delta, normal_delta) = extract_delta_pair_from_if(body, escape_idx)?; // Phase 92 P0-3: Now also extracts condition expression
let (counter_name, escape_delta, normal_delta, escape_cond) = extract_delta_pair_from_if(body, escape_idx)?;
// Extract body statements before break check // Extract body statements before break check
let body_stmts = body[..break_idx].to_vec(); let body_stmts = body[..break_idx].to_vec();
@ -70,6 +74,7 @@ pub fn detect_escape_skip_pattern(body: &[ASTNode]) -> Option<EscapeSkipPatternI
quote_char: '"', // Default for JSON/CSV (Phase 91 MVP) quote_char: '"', // Default for JSON/CSV (Phase 91 MVP)
escape_char: '\\', // Default for JSON/CSV (Phase 91 MVP) escape_char: '\\', // Default for JSON/CSV (Phase 91 MVP)
body_stmts, body_stmts,
escape_cond, // Phase 92 P0-3: Condition for JoinIR Select
}) })
} }
@ -139,17 +144,20 @@ fn find_escape_in_if(body: &[ASTNode], after_idx: usize) -> Option<usize> {
None None
} }
/// Extract both escape_delta and normal_delta from if statement /// Extract escape_delta, normal_delta, and condition from if statement
/// ///
/// Handles both: /// Handles both:
/// - if ch == escape_char { i = i + 2 } else { i = i + 1 } /// - if ch == escape_char { i = i + 2 } else { i = i + 1 }
/// - if ch == escape_char { i = i + 2 } (followed by separate increment) /// - if ch == escape_char { i = i + 2 } (followed by separate increment)
fn extract_delta_pair_from_if(body: &[ASTNode], idx: usize) -> Option<(String, i64, i64)> { ///
/// Phase 92 P0-3: Now returns the condition expression for JoinIR Select generation
fn extract_delta_pair_from_if(body: &[ASTNode], idx: usize) -> Option<(String, i64, i64, Box<ASTNode>)> {
if idx >= body.len() { if idx >= body.len() {
return None; return None;
} }
if let ASTNode::If { if let ASTNode::If {
condition,
then_body, then_body,
else_body, else_body,
.. ..
@ -204,7 +212,8 @@ fn extract_delta_pair_from_if(body: &[ASTNode], idx: usize) -> Option<(String, i
found_delta? found_delta?
}; };
Some((counter_name, escape_delta, normal_delta)) // Phase 92 P0-3: Return condition along with deltas
Some((counter_name, escape_delta, normal_delta, condition.clone()))
} else { } else {
None None
} }

View File

@ -717,6 +717,7 @@ pub(crate) fn can_lower(builder: &MirBuilder, ctx: &super::router::LoopPatternCo
/// ///
/// Wrapper around cf_loop_pattern2_with_break to match router signature /// Wrapper around cf_loop_pattern2_with_break to match router signature
/// Phase 200-C: Pass fn_body to cf_loop_pattern2_with_break /// Phase 200-C: Pass fn_body to cf_loop_pattern2_with_break
/// Phase 92 P0-3: Pass skeleton for ConditionalStep support
pub(crate) fn lower( pub(crate) fn lower(
builder: &mut MirBuilder, builder: &mut MirBuilder,
ctx: &super::router::LoopPatternContext, ctx: &super::router::LoopPatternContext,
@ -727,6 +728,7 @@ pub(crate) fn lower(
ctx.func_name, ctx.func_name,
ctx.debug, ctx.debug,
ctx.fn_body, ctx.fn_body,
ctx.skeleton, // Phase 92 P0-3: Pass skeleton for ConditionalStep
) )
} }
@ -735,6 +737,7 @@ impl MirBuilder {
/// ///
/// **Refactored**: Now uses PatternPipelineContext for unified preprocessing /// **Refactored**: Now uses PatternPipelineContext for unified preprocessing
/// **Phase 200-C**: Added fn_body parameter for capture analysis /// **Phase 200-C**: Added fn_body parameter for capture analysis
/// **Phase 92 P0-3**: Added skeleton parameter for ConditionalStep support
/// ///
/// # Pipeline (Phase 179-B) /// # Pipeline (Phase 179-B)
/// 1. Build preprocessing context → PatternPipelineContext /// 1. Build preprocessing context → PatternPipelineContext
@ -754,10 +757,12 @@ impl MirBuilder {
debug: bool, debug: bool,
) -> Result<Option<ValueId>, String> { ) -> Result<Option<ValueId>, String> {
// Phase 200-C: Delegate to impl function with fn_body=None for backward compatibility // Phase 200-C: Delegate to impl function with fn_body=None for backward compatibility
self.cf_loop_pattern2_with_break_impl(condition, _body, _func_name, debug, None) // Phase 92 P0-3: skeleton=None for backward compatibility
self.cf_loop_pattern2_with_break_impl(condition, _body, _func_name, debug, None, None)
} }
/// Phase 200-C: Pattern 2 implementation with optional fn_body for capture analysis /// Phase 200-C: Pattern 2 implementation with optional fn_body for capture analysis
/// Phase 92 P0-3: Added skeleton parameter for ConditionalStep support
fn cf_loop_pattern2_with_break_impl( fn cf_loop_pattern2_with_break_impl(
&mut self, &mut self,
condition: &ASTNode, condition: &ASTNode,
@ -765,6 +770,7 @@ impl MirBuilder {
_func_name: &str, _func_name: &str,
debug: bool, debug: bool,
fn_body: Option<&[ASTNode]>, fn_body: Option<&[ASTNode]>,
skeleton: Option<&crate::mir::loop_canonicalizer::LoopSkeleton>,
) -> Result<Option<ValueId>, String> { ) -> Result<Option<ValueId>, String> {
use crate::mir::join_ir::lowering::loop_with_break_minimal::lower_loop_with_break_minimal; use crate::mir::join_ir::lowering::loop_with_break_minimal::lower_loop_with_break_minimal;
@ -869,6 +875,7 @@ impl MirBuilder {
analysis_body, // Phase 191/192: Pass normalized body AST for init lowering analysis_body, // Phase 191/192: Pass normalized body AST for init lowering
Some(&mut inputs.body_local_env), // Phase 191: Pass mutable body-local environment Some(&mut inputs.body_local_env), // Phase 191: Pass mutable body-local environment
&mut inputs.join_value_space, // Phase 201: Unified ValueId allocation (Local region) &mut inputs.join_value_space, // Phase 201: Unified ValueId allocation (Local region)
skeleton, // Phase 92 P0-3: Pass skeleton for ConditionalStep support
) { ) {
Ok((module, meta)) => (module, meta), Ok((module, meta)) => (module, meta),
Err(e) => { Err(e) => {

View File

@ -362,6 +362,101 @@ pub fn emit_carrier_update(
} }
} }
// ============================================================================
// Phase 92 P0-3: ConditionalStep Support
// ============================================================================
use crate::ast::ASTNode;
use crate::mir::join_ir::lowering::condition_lowerer::lower_condition_to_joinir;
use crate::mir::join_ir::VarId;
use crate::mir::MirType;
/// Emit JoinIR instructions for conditional step update (Phase 92 P0-3)
///
/// Handles the P5b escape sequence pattern where carrier update depends on a condition:
/// ```text
/// if escape_cond { carrier = carrier + then_delta }
/// else { carrier = carrier + else_delta }
/// ```
///
/// This generates:
/// 1. Lower condition expression to get cond_id
/// 2. Compute then_result = carrier + then_delta
/// 3. Compute else_result = carrier + else_delta
/// 4. JoinInst::Select { dst: carrier_new, cond: cond_id, then_val: then_result, else_val: else_result }
///
/// # Arguments
///
/// * `carrier` - Carrier variable information (name, ValueId)
/// * `cond_ast` - AST node for the condition expression (e.g., `ch == '\\'`)
/// * `then_delta` - Delta to add when condition is true
/// * `else_delta` - Delta to add when condition is false
/// * `alloc_value` - ValueId allocator closure
/// * `env` - ConditionEnv for variable resolution
/// * `instructions` - Output vector to append instructions to
///
/// # Returns
///
/// ValueId of the computed update result (the dst of Select)
pub fn emit_conditional_step_update(
carrier: &CarrierVar,
cond_ast: &ASTNode,
then_delta: i64,
else_delta: i64,
alloc_value: &mut dyn FnMut() -> ValueId,
env: &ConditionEnv,
instructions: &mut Vec<JoinInst>,
) -> Result<ValueId, String> {
// Step 1: Lower the condition expression
let (cond_id, cond_insts) = lower_condition_to_joinir(cond_ast, alloc_value, env)?;
instructions.extend(cond_insts);
// Step 2: Get carrier parameter ValueId from env
let carrier_param = env
.get(&carrier.name)
.ok_or_else(|| format!("Carrier '{}' not found in ConditionEnv", carrier.name))?;
// Step 3: Compute then_result = carrier + then_delta
let then_const_id = alloc_value();
instructions.push(JoinInst::Compute(MirLikeInst::Const {
dst: then_const_id,
value: ConstValue::Integer(then_delta),
}));
let then_result = alloc_value();
instructions.push(JoinInst::Compute(MirLikeInst::BinOp {
dst: then_result,
op: BinOpKind::Add,
lhs: carrier_param,
rhs: then_const_id,
}));
// Step 4: Compute else_result = carrier + else_delta
let else_const_id = alloc_value();
instructions.push(JoinInst::Compute(MirLikeInst::Const {
dst: else_const_id,
value: ConstValue::Integer(else_delta),
}));
let else_result = alloc_value();
instructions.push(JoinInst::Compute(MirLikeInst::BinOp {
dst: else_result,
op: BinOpKind::Add,
lhs: carrier_param,
rhs: else_const_id,
}));
// Step 5: Emit Select instruction
let carrier_new: VarId = alloc_value();
instructions.push(JoinInst::Select {
dst: carrier_new,
cond: cond_id,
then_val: then_result,
else_val: else_result,
type_hint: Some(MirType::Integer), // Carrier is always Integer
});
Ok(carrier_new)
}
#[cfg(test)] #[cfg(test)]
mod tests { mod tests {
use super::*; use super::*;

View File

@ -63,9 +63,10 @@ mod tests;
use crate::mir::join_ir::lowering::carrier_info::{CarrierInfo, JoinFragmentMeta}; use crate::mir::join_ir::lowering::carrier_info::{CarrierInfo, JoinFragmentMeta};
use crate::mir::join_ir::lowering::carrier_update_emitter::{ use crate::mir::join_ir::lowering::carrier_update_emitter::{
emit_carrier_update, emit_carrier_update_with_env, emit_carrier_update, emit_carrier_update_with_env, emit_conditional_step_update,
}; };
use crate::mir::join_ir::lowering::condition_to_joinir::ConditionEnv; use crate::mir::join_ir::lowering::condition_to_joinir::ConditionEnv;
use crate::mir::loop_canonicalizer::UpdateKind;
use crate::mir::join_ir::lowering::join_value_space::JoinValueSpace; use crate::mir::join_ir::lowering::join_value_space::JoinValueSpace;
use crate::mir::join_ir::lowering::loop_body_local_env::LoopBodyLocalEnv; use crate::mir::join_ir::lowering::loop_body_local_env::LoopBodyLocalEnv;
use crate::mir::join_ir::lowering::loop_scope_shape::LoopScopeShape; use crate::mir::join_ir::lowering::loop_scope_shape::LoopScopeShape;
@ -145,6 +146,7 @@ use std::collections::BTreeMap; // Phase 222.5-D: HashMap → BTreeMap for deter
/// * `body_ast` - Phase 191: Loop body AST for body-local init lowering /// * `body_ast` - Phase 191: Loop body AST for body-local init lowering
/// * `body_local_env` - Phase 185-2: Optional mutable body-local variable environment for init expressions /// * `body_local_env` - Phase 185-2: Optional mutable body-local variable environment for init expressions
/// * `join_value_space` - Phase 201: Unified JoinIR ValueId allocator (Local region: 1000+) /// * `join_value_space` - Phase 201: Unified JoinIR ValueId allocator (Local region: 1000+)
/// * `skeleton` - Phase 92 P0-3: Optional LoopSkeleton for ConditionalStep support
pub(crate) fn lower_loop_with_break_minimal( pub(crate) fn lower_loop_with_break_minimal(
_scope: LoopScopeShape, _scope: LoopScopeShape,
condition: &ASTNode, condition: &ASTNode,
@ -155,6 +157,7 @@ pub(crate) fn lower_loop_with_break_minimal(
body_ast: &[ASTNode], body_ast: &[ASTNode],
mut body_local_env: Option<&mut LoopBodyLocalEnv>, mut body_local_env: Option<&mut LoopBodyLocalEnv>,
join_value_space: &mut JoinValueSpace, join_value_space: &mut JoinValueSpace,
skeleton: Option<&crate::mir::loop_canonicalizer::LoopSkeleton>,
) -> Result<(JoinModule, JoinFragmentMeta), String> { ) -> Result<(JoinModule, JoinFragmentMeta), String> {
// Phase 170-D-impl-3: Validate that conditions only use supported variable scopes // Phase 170-D-impl-3: Validate that conditions only use supported variable scopes
// LoopConditionScopeBox checks that loop conditions don't reference loop-body-local variables // LoopConditionScopeBox checks that loop conditions don't reference loop-body-local variables
@ -585,6 +588,34 @@ pub(crate) fn lower_loop_with_break_minimal(
continue; continue;
} }
// Phase 92 P0-3: Check if skeleton has ConditionalStep for this carrier
if let Some(skel) = skeleton {
if let Some(carrier_slot) = skel.carriers.iter().find(|c| c.name == *carrier_name) {
if let UpdateKind::ConditionalStep { cond, then_delta, else_delta } = &carrier_slot.update_kind {
// Use emit_conditional_step_update instead of emit_carrier_update
eprintln!(
"[joinir/pattern2] Phase 92 P0-3: ConditionalStep detected for carrier '{}': then={}, else={}",
carrier_name, then_delta, else_delta
);
let updated_value = emit_conditional_step_update(
carrier,
&*cond,
*then_delta,
*else_delta,
&mut alloc_value,
env,
&mut carrier_update_block,
)?;
updated_carrier_values.push(updated_value);
eprintln!(
"[joinir/pattern2] Phase 92 P0-3: ConditionalStep carrier '{}' updated → {:?}",
carrier_name, updated_value
);
continue; // Skip normal carrier update
}
}
}
// Get the update expression for this carrier // Get the update expression for this carrier
let update_expr = carrier_updates.get(carrier_name).ok_or_else(|| { let update_expr = carrier_updates.get(carrier_name).ok_or_else(|| {
format!( format!(

View File

@ -144,6 +144,7 @@ fn test_pattern2_header_condition_via_exprlowerer() {
&[], &[],
None, None,
&mut join_value_space, &mut join_value_space,
None, // Phase 92 P0-3: skeleton=None for backward compatibility
); );
assert!(result.is_ok(), "ExprLowerer header path should succeed"); assert!(result.is_ok(), "ExprLowerer header path should succeed");

View File

@ -132,6 +132,7 @@ pub fn build_pattern2_minimal_structured() -> JoinModule {
&[], &[],
None, None,
&mut join_value_space, &mut join_value_space,
None, // Phase 92 P0-3: skeleton=None for backward compatibility
) )
.expect("pattern2 minimal lowering should succeed"); .expect("pattern2 minimal lowering should succeed");

View File

@ -314,7 +314,8 @@ pub fn canonicalize_loop_expr(
// Chosen: Pattern2Break (same as skip_whitespace, but with richer Skeleton) // Chosen: Pattern2Break (same as skip_whitespace, but with richer Skeleton)
// Notes: Added for parity/observability, lowering deferred to Phase 92 // Notes: Added for parity/observability, lowering deferred to Phase 92
if let Some((counter_name, normal_delta, escape_delta, _quote_char, _escape_char, body_stmts)) = // Phase 92 P0-3: Now also extracts escape_cond for JoinIR Select generation
if let Some((counter_name, normal_delta, escape_delta, _quote_char, _escape_char, body_stmts, escape_cond)) =
try_extract_escape_skip_pattern(body) try_extract_escape_skip_pattern(body)
{ {
// Build skeleton for escape skip pattern (P5b) // Build skeleton for escape skip pattern (P5b)
@ -334,10 +335,12 @@ pub fn canonicalize_loop_expr(
// Step 3: Update step with ConditionalStep (escape_delta vs normal_delta) // Step 3: Update step with ConditionalStep (escape_delta vs normal_delta)
// Pattern: normal i = i + 1, escape i = i + escape_delta (e.g., +2) // Pattern: normal i = i + 1, escape i = i + escape_delta (e.g., +2)
// Represented as UpdateKind::ConditionalStep with both deltas // Represented as UpdateKind::ConditionalStep with both deltas and condition
// Phase 92 P0-3: Now includes escape_cond for JoinIR Select generation
skeleton.steps.push(SkeletonStep::Update { skeleton.steps.push(SkeletonStep::Update {
carrier_name: counter_name.clone(), carrier_name: counter_name.clone(),
update_kind: UpdateKind::ConditionalStep { update_kind: UpdateKind::ConditionalStep {
cond: escape_cond.clone(), // Phase 92 P0-3: Condition for Select
then_delta: escape_delta, // Escape branch: +2 or other then_delta: escape_delta, // Escape branch: +2 or other
else_delta: normal_delta, // Normal branch: +1 else_delta: normal_delta, // Normal branch: +1
}, },
@ -348,6 +351,7 @@ pub fn canonicalize_loop_expr(
name: counter_name, name: counter_name,
role: CarrierRole::Counter, role: CarrierRole::Counter,
update_kind: UpdateKind::ConditionalStep { update_kind: UpdateKind::ConditionalStep {
cond: escape_cond, // Phase 92 P0-3: Condition for Select
then_delta: escape_delta, then_delta: escape_delta,
else_delta: normal_delta, else_delta: normal_delta,
}, },
@ -1787,8 +1791,10 @@ mod tests {
assert_eq!(carrier.role, CarrierRole::Counter, "Carrier should be a Counter"); assert_eq!(carrier.role, CarrierRole::Counter, "Carrier should be a Counter");
// Verify ConditionalStep with escape_delta=2, normal_delta=1 // Verify ConditionalStep with escape_delta=2, normal_delta=1
// Phase 92 P0-3: ConditionalStep now includes cond
match &carrier.update_kind { match &carrier.update_kind {
UpdateKind::ConditionalStep { UpdateKind::ConditionalStep {
cond: _, // Phase 92 P0-3: Condition for Select (don't check exact AST)
then_delta, then_delta,
else_delta, else_delta,
} => { } => {

View File

@ -310,16 +310,21 @@ mod tests {
/// } /// }
/// ``` /// ```
/// ///
/// Returns (counter_name, normal_delta, escape_delta, quote_char, escape_char, body_stmts) /// Returns (counter_name, normal_delta, escape_delta, quote_char, escape_char, body_stmts, escape_cond)
/// if pattern matches. /// if pattern matches.
/// ///
/// # Phase 91 P5b: Escape Sequence Pattern Detection /// # Phase 91 P5b: Escape Sequence Pattern Detection
/// ///
/// This function delegates to `ast_feature_extractor::detect_escape_skip_pattern` /// This function delegates to `ast_feature_extractor::detect_escape_skip_pattern`
/// for SSOT implementation. /// for SSOT implementation.
///
/// # Phase 92 P0-3: Added escape_cond
///
/// The escape_cond is the condition expression for the conditional increment
/// (e.g., `ch == '\\'`). This is needed for JoinIR Select generation.
pub fn try_extract_escape_skip_pattern( pub fn try_extract_escape_skip_pattern(
body: &[ASTNode], body: &[ASTNode],
) -> Option<(String, i64, i64, char, char, Vec<ASTNode>)> { ) -> Option<(String, i64, i64, char, char, Vec<ASTNode>, Box<ASTNode>)> {
ast_detect_escape(body).map(|info| { ast_detect_escape(body).map(|info| {
( (
info.counter_name, info.counter_name,
@ -328,6 +333,7 @@ pub fn try_extract_escape_skip_pattern(
info.quote_char, info.quote_char,
info.escape_char, info.escape_char,
info.body_stmts, info.body_stmts,
info.escape_cond, // Phase 92 P0-3: Condition for JoinIR Select
) )
}) })
} }

View File

@ -106,7 +106,15 @@ pub enum UpdateKind {
/// ``` /// ```
/// ///
/// Phase 91 P5b: Used for escape sequence handling and similar conditional increments /// Phase 91 P5b: Used for escape sequence handling and similar conditional increments
ConditionalStep { then_delta: i64, else_delta: i64 }, /// Phase 92 P0-3: Added condition expression for JoinIR Select generation
ConditionalStep {
/// The condition expression (e.g., `ch == '\\'`)
cond: Box<ASTNode>,
/// Delta for then branch (when condition is true)
then_delta: i64,
/// Delta for else branch (when condition is false)
else_delta: i64,
},
/// Conditional update with AST expressions (`if cond { x = a } else { x = b }`) /// Conditional update with AST expressions (`if cond { x = a } else { x = b }`)
Conditional { Conditional {