fix(extraction): a function in an initializer is referenced once, by what the initializer declares (#2743)

PRs #693 and #1802 walk a declaration's initializer under the symbol it declares,
so `const routes = { home }` references `home` from `routes`. The
dispatcher's declaration-wide function-as-value scan then took the same
values again for the enclosing file (or class). Flush-time dedup keys on
the source, so both stayed: a second reference to the same function from
the same spot.

Each initializer walk now records what it walked, and the scan skips it.
PR #2738 added that mechanism for component wrappers, and C#, Scala and Dart
already had their own. The newly covered paths:
- TS/JS module-level constants and variables, owned object literals
  included (#2300), and TS/JS class properties;
- Python assignments, Go var/const specs, Rust const/static items;
- Java fields, and Kotlin properties (walkInitializer takes the property's
  id, so their walk is unchanged).
Mirrored in the kernel: tsjs, python, go, rustlang, java and kotlin.

References no initializer walk takes are unchanged: a module-level call
argument `register(fn)`, C-family tables, Rust's `static CB: fn() = handler`
(the declaration's own value, which the scan still reads).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This commit is contained in:
Colby Mchenry
2026-10-21 11:41:45 +00:00
committed by GitHub
co-authored by Claude Opus 5.5
parent ae0fc631e0
commit 51be28d15b
17 changed files with 233 additions and 79 deletions
+1
View File
@@ -269,6 +269,7 @@ and adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
- In TypeScript, a type written with a qualifier, like `React.ReactNode`, `Y.Doc`, `TChat.ChatMessage` or `EQueryType.CLICKHOUSE`, now links to the member of what the qualifier names: a type from a package, like React's, no longer links to a class or component of your own that shares its name, and a member of an enum or a namespace your project declares links to that member instead of to another enum's or a same-named component. Re-index TypeScript projects after upgrading.
- In JavaScript and TypeScript, an import written from the app's root with a leading slash, like bigbluebutton's `import Icon from '/imports/ui/components/common/icon/component'`, now links to the file it names when the app's `tsconfig.json` or `jsconfig.json` maps it, as `"paths": { "/*": ["*"] }` does. Before, such an import was taken for a package, so the module looked unused and every name it imports linked by name instead, often to the wrong one of several components of the same name, or to nothing. Re-index projects that import this way after upgrading.
- In JavaScript and TypeScript, a call whose callee is not a name, like an immediately invoked function `(() => { … })()`, a curried call such as `connect(mapState)(View)`, `handlers[type]()` or `(0, fn)()`, no longer stores the callee's whole source text as the name of a reference that nothing in the project can answer. A single immediately invoked function could leave its entire body in the index that way. The calls inside such a callee, and the inner call of a curried one, are still recorded. A call through parentheses, a cast or a non-null assertion, like `(fn)()`, `(fn as F)()` or `fn!()`, now links to `fn` exactly as `fn()` does, so a parameter named `fn` no longer links to a same-named property elsewhere in the project. In TypeScript, a call with type arguments behind an operator, like `!settings.get<boolean>('X')` or `void apiRequest<Status>()`, now links to what it calls, as `await load<T>()` already did.
- In JavaScript, TypeScript, Python, Go, Rust, Java and Kotlin, a function named in the value of a top-level constant or variable, or of a class field, like `const routes = { home }`, `HANDLERS = {"home": home}`, a Go `map[string]func()` table, a Rust `static` list of functions, `static handlers = { click: onClick }`, a Java `List.of(Registry::home)` field or a Kotlin `listOf(::home)` property, is now referenced once, by the constant, variable, field or property that holds it. The file or class around it no longer also shows up as a second reference from the same place, so a function's callers and impact list each table once.
## [1.6.2] - 2026-10-03
@@ -31,6 +31,7 @@ public class TortureService extends BaseService implements Runnable, AutoCloseab
}
};
private final Runnable fieldMethodRef = TortureService::compute;
private static final List<Runnable> FIELD_TABLE = List.of(TortureService::compute);
/** Ctor javadoc. */
public TortureService(List<String> names) {
@@ -14,6 +14,7 @@ const arrow = (x) => legacyHelper(x, 1);
class Widget extends EventEmitter {
static registry = new Map();
static hooks = { tick: legacyHelper };
#privateField = 1;
label = 'w';
onTick = () => {
@@ -78,6 +79,7 @@ export default {
// leaked to the FILE node; a non-exported object literal was skipped outright.
const eagerConfig = loadConfig();
const handlerMap = { onSave: () => persist(eagerConfig), onLoad: loadConfig() };
const registry = { legacyHelper, run: arrow, nested: [legacyHelper] };
const lazyList = [() => persist(eagerConfig)];
// --- CommonJS export assignments (#1675) -----------------------------------
exports.getItems = async (req, res) => { res.json(await findItems()); };
@@ -52,6 +52,7 @@ val withGetter: Int
get() = 42
val initLambda: () -> Unit = { caller() }
val initSam = Runnable { caller() }
val handlerList = listOf(::caller)
val initObject = object : Runnable {
override fun run() {
caller()
@@ -60,6 +61,7 @@ val initObject = object : Runnable {
class WidgetK(val size: Int, private var name: String = defaultName()) {
val area: Int = size * size
val hooks = listOf(::caller)
var label: String? = null
val computed: Int
get() = size * 2
@@ -305,6 +305,8 @@ routes![top_level_h];
const INIT_CONST: usize = compute_len();
static INIT_LAZY: Lazy<Cfg> = Lazy::new(|| build_cfg());
static INIT_ALIAS: fn() = free_fn;
// A table the walk above takes for its symbol, and the scan no longer for the file.
static INIT_TABLE: &[fn()] = &[handler, handler2];
pub union Reg {
pub raw: u32,
@@ -23,6 +23,7 @@ var oldVar = 'x';
@scoped.Registry<Config>('name')
export abstract class BaseService extends EventTarget implements Disposable, Serializable {
static instances = 0;
static handlers = { refresh: helperFn, table: CONFIG_TABLE };
private readonly cache: Map<string, Config> = new Map();
public fonts: FontConfig;
count = 0;
@@ -189,6 +190,7 @@ export const actionsMap = {
update(input: UpdateInput) {
return persistExisting(input);
},
fallback: helperFn,
};
// Zustand-style store through middleware wrappers.
+70 -3
View File
@@ -779,12 +779,11 @@ describe('Function-as-value capture (#756)', () => {
// The DRF wiring: get_serializer_class → the imported serializer class,
// via `return` — the issue's headline gap. The module-level registry
// dict rides BOTH the assigned name (the initializer walk, #693) and the
// file node (the dispatcher's own scan, which runs either way).
// dict rides the assigned name (the initializer walk, #693), and not the
// file node as well: the dispatcher's scan skips what that walk took.
expect(sourceNames(cg, fnRefEdgesInto(cg, 'OrgSerializerFull'))).toEqual([
'SERIALIZER_REGISTRY',
'get_serializer_class',
'views.py',
]);
// Second branch return + a module-level call argument.
expect(sourceNames(cg, fnRefEdgesInto(cg, 'OrgSerializerBrief'))).toEqual([
@@ -1581,4 +1580,72 @@ func Unknown(obj interface{}) { Submit(obj.Fetch) }
cg.destroy();
}
});
// An initializer walked under the symbol it declares (#693) takes its
// function values for that symbol. The declaration-wide scan took each one
// again for the enclosing file or class: a second reference to the same
// function from the same spot (TS/JS constants and class fields, Python,
// Go, Rust, Java fields, Kotlin properties). Each function here is
// referenced only from an initializer, so a second source shows.
it('INITIALIZER WALKS: a function in an initializer is referenced once, by what it declares', async () => {
tmpDir = fs.mkdtempSync(path.join(os.tmpdir(), 'cg-fnref-init-'));
fs.writeFileSync(
path.join(tmpDir, 'app.ts'),
[
'export function tsHome() {}',
'export function tsAbout() {}',
'export function tsClick() {}',
'export function tsBoot() {}',
'function register(fn: () => void) {}',
'const routes = { tsHome, about: tsAbout, list: [tsHome] };',
'const api = { load() {}, tsAbout };',
'export class Widget {',
' static handlers = { click: tsClick };',
'}',
'register(tsBoot);',
].join('\n')
);
fs.writeFileSync(
path.join(tmpDir, 'views.py'),
['def py_home():', ' pass', '', 'HANDLERS = {"home": py_home}', 'ROUTES = [py_home]'].join('\n')
);
fs.writeFileSync(
path.join(tmpDir, 'handlers.go'),
['package main', '', 'func goHome() {}', '', 'var handlers = map[string]func(){"home": goHome}'].join('\n')
);
fs.writeFileSync(
path.join(tmpDir, 'lib.rs'),
['fn rust_home() {}', 'static HANDLERS: &[fn()] = &[rust_home];'].join('\n')
);
fs.writeFileSync(
path.join(tmpDir, 'Registry.java'),
[
'class Registry {',
' static void javaHome() {}',
' static final java.util.List<Runnable> TABLE = java.util.List.of(Registry::javaHome);',
'}',
].join('\n')
);
fs.writeFileSync(
path.join(tmpDir, 'handlers.kt'),
['fun ktHome() {}', 'val ktHandlers = listOf(::ktHome)', 'class KtWidget {', ' val hooks = listOf(::ktHome)', '}'].join('\n')
);
const cg = CodeGraph.initSync(tmpDir);
try {
await cg.indexAll();
expect(sourceNames(cg, fnRefEdgesInto(cg, 'tsHome'))).toEqual(['routes']);
expect(sourceNames(cg, fnRefEdgesInto(cg, 'tsAbout'))).toEqual(['api', 'routes']);
expect(sourceNames(cg, fnRefEdgesInto(cg, 'tsClick'))).toEqual(['handlers']);
// A call argument at module scope is the file's, as before.
expect(sourceNames(cg, fnRefEdgesInto(cg, 'tsBoot'))).toEqual(['app.ts']);
expect(sourceNames(cg, fnRefEdgesInto(cg, 'py_home'))).toEqual(['HANDLERS', 'ROUTES']);
expect(sourceNames(cg, fnRefEdgesInto(cg, 'goHome'))).toEqual(['handlers']);
expect(sourceNames(cg, fnRefEdgesInto(cg, 'rust_home'))).toEqual(['HANDLERS']);
expect(sourceNames(cg, fnRefEdgesInto(cg, 'javaHome'))).toEqual(['TABLE']);
expect(sourceNames(cg, fnRefEdgesInto(cg, 'ktHome'))).toEqual(['hooks', 'ktHandlers']);
} finally {
cg.destroy();
}
});
});
+16 -5
View File
@@ -428,8 +428,10 @@ impl<'t> Walker<'t> {
} else if matches!(kind, "var_declaration" | "short_var_declaration" | "const_declaration")
&& !self.inside_class_like()
{
self.extract_variable(node);
self.scan_fn_ref_subtree(node, 0);
// The initializers extract_variable walked under their specs'
// symbols took their own candidates; the scan covers the rest.
let walked = self.extract_variable(node);
self.scan_fn_ref_subtree(node, 0, &walked);
skip_children = true;
} else if kind == "import_declaration" {
self.extract_import(node);
@@ -678,7 +680,10 @@ impl<'t> Walker<'t> {
}
/// extractVariable's Go branch: var/const specs + short_var_declaration.
fn extract_variable(&mut self, node: Node<'t>) {
/// Returns the initializers it walked under their specs' symbols, for the
/// declaration scan to skip.
fn extract_variable(&mut self, node: Node<'t>) -> Vec<usize> {
let mut walked = Vec::new();
let docstring = self.docstring_of(node);
let is_const_decl = node.kind() == "const_declaration";
@@ -726,6 +731,7 @@ impl<'t> Walker<'t> {
self.stack.push(Scope { row, kind: "variable", name });
self.visit_function_body(value_field);
self.stack.pop();
walked.push(value_field.id());
} else {
self.visit_function_body(value_field);
}
@@ -756,6 +762,7 @@ impl<'t> Walker<'t> {
}
}
}
walked
}
/// extractImport's Go branch: one import node + ref per import_spec.
@@ -1241,11 +1248,15 @@ impl<'t> Walker<'t> {
}
}
fn scan_fn_ref_subtree(&mut self, node: Node<'t>, depth: u32) {
fn scan_fn_ref_subtree(&mut self, node: Node<'t>, depth: u32, walked: &[usize]) {
stack_guard!();
if depth > 12 {
return;
}
// Subtrees the body walker has already been through.
if walked.contains(&node.id()) {
return;
}
if depth > 0
&& matches!(
node.kind(),
@@ -1258,7 +1269,7 @@ impl<'t> Walker<'t> {
self.maybe_capture_fn_refs(node);
for i in 0..node.named_child_count() {
if let Some(c) = node.named_child(i) {
self.scan_fn_ref_subtree(c, depth + 1);
self.scan_fn_ref_subtree(c, depth + 1, walked);
}
}
}
+18 -7
View File
@@ -507,12 +507,14 @@ impl<'t> Walker<'t> {
self.extract_enum(node);
skip_children = true;
} else if kind == "field_declaration" && self.inside_class_like() {
self.extract_field(node);
self.scan_fn_ref_subtree(node, 0);
// The initializers extract_field walked under their fields took
// their own candidates; the scan covers the rest.
let walked = self.extract_field(node);
self.scan_fn_ref_subtree(node, 0, &walked);
skip_children = true;
} else if kind == "local_variable_declaration" && !self.inside_class_like() {
self.extract_variable(node);
self.scan_fn_ref_subtree(node, 0);
self.scan_fn_ref_subtree(node, 0, &[]);
skip_children = true;
} else if kind == "import_declaration" {
self.extract_import(node);
@@ -726,8 +728,11 @@ impl<'t> Walker<'t> {
self.stack.pop();
}
/// extractField — each declarator becomes a field/constant node.
fn extract_field(&mut self, node: Node<'t>) {
/// extractField — each declarator becomes a field/constant node. Returns
/// the initializers walked under their fields, for the declaration scan to
/// skip.
fn extract_field(&mut self, node: Node<'t>) -> Vec<usize> {
let mut walked = Vec::new();
let docstring = preceding_docstring(node, self.src);
let visibility = self.visibility_of(node);
let is_static = Some(self.is_static(node));
@@ -786,6 +791,7 @@ impl<'t> Walker<'t> {
self.stack.push(Scope { row, kind: field_kind, name: name.clone() });
self.visit_function_body(value);
self.stack.pop();
walked.push(value.id());
}
}
}
@@ -805,6 +811,7 @@ impl<'t> Walker<'t> {
);
}
}
walked
}
/// extractVariable's generic fallback (top-level locals — rare in Java).
@@ -1230,11 +1237,15 @@ impl<'t> Walker<'t> {
}
}
fn scan_fn_ref_subtree(&mut self, node: Node<'t>, depth: u32) {
fn scan_fn_ref_subtree(&mut self, node: Node<'t>, depth: u32, walked: &[usize]) {
stack_guard!();
if depth > 12 {
return;
}
// Subtrees the body walker has already been through.
if walked.contains(&node.id()) {
return;
}
// (functionTypes is empty for Java; lambda_expression halts the scan)
if depth > 0 && matches!(node.kind(), "lambda_literal" | "lambda_expression") {
return;
@@ -1242,7 +1253,7 @@ impl<'t> Walker<'t> {
self.maybe_capture_fn_refs(node);
for i in 0..node.named_child_count() {
if let Some(c) = node.named_child(i) {
self.scan_fn_ref_subtree(c, depth + 1);
self.scan_fn_ref_subtree(c, depth + 1, walked);
}
}
}
+24 -13
View File
@@ -733,7 +733,10 @@ impl<'t> Walker<'t> {
})
}
fn try_visit_hook(&mut self, node: Node<'t>) -> bool {
/// The Kotlin visitNode hook: `Some` when it consumed the node, holding the
/// initializers it walked as such (`walkInitializer`), which the caller's
/// fn-ref scan of the node skips.
fn try_visit_hook(&mut self, node: Node<'t>) -> Option<Vec<usize>> {
// An own-line accessor already walked by its owning property below. The
// ownership test re-derives the property's kind rather than remembering
// it: a destructured or local declaration mints no node, so its
@@ -741,10 +744,10 @@ impl<'t> Walker<'t> {
if matches!(node.kind(), "getter" | "setter") {
return accessor_owner(node)
.and_then(|owner| self.property_kind(owner))
.is_some();
.map(|_| Vec::new());
}
if node.kind() != "property_declaration" {
return false;
return None;
}
let var_decl = (0..node.named_child_count())
.filter_map(|i| node.named_child(i))
@@ -764,11 +767,11 @@ impl<'t> Walker<'t> {
for init in property_initializers(node) {
self.visit_function_body(init);
}
return true;
return Some(Vec::new());
};
let name = self.text(name_node).to_string();
if name.is_empty() {
return false;
return None;
}
let Some(kind) = self.property_kind(node) else {
// A local — no node is minted, but the initializer is still code.
@@ -778,7 +781,7 @@ impl<'t> Walker<'t> {
for init in property_initializers(node) {
self.visit_function_body(init);
}
return true;
return Some(Vec::new());
};
// The `type`-field signature read is dead (zero fields) → signature
// undefined; NO docstring/visibility/isStatic — the modifiers merge in
@@ -790,26 +793,30 @@ impl<'t> Walker<'t> {
// the idiomatic Android callback field) contributed NO call edge at all.
// The property also OWNS any accessor written on its own line, which the
// grammar makes a following SIBLING rather than a child; those bodies
// used to attribute to the enclosing class.
// used to attribute to the enclosing class. The initializers are
// walked as such: the fn-ref scan skips them, or it would take each
// function value again for the enclosing file or class.
let mut walked = Vec::new();
if let Some(row) = row {
self.stack.push(Scope { row, kind, name: name.clone() });
for init in property_initializers(node) {
self.visit_function_body(init);
walked.push(init.id());
}
for acc in following_accessors(node) {
self.visit_function_body(acc);
}
self.stack.pop();
}
true
Some(walked)
}
// --- the dispatcher (visitNode, Kotlin-relevant branches) -----------------------
fn visit_node(&mut self, node: Node<'t>) {
stack_guard!();
if self.try_visit_hook(node) {
self.scan_fn_ref_subtree(node, 0);
if let Some(walked) = self.try_visit_hook(node) {
self.scan_fn_ref_subtree(node, 0, &walked);
return;
}
@@ -856,7 +863,7 @@ impl<'t> Walker<'t> {
// Hook-declined destructuring: extractField/extractVariable both
// find no matching children for kotlin — NOTHING minted, RHS
// invisible; candidates-only scan.
self.scan_fn_ref_subtree(node, 0);
self.scan_fn_ref_subtree(node, 0, &[]);
skip_children = true;
} else if kind == "import_header" {
self.extract_import(node);
@@ -1574,11 +1581,15 @@ impl<'t> Walker<'t> {
});
}
fn scan_fn_ref_subtree(&mut self, node: Node<'t>, depth: u32) {
fn scan_fn_ref_subtree(&mut self, node: Node<'t>, depth: u32, walked: &[usize]) {
stack_guard!();
if depth > 12 {
return;
}
// Subtrees the body walker has already been through.
if walked.contains(&node.id()) {
return;
}
// Halts at functionTypes (function_declaration) + the fixed list —
// lambda_literal halts (no captures inside `by lazy { }` under a
// hook-consumed property).
@@ -1594,7 +1605,7 @@ impl<'t> Walker<'t> {
self.maybe_capture_fn_refs(node);
for i in 0..node.named_child_count() {
if let Some(c) = node.named_child(i) {
self.scan_fn_ref_subtree(c, depth + 1);
self.scan_fn_ref_subtree(c, depth + 1, walked);
}
}
}
+16 -5
View File
@@ -431,8 +431,10 @@ impl<'t> Walker<'t> {
self.extract_class(node);
skip_children = true;
} else if kind == "assignment" && !self.inside_class_like() {
self.extract_variable(node);
self.scan_fn_ref_subtree(node, 0);
// The right-hand side extract_variable walked under the assigned
// name took its own candidates; the scan covers the rest.
let walked = self.extract_variable(node);
self.scan_fn_ref_subtree(node, 0, &walked);
skip_children = true;
} else if kind == "import_statement" || kind == "import_from_statement" {
self.extract_import(node);
@@ -564,7 +566,9 @@ impl<'t> Walker<'t> {
}
/// extractVariable's python branch: `left = right` at module scope.
fn extract_variable(&mut self, node: Node<'t>) {
/// Returns the right-hand side when it was walked under the assigned
/// name, for the declaration scan to skip.
fn extract_variable(&mut self, node: Node<'t>) -> Vec<usize> {
let docstring = preceding_docstring(node, self.src);
let left = node.child_by_field_name("left").or_else(|| node.named_child(0));
let right = node.child_by_field_name("right").or_else(|| node.named_child(1));
@@ -589,16 +593,19 @@ impl<'t> Walker<'t> {
// module-level `app = FastAPI()` / `handler = lambda: run()` dropped
// every call on the right-hand side. A tuple target mints no symbol, so
// its RHS is walked at the enclosing scope rather than lost.
let mut walked = Vec::new();
if let Some(right) = right {
match assigned {
Some((row, name)) => {
self.stack.push(Scope { row, kind: "variable", name });
self.visit_function_body(right);
self.stack.pop();
walked.push(right.id());
}
None => self.visit_function_body(right),
}
}
walked
}
fn extract_import(&mut self, node: Node<'t>) {
@@ -955,11 +962,15 @@ impl<'t> Walker<'t> {
}
}
fn scan_fn_ref_subtree(&mut self, node: Node<'t>, depth: u32) {
fn scan_fn_ref_subtree(&mut self, node: Node<'t>, depth: u32, walked: &[usize]) {
stack_guard!();
if depth > 12 {
return;
}
// Subtrees the body walker has already been through.
if walked.contains(&node.id()) {
return;
}
// Halts at functionTypes ∪ the fixed arrow/lambda list — python's
// `lambda` kind is NOT in that list (mirrored).
if depth > 0
@@ -974,7 +985,7 @@ impl<'t> Walker<'t> {
self.maybe_capture_fn_refs(node);
for i in 0..node.named_child_count() {
if let Some(c) = node.named_child(i) {
self.scan_fn_ref_subtree(c, depth + 1);
self.scan_fn_ref_subtree(c, depth + 1, walked);
}
}
}
+17 -7
View File
@@ -500,8 +500,10 @@ impl<'t> Walker<'t> {
// Inside a class-like scope the gate fails and the else-ladder
// falls through with children VISITED — a trait const's value
// expression emits calls refs from the trait node.
self.extract_variable(node);
self.scan_fn_ref_subtree(node, 0);
// The initializer extract_variable walked under the declared
// symbol took its own candidates; the scan covers the rest.
let walked = self.extract_variable(node);
self.scan_fn_ref_subtree(node, 0, &walked);
skip_children = true;
} else if kind == "use_declaration" {
self.extract_import(node);
@@ -721,9 +723,10 @@ impl<'t> Walker<'t> {
/// extractVariable's generic fallback: kind is ALWAYS `variable` (no
/// isConst hook), every direct `identifier` child mints a node positioned
/// at the CHILD, docstring shared, isExported present-false, no signature,
/// and the initializer value is never body-walked.
fn extract_variable(&mut self, node: Node<'t>) {
/// at the CHILD, docstring shared, isExported present-false, no signature.
/// The initializer is walked under the declared symbol (#693), and returned
/// then for the declaration scan to skip.
fn extract_variable(&mut self, node: Node<'t>) -> Vec<usize> {
let docstring = preceding_docstring(node, self.src);
let name_field = node.child_by_field_name("name");
// An impl's associated const is its type's (`Limits::MAX`), as its methods are.
@@ -759,16 +762,19 @@ impl<'t> Walker<'t> {
// `static REGISTRY: Lazy<T> = Lazy::new(|| build())` dropped every call
// inside the initializer, so a handler table or a lazily-built
// singleton linked to nothing.
let mut walked = Vec::new();
if let Some(value) = node.child_by_field_name("value") {
match declared {
Some((row, name)) => {
self.stack.push(Scope { row, kind: "variable", name });
self.visit_function_body(value);
self.stack.pop();
walked.push(value.id());
}
None => self.visit_function_body(value),
}
}
walked
}
/// extractImport via the rust hook: import node named by the ROOT module +
@@ -1430,11 +1436,15 @@ impl<'t> Walker<'t> {
}
}
fn scan_fn_ref_subtree(&mut self, node: Node<'t>, depth: u32) {
fn scan_fn_ref_subtree(&mut self, node: Node<'t>, depth: u32, walked: &[usize]) {
stack_guard!();
if depth > 12 {
return;
}
// Subtrees the body walker has already been through.
if walked.contains(&node.id()) {
return;
}
if depth > 0
&& matches!(
node.kind(),
@@ -1447,7 +1457,7 @@ impl<'t> Walker<'t> {
self.maybe_capture_fn_refs(node);
for i in 0..node.named_child_count() {
if let Some(c) = node.named_child(i) {
self.scan_fn_ref_subtree(c, depth + 1);
self.scan_fn_ref_subtree(c, depth + 1, walked);
}
}
}
+7 -3
View File
@@ -530,9 +530,11 @@ impl<'t> Walker<'t> {
}
}
/// Returns the initializers it walked under a component wrapper, whose
/// fn-ref candidates that walk already took (the caller's scan of the
/// declaration skips them). Mirrors TreeSitterExtractor.extractVariable.
/// Returns the initializers it walked under the symbol they declare — a
/// constant's or a variable's (#693), a component wrapper's — whose fn-ref
/// candidates that walk already took for the symbol (the caller's scan of
/// the declaration skips them; it would take each again for the file).
/// Mirrors TreeSitterExtractor.extractVariable.
pub(super) fn extract_variable(&mut self, node: Node<'t>) -> Option<HashSet<usize>> {
let mut walked: Option<HashSet<usize>> = None;
let is_const = self.is_const_decl(node);
@@ -619,6 +621,7 @@ impl<'t> Walker<'t> {
// the ownedObject branch of TreeSitterExtractor.extractVariable.
if let (Some(row), Some(obj)) = (var_row, self.owned_object_value(value)) {
self.extract_owned_object_members(obj, row, kind, &name, true, false);
walked.get_or_insert_with(HashSet::new).insert(obj.id());
continue;
}
@@ -678,6 +681,7 @@ impl<'t> Walker<'t> {
self.stack.push(Scope { row, kind, name: name.clone() });
self.visit_function_body(v);
self.stack.pop();
walked.get_or_insert_with(HashSet::new).insert(v.id());
}
None => self.visit_function_body(v),
}
+8 -3
View File
@@ -700,12 +700,16 @@ impl<'t> Walker<'t> {
{
if classify_ts_class_member(node) == Member::Property {
let prop = self.extract_property(node);
// The initializer walked under the property took its own
// candidates; the scan covers the rest of the field.
let mut walked: Option<HashSet<usize>> = None;
if let (Some((row, name)), Some(value)) = (prop, node.child_by_field_name("value")) {
self.stack.push(Scope { row, kind: "property", name });
self.visit_function_body(value);
self.stack.pop();
walked = Some(HashSet::from([value.id()]));
}
self.scan_fn_ref_subtree(node, 0, None);
self.scan_fn_ref_subtree(node, 0, walked.as_ref());
} else {
self.extract_method(node);
}
@@ -719,8 +723,9 @@ impl<'t> Walker<'t> {
} else if self.variant.is_ts() && kind == "type_alias_declaration" {
skip_children = self.extract_type_alias(node);
} else if is_variable_type(kind) && !self.inside_class_like() {
// A component wrapper's initializer extract_variable walked under
// the component captured its own candidates; the scan skips it.
// An initializer extract_variable walked under the symbol it
// declares (a constant's, a component wrapper's) captured its own
// candidates; the scan skips it.
let walked = self.extract_variable(node);
self.scan_fn_ref_subtree(node, 0, walked.as_ref());
skip_children = true;
+6 -5
View File
@@ -293,12 +293,13 @@ export const kotlinExtractor: LanguageExtractor = {
// grammar makes a following SIBLING rather than a child; those bodies used
// to attribute to the enclosing class. Consumed here so the accessor
// branch below can skip them without any cross-node state.
const inits = created
? [...kotlinPropertyInitializers(node), ...kotlinFollowingAccessors(node)]
: [];
if (created && inits.length > 0) {
// The initializers are walked as such, so the dispatcher's
// function-as-value scan of the property skips them: it would take each
// function value again for the enclosing file or class.
if (created) {
ctx.pushScope(created.id);
for (const init of inits) ctx.visitFunctionBody(init, created.id);
for (const init of kotlinPropertyInitializers(node)) ctx.walkInitializer(init, created.id);
for (const accessor of kotlinFollowingAccessors(node)) ctx.visitFunctionBody(accessor, created.id);
ctx.popScope();
}
return true;
+5 -4
View File
@@ -56,11 +56,12 @@ export interface ExtractorContext {
visitFunctionBody(body: SyntaxNode, functionId: string): void;
/**
* Walk a declaration's initializer as code the scope on top of the stack
* runs, the way a function body is walked. When the hook then reports the
* node handled, the function-as-value scan the dispatcher runs over it skips
* what was walked here: the walk captured those candidates already.
* runs, the way a function body is walked (`functionId` as for
* visitFunctionBody). When the hook then reports the node handled, the
* function-as-value scan the dispatcher runs over it skips what was walked
* here: the walk captured those candidates already.
*/
walkInitializer(node: SyntaxNode): void;
walkInitializer(node: SyntaxNode, functionId?: string): void;
/** Add an unresolved reference */
addUnresolvedReference(ref: UnresolvedReference): void;
/** Push a node ID onto the scope stack (for containment/qualified name building) */
+36 -24
View File
@@ -1521,9 +1521,12 @@ export class TreeSitterExtractor {
this.nodeStack.pop();
}
// A field initializer can also register callbacks
// (`static handlers = { click: onClick }`) — scan it for
// function-as-value candidates (capture-only, halts at functions).
this.scanFnRefSubtree(node, 0);
// (`static handlers = { click: onClick }`). The walk above took those
// for the property, so the scan for function-as-value candidates
// (capture-only, halts at functions) skips the initializer it walked
// and covers the rest of the field: the class would keep a second
// reference to each function from the same spot.
this.scanFnRefSubtree(node, 0, propNode && valueNode ? new Set([valueNode.id]) : undefined);
skipChildren = true;
} else {
this.extractMethod(node);
@@ -1616,16 +1619,18 @@ export class TreeSitterExtractor {
const definedType = declaredType && isClassLikeDefinition(declaredType) ? declaredType : null;
if (definedType) this.visitNode(definedType);
const walked = this.extractVariable(node);
// extractVariable doesn't walk every initializer shape (object literals
// are deliberately skipped; Python/Ruby don't walk at all), so scan the
// declaration subtree for function-as-value candidates — `const routes =
// { home: renderHome }`, `handlers = {"recv": target_cb}`. The scan halts
// at nested function definitions (their bodies are walked — and
// attributed — separately) and flush-time dedup absorbs any overlap with
// initializers extractVariable DOES walk. A type walked above, and a
// component wrapper's initializer extractVariable walked under the
// component (which must not hand the component it wraps to the file),
// captured their own.
// extractVariable doesn't walk every initializer (a store's members are
// extracted one by one, and only TS/JS, Python, Go and Rust walk one at
// all), so scan the declaration subtree for function-as-value candidates
// — `const routes = { home: renderHome }`, `handlers = {"recv":
// target_cb}`. The scan halts at nested function definitions (their
// bodies are walked — and attributed — separately). It skips what was
// walked already: a type walked above, and every initializer
// extractVariable walked under the symbol it declares — a constant's, a
// component wrapper's (which must not hand the component it wraps to the
// file). Those captured their own, and flush-time dedup keys on the
// source, so the file would keep a second reference to each function
// from the same spot.
this.scanFnRefSubtree(node, 0, definedType ? new Set([definedType.id]) : walked);
skipChildren = true; // extractVariable handles children
}
@@ -2016,9 +2021,9 @@ export class TreeSitterExtractor {
createNode: (kind, name, node, extra) => self.createNode(kind, name, node, extra),
visitNode: (node) => self.visitNode(node),
visitFunctionBody: (body, functionId) => self.visitFunctionBody(body, functionId),
walkInitializer: (node) => {
walkInitializer: (node, functionId = '') => {
(ctx.walked ??= new Set()).add(node.id);
self.visitFunctionBody(node, '');
self.visitFunctionBody(node, functionId);
},
addUnresolvedReference: (ref) => self.unresolvedReferences.push(ref),
pushScope: (nodeId) => self.nodeStack.push(nodeId),
@@ -2869,8 +2874,8 @@ export class TreeSitterExtractor {
/**
* Extract a class field declaration (e.g. Java field_declaration, C# field_declaration).
* Extracts each declarator as a 'field' kind node inside the owning class.
* Returns the C# declarators it walked, for the function-as-value scan to
* skip.
* Returns the initializers it walked (a C# declarator, a Java `value`), for
* the function-as-value scan to skip.
*/
private extractField(node: SyntaxNode): Set<number> {
const walked = new Set<number>();
@@ -2988,11 +2993,11 @@ export class TreeSitterExtractor {
decl.namedChild(decl.namedChildCount - 1),
varDecl ? getChildByField(varDecl, 'type') : null,
);
// Its function-as-value candidates are the field's, captured
// here once. (VB.NET captures none; Java's scan still takes
// its initializers for the class as well.)
walked.add(decl.id);
}
// Its function-as-value candidates are the field's, captured here
// once (VB.NET captures none): the scan would take Java's again for
// the class.
walked.add(valueNode.id);
this.visitFunctionBody(valueNode, fieldNode.id);
this.nodeStack.pop();
}
@@ -3584,9 +3589,11 @@ export class TreeSitterExtractor {
* Extracts top-level and module-level variable declarations.
* Captures the variable name and first 100 chars of initializer in signature for searchability.
*
* Returns the initializers it walked under a component wrapper, whose
* function-as-value candidates that walk already took (the caller's scan of
* the declaration skips them).
* Returns the initializers it walked under the symbol they declare — a
* constant's or a variable's (#693), a component wrapper's — whose
* function-as-value candidates that walk already took for the symbol. The
* caller's scan of the declaration skips them: it would take each again for
* the file, as a second reference to the same function from the same spot.
*/
private extractVariable(node: SyntaxNode): Set<number> | undefined {
if (!this.extractor) return undefined;
@@ -3686,6 +3693,7 @@ export class TreeSitterExtractor {
const ownedObject = varNode ? this.ownedObjectValue(valueNode) : null;
if (varNode && ownedObject) {
this.extractOwnedObjectMembers(ownedObject, varNode.id, true, false);
(walked ??= new Set()).add(ownedObject.id);
continue;
}
@@ -3772,6 +3780,7 @@ export class TreeSitterExtractor {
if (varNode) this.nodeStack.push(varNode.id);
this.visitFunctionBody(valueNode, '');
if (varNode) this.nodeStack.pop();
if (varNode) (walked ??= new Set()).add(valueNode.id);
}
if (extractObjectMethods && objectOfFns) {
@@ -3820,6 +3829,7 @@ export class TreeSitterExtractor {
if (assigned) this.nodeStack.push(assigned.id);
this.visitFunctionBody(right, '');
if (assigned) this.nodeStack.pop();
if (assigned) (walked ??= new Set()).add(right.id);
}
} else if (this.language === 'go') {
// Go: var_declaration, short_var_declaration, const_declaration
@@ -3860,6 +3870,7 @@ export class TreeSitterExtractor {
if (varNode) this.nodeStack.push(varNode.id);
this.visitFunctionBody(valueField, varNode?.id ?? '');
if (varNode) this.nodeStack.pop();
if (varNode) (walked ??= new Set()).add(valueField.id);
}
}
@@ -4016,6 +4027,7 @@ export class TreeSitterExtractor {
if (declared) this.nodeStack.push(declared.id);
this.visitFunctionBody(valueNode, '');
if (declared) this.nodeStack.pop();
if (declared) (walked ??= new Set()).add(valueNode.id);
}
}
}