Unified: Only translate conditional compilation blocks to if expressions in statement context

Author: paldepind

unified/extractor/ast_types.yml

@@ -68,8 +68,6 @@ supertypes:
     - do_while_stmt
     - defer_stmt
     - labeled_stmt
-    - associated_type_declaration
-    - initializer_declaration
   callable:
     subtypes:
       - top_level
@@ -95,7 +93,6 @@ supertypes:
       - class_like_declaration
       - type_alias_declaration
       - associated_type_declaration
-      - if_expr # For conditional compilation
       - unsupported_node
       - unhandled_node
     fields:

unified/extractor/src/languages/swift/swift.rs

@@ -38,6 +38,9 @@ struct SwiftContext {
     /// True while translating a pattern. Optional chaining in this context is
     /// represented as an `Optional.some` call rather than being unwrapped.
     in_pattern: bool,
+    /// True when a conditional compilation block should be translated into an
+    /// if expression.
+    conditional_compilation_as_expr: bool,
     /// When iterating over the clauses of a conditional compilation block, this
     /// holds the tail of the clauses that have already been translated.
     conditional_compilation_tail: Option<yeast::OutputId>,
@@ -67,29 +70,11 @@ fn implicit_parameterless_initializer(
     ctx: &mut yeast::build::BuildCtx<'_, SwiftContext>,
     members: &[yeast::OutputId],
 ) -> Option<yeast::OutputId> {
-    // Search for an initializer while recursing into nodes for conditional
-    // compilation blocks.
-    fn find_initializer(ast: &yeast::OutputAst, id: yeast::OutputId) -> Option<yeast::OutputId> {
-        let node = ast.get_node(id)?;
-        match node.kind_name() {
-            "constructor_declaration" => Some(id),
-            name => {
-                let fields: &[&str] = match name {
-                    "if_expr" => &["then", "else"], // look into conditional compilation
-                    "block" => &["stmt"],           // look into blocks in branches
-                    _ => &[],
-                };
-                fields.iter().find_map(|field| {
-                    node.field_children(ast.field_id_for_name(field)?)
-                        .iter()
-                        .find_map(|child| find_initializer(ast, *child))
-                })
-            }
-        }
-    }
-    let has_explicit_initializer = members
-        .iter()
-        .any(|member| find_initializer(ctx.output, *member).is_some());
+    let has_explicit_initializer = members.iter().any(|member| {
+        ctx.output
+            .get_node(*member)
+            .is_some_and(|node| node.kind_name() == "constructor_declaration")
+    });
     (!has_explicit_initializer).then(|| {
         tree!((constructor_declaration
             modifier: (modifier "generated")
@@ -174,6 +159,14 @@ fn translation_rules() -> Vec<Rule<SwiftContext>> {
         // but a single wrapped `variableDecl` can translate to *several*
         // declarations (`let x = 1, y = 2`), so the wrapped node is captured with
         // `_*` and the result annotated `stmt*` to splice all of them.
+        rule!(
+            (codeBlockItem item: (ifConfigDecl) @@item)
+            =>
+            stmt* {
+                ctx.conditional_compilation_as_expr = true;
+                ctx.translate(item)?
+            }
+        ),
         rule!((codeBlockItem item: _* @item) => stmt* { item }),
         // ---- Literals ----
         // swift-syntax does not distinguish the lexical integer/string forms
@@ -897,25 +890,33 @@ fn translation_rules() -> Vec<Rule<SwiftContext>> {
             =>
             (parameter pattern: (identifier #{name}))
         ),
-        // ---- Conditioal compilation blocks ----
+        // ---- Conditional compilation blocks ----
         // Conditional compilation blocks (`#if … #elseif … #else … #endif`) are
         // represented with a `ifConfigDecl` node for the entire block and an
-        // `ifConfigClause` for each branch. We translate them into nested
-        // `if_expr` nodes.
+        // `ifConfigClause` for each branch.
 
-        // When a conditional compilation block contains a switch case it
-        // appears within a `switch` -- we do not support this yet.
+        // When a conditional compilation block is not in a context where it
+        // should be translated into an if expression we simply extract the
+        // first branch (for now).
         rule!(
-            (ifConfigDecl clauses: (ifConfigClause elements: (switchCase)))
+            (ifConfigDecl
+                clauses: (ifConfigClause elements: _* @items)
+                clauses: _* @@rest)
+            where !ctx.conditional_compilation_as_expr
             =>
-            (unsupported_node)
+            _* { items }
         ),
         // A conditional compilation block is translated into `if_expr` nodes
         // with branches occurring nested within the else branch.
         rule!(
             (ifConfigDecl clauses: _* @@clauses) @@declaration
             =>
             expr {
+                // Revert `conditional_compilation_as_expr` before processing
+                // clauses. Note that if a clause contains a nested conditional
+                // compilation block, then that block will be wrapped in another
+                // `codeBlockItem` which will flip this back to `true`.
+                ctx.conditional_compilation_as_expr = false;
                 ctx.conditional_compilation_tail = None;
                 for clause in clauses.into_iter().rev() {
                     ctx.conditional_compilation_tail = ctx.translate(clause)?.into_iter().next();

unified/extractor/tests/corpus/swift/conditional-compilation/conditional-compilation-in-attributes.output

@@ -98,19 +98,14 @@ sourceFile
 
 top_level source="⟨body⟩"
   body:
-    block source="⟨stmt⟩\n\n⟨stmt⟩"
+    block source="⟨stmt⟩\n\n#if DEBUG\n    ⟨stmt⟩"
       stmt:
         function_declaration source="#if DEBUG\n    @available(*, deprecated)\n#endif\nfunc ⟨name_node⟩() ⟨body⟩"
           name_node: identifier "f" source="f"
           body: block "{}" source="{}"
-        variable_declaration source="⟨modifier⟩\n⟨modifier⟩ ⟨pattern⟩ = ⟨value⟩"
+        variable_declaration source="⟨modifier⟩\n#endif\n⟨modifier⟩ ⟨pattern⟩ = ⟨value⟩"
           modifier:
             modifier "let" source="let"
-            if_expr source="#if ⟨condition⟩\n    ⟨then⟩\n#endif" <-- ERROR: The field variable_declaration.modifier should contain modifier, but got if_expr
-              modifier: modifier "conditional_compilation" source=""
-              condition: identifier "DEBUG" source="DEBUG"
-              then:
-                block source="⟨stmt⟩"
-                  stmt: modifier "@available(*, deprecated)" source="@available(*, deprecated)" <-- ERROR: The field block.stmt should contain stmt, but got modifier
+            modifier "@available(*, deprecated)" source="@available(*, deprecated)"
           pattern: identifier "x" source="x"
           value: int_literal "1" source="1"

unified/extractor/tests/corpus/swift/conditional-compilation/conditional-compilation-in-class-body.output

@@ -78,23 +78,17 @@ top_level source="⟨body⟩"
   body:
     block source="⟨stmt⟩"
       stmt:
-        class_like_declaration source="⟨modifier⟩ ⟨name_node⟩ {\n    ⟨member⟩\n}"
+        class_like_declaration source="⟨modifier⟩ ⟨name_node⟩ {\n    #if DEBUG\n        ⟨member⟩\n        ⟨member⟩\n    #endif\n}"
           modifier: modifier "class" source="class"
           name_node: identifier "C" source="C"
           member:
-            if_expr source="#if ⟨condition⟩\n        ⟨then⟩\n    #endif"
-              modifier: modifier "conditional_compilation" source=""
-              condition: identifier "DEBUG" source="DEBUG"
-              then:
-                block source="⟨stmt⟩\n        ⟨stmt⟩"
-                  stmt:
-                    constructor_declaration source="⟨name_node⟩(⟨parameter⟩) ⟨body⟩"
-                      name_node: identifier "init" source="init"
-                      parameter:
-                        parameter source="⟨external_name_node⟩⟨pattern⟩: ⟨type⟩"
-                          external_name_node: identifier "x" source="x"
-                          type: identifier "Int" source="Int"
-                          pattern: identifier "x" source="x"
-                      body: block "{}" source="{}"
-                    destructor_declaration source="deinit ⟨body⟩"
-                      body: block "{}" source="{}"
+            constructor_declaration source="⟨name_node⟩(⟨parameter⟩) ⟨body⟩"
+              name_node: identifier "init" source="init"
+              parameter:
+                parameter source="⟨external_name_node⟩⟨pattern⟩: ⟨type⟩"
+                  external_name_node: identifier "x" source="x"
+                  type: identifier "Int" source="Int"
+                  pattern: identifier "x" source="x"
+              body: block "{}" source="{}"
+            destructor_declaration source="deinit ⟨body⟩"
+              body: block "{}" source="{}"

unified/extractor/tests/corpus/swift/conditional-compilation/conditional-compilation-in-switch-cases.output

@@ -123,10 +123,19 @@ top_level source="⟨body⟩"
           body:
             block source="{\n    ⟨stmt⟩\n}"
               stmt:
-                switch_expr source="switch ⟨value⟩ {\n    ⟨case⟩\n    ⟨case⟩\n    }"
+                switch_expr source="switch ⟨value⟩ {\n    #if DEBUG\n        ⟨case⟩\n        ⟨case⟩\n    #endif\n    ⟨case⟩\n    }"
                   value: identifier "x" source="x"
                   case:
-                    unsupported_node "#if DEBUG\n        case 0: break\n        case 1: break\n    #endif" source="#if DEBUG\n        case 0: break\n        case 1: break\n    #endif" <-- ERROR: The field switch_expr.case should contain switch_case, but got unsupported_node
+                    switch_case source="case ⟨pattern⟩: ⟨body⟩"
+                      pattern: int_literal "0" source="0"
+                      body:
+                        block source="⟨stmt⟩"
+                          stmt: break_expr "break" source="break"
+                    switch_case source="case ⟨pattern⟩: ⟨body⟩"
+                      pattern: int_literal "1" source="1"
+                      body:
+                        block source="⟨stmt⟩"
+                          stmt: break_expr "break" source="break"
                     switch_case source="default: ⟨body⟩"
                       body:
                         block source="⟨stmt⟩"