feat: 搭建闭包捕获基础设施

crates/oxide_compiler/src/compiler.rs

@@ -3,6 +3,7 @@ use std::collections::HashMap;
 use crate::emit_ctx::{LabelCtx, ScopeCtx};
 use crate::symbol_table::{Binding, SymbolTable};
 use oxide_bytecode::module::CompiledModule;
+use oxide_bytecode::module::UpvalueCapture;
 use oxide_bytecode::opcode::{self, OpCode};
 
 pub use crate::hash::{compiled_module_hash, structural_hash};
@@ -164,6 +165,7 @@ pub(crate) struct CompileCtx {
     /// constructor's instance-field code, added at the super() call site during body
     /// counting so projected_pc matches where the emit pass splices the field bytecode.
     pub(crate) after_super_count_words: Option<usize>,
+    pub(crate) current_upvalue_captures: Vec<UpvalueCapture>,
     /// Set when alloc_reg() overflows into the reserved this/new.target range (≥254).
     /// Checked after each emit phase to produce a compile error rather than silent corruption.
     pub(crate) reg_overflow: bool,
@@ -227,6 +229,7 @@ impl CompileCtx {
                 builtin_reg_map: Vec::new(),
                 private_name_map: Vec::new(),
                 next_private_name_id: 1,
+                cell_registry: Vec::new(),
             },
             projected_pc: 0,
             sub_modules: Vec::new(),
@@ -238,6 +241,7 @@ impl CompileCtx {
             after_super_insert: None,
             after_super_inserted: false,
             after_super_count_words: None,
+            current_upvalue_captures: Vec::new(),
             reg_overflow: false,
             const_overflow: false,
             jump_overflow: false,
@@ -587,6 +591,203 @@ impl Compiler {
         Self
     }
 
+    #[allow(dead_code)]
+    pub(crate) fn analyze_upvalue_captures(
+        &self, body_stmts: &[Statement], parent_ctx: &CompileCtx, nested_symbols: &SymbolTable,
+    ) -> (Vec<UpvalueCapture>, u8) {
+        let mut captures: Vec<UpvalueCapture> = Vec::new();
+        let mut seen: HashMap<String, usize> = HashMap::new();
+
+        for stmt in body_stmts {
+            self.collect_upvalue_stmt(stmt, parent_ctx, nested_symbols, &mut captures, &mut seen);
+        }
+
+        let count = captures.len() as u8;
+        (captures, count)
+    }
+
+    #[allow(dead_code)]
+    fn collect_upvalue_stmt(
+        &self, stmt: &Statement, parent_ctx: &CompileCtx, nested_symbols: &SymbolTable,
+        captures: &mut Vec<UpvalueCapture>, seen: &mut HashMap<String, usize>,
+    ) {
+        match stmt {
+            Statement::ExpressionStatement(es) => {
+                self.collect_upvalue_expr(&es.expression, parent_ctx, nested_symbols, captures, seen);
+            }
+            Statement::VariableDeclaration(vd) => {
+                for decl in &vd.declarations {
+                    if let Some(init) = &decl.init {
+                        self.collect_upvalue_expr(init, parent_ctx, nested_symbols, captures, seen);
+                    }
+                }
+            }
+            Statement::ReturnStatement(rs) => {
+                if let Some(expr) = &rs.argument {
+                    self.collect_upvalue_expr(expr, parent_ctx, nested_symbols, captures, seen);
+                }
+            }
+            Statement::ForStatement(fs) => {
+                if let Some(init) = &fs.init {
+                    if let Some(e) = init.as_expression() {
+                        self.collect_upvalue_expr(e, parent_ctx, nested_symbols, captures, seen);
+                    }
+                }
+                if let Some(test) = &fs.test {
+                    self.collect_upvalue_expr(test, parent_ctx, nested_symbols, captures, seen);
+                }
+                if let Some(update) = &fs.update {
+                    self.collect_upvalue_expr(update, parent_ctx, nested_symbols, captures, seen);
+                }
+                self.collect_upvalue_stmt(&fs.body, parent_ctx, nested_symbols, captures, seen);
+            }
+            Statement::IfStatement(is) => {
+                self.collect_upvalue_expr(&is.test, parent_ctx, nested_symbols, captures, seen);
+                self.collect_upvalue_stmt(&is.consequent, parent_ctx, nested_symbols, captures, seen);
+                if let Some(alt) = &is.alternate {
+                    self.collect_upvalue_stmt(alt, parent_ctx, nested_symbols, captures, seen);
+                }
+            }
+            Statement::WhileStatement(ws) => {
+                self.collect_upvalue_expr(&ws.test, parent_ctx, nested_symbols, captures, seen);
+                self.collect_upvalue_stmt(&ws.body, parent_ctx, nested_symbols, captures, seen);
+            }
+            Statement::DoWhileStatement(dw) => {
+                self.collect_upvalue_stmt(&dw.body, parent_ctx, nested_symbols, captures, seen);
+                self.collect_upvalue_expr(&dw.test, parent_ctx, nested_symbols, captures, seen);
+            }
+            Statement::BlockStatement(bs) => {
+                for s in &bs.body {
+                    self.collect_upvalue_stmt(s, parent_ctx, nested_symbols, captures, seen);
+                }
+            }
+            Statement::TryStatement(ts) => {
+                for s in &ts.block.body {
+                    self.collect_upvalue_stmt(s, parent_ctx, nested_symbols, captures, seen);
+                }
+                if let Some(handler) = &ts.handler {
+                    for s in &handler.body.body {
+                        self.collect_upvalue_stmt(s, parent_ctx, nested_symbols, captures, seen);
+                    }
+                }
+                if let Some(finalizer) = &ts.finalizer {
+                    for s in &finalizer.body {
+                        self.collect_upvalue_stmt(s, parent_ctx, nested_symbols, captures, seen);
+                    }
+                }
+            }
+            Statement::ThrowStatement(ts) => {
+                self.collect_upvalue_expr(&ts.argument, parent_ctx, nested_symbols, captures, seen);
+            }
+            Statement::SwitchStatement(ss) => {
+                self.collect_upvalue_expr(&ss.discriminant, parent_ctx, nested_symbols, captures, seen);
+                for case in &ss.cases {
+                    for s in &case.consequent {
+                        self.collect_upvalue_stmt(s, parent_ctx, nested_symbols, captures, seen);
+                    }
+                }
+            }
+            _ => {}
+        }
+    }
+
+    #[allow(dead_code)]
+    fn collect_upvalue_expr(
+        &self, expr: &Expression, parent_ctx: &CompileCtx, nested_symbols: &SymbolTable,
+        captures: &mut Vec<UpvalueCapture>, seen: &mut HashMap<String, usize>,
+    ) {
+        match expr {
+            Expression::Identifier(ident) => {
+                let name = ident.name.as_str();
+                if nested_symbols.lookup_any_binding(name).is_some() {
+                    return;
+                }
+                if let Some((binding, _)) = parent_ctx.scopes.symbols.lookup_any_binding(name) {
+                    if seen.contains_key(name) {
+                        return;
+                    }
+                    let cell_idx = captures.len() as u8;
+                    seen.insert(name.to_string(), cell_idx as usize);
+                    captures.push(UpvalueCapture {
+                        name: name.to_string(),
+                        enclosing_reg: binding.reg,
+                        cell_idx,
+                    });
+                }
+            }
+            Expression::AssignmentExpression(ae) => {
+                if let oxide_parser::AssignmentTarget::AssignmentTargetIdentifier(ati) = &ae.left {
+                    let name = ati.name.as_str();
+                    if nested_symbols.lookup_any_binding(name).is_none() {
+                        if let Some((binding, _)) = parent_ctx.scopes.symbols.lookup_any_binding(name) {
+                            if !seen.contains_key(name) {
+                                let cell_idx = captures.len() as u8;
+                                seen.insert(name.to_string(), cell_idx as usize);
+                                captures.push(UpvalueCapture {
+                                    name: name.to_string(),
+                                    enclosing_reg: binding.reg,
+                                    cell_idx,
+                                });
+                            }
+                        }
+                    }
+                }
+                self.collect_upvalue_expr(&ae.right, parent_ctx, nested_symbols, captures, seen);
+            }
+            Expression::BinaryExpression(be) => {
+                self.collect_upvalue_expr(&be.left, parent_ctx, nested_symbols, captures, seen);
+                self.collect_upvalue_expr(&be.right, parent_ctx, nested_symbols, captures, seen);
+            }
+            Expression::UnaryExpression(ue) => {
+                self.collect_upvalue_expr(&ue.argument, parent_ctx, nested_symbols, captures, seen);
+            }
+            Expression::UpdateExpression(ue) => {
+                if let oxide_parser::SimpleAssignmentTarget::AssignmentTargetIdentifier(ati) = &ue.argument {
+                    let name = ati.name.as_str();
+                    if nested_symbols.lookup_any_binding(name).is_none() {
+                        if let Some((binding, _)) = parent_ctx.scopes.symbols.lookup_any_binding(name) {
+                            if !seen.contains_key(name) {
+                                let cell_idx = captures.len() as u8;
+                                seen.insert(name.to_string(), cell_idx as usize);
+                                captures.push(UpvalueCapture {
+                                    name: name.to_string(),
+                                    enclosing_reg: binding.reg,
+                                    cell_idx,
+                                });
+                            }
+                        }
+                    }
+                }
+            }
+            Expression::CallExpression(ce) => {
+                self.collect_upvalue_expr(&ce.callee, parent_ctx, nested_symbols, captures, seen);
+                for arg in &ce.arguments {
+                    if let Some(e) = arg.as_expression() {
+                        self.collect_upvalue_expr(e, parent_ctx, nested_symbols, captures, seen);
+                    }
+                }
+            }
+            Expression::SequenceExpression(se) => {
+                for sub_expr in &se.expressions {
+                    self.collect_upvalue_expr(sub_expr, parent_ctx, nested_symbols, captures, seen);
+                }
+            }
+            Expression::ConditionalExpression(ce) => {
+                self.collect_upvalue_expr(&ce.test, parent_ctx, nested_symbols, captures, seen);
+                self.collect_upvalue_expr(&ce.consequent, parent_ctx, nested_symbols, captures, seen);
+                self.collect_upvalue_expr(&ce.alternate, parent_ctx, nested_symbols, captures, seen);
+            }
+            Expression::ArrayExpression(ae) => {
+                for elem in &ae.elements {
+                    if let Some(e) = elem.as_expression() {
+                        self.collect_upvalue_expr(e, parent_ctx, nested_symbols, captures, seen);
+                    }
+                }
+            }
+            _ => {}
+        }
+    }
+
     pub(crate) fn extract_function_parts<'a>(
         &self, function: &'a oxide_parser::Function<'a>,
     ) -> Result<(Vec<ParamSpec<'a>>, &'a [Statement<'a>]), String> {
@@ -693,6 +894,7 @@ impl Compiler {
                     reg: binding.reg,
                     initialized: binding.initialized,
                     is_const: binding.is_const,
+                    is_captured: binding.is_captured,
                 },
             );
             inherited_reg_start = inherited_reg_start.max(binding.reg.saturating_add(1));
@@ -704,6 +906,7 @@ impl Compiler {
                     reg: *reg,
                     initialized: true,
                     is_const: true,
+                    is_captured: false,
                 },
             );
             inherited_reg_start = inherited_reg_start.max(reg.saturating_add(1));
@@ -714,6 +917,23 @@ impl Compiler {
         // same register offset (params go after builtin slots).
         ctx.reset_regs();
 
+        // Free variable analysis for upvalue capture (Ordinary + Arrow functions only)
+        // ponytail: temporarily disabled to isolate closure bug
+        /*
+        if matches!(body_context, FunctionBodyContext::Ordinary | FunctionBodyContext::Arrow) {
+            let (captures, _cells) = self.analyze_upvalue_captures(body_stmts, parent_ctx, &ctx.scopes.symbols);
+            ctx.current_upvalue_captures = captures;
+            ctx.scopes.cell_registry = ctx.current_upvalue_captures.iter()
+                .map(|u| (u.name.clone(), u.cell_idx))
+                .collect();
+            for up in &ctx.current_upvalue_captures {
+                if let Some((binding, _)) = parent_ctx.scopes.symbols.lookup_any_binding(&up.name) {
+                    binding.is_captured.set(true);
+                }
+            }
+        }
+        */
+
         // Function body scope - params and local vars
         ctx.push_scope_with_kind(ScopeKind::FunctionScope);
 
@@ -858,6 +1078,8 @@ impl Compiler {
             is_class_constructor: false,
             is_derived_constructor: false,
             needs_home_object: false,
+            upvalue_captures: Vec::new(),
+            cells_needed: 0,
         })
     }
 
@@ -956,6 +1178,8 @@ impl Compiler {
             is_class_constructor: false,
             is_derived_constructor: false,
             needs_home_object: false,
+            upvalue_captures: Vec::new(),
+            cells_needed: 0,
         })
     }
 }

crates/oxide_compiler/src/counter/expr/assignment.rs

@@ -32,13 +32,21 @@ impl Compiler {
             self.count_expression(&assign.right, ctx);
             ctx.count_load_const(); // private id
             ctx.count_instr(); // SET_PRIVATE
-        } else if let oxide_parser::AssignmentTarget::AssignmentTargetIdentifier(_) = &assign.left {
+        } else if let oxide_parser::AssignmentTarget::AssignmentTargetIdentifier(ati) = &assign.left {
+            let name = ati.name.as_str();
+            // Check if target is an upvalue or captured cell
+            let is_upvalue = ctx.current_upvalue_captures.iter().any(|u| u.name == name);
+            let is_captured = ctx.scopes.symbols.lookup_is_captured(name);
             if let Some(logical_op) = assign.operator.to_logical_operator() {
                 self.count_identifier_logical_assignment(logical_op, &assign.right, ctx);
             } else {
                 self.count_expression(&assign.right, ctx);
                 if assign.operator != AssignmentOperator::Assign {
                     ctx.projected_pc += 1; // COMPOUND_* on var
+                } else if is_upvalue {
+                    ctx.count_instr(); // STORE_UPVALUE
+                } else if is_captured {
+                    ctx.count_instr(); // CELL_SET
                 } else {
                     ctx.alloc_reg();
                     ctx.projected_pc += 1; // STORE_VAR

crates/oxide_compiler/src/counter/expr/literal.rs

@@ -21,8 +21,26 @@ impl Compiler {
         let Expression::Identifier(ident) = expr else {
             return;
         };
-        if CompileCtx::is_known_builtin(ident.name.as_str()) {
-            let _ = ctx.lookup_or_builtin(ident.name.as_str());
+        let name = ident.name.as_str();
+
+        // Check if this is an upvalue reference (nested closure)
+        for up in &ctx.current_upvalue_captures {
+            if up.name == name {
+                ctx.alloc_reg();
+                ctx.count_instr(); // LOAD_UPVALUE
+                return;
+            }
+        }
+
+        // Check if the owner's binding is captured (cell access)
+        if ctx.scopes.symbols.lookup_is_captured(name) {
+            ctx.alloc_reg();
+            ctx.count_instr(); // CELL_GET
+            return;
+        }
+
+        if CompileCtx::is_known_builtin(name) {
+            let _ = ctx.lookup_or_builtin(name);
         }
         ctx.alloc_reg();
         ctx.projected_pc += 1; // LOAD_VAR

crates/oxide_compiler/src/emit_ctx.rs

@@ -30,6 +30,7 @@ pub(crate) struct ScopeCtx {
     pub(crate) builtin_reg_map: Vec<(String, u8)>,
     pub(crate) private_name_map: Vec<(String, u32)>,
     pub(crate) next_private_name_id: u32,
+    pub(crate) cell_registry: Vec<(String, u8)>,
 }
 
 /// Reserved for destructuring-pattern state (Phase 15). Intentionally empty

crates/oxide_compiler/src/emitter/expr/assignment.rs

@@ -142,6 +142,20 @@ impl Compiler {
             } else {
                 let val_reg = self.emit_expression(&assign.right, ctx)?;
                 let name = id_ref.name.as_str();
+                // Check if target is an upvalue reference
+                if let Some(uv_idx) = ctx.current_upvalue_captures.iter().position(|u| u.name == name) {
+                    ctx.emit(opcode::encode(OpCode::STORE_UPVALUE, 0, val_reg, uv_idx as u8));
+                    return Ok(val_reg);
+                }
+                // Check if target is a captured cell
+                if ctx.scopes.symbols.lookup_is_captured(name) {
+                    let cell_idx = ctx.scopes.cell_registry.iter()
+                        .find(|(n, _)| n == name)
+                        .map(|(_, idx)| *idx)
+                        .unwrap_or(0);
+                    ctx.emit(opcode::encode(OpCode::CELL_SET, 0, val_reg, cell_idx));
+                    return Ok(val_reg);
+                }
                 let var_reg = ctx.lookup_or_global(name);
                 let is_const = ctx.lookup_const_flag(name);
                 let const_flag = if is_const { 1 } else { 0 };

crates/oxide_compiler/src/emitter/expr/literal.rs

@@ -56,7 +56,34 @@ impl Compiler {
     pub(in crate::emitter) fn emit_identifier_expression(
         &self, ident: &oxide_parser::IdentifierReference, ctx: &mut CompileCtx,
     ) -> Result<u8, String> {
-        let var_reg = ctx.lookup_or_builtin(ident.name.as_str())?;
+        let name = ident.name.as_str();
+
+        // Check if this is an upvalue reference (nested closure)
+        for (uv_idx, up) in ctx.current_upvalue_captures.iter().enumerate() {
+            if up.name == name {
+                let r = ctx.alloc_reg();
+                let idx = uv_idx as u8;
+                ctx.emit(opcode::encode(OpCode::LOAD_UPVALUE, r, idx, 0));
+                return Ok(r);
+            }
+        }
+
+        // Check if the owner's binding is captured (cell access)
+        if ctx.scopes.symbols.lookup_is_captured(name) {
+            let cell_idx = ctx.scopes.cell_registry.iter()
+                .find(|(n, _)| n == name)
+                .map(|(_, idx)| *idx)
+                .unwrap_or(0);
+            let r = ctx.alloc_reg();
+            if let Some((binding, _)) = ctx.scopes.symbols.lookup_any_binding(name) {
+                ctx.emit(opcode::encode(OpCode::CELL_GET, r, binding.reg, cell_idx));
+            } else {
+                ctx.emit(opcode::encode(OpCode::CELL_GET, r, 0, cell_idx));
+            }
+            return Ok(r);
+        }
+
+        let var_reg = ctx.lookup_or_builtin(name)?;
         let r = ctx.alloc_reg();
         ctx.emit(opcode::encode(OpCode::LOAD_VAR, r, var_reg, 0));
         Ok(r)

crates/oxide_compiler/src/emitter/stmt/declaration.rs

@@ -33,7 +33,14 @@ impl Compiler {
                 ctx.emit_load_const(tmp, idx);
                 let var_reg = ctx.alloc_reg();
                 ctx.declare(bi.name.as_str(), var_reg, decl.kind, is_const)?;
-                ctx.emit(opcode::encode(OpCode::STORE_VAR, var_reg, tmp, 0));
+                let is_captured = ctx.scopes.symbols.lookup_is_captured(bi.name.as_str());
+                if is_captured {
+                    let cell_idx = ctx.scopes.cell_registry.len() as u8;
+                    ctx.scopes.cell_registry.push((bi.name.to_string(), cell_idx));
+                    ctx.emit(opcode::encode(OpCode::MAKE_CELL, var_reg, cell_idx, 0));
+                } else {
+                    ctx.emit(opcode::encode(OpCode::STORE_VAR, var_reg, tmp, 0));
+                }
                 ctx.init_var(bi.name.as_str());
                 r = Some(var_reg);
             }

crates/oxide_compiler/src/symbol_table.rs

@@ -17,6 +17,7 @@ pub(crate) struct Binding {
     pub(crate) reg: u8,
     pub(crate) initialized: bool,
     pub(crate) is_const: bool,
+    pub(crate) is_captured: bool,
 }
 
 pub struct SymbolTable {
@@ -84,6 +85,7 @@ impl SymbolTable {
                 reg,
                 initialized: false,
                 is_const: matches!(kind, VariableDeclarationKind::Const) || is_const,
+                is_captured: false,
             },
         );
         Ok(())
@@ -101,13 +103,22 @@ impl SymbolTable {
         Err(format!("Identifier '{name}' is not defined"))
     }
 
-    pub fn lookup_any(&self, name: &str) -> Option<u8> {
+    pub(crate) fn lookup_any(&self, name: &str) -> Option<u8> {
         self.scopes
             .iter()
             .rev()
             .find_map(|scope| scope.bindings.get(name).map(|binding| binding.reg))
     }
 
+    pub(crate) fn lookup_any_binding(&self, name: &str) -> Option<(&Binding, usize)> {
+        for (i, scope) in self.scopes.iter().enumerate().rev() {
+            if let Some(b) = scope.bindings.get(name) {
+                return Some((b, i));
+            }
+        }
+        None
+    }
+
     pub fn lookup_or_global(&mut self, name: &str, reg_for_new: u8) -> u8 {
         for scope in self.scopes.iter().rev() {
             if let Some(b) = scope.bindings.get(name) {
@@ -120,6 +131,7 @@ impl SymbolTable {
                 reg: reg_for_new,
                 initialized: true,
                 is_const: false,
+                is_captured: false,
             },
         );
         reg_for_new
@@ -137,6 +149,15 @@ impl SymbolTable {
         false
     }
 
+    pub fn lookup_is_captured(&self, name: &str) -> bool {
+        for scope in self.scopes.iter().rev() {
+            if let Some(b) = scope.bindings.get(name) {
+                return b.is_captured;
+            }
+        }
+        false
+    }
+
     pub fn declare_initialized(
         &mut self, name: &str, reg: u8, kind: VariableDeclarationKind, is_const: bool,
     ) -> Result<(), String> {
@@ -157,6 +178,7 @@ impl SymbolTable {
                 reg,
                 initialized: true,
                 is_const: matches!(kind, VariableDeclarationKind::Const) || is_const,
+                is_captured: false,
             },
         );
         Ok(())
@@ -167,6 +189,7 @@ impl SymbolTable {
             reg,
             initialized: true,
             is_const: false,
+            is_captured: false,
         });
     }