Merge branch 'main' into fix/1168-libc-verifier-derive

Author: carlos-alm

crates/codegraph-core/src/extractors/fsharp.rs

@@ -19,20 +19,51 @@ impl SymbolExtractor for FSharpExtractor {
 fn match_fsharp_node(node: &Node, source: &[u8], symbols: &mut FileSymbols, _depth: usize) {
     match node.kind() {
         "named_module" => handle_named_module(node, source, symbols),
+        "module_defn" => handle_module_defn(node, source, symbols),
         "function_declaration_left" => handle_function_decl(node, source, symbols),
         "type_definition" => handle_type_def(node, source, symbols),
         "import_decl" => handle_import_decl(node, source, symbols),
         "application_expression" => handle_application(node, source, symbols),
         "dot_expression" => handle_dot_expression(node, source, symbols),
+        "value_definition" => handle_value_definition(node, source, symbols),
         _ => {}
     }
 }
 
-/// Find the enclosing `named_module` and return its identifier text.
+/// Find the enclosing module name, walking up through any number of
+/// `module_defn` (nested signature modules) wrappers before reaching the
+/// top-level `named_module`. Returns the dotted path, e.g. `Outer.Inner`.
+///
+/// Source files use `named_module` for the top-level `module M = …` and
+/// the signature grammar (cargo 0.3.0) wraps nested signature modules in
+/// `module_defn` nodes. The WASM signature grammar currently emits ERROR
+/// nodes for nested signature modules so we cannot recover qualification
+/// there — tracked under #1161.
 fn enclosing_module_name(node: &Node, source: &[u8]) -> Option<String> {
-    let module = find_parent_of_type(node, "named_module")?;
-    let id = find_child(&module, "long_identifier")?;
-    Some(node_text(&id, source).to_string())
+    let mut parts: Vec<String> = Vec::new();
+    let mut current = node.parent();
+    while let Some(p) = current {
+        match p.kind() {
+            "module_defn" => {
+                if let Some(id) = find_child(&p, "identifier") {
+                    parts.push(node_text(&id, source).to_string());
+                }
+            }
+            "named_module" => {
+                if let Some(id) = find_child(&p, "long_identifier") {
+                    parts.push(node_text(&id, source).to_string());
+                }
+                break;
+            }
+            _ => {}
+        }
+        current = p.parent();
+    }
+    if parts.is_empty() {
+        return None;
+    }
+    parts.reverse();
+    Some(parts.join("."))
 }
 
 fn handle_named_module(node: &Node, source: &[u8], symbols: &mut FileSymbols) {
@@ -52,6 +83,36 @@ fn handle_named_module(node: &Node, source: &[u8], symbols: &mut FileSymbols) {
     });
 }
 
+/// Handle nested signature modules (`module Foo = ...`) emitted by the
+/// cargo 0.3.0 grammar as `module_defn`. Emits a `module` definition with
+/// the dotted parent path (e.g. `Outer.Foo`) and lets the DFS walker
+/// continue into child `val` declarations, which pick up the same path via
+/// `enclosing_module_name`.
+fn handle_module_defn(node: &Node, source: &[u8], symbols: &mut FileSymbols) {
+    let name_node = match find_child(node, "identifier") {
+        Some(n) => n,
+        None => return,
+    };
+    let raw = node_text(&name_node, source).to_string();
+    // `enclosing_module_name` walks `node.parent()` upward, so calling it on
+    // the `module_defn` itself yields the dotted prefix of its enclosing
+    // module(s) without including this module's own name.
+    let qualified = match enclosing_module_name(node, source) {
+        Some(prefix) if !prefix.is_empty() => format!("{}.{}", prefix, raw),
+        _ => raw,
+    };
+    symbols.definitions.push(Definition {
+        name: qualified,
+        kind: "module".to_string(),
+        line: start_line(node),
+        end_line: Some(end_line(node)),
+        decorators: None,
+        complexity: None,
+        cfg: None,
+        children: None,
+    });
+}
+
 fn handle_function_decl(node: &Node, source: &[u8], symbols: &mut FileSymbols) {
     // function_declaration_left: first child is the function name identifier,
     // followed by argument_patterns.
@@ -300,3 +361,171 @@ fn handle_dot_expression(node: &Node, source: &[u8], symbols: &mut FileSymbols)
         });
     }
 }
+
+/// Handle `val name : type` declarations in `.fsi` signature files.
+///
+/// The signature grammar reuses the `value_definition` node kind for `val`
+/// declarations, distinguished from the source grammar's `let` bindings by
+/// the first child being the literal `val` keyword. Source-file
+/// `value_definition` nodes (which start with `let`) are intentionally
+/// ignored here to preserve `.fs` extractor parity.
+fn handle_value_definition(node: &Node, source: &[u8], symbols: &mut FileSymbols) {
+    let first = match node.child(0) {
+        Some(c) => c,
+        None => return,
+    };
+    if first.kind() != "val" {
+        return;
+    }
+
+    let decl_left = match find_child(node, "value_declaration_left") {
+        Some(n) => n,
+        None => return,
+    };
+    let name = match extract_value_name(&decl_left, source) {
+        Some(n) => n,
+        None => return,
+    };
+
+    let kind = if has_function_type(node) { "function" } else { "variable" };
+    let module_name = enclosing_module_name(node, source);
+    let qualified = match module_name {
+        Some(m) => format!("{}.{}", m, name),
+        None => name,
+    };
+
+    symbols.definitions.push(Definition {
+        name: qualified,
+        kind: kind.to_string(),
+        line: start_line(node),
+        end_line: Some(end_line(node)),
+        decorators: None,
+        complexity: None,
+        cfg: None,
+        children: None,
+    });
+}
+
+fn extract_value_name(decl_left: &Node, source: &[u8]) -> Option<String> {
+    let pattern = find_child(decl_left, "identifier_pattern")?;
+    let ident = find_child(&pattern, "long_identifier_or_op")
+        .and_then(|n| find_child(&n, "identifier"))
+        .or_else(|| find_child(&pattern, "identifier"))?;
+    Some(node_text(&ident, source).to_string())
+}
+
+fn has_function_type(node: &Node) -> bool {
+    // The grammar wraps every type signature in `curried_spec`. A function type
+    // (e.g. `val add : int -> int -> int`) contains one or more `arguments_spec`
+    // children; a plain value (e.g. `val pi : float`) wraps a single `simple_type`.
+    let Some(curried) = find_child(node, "curried_spec") else { return false };
+    for i in 0..curried.child_count() {
+        if let Some(child) = curried.child(i) {
+            if child.kind() == "arguments_spec" {
+                return true;
+            }
+        }
+    }
+    false
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+    use crate::extractors::SymbolExtractor;
+    use tree_sitter::Parser;
+
+    fn parse_source(code: &str) -> FileSymbols {
+        let mut parser = Parser::new();
+        parser
+            .set_language(&tree_sitter_fsharp::LANGUAGE_FSHARP.into())
+            .unwrap();
+        let tree = parser.parse(code.as_bytes(), None).unwrap();
+        FSharpExtractor.extract(&tree, code.as_bytes(), "test.fs")
+    }
+
+    fn parse_signature(code: &str) -> FileSymbols {
+        let mut parser = Parser::new();
+        parser
+            .set_language(&tree_sitter_fsharp::LANGUAGE_SIGNATURE.into())
+            .unwrap();
+        let tree = parser.parse(code.as_bytes(), None).unwrap();
+        FSharpExtractor.extract(&tree, code.as_bytes(), "test.fsi")
+    }
+
+    #[test]
+    fn signature_extracts_val_declarations() {
+        let s = parse_signature("namespace MyApp.Domain\n\nval add : int -> int -> int\nval pi : float\n");
+        let add = s
+            .definitions
+            .iter()
+            .find(|d| d.name == "add")
+            .expect("val add should be extracted");
+        assert_eq!(add.kind, "function");
+        let pi = s
+            .definitions
+            .iter()
+            .find(|d| d.name == "pi")
+            .expect("val pi should be extracted");
+        assert_eq!(pi.kind, "variable");
+    }
+
+    #[test]
+    fn signature_extracts_bare_val_declarations() {
+        let s = parse_signature("val negate : int -> int\nval count : int\n");
+        assert!(s
+            .definitions
+            .iter()
+            .any(|d| d.name == "negate" && d.kind == "function"));
+        assert!(s
+            .definitions
+            .iter()
+            .any(|d| d.name == "count" && d.kind == "variable"));
+    }
+
+    #[test]
+    fn source_grammar_does_not_extract_let_bindings_as_val() {
+        // `let x = 5` is a value_definition in the source grammar but its
+        // first child is `let`, not `val`. Our handler must not extract it
+        // (preserves prior `.fs` extraction parity — only function_declaration_left
+        // produces definitions in source files).
+        let s = parse_source("module M\n\nlet x = 5\n");
+        assert!(
+            s.definitions.iter().all(|d| d.name != "x"),
+            "let bindings in .fs files must not be extracted as val definitions"
+        );
+    }
+
+    #[test]
+    fn signature_qualifies_val_inside_nested_module_defn() {
+        // The cargo 0.3.0 signature grammar wraps `module Foo = ...` as a
+        // `module_defn` node (the WASM 0.1.0 grammar emits ERROR for this
+        // construct — tracked under #1161). The `val` declarations inside
+        // must be qualified with the module path.
+        let s = parse_signature("namespace X\n\nmodule Foo =\n  val add : int -> int\n");
+        assert!(
+            s.definitions.iter().any(|d| d.name == "Foo.add" && d.kind == "function"),
+            "val add nested under `module Foo =` must be indexed as `Foo.add`, got: {:?}",
+            s.definitions.iter().map(|d| &d.name).collect::<Vec<_>>(),
+        );
+        assert!(
+            s.definitions.iter().any(|d| d.name == "Foo" && d.kind == "module"),
+            "module Foo must be indexed as a module definition"
+        );
+    }
+
+    #[test]
+    fn source_grammar_does_not_extract_val_mutable_class_fields() {
+        // `val mutable count: int = 0` inside a class is parsed as a `member_defn`
+        // node in the source grammar — NOT a `value_definition` — so our
+        // `value_definition`/`val`-first-child handler does not see it.
+        // This regression guard makes that empirical fact explicit.
+        let s = parse_source(
+            "module M\n\ntype C() =\n    val mutable count: int = 0\n",
+        );
+        assert!(
+            s.definitions.iter().all(|d| d.name != "count"),
+            "val mutable class fields must not be extracted by the signature value_definition handler"
+        );
+    }
+}

crates/codegraph-core/src/extractors/mod.rs

@@ -140,7 +140,7 @@ pub fn extract_symbols_with_opts(
         LanguageKind::Ocaml | LanguageKind::OcamlInterface => {
             ocaml::OcamlExtractor.extract_with_opts(tree, source, file_path, include_ast_nodes)
         }
-        LanguageKind::FSharp => {
+        LanguageKind::FSharp | LanguageKind::FSharpSignature => {
             fsharp::FSharpExtractor.extract_with_opts(tree, source, file_path, include_ast_nodes)
         }
         LanguageKind::ObjC => {

crates/codegraph-core/src/parser_registry.rs

@@ -28,6 +28,7 @@ pub enum LanguageKind {
     Ocaml,
     OcamlInterface,
     FSharp,
+    FSharpSignature,
     ObjC,
     Gleam,
     Julia,
@@ -70,6 +71,7 @@ impl LanguageKind {
             Self::Ocaml => "ocaml",
             Self::OcamlInterface => "ocaml-interface",
             Self::FSharp => "fsharp",
+            Self::FSharpSignature => "fsharp-signature",
             Self::ObjC => "objc",
             Self::Gleam => "gleam",
             Self::Julia => "julia",
@@ -120,7 +122,8 @@ impl LanguageKind {
             "hs" => Some(Self::Haskell),
             "ml" => Some(Self::Ocaml),
             "mli" => Some(Self::OcamlInterface),
-            "fs" | "fsx" | "fsi" => Some(Self::FSharp),
+            "fs" | "fsx" => Some(Self::FSharp),
+            "fsi" => Some(Self::FSharpSignature),
             "m" => Some(Self::ObjC),
             "gleam" => Some(Self::Gleam),
             "jl" => Some(Self::Julia),
@@ -165,6 +168,7 @@ impl LanguageKind {
             "ocaml" => Some(Self::Ocaml),
             "ocaml-interface" => Some(Self::OcamlInterface),
             "fsharp" => Some(Self::FSharp),
+            "fsharp-signature" => Some(Self::FSharpSignature),
             "objc" => Some(Self::ObjC),
             "gleam" => Some(Self::Gleam),
             "julia" => Some(Self::Julia),
@@ -207,6 +211,7 @@ impl LanguageKind {
             Self::Ocaml => tree_sitter_ocaml::LANGUAGE_OCAML.into(),
             Self::OcamlInterface => tree_sitter_ocaml::LANGUAGE_OCAML_INTERFACE.into(),
             Self::FSharp => tree_sitter_fsharp::LANGUAGE_FSHARP.into(),
+            Self::FSharpSignature => tree_sitter_fsharp::LANGUAGE_SIGNATURE.into(),
             Self::ObjC => tree_sitter_objc::LANGUAGE.into(),
             Self::Gleam => tree_sitter_gleam::LANGUAGE.into(),
             Self::Julia => tree_sitter_julia::LANGUAGE.into(),
@@ -232,8 +237,8 @@ impl LanguageKind {
         &[
             JavaScript, TypeScript, Tsx, Python, Go, Rust, Java, CSharp, Ruby, Php, Hcl, C,
             Cpp, Kotlin, Swift, Scala, Bash, Elixir, Lua, Dart, Zig, Haskell, Ocaml,
-            OcamlInterface, FSharp, ObjC, Gleam, Julia, Cuda, Clojure, Erlang, Groovy, R, Solidity,
-            Verilog,
+            OcamlInterface, FSharp, FSharpSignature, ObjC, Gleam, Julia, Cuda, Clojure, Erlang,
+            Groovy, R, Solidity, Verilog,
         ]
     }
 }
@@ -304,6 +309,7 @@ mod tests {
             | LanguageKind::Ocaml
             | LanguageKind::OcamlInterface
             | LanguageKind::FSharp
+            | LanguageKind::FSharpSignature
             | LanguageKind::ObjC
             | LanguageKind::Gleam
             | LanguageKind::Julia
@@ -320,7 +326,7 @@ mod tests {
         // Because both checks require the same manual update, they reinforce
         // each other: a developer who updates the match is reminded to also
         // update `all()` and this count.
-        const EXPECTED_LEN: usize = 35;
+        const EXPECTED_LEN: usize = 36;
         assert_eq!(
             LanguageKind::all().len(),
             EXPECTED_LEN,

package-lock.json

@@ -162,7 +162,7 @@
     "tree-sitter-dart": "^1.0.0",
     "tree-sitter-elixir": "^0.3.5",
     "tree-sitter-erlang": "github:WhatsApp/tree-sitter-erlang#semver:*",
-    "tree-sitter-fsharp": "^0.1.0",
+    "tree-sitter-fsharp": "https://github.com/ionide/tree-sitter-fsharp/archive/refs/tags/0.3.0.tar.gz",
     "tree-sitter-gleam": "github:gleam-lang/tree-sitter-gleam",
     "tree-sitter-go": "^0.25.0",
     "tree-sitter-groovy": "^0.1.2",

package.json

@@ -206,6 +206,7 @@ const grammars = [
   { name: 'tree-sitter-ocaml', pkg: 'tree-sitter-ocaml', sub: 'grammars/ocaml' },
   { name: 'tree-sitter-ocaml_interface', pkg: 'tree-sitter-ocaml', sub: 'grammars/interface' },
   { name: 'tree-sitter-fsharp', pkg: 'tree-sitter-fsharp', sub: 'fsharp' },
+  { name: 'tree-sitter-fsharp_signature', pkg: 'tree-sitter-fsharp', sub: 'fsharp_signature' },
   { name: 'tree-sitter-gleam', pkg: 'tree-sitter-gleam', sub: null },
   { name: 'tree-sitter-clojure', pkg: 'tree-sitter-clojure', sub: null },
   { name: 'tree-sitter-julia', pkg: 'tree-sitter-julia', sub: null },