Filter JSX tag reference false positives

Author: jayhack

crates/graph-sitter-engine/src/lib.rs

@@ -3168,6 +3168,86 @@ fn collect_typescript_identifier_candidates(
             }
             return;
         }
+        "jsx_element" => {
+            if let Some(open_tag) = node.child_by_field_name("open_tag") {
+                collect_typescript_jsx_tag_candidates(
+                    file_id,
+                    source,
+                    open_tag,
+                    symbol_ranges,
+                    local_bindings_by_symbol_id,
+                    local_binding_scopes,
+                    excluded_ranges,
+                    indexed_local_symbols,
+                    out,
+                );
+            }
+            let open_range = node
+                .child_by_field_name("open_tag")
+                .map(|tag| SourceRange::from(tag.range()));
+            let close_range = node
+                .child_by_field_name("close_tag")
+                .map(|tag| SourceRange::from(tag.range()));
+            let mut cursor = node.walk();
+            for child in node.named_children(&mut cursor) {
+                let child_range = SourceRange::from(child.range());
+                if open_range.is_some_and(|range| range == child_range)
+                    || close_range.is_some_and(|range| range == child_range)
+                {
+                    continue;
+                }
+                collect_typescript_identifier_candidates(
+                    file_id,
+                    source,
+                    child,
+                    symbol_ranges,
+                    local_bindings_by_symbol_id,
+                    local_binding_scopes,
+                    excluded_ranges,
+                    indexed_local_symbols,
+                    out,
+                );
+            }
+            return;
+        }
+        "jsx_opening_element" | "jsx_self_closing_element" => {
+            collect_typescript_jsx_tag_candidates(
+                file_id,
+                source,
+                node,
+                symbol_ranges,
+                local_bindings_by_symbol_id,
+                local_binding_scopes,
+                excluded_ranges,
+                indexed_local_symbols,
+                out,
+            );
+            return;
+        }
+        "jsx_closing_element" => return,
+        "jsx_attribute" => {
+            let mut cursor = node.walk();
+            for child in node.named_children(&mut cursor) {
+                if matches!(
+                    child.kind(),
+                    "property_identifier" | "jsx_namespace_name" | "string"
+                ) {
+                    continue;
+                }
+                collect_typescript_identifier_candidates(
+                    file_id,
+                    source,
+                    child,
+                    symbol_ranges,
+                    local_bindings_by_symbol_id,
+                    local_binding_scopes,
+                    excluded_ranges,
+                    indexed_local_symbols,
+                    out,
+                );
+            }
+            return;
+        }
         "member_expression" => {
             if let (Some(object), Some(property)) = (
                 node.child_by_field_name("object"),
@@ -3262,6 +3342,92 @@ fn collect_typescript_identifier_candidates(
     }
 }
 
+fn collect_typescript_jsx_tag_candidates(
+    file_id: u32,
+    source: &str,
+    tag_node: Node<'_>,
+    symbol_ranges: &[(u32, SourceRange)],
+    local_bindings_by_symbol_id: &HashMap<u32, HashSet<String>>,
+    local_binding_scopes: &[LocalBindingScope],
+    excluded_ranges: &[SourceRange],
+    indexed_local_symbols: &IndexedLocalSymbols,
+    out: &mut Vec<ReferenceCandidate>,
+) {
+    let Some(name_node) = tag_node.child_by_field_name("name") else {
+        return;
+    };
+
+    match name_node.kind() {
+        "identifier" => {
+            let range = SourceRange::from(name_node.range());
+            if !range_matches_any(range, excluded_ranges) {
+                if let Ok(name) = name_node.utf8_text(source.as_bytes()) {
+                    if typescript_jsx_tag_identifier_can_reference(name) {
+                        let source_symbol_id = innermost_symbol_for_range(symbol_ranges, range);
+                        if !typescript_reference_is_shadowed(
+                            source_symbol_id,
+                            name,
+                            range,
+                            local_bindings_by_symbol_id,
+                            local_binding_scopes,
+                            indexed_local_symbols,
+                        ) {
+                            out.push(ReferenceCandidate {
+                                source_file_id: file_id,
+                                source_symbol_id,
+                                name: name.to_owned(),
+                                qualifier: None,
+                                range,
+                                is_subclass: false,
+                                call_range: None,
+                            });
+                        }
+                    }
+                }
+            }
+        }
+        "member_expression" => {
+            collect_typescript_identifier_candidates(
+                file_id,
+                source,
+                name_node,
+                symbol_ranges,
+                local_bindings_by_symbol_id,
+                local_binding_scopes,
+                excluded_ranges,
+                indexed_local_symbols,
+                out,
+            );
+        }
+        _ => {}
+    }
+
+    let name_range = SourceRange::from(name_node.range());
+    let mut cursor = tag_node.walk();
+    for child in tag_node.named_children(&mut cursor) {
+        if SourceRange::from(child.range()) == name_range || child.kind() == "type_arguments" {
+            continue;
+        }
+        collect_typescript_identifier_candidates(
+            file_id,
+            source,
+            child,
+            symbol_ranges,
+            local_bindings_by_symbol_id,
+            local_binding_scopes,
+            excluded_ranges,
+            indexed_local_symbols,
+            out,
+        );
+    }
+}
+
+fn typescript_jsx_tag_identifier_can_reference(name: &str) -> bool {
+    name.chars()
+        .next()
+        .is_some_and(|first| first == '_' || first == '$' || first.is_ascii_uppercase())
+}
+
 fn collect_typescript_call_function_operands(
     file_id: u32,
     source: &str,
@@ -8641,6 +8807,101 @@ export function run() {\n  local(helper(1));\n  return run();\n}\n",
         }));
     }
 
+    #[test]
+    fn resolves_jsx_component_tag_references_without_intrinsic_or_prop_false_edges() {
+        let repo = temp_repo_path("resolve-typescript-jsx-tag-references");
+        fs::create_dir_all(repo.join("src")).unwrap();
+        fs::write(
+            repo.join("src/components.tsx"),
+            "export function Button() { return null; }\n\
+export function div() { return null; }\n\
+export namespace UI { export function Card() { return null; } }\n",
+        )
+        .unwrap();
+        fs::write(
+            repo.join("src/app.tsx"),
+            "import { Button, UI, div } from './components';\n\n\
+export function App() {\n\
+  const local = Button;\n\
+  return <main Button={Button}><Button /><UI.Card /><div /></main>;\n\
+}\n",
+        )
+        .unwrap();
+
+        let index = index_typescript_path(&repo).unwrap();
+        fs::remove_dir_all(&repo).unwrap();
+
+        let app_file_id = index
+            .files
+            .iter()
+            .find(|file| file.path == "src/app.tsx")
+            .unwrap()
+            .id;
+        let components_file_id = index
+            .files
+            .iter()
+            .find(|file| file.path == "src/components.tsx")
+            .unwrap()
+            .id;
+        let app_symbol_id = index
+            .symbols
+            .iter()
+            .find(|symbol| symbol.file_id == app_file_id && symbol.name == "App")
+            .unwrap()
+            .id;
+        let button_symbol_id = index
+            .symbols
+            .iter()
+            .find(|symbol| symbol.file_id == components_file_id && symbol.name == "Button")
+            .unwrap()
+            .id;
+        let div_symbol_id = index
+            .symbols
+            .iter()
+            .find(|symbol| symbol.file_id == components_file_id && symbol.name == "div")
+            .unwrap()
+            .id;
+        let ui_symbol_id = index
+            .symbols
+            .iter()
+            .find(|symbol| symbol.file_id == components_file_id && symbol.name == "UI")
+            .unwrap()
+            .id;
+        let card_symbol_id = index
+            .symbols
+            .iter()
+            .find(|symbol| symbol.file_id == components_file_id && symbol.name == "Card")
+            .unwrap()
+            .id;
+
+        let button_references = index
+            .references
+            .iter()
+            .filter(|reference| {
+                reference.source_symbol_id == Some(app_symbol_id)
+                    && reference.target_symbol_id == button_symbol_id
+            })
+            .count();
+        assert_eq!(
+            button_references, 3,
+            "local assignment, prop expression, and component tag should resolve without counting the prop name"
+        );
+        assert!(index.references.iter().any(|reference| {
+            reference.source_symbol_id == Some(app_symbol_id)
+                && reference.target_symbol_id == ui_symbol_id
+                && reference.name == "UI"
+        }));
+        assert!(index.references.iter().any(|reference| {
+            reference.source_symbol_id == Some(app_symbol_id)
+                && reference.target_symbol_id == card_symbol_id
+                && reference.name == "Card"
+        }));
+        assert!(!index.references.iter().any(|reference| {
+            reference.source_symbol_id == Some(app_symbol_id)
+                && reference.target_symbol_id == div_symbol_id
+        }));
+    }
+
     #[test]
     fn extracts_typescript_promise_chain_records() {
         let repo = temp_repo_path("typescript-promise-chain-records");

docs/benchmarks/large-repos.mdx

@@ -32,13 +32,14 @@ mode, the old eager Python graph is blocked after the compact Rust index builds.
 | Repository | Python wall | Python max RSS | Rust wall | Rust max RSS | Wall improvement | RSS improvement |
 | --- | ---: | ---: | ---: | ---: | ---: | ---: |
 | Apache Airflow `2.10.5` | 18.940s | 3469.5 MB | 4.085s | 266.2 MB | 4.637x | 13.031x |
-| Next.js `v15.0.0` | 24.959s | 3100.1 MB | 10.861s | 439.8 MB | 2.298x | 7.049x |
+| Next.js `v15.0.0` | 24.959s | 3100.1 MB | 10.771s | 435.2 MB | 2.317x | 7.123x |
 
 The Airflow row exercises compact Python files, symbols, imports, import
 resolution, references, dependencies, and Python compatibility handles. The
 Next.js row exercises TypeScript/JavaScript files, symbols, imports, exports,
-relative and tsconfig path resolution, references, dependencies, subclass edges,
-read-only function call records, and read-only Promise-chain records.
+relative and tsconfig path resolution, JSX component tag references,
+dependencies, subclass edges, read-only function call records, and read-only
+Promise-chain records.
 
 ## Installed-Wheel uvx Proof
 

rust-rewrite/benchmarks.md

@@ -257,9 +257,9 @@ uv run python rust-rewrite/tools/check_pinned_typescript_codebase.py \
 
 | Input | Rust `Codebase` wall | Rust `Codebase` max RSS | Files | Symbols | Imports | Exports | Import resolutions | References | Dependencies | Function calls | Promise chains | Python graph blocked |
 | --- | ---: | ---: | ---: | ---: | ---: | ---: | ---: | ---: | ---: | ---: | ---: | --- |
-| Next.js `v15.0.0` (`51bfe3c1863b191f4b039bc230e8ed5c57b0baf3`) | 10.861s | 439.8 MB | 13688 | 44871 | 28210 | 16027 | 13462 | 114465 | 49287 | 197581 | 878 | yes |
+| Next.js `v15.0.0` (`51bfe3c1863b191f4b039bc230e8ed5c57b0baf3`) | 10.771s | 435.2 MB | 13688 | 44871 | 28210 | 16027 | 13462 | 113809 | 49285 | 197581 | 878 | yes |
 
-Compared with the Python TypeScript parse/object-materialization baseline above, the current Rust `Codebase` TypeScript shell is about 2.298x faster and about 7.049x lower max RSS while exposing compact export, call, and Promise-chain handles and keeping the eager Python graph unbuilt. The pinned proof also validates a real `packages/next/src/cli/next-lint.ts` file/symbol lookup for 27 file-local call records, 16 `nextLint` symbol call records, and one `.then/.catch` Promise chain without materializing the full call or chain caches. A parser fallback now tries the TS grammar for `.ts`/`.js` files and keeps the lower-error parse, reducing pinned Next.js parser-error files from 114 to 113 by recovering `test/integration/typescript/components/angle-bracket-type-assertions.ts`.
+Compared with the Python TypeScript parse/object-materialization baseline above, the current Rust `Codebase` TypeScript shell is about 2.317x faster and about 7.123x lower max RSS while exposing compact export, call, and Promise-chain handles and keeping the eager Python graph unbuilt. The pinned proof also validates a real `packages/next/src/cli/next-lint.ts` file/symbol lookup for 27 file-local call records, 16 `nextLint` symbol call records, and one `.then/.catch` Promise chain without materializing the full call or chain caches. JSX traversal now records component tag references while skipping lowercase intrinsic tags and prop-name false positives. A parser fallback tries the TS grammar for `.ts`/`.js` files and keeps the lower-error parse, reducing pinned Next.js parser-error files from 114 to 113 by recovering `test/integration/typescript/components/angle-bracket-type-assertions.ts`.
 
 The same proof is now available as an opt-in test gate:
 
@@ -269,7 +269,7 @@ uv run python rust-rewrite/tools/check_pinned_typescript_codebase.py \
   --skip-fetch
 ```
 
-On 2026-06-19, that checker validated exact pinned Next.js `Codebase` handle counts plus compact function-call and Promise-chain counts, confirmed the Python graph stayed blocked, and measured 10.861s wall / 439.8 MB max RSS. Against the recorded Python TypeScript parse/object-materialization baseline above, that is 2.298x faster wall time and 7.049x lower max RSS with conservative CI-style ceilings.
+On 2026-06-19, the full pinned large-repo gate validated exact pinned Next.js `Codebase` handle counts plus compact function-call and Promise-chain counts, confirmed the Python graph stayed blocked, and measured 10.771s wall / 435.2 MB max RSS. Against the recorded Python TypeScript parse/object-materialization baseline above, that is 2.317x faster wall time and 7.123x lower max RSS with conservative CI-style ceilings.
 
 ## Installed-Wheel `uvx` Airflow Evidence
 

rust-rewrite/golden/next.js-v15.0.0-rust-compact-typescript.json

@@ -1,7 +1,7 @@
 {
   "graphs": {
     "dependencies": {
-      "count": 49287,
+      "count": 49285,
       "samples": [
         {
           "reference_count": 2,
@@ -210,7 +210,7 @@
           "target_symbol": ".github/actions/next-integration-stat/src/index.ts:type_alias:Octokit@284"
         }
       ],
-      "sha256": "2b4504dada98ee8d1c67f44555c126db3afb6995cbe9d3a3fd8ef98a90486da3"
+      "sha256": "72d6c7c2ee1ec459b47b109e4bfdef8b935a0446ea0ca5c8c3990e088185da2d"
     },
     "exports": {
       "count": 16027,
@@ -905,7 +905,7 @@
       "sha256": "75176bcfeb2ad12c09a7dfddc5de0f357b3f760e0b16b6cd05c9cf8260113b33"
     },
     "external_references": {
-      "count": 25317,
+      "count": 22427,
       "samples": [
         {
           "import": ".github/actions/needs-triage/src/index.ts:namespace_import:@actions/github:*:github@7",
@@ -1208,7 +1208,7 @@
           "source_symbol": ".github/actions/next-integration-stat/src/index.ts:function:getTestResultDiffBase@10615"
         }
       ],
-      "sha256": "30bdfc999d06a7b426261e850bdbec296db2b5dac2ad482c73fd87f06181579f"
+      "sha256": "911fa8cb79987fc29b94fd6d6fe19b3d188af26dc834a37faaf685690db8a768"
     },
     "files": {
       "count": 13688,
@@ -1969,7 +1969,7 @@
       "sha256": "701c5aeb3c9738bd1051ddb1c7a97afe67422e303b25dc50e4c2d028f739856f"
     },
     "references": {
-      "count": 114465,
+      "count": 113809,
       "samples": [
         {
           "import": null,
@@ -2292,7 +2292,7 @@
           "target_symbol": ".github/actions/next-integration-stat/src/index.ts:function:fetchJobLogsFromWorkflow@2452"
         }
       ],
-      "sha256": "51ecfc2eab6601e947f7f64944f162c63e304c0e00c256c2ac497fc28dd95202"
+      "sha256": "273b2c2b48d09aa7fabdc027ba0bc6f9d2124ce1003ff0b359f31209a00fc4ca"
     },
     "subclass_edges": {
       "count": 160,
@@ -2991,18 +2991,18 @@
   "summary": {
     "bytes": 25421217,
     "classes": 502,
-    "dependencies": 49287,
+    "dependencies": 49285,
     "exports": 16027,
     "external_modules": 13525,
-    "external_references": 25317,
+    "external_references": 22427,
     "files": 13688,
     "files_with_errors": 113,
     "functions": 13497,
     "global_variables": 28742,
     "import_resolutions": 13462,
     "imports": 28210,
     "lines": 634891,
-    "references": 114465,
+    "references": 113809,
     "subclass_edges": 160,
     "symbols": 44871
   }