fix(resolve): find the token that starts at a point, not the one ending there

CHANGELOG.md

@@ -99,6 +99,7 @@ and adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
 ### Fixes
 
+- **Code written without a space before a name resolves like spaced code.** Where one token ends exactly where the next begins, the resolver read the syntax at that point as the earlier token: in `x=1;local function helper()` it saw the `;`, so the Lua local read as global and calls from other files linked to it, and in `if ($x instanceof Foo)$x->bar();` it saw the `)`, so the PHP guard did not type `$x`. It now reads the token that starts there, the same way the TypeScript side already did. On json, jq and CPython the graph is unchanged.
 - **React Router's own `flatRoutes() satisfies RouteConfig` template now gets file routes, and RSC route modules link to their `ServerComponent`.** An `app/routes.ts` written as `export default flatRoutes() satisfies RouteConfig;` (or `as RouteConfig`), the form React Router's templates use, was taken for custom configuration, so the app got no file routes at all. In React Router's RSC framework mode a route module exports `ServerComponent` instead of a default component; such a page now gets its route and links to that component, both for files under `app/routes/` and for modules listed in `app/routes.ts`. A module that exports both keeps its default component.
 - **An inherited method no longer links to a same-named class in another package.** When a call on a `Child` reached `m()` through `Child extends Base`, the lookup past the first parent matched any class called `Base` in the project, so with an unrelated `Base` in another package the call could land on that package's `Base.m()`, or skip a grandparent's `m()` for it. The lookup now follows the parent the declaration names, for Java, Kotlin, C#, Scala and Swift, where a Java or Kotlin receiver comes from a factory call, or a C#, Scala or Swift receiver is typed. In Scala, a same-named object or a package block in the same file is told apart too; in Swift the other `Base` is one declared in another file, since modules are not modeled. Dart, TypeScript, Python and C++ are unchanged: their wrong links did not reproduce, or Dart's member lookup matches parents by name elsewhere.
 - **Python calls on a local built through a module, or through one of its fields, now resolve.** `bp = pkg.Blueprint()`, `c.Blueprint()` and `core.Blueprint()` type the local like a class imported by name, and `app.config.from_mapping()` resolves on the class the field holds, as `self.config.from_mapping()` already did. On Flask and requests this adds 98 and 103 call edges (almost all from tests, none removed); a 30-edge sample was all correct.

__tests__/kernel-point-lookup-boundary.test.ts

@@ -0,0 +1,55 @@
+/**
+ * The kernel finds the syntax node at a reference or declaration's start
+ * position. Where the previous token ends exactly there (`x=1;local function`,
+ * `instanceof Foo)$x->bar()`), the node at that position is the token that
+ * starts there, not the `;` or `)` before it. Finding the earlier token made a
+ * Lua `local` function read as global and hid a PHP `instanceof` guard.
+ * The spaced forms are the controls: they resolved the same way before.
+ */
+import { describe, it, expect, afterAll } from 'vitest';
+import * as fs from 'fs';
+import * as os from 'os';
+import * as path from 'path';
+import { CodeGraph } from '../src';
+
+const roots: string[] = [];
+afterAll(() => {
+  for (const r of roots.splice(0)) fs.rmSync(r, { recursive: true, force: true });
+});
+
+async function callTargets(files: Record<string, string>, from: string): Promise<string[]> {
+  const root = fs.mkdtempSync(path.join(os.tmpdir(), 'cg-point-lookup-'));
+  roots.push(root);
+  for (const [rel, content] of Object.entries(files)) fs.writeFileSync(path.join(root, rel), content);
+  const cg = await CodeGraph.init(root, { index: true });
+  try {
+    return cg
+      .getOutgoingEdgesFrom(cg.getNodesInFile(from).map((n) => n.id), ['calls'])
+      .map((e) => cg.getNode(e.target)!.qualifiedName)
+      .sort();
+  } finally {
+    cg.close();
+  }
+}
+
+describe('point lookup at a token boundary', () => {
+  it('reads a Lua local function written right after `;` as local', async () => {
+    const files = {
+      'a.lua': 'x=1;local function helper() return 1 end\nlocal y = 2; local function spaced() return 2 end\nreturn x\n',
+      'b.lua': 'local function main()\n  helper()\n  spaced()\nend\nreturn main\n',
+    };
+    expect(await callTargets(files, 'b.lua')).toEqual([]);
+  });
+
+  it('applies a PHP instanceof guard to a call written right after `)`', async () => {
+    const files = {
+      'a.php': `<?php
+class Foo { public function bar() { return 1; } }
+class Qux { public function bar() { return 2; } }
+function tight($x) { if ($x instanceof Foo)$x->bar(); }
+function spaced($x) { if ($x instanceof Foo) $x->bar(); }
+`,
+    };
+    expect(await callTargets(files, 'a.php')).toEqual(['Foo::bar', 'Foo::bar']);
+  });
+});

codegraph-kernel/src/resolve/call_shape.rs

@@ -208,7 +208,7 @@ impl KernelResolver {
         let lines = self.read_file(&r.file_path)?;
         if r.language=="rust" && r.reference_kind=="calls" {
             if let Some(tree)=self.parsed_tree(&lines,r) {
-                let mut node=super::iteration::descendant_for_position(tree.root_node(),lines.text(),((r.line-1).max(0) as usize,r.column.max(0) as usize+1));
+                let mut node=super::iteration::descendant_for_position(tree.root_node(),lines.text(),((r.line-1).max(0) as usize,r.column.max(0) as usize));
                 loop {
                     if node.kind()=="field_expression" {
                         if let (Some(field),Some(value))=(node.child_by_field_name("field"),node.child_by_field_name("value")) {

codegraph-kernel/src/resolve/cpp_aliases.rs

@@ -152,7 +152,7 @@ impl KernelResolver {
                             let Some(tree) = self.parsed_tree(&source, r) else { continue; };
                             for keyword in re!(r"\b(?:using|typedef)\b").find_iter(line) {
                                 let column = line[..keyword.start()].encode_utf16().count();
-                                let mut node = super::iteration::descendant_for_position(tree.root_node(), source.text(), (i, column + 1));
+                                let mut node = super::iteration::descendant_for_position(tree.root_node(), source.text(), (i, column));
                                 let mut code = true;
                                 loop {
                                     if matches!(node.kind(), "comment" | "string_literal" | "raw_string_literal" | "char_literal") { code = false; break; }
@@ -312,7 +312,7 @@ impl KernelResolver {
         let Some(source) = self.read_file(&owner.file_path) else { return Ok(false); };
         let site = r.clone().at(owner);
         let Some(tree) = self.parsed_tree(&source, &site) else { return Ok(false); };
-        let mut node = super::iteration::descendant_for_position(tree.root_node(), source.text(), ((owner.start_line - 1).max(0) as usize, owner.start_column.max(0) as usize + 1));
+        let mut node = super::iteration::descendant_for_position(tree.root_node(), source.text(), ((owner.start_line - 1).max(0) as usize, owner.start_column.max(0) as usize));
         while let Some(parent) = node.parent() {
             if parent.kind() == "template_declaration" { return Ok(true); }
             node = parent;

codegraph-kernel/src/resolve/cpp_range_for.rs

@@ -174,7 +174,7 @@ impl KernelResolver {
         if r.language != "cpp" { return Ok(None); }
         let Some(source) = self.read_file(&r.file_path) else { return Ok(None) };
         let Some(tree) = self.parsed_tree(&source, r) else { return Ok(None) };
-        let mut at = descendant_for_position(tree.root_node(), source.text(), ((r.line - 1).max(0) as usize, r.column.max(0) as usize + 1));
+        let mut at = descendant_for_position(tree.root_node(), source.text(), ((r.line - 1).max(0) as usize, r.column.max(0) as usize));
         while let Some(parent) = at.parent() {
             // A nearer declaration of the name, `{ Gadget* w = g; w->Self(); }` in the body, hides the loop's.
             if cpp_binding_type(parent, at, receiver, &source).is_some() { return Ok(None); }
@@ -270,7 +270,7 @@ impl KernelResolver {
         let Some(source) = self.read_file(&function.file_path) else { return Ok(None) };
         let site = site_at(r, function);
         let Some(tree) = self.parsed_tree(&source, &site) else { return Ok(None) };
-        let mut node = descendant_for_position(tree.root_node(), source.text(), ((function.start_line - 1).max(0) as usize, function.start_column.max(0) as usize + 1));
+        let mut node = descendant_for_position(tree.root_node(), source.text(), ((function.start_line - 1).max(0) as usize, function.start_column.max(0) as usize));
         let declaration = loop {
             match node.kind() {
                 "function_definition" | "field_declaration" | "declaration" => break node,

codegraph-kernel/src/resolve/cpp_receivers.rs

@@ -10,7 +10,7 @@ impl KernelResolver {
         if !re!(r"^[A-Za-z_]\w*$").is_match(method) { return Ok(None); }
         let Some(source) = self.read_file(&r.file_path) else { return Ok(None) };
         let Some(tree) = self.parsed_tree(&source, r) else { return Ok(None) };
-        let mut at = descendant_for_position(tree.root_node(), source.text(), ((r.line - 1).max(0) as usize, r.column.max(0) as usize + 1));
+        let mut at = descendant_for_position(tree.root_node(), source.text(), ((r.line - 1).max(0) as usize, r.column.max(0) as usize));
         while at.kind() != "call_expression" {
             let Some(parent) = at.parent() else { return Ok(None) }; at = parent;
         }
@@ -149,7 +149,7 @@ impl KernelResolver {
         if r.language != "cpp" || r.reference_kind != "calls" || !re!(r"^[A-Za-z_]\w*$").is_match(&r.reference_name) { return Ok(None); }
         let Some(source) = self.read_file(&r.file_path) else { return Ok(None) };
         let Some(tree) = self.parsed_tree(&source, r) else { return Ok(None) };
-        let mut node = descendant_for_position(tree.root_node(), source.text(), ((r.line - 1).max(0) as usize, r.column.max(0) as usize + 1));
+        let mut node = descendant_for_position(tree.root_node(), source.text(), ((r.line - 1).max(0) as usize, r.column.max(0) as usize));
         loop {
             if node.kind() == "call_expression" {
                 if let Some(function) = node.child_by_field_name("function").filter(|n| n.kind() == "field_expression") {
@@ -240,7 +240,7 @@ impl KernelResolver {
         let Some(source) = self.read_file(&method.file_path) else { return Ok(None) };
         let mut site = r.clone().at(method); site.from_node_id = method.id.clone(); site.column = method.start_column;
         let Some(tree) = self.parsed_tree(&source, &site) else { return Ok(None) };
-        let mut node = descendant_for_position(tree.root_node(), source.text(), ((method.start_line - 1).max(0) as usize, method.start_column.max(0) as usize + 1));
+        let mut node = descendant_for_position(tree.root_node(), source.text(), ((method.start_line - 1).max(0) as usize, method.start_column.max(0) as usize));
         loop {
             if matches!(node.kind(), "function_definition" | "field_declaration" | "declaration") {
                 return Ok(node.child_by_field_name("type").map(|ty| source.text()[ty.start_byte()..ty.end_byte()].to_string()));
@@ -279,7 +279,7 @@ impl KernelResolver {
         let Some(source) = self.read_file(&owner.file_path) else { return Ok(None) };
         let mut site = r.clone().at(owner); site.from_node_id = owner.id.clone(); site.column = owner.start_column;
         let Some(tree) = self.cpp_field_tree(&source, &site) else { return Ok(None) };
-        let mut node = descendant_for_position(tree.root_node(), source.text(), ((owner.start_line - 1).max(0) as usize, owner.start_column.max(0) as usize + 1));
+        let mut node = descendant_for_position(tree.root_node(), source.text(), ((owner.start_line - 1).max(0) as usize, owner.start_column.max(0) as usize));
         while !matches!(node.kind(), "class_specifier" | "struct_specifier" | "union_specifier") {
             let Some(parent) = node.parent() else { break }; node = parent;
         }
@@ -313,7 +313,7 @@ impl KernelResolver {
     fn cpp_container_element(&mut self, name: &str, r: &ResolveRefIn) -> Res<Option<Arc<KNode>>> {
         let Some(source) = self.read_file(&r.file_path) else { return Ok(None) };
         if let Some(tree) = self.parsed_tree(&source, r) {
-            let mut at = descendant_for_position(tree.root_node(), source.text(), ((r.line - 1).max(0) as usize, r.column.max(0) as usize + 1));
+            let mut at = descendant_for_position(tree.root_node(), source.text(), ((r.line - 1).max(0) as usize, r.column.max(0) as usize));
             while let Some(parent) = at.parent() {
                 if let Some(raw) = cpp_binding_type(parent, at, name, &source) {
                     let Some(element) = cpp_element_type(&raw) else { return Ok(None) };

codegraph-kernel/src/resolve/cpp_types.rs

@@ -54,7 +54,7 @@ impl KernelResolver {
         let mut directives = Vec::new();
         if let Some(tree) = self.parsed_tree(&lines, r) {
             let mut site = super::iteration::descendant_for_position(tree.root_node(), lines.text(),
-                ((r.line - 1).max(0) as usize, r.column.max(0) as usize + 1));
+                ((r.line - 1).max(0) as usize, r.column.max(0) as usize));
             let before = site.start_byte();
             loop {
                 if matches!(site.kind(), "translation_unit" | "declaration_list" | "field_declaration_list" | "compound_statement") {
@@ -174,7 +174,7 @@ impl KernelResolver {
         let Some(source) = self.read_file(&owner.file_path) else { return Ok(false) };
         let mut site = r.clone().at(owner); site.from_node_id = owner.id.clone(); site.column = owner.start_column;
         let Some(tree) = self.parsed_tree(&source, &site) else { return Ok(false) };
-        let mut node = super::iteration::descendant_for_position(tree.root_node(), source.text(), ((owner.start_line - 1).max(0) as usize, owner.start_column.max(0) as usize + 1));
+        let mut node = super::iteration::descendant_for_position(tree.root_node(), source.text(), ((owner.start_line - 1).max(0) as usize, owner.start_column.max(0) as usize));
         while let Some(parent) = node.parent() {
             if parent.kind() == "template_declaration" {
                 if let Some(parameters) = parent.child_by_field_name("parameters") {

codegraph-kernel/src/resolve/fields.rs

@@ -178,7 +178,7 @@ impl KernelResolver {
             let Some(tree) = self.parsed_tree(&source, &site) else { return Ok(false); };
             let mut declaration = super::iteration::descendant_for_position(
                 tree.root_node(), source.text(),
-                ((target.start_line - 1).max(0) as usize, target.start_column.max(0) as usize + 1),
+                ((target.start_line - 1).max(0) as usize, target.start_column.max(0) as usize),
             );
             while !matches!(declaration.kind(), "type_spec" | "type_alias") {
                 let Some(parent) = declaration.parent() else { return Ok(false); };
@@ -231,7 +231,7 @@ impl KernelResolver {
                 if let Some(tree) = self.parsed_tree(&lines, &site) {
                     let mut declaration = super::iteration::descendant_for_position(
                         tree.root_node(), lines.text(),
-                        ((owner.start_line - 1).max(0) as usize, owner.start_column.max(0) as usize + 1),
+                        ((owner.start_line - 1).max(0) as usize, owner.start_column.max(0) as usize),
                     );
                     while !matches!(declaration.kind(), "type_alias" | "type_spec") {
                         let Some(parent) = declaration.parent() else { break; };

codegraph-kernel/src/resolve/go_asserted.rs

@@ -151,7 +151,7 @@ impl KernelResolver {
         if !text.contains(".(") { return None; }
         let tree = self.parsed_tree(&file, r)?;
         let site = super::iteration::descendant_for_position(tree.root_node(), text,
-            ((r.line - 1).max(0) as usize, r.column.max(0) as usize + 1));
+            ((r.line - 1).max(0) as usize, r.column.max(0) as usize));
         let at = site.start_byte();
         let mut scope = Some(site);
         while let Some(node) = scope {

codegraph-kernel/src/resolve/iteration.rs

@@ -27,15 +27,18 @@ fn before(a: (usize, usize), b: (usize, usize)) -> bool {
 }
 
 /// descendantForPosition (tree.ts): descend into the first child, named or
-/// not, whose span contains the point, until none does.
+/// not, whose span contains the point, until none does. A child must end
+/// after the point, so where one token ends and the next begins (`)` then
+/// `Foo` in `if(ok)Foo f;`) the next one is found: pass a token's start to
+/// find that token.
 pub(super) fn descendant_for_position<'t>(root: TsNode<'t>, text: &str, at: (usize, usize)) -> TsNode<'t> {
     let mut node = root;
     'down: loop {
         let mut cursor = node.walk();
         for c in node.children(&mut cursor) {
             let start = point16(text, c.start_byte(), c.start_position());
             let end = point16(text, c.end_byte(), c.end_position());
-            if before(start, at) && before(at, end) {
+            if before(start, at) && !before(end, at) {
                 node = c;
                 continue 'down;
             }
@@ -266,7 +269,7 @@ impl KernelResolver {
     pub(super) fn kotlin_chain_receiver_call(&mut self, r: &ResolveRefIn) -> Option<String> {
         let file = self.read_file(&r.file_path)?;
         let tree = self.parsed_tree(&file, r)?;
-        let mut node = descendant_for_position(tree.root_node(), file.text(), ((r.line - 1).max(0) as usize, r.column.max(0) as usize + 1));
+        let mut node = descendant_for_position(tree.root_node(), file.text(), ((r.line - 1).max(0) as usize, r.column.max(0) as usize));
         if node.kind() != "simple_identifier" { return None; }
         while let Some(parent) = node.parent() {
             if parent.kind() == "navigation_expression" {
@@ -753,6 +756,27 @@ impl KernelResolver {
 mod tests {
     use super::kotlin_takes_trailing_lambda as trailing;
 
+    /// The node `descendant_for_position` finds at (0, `column`) of `code`.
+    fn token_at(code: &str, language: &str, column: usize) -> (String, String) {
+        let mut parser = tree_sitter::Parser::new();
+        parser.set_language(&crate::langs::grammar_for(language).unwrap()).unwrap();
+        let tree = parser.parse(code, None).unwrap();
+        let node = super::descendant_for_position(tree.root_node(), code, (0, column));
+        (node.kind().to_string(), code[node.start_byte()..node.end_byte()].to_string())
+    }
+
+    #[test]
+    fn a_point_where_one_token_ends_and_the_next_begins_finds_the_next() {
+        // `)` ends where the name starts; the name is the token at that point.
+        let code = "void g(){if(ok)Foo f;}";
+        assert_eq!(token_at(code, "cpp", code.find("Foo").unwrap()).1, "Foo");
+        let code = "void g(){if(ok)target();}";
+        assert_eq!(token_at(code, "cpp", code.find("target").unwrap()), ("identifier".into(), "target".into()));
+        // A one-character name followed by `(`: its start finds the name.
+        let code = "void g(){f(x);}";
+        assert_eq!(token_at(code, "cpp", code.find("f(").unwrap()), ("identifier".into(), "f".into()));
+    }
+
     #[test]
     fn a_trailing_lambda_needs_a_function_typed_last_parameter() {
         for sig in [