codegraph/src/resolution/name-matcher.ts

/**
 * Name Matcher
 *
 * Handles symbol name matching for reference resolution.
 */

import { Node } from '../types';
import { UnresolvedRef, ResolvedRef, ResolutionContext } from './types';

/**
 * Try to resolve a path-like reference (e.g., "snippets/drawer-menu.liquid")
 * by matching the filename against file nodes.
 */
export function matchByFilePath(
  ref: UnresolvedRef,
  context: ResolutionContext
): ResolvedRef | null {
  // Path-like (`a/b.liquid`) OR a bare filename ending in a short extension
  // (`Foo.h` — an Objective-C `#import "Foo.h"`, resolved to the header by
  // basename). A bare ref WITHOUT an extension is a symbol name, not a file, so
  // leave it to the symbol-matching strategies.
  if (!ref.referenceName.includes('/') && !/\.[A-Za-z][A-Za-z0-9]{0,3}$/.test(ref.referenceName)) {
    return null;
  }

  // Extract the filename from the path
  const fileName = ref.referenceName.split('/').pop();
  if (!fileName) return null;

  // Search for file nodes with this name
  const candidates = context.getNodesByName(fileName);
  const fileNodes = candidates.filter(n => n.kind === 'file');

  if (fileNodes.length === 0) return null;

  // Prefer exact path match on qualified_name
  const exactMatch = fileNodes.find(n => n.qualifiedName === ref.referenceName || n.filePath === ref.referenceName);
  if (exactMatch) {
    return {
      original: ref,
      targetNodeId: exactMatch.id,
      confidence: 0.95,
      resolvedBy: 'file-path',
    };
  }

  // Fall back to suffix match (e.g., ref="snippets/foo.liquid" matches
  // "src/snippets/foo.liquid"). When several files share the basename — a
  // `#include "RNCAsyncStorage.h"` with a same-named header on another platform
  // (windows/code/ vs apple/) — prefer the one in the includer's own directory,
  // then by directory proximity / same language family. A C/C++ include (and any
  // bare-filename import) resolves relative to the including file, not to an
  // arbitrary same-named header elsewhere in the tree.
  const suffixMatches = fileNodes.filter(
    n => n.qualifiedName.endsWith(ref.referenceName) || n.filePath.endsWith(ref.referenceName)
  );
  if (suffixMatches.length > 0) {
    return {
      original: ref,
      targetNodeId: pickClosestFileNode(suffixMatches, ref).id,
      confidence: 0.85,
      resolvedBy: 'file-path',
    };
  }

  // If only one file node with this name, use it with lower confidence
  if (fileNodes.length === 1) {
    return {
      original: ref,
      targetNodeId: fileNodes[0]!.id,
      confidence: 0.7,
      resolvedBy: 'file-path',
    };
  }

  return null;
}

/**
 * Among several file nodes that all match a bare include/import by basename,
 * pick the one closest to the referencing file: same directory first, then by
 * directory-tree proximity, with the same language family as a tiebreak. A
 * C/C++ `#include "X.h"` (and any bare-filename import) resolves relative to the
 * including file — not to an arbitrary same-named header on another platform.
 */
function pickClosestFileNode(candidates: Node[], ref: UnresolvedRef): Node {
  const dirOf = (p: string): string => {
    const i = p.lastIndexOf('/');
    return i >= 0 ? p.slice(0, i) : '';
  };
  const refDir = dirOf(ref.filePath);
  const sameDir = candidates.filter((c) => dirOf(c.filePath) === refDir);
  const pool = sameDir.length > 0 ? sameDir : candidates;
  let best = pool[0]!;
  let bestScore = -Infinity;
  for (const c of pool) {
    const score =
      computePathProximity(ref.filePath, c.filePath) +
      (sameLanguageFamily(c.language, ref.language) ? 5 : 0);
    if (score > bestScore) {
      bestScore = score;
      best = c;
    }
  }
  return best;
}

/**
 * Language families that share a type system / runtime, so a same-language-only
 * reference may still resolve across them (a Kotlin `Foo.BAR` can name a Java
 * `Foo`). Anything not listed forms its own singleton family.
 */
const LANGUAGE_FAMILY: Record<string, string> = {
  java: 'jvm', kotlin: 'jvm', scala: 'jvm',
  swift: 'apple', objc: 'apple',
  typescript: 'web', tsx: 'web', javascript: 'web', jsx: 'web',
  c: 'c', cpp: 'c',
  // Razor/Blazor markup names C# types — same family so `@model Foo` /
  // `<MyComponent/>` resolve to their `.cs` class through the cross-family gate.
  csharp: 'dotnet', razor: 'dotnet',
};
export function sameLanguageFamily(a: string, b: string): boolean {
  if (a === b) return true;
  const fa = LANGUAGE_FAMILY[a];
  return fa !== undefined && fa === LANGUAGE_FAMILY[b];
}
/**
 * True when `lang` belongs to a known multi-language family (jvm/apple/web/c).
 * Languages not listed (php, python, go, ruby, rust, dart, …) and config
 * formats (yaml/xml/blade) form their own singleton families and return
 * `false` — used to leave config↔code framework bridges (whose config side is
 * never a known programming-language family) out of the cross-family gate.
 */
export function isKnownLanguageFamily(lang: string): boolean {
  return LANGUAGE_FAMILY[lang] !== undefined;
}
/**
 * True when `a` and `b` are two DIFFERENT *known* language families — the
 * signature of a coincidental cross-language name collision (a TS `import
 * React` matching a Swift `import React`, a C++ `#include "X.h"` matching a
 * same-named ObjC header on another platform). The both-*known* test is
 * deliberately weaker than {@link sameLanguageFamily}'s negation: a
 * single-file-component language that carries its own tag (`vue`/`svelte`)
 * importing a `.ts` module, or any singleton-family language (php/go/ruby/…),
 * returns `false` here and is left alone.
 */
export function crossesKnownFamily(a: string, b: string): boolean {
  return isKnownLanguageFamily(a) && isKnownLanguageFamily(b) && !sameLanguageFamily(a, b);
}
/**
 * Drop cross-language candidates from a name lookup. Two regimes:
 *  - `references` (type-usage): a type named in language X resolves to a
 *    SAME-family type, never a coincidentally same-named symbol in another
 *    language (the Android `BatteryManager` system class vs a JS one). Strict
 *    same-family filter — cross-language communication is `calls`, not refs.
 *  - `imports` (import binding): an `import`/`#include` never crosses two
 *    KNOWN families (TS `import React` ↮ Swift `import React`). Weaker
 *    both-known filter so `.vue`/`.svelte` (own tag) importing `.ts` survives.
 */
function applyLanguageGate(candidates: Node[], ref: UnresolvedRef): Node[] {
  if (ref.referenceKind === 'references') {
    return candidates.filter((c) => sameLanguageFamily(c.language, ref.language));
  }
  if (ref.referenceKind === 'imports') {
    return candidates.filter((c) => !crossesKnownFamily(c.language, ref.language));
  }
  return candidates;
}

/**
 * Try to resolve a reference by exact name match
 */
export function matchByExactName(
  ref: UnresolvedRef,
  context: ResolutionContext
): ResolvedRef | null {
  const candidates = applyLanguageGate(context.getNodesByName(ref.referenceName), ref);

  if (candidates.length === 0) {
    return null;
  }

  // If only one match, use it — but penalize cross-language matches
  if (candidates.length === 1) {
    const isCrossLanguage = candidates[0]!.language !== ref.language;
    return {
      original: ref,
      targetNodeId: candidates[0]!.id,
      confidence: isCrossLanguage ? 0.5 : 0.9,
      resolvedBy: 'exact-match',
    };
  }

  // Multiple matches - try to narrow down
  const bestMatch = findBestMatch(ref, candidates, context);
  if (bestMatch) {
    // Lower confidence when the match is from a distant/unrelated module
    const proximity = computePathProximity(ref.filePath, bestMatch.filePath);
    const confidence = proximity >= 30 ? 0.7 : 0.4;
    return {
      original: ref,
      targetNodeId: bestMatch.id,
      confidence,
      resolvedBy: 'exact-match',
    };
  }

  return null;
}

/**
 * Try to resolve by qualified name
 */
export function matchByQualifiedName(
  ref: UnresolvedRef,
  context: ResolutionContext
): ResolvedRef | null {
  // Check if the reference name looks qualified (contains :: or .)
  if (!ref.referenceName.includes('::') && !ref.referenceName.includes('.')) {
    return null;
  }

  const candidates = context.getNodesByQualifiedName(ref.referenceName);

  if (candidates.length === 1) {
    return {
      original: ref,
      targetNodeId: candidates[0]!.id,
      confidence: 0.95,
      resolvedBy: 'qualified-name',
    };
  }

  // Try partial qualified name match
  const parts = ref.referenceName.split(/[:.]/);
  const lastName = parts[parts.length - 1];
  if (lastName) {
    const partialCandidates = context.getNodesByName(lastName);
    for (const candidate of partialCandidates) {
      if (candidate.qualifiedName.endsWith(ref.referenceName)) {
        return {
          original: ref,
          targetNodeId: candidate.id,
          confidence: 0.85,
          resolvedBy: 'qualified-name',
        };
      }
    }
  }

  return null;
}

function resolveMethodOnType(
  typeName: string,
  methodName: string,
  ref: UnresolvedRef,
  context: ResolutionContext,
  confidence: number,
  resolvedBy: ResolvedRef['resolvedBy'],
  /**
   * Optional FQN that identifies WHICH class declaration `typeName`
   * refers to in the caller's file. When multiple candidates share
   * the same qualifiedName (`FooConverter::convert` in both
   * `dao/converter/` and `service/converter/`), the FQN's
   * file-path-suffix picks the right one — the disambiguation
   * signal Java imports carry but the call site doesn't (#314).
   */
  preferredFqn?: string,
  /** Recursion guard for the supertype/conformance walk. */
  depth = 0,
): ResolvedRef | null {
  // Look up methods by name and match by qualifiedName ending in
  // `<typeName>::<methodName>`. This works whether the method is defined
  // in-class (`class Foo { int bar() { ... } }`) or out-of-line in a separate
  // file (`int Foo::bar() { ... }` in foo.cpp while class Foo is in foo.hpp).
  // The previous same-file approach missed the latter — the typical C++ layout.
  const methodCandidates = context.getNodesByName(methodName);
  const want = `${typeName}::${methodName}`;
  const matches: Node[] = [];
  for (const m of methodCandidates) {
    if (m.kind !== 'method') continue;
    if (m.language !== ref.language) continue;
    const qn = m.qualifiedName;
    if (qn === want || qn.endsWith(`::${want}`)) {
      matches.push(m);
    }
  }
  if (matches.length === 0) {
    // Conformance fallback: the method may be defined on a supertype `typeName`
    // extends, or on a protocol / trait it conforms to (e.g. a Swift protocol-
    // extension method, a C# default-interface or extension method, a Kotlin
    // extension on a supertype). Walk supertypes transitively (depth-capped) via
    // the resolved implements/extends edges — empty in the first resolution pass,
    // populated in the conformance pass. Still VALIDATED (the method must exist on
    // a supertype), so a wrong inference produces no edge.
    if (depth < 4 && context.getSupertypes) {
      for (const supertype of context.getSupertypes(typeName, ref.language)) {
        const via = resolveMethodOnType(
          supertype, methodName, ref, context, confidence, resolvedBy, preferredFqn, depth + 1,
        );
        if (via) return via;
      }
    }
    return null;
  }

  if (matches.length > 1 && preferredFqn) {
    const ext = ref.language === 'kotlin' ? '.kt' : '.java';
    const fqnPath = preferredFqn.replace(/\./g, '/') + ext;
    const chosen = matches.find((m) => {
      const fp = m.filePath.replace(/\\/g, '/');
      return fp.endsWith(fqnPath) || fp.endsWith('/' + fqnPath);
    });
    if (chosen) {
      return {
        original: ref,
        targetNodeId: chosen.id,
        confidence,
        resolvedBy,
      };
    }
  }

  return {
    original: ref,
    targetNodeId: matches[0]!.id,
    confidence,
    resolvedBy,
  };
}

// C++ keywords/control-flow tokens that can appear right before a receiver
// (e.g. `return ptr->m()`) and must NOT be treated as a type.
const CPP_NON_TYPE_TOKENS = new Set([
  'return', 'if', 'else', 'for', 'while', 'do', 'switch', 'case', 'default',
  'break', 'continue', 'goto', 'throw', 'new', 'delete', 'co_await', 'co_yield',
  'co_return', 'static_cast', 'const_cast', 'dynamic_cast', 'reinterpret_cast',
  'sizeof', 'alignof', 'typeid', 'and', 'or', 'not', 'xor',
]);

function normalizeCppTypeName(typeName: string): string | null {
  const normalized = typeName
    .replace(/\b(const|volatile|mutable|typename|class|struct)\b/g, ' ')
    .replace(/[&*]+/g, ' ')
    .replace(/<[^>]*>/g, ' ')
    .replace(/\s+/g, ' ')
    .trim();

  if (!normalized) return null;
  const parts = normalized.split(/::/).filter(Boolean);
  const last = parts[parts.length - 1];
  if (!last) return null;
  if (CPP_NON_TYPE_TOKENS.has(last)) return null;
  return last;
}

// Declarator regex: matches `Type receiver`, `Type* receiver`, `Type *receiver`,
// `Type*receiver`, `Type<X> receiver`, etc., REQUIRING a declarator terminator
// (`;`, `=`, `,`, `)`, `[`, `{`, `(`, or end-of-line) after the receiver. The
// terminator rules out uses like `return receiver->m()` where the preceding
// token is a keyword, not a type.
function buildDeclaratorRegex(escapedReceiver: string): RegExp {
  return new RegExp(
    `([A-Za-z_][\\w:]*(?:\\s*<[^;=(){}]+>)?(?:\\s*[*&]+)?)\\s*\\b${escapedReceiver}\\b\\s*(?=[;=,)\\[{(]|$)`,
  );
}

function inferCppReceiverType(
  receiverName: string,
  ref: UnresolvedRef,
  context: ResolutionContext,
  depth = 0,
): string | null {
  const source = context.readFile(ref.filePath);
  if (!source) return null;

  const lines = source.split(/\r?\n/);
  const callLineIndex = Math.max(0, Math.min(lines.length - 1, ref.line - 1));
  const escapedReceiver = receiverName.replace(/[.*+?^${}()|[\]\\]/g, '\\$&');
  const receiverPattern = new RegExp(`\\b${escapedReceiver}\\b`);
  const declaratorRegex = buildDeclaratorRegex(escapedReceiver);

  for (let i = callLineIndex; i >= 0; i--) {
    const line = lines[i];
    if (!line || !receiverPattern.test(line)) continue;

    const declaratorMatch = line.match(declaratorRegex);
    if (declaratorMatch) {
      const normalized = normalizeCppTypeName(declaratorMatch[1] ?? '');
      if (normalized === 'auto') {
        // `auto x = Foo::instance();` — the declared type is deduced; recover it
        // from the initializer (call return type / construction) (#645).
        const initType = inferCppAutoInitializerType(line, receiverName, ref, context, depth);
        if (initType) return initType;
        // No usable initializer on this line — keep scanning earlier ones.
      } else if (normalized) {
        return normalized;
      }
    }
  }

  const headerCandidates = [
    ref.filePath.replace(/\.(?:c|cc|cpp|cxx)$/i, '.h'),
    ref.filePath.replace(/\.(?:c|cc|cpp|cxx)$/i, '.hpp'),
    ref.filePath.replace(/\.(?:c|cc|cpp|cxx)$/i, '.hxx'),
  ].filter((candidate, index, arr) => arr.indexOf(candidate) === index && candidate !== ref.filePath);

  for (const headerPath of headerCandidates) {
    if (!context.fileExists(headerPath)) continue;
    const headerSource = context.readFile(headerPath);
    if (!headerSource) continue;

    for (const line of headerSource.split(/\r?\n/)) {
      if (!receiverPattern.test(line)) continue;
      const declaratorMatch = line.match(declaratorRegex);
      if (!declaratorMatch) continue;
      const normalized = normalizeCppTypeName(declaratorMatch[1] ?? '');
      if (normalized && normalized !== 'auto') return normalized;
    }
  }

  return null;
}

/**
 * Last `::`-separated segment of a (possibly namespace-qualified) C++ name.
 */
function cppLastSegment(name: string): string {
  const parts = name.split('::').filter(Boolean);
  return parts[parts.length - 1] ?? name;
}

/**
 * Return type captured at extraction for `Class::method` (or a free function),
 * read off the indexed node's `returnType` — used by the C++ (#645) and PHP
 * (#608) chained-call resolvers. Language-filtered. Null when not indexed or no
 * return type was recorded (a `void`/primitive return).
 */
function lookupCalleeReturnType(
  callee: string,
  ref: UnresolvedRef,
  context: ResolutionContext,
): string | null {
  let method = callee;
  let cls: string | null = null;
  if (callee.includes('::')) {
    const parts = callee.split('::').filter(Boolean);
    method = parts[parts.length - 1] ?? callee;
    cls = parts.slice(0, -1).join('::');
  }
  const candidates = context.getNodesByName(method).filter(
    (n) =>
      (n.kind === 'method' || n.kind === 'function') &&
      n.language === ref.language &&
      !!n.returnType,
  );
  if (cls) {
    const want = `${cls}::${method}`;
    // The call site may name the class with MORE namespace qualification than
    // the stored node (`details::registry::instance` at the call vs
    // `registry::instance` on the node — the receiver type only carries the
    // immediate class), or LESS. Accept an exact match or either being a
    // namespace-suffix of the other; the shared `::<class>::<method>` tail keeps
    // it specific.
    const m = candidates.find(
      (n) =>
        n.qualifiedName === want ||
        n.qualifiedName.endsWith(`::${want}`) ||
        want.endsWith(`::${n.qualifiedName}`),
    );
    return m?.returnType ?? null;
  }
  return candidates.find((n) => n.kind === 'function')?.returnType ?? null;
}

/** Does the graph contain a class/struct named `name`'s last segment? */
function cppClassExists(name: string, ref: UnresolvedRef, context: ResolutionContext): boolean {
  const last = cppLastSegment(name);
  return context
    .getNodesByName(last)
    .some((n) => (n.kind === 'class' || n.kind === 'struct') && n.language === ref.language);
}

/**
 * Infer the class produced by a C++ call/construction expression, using return
 * types captured at extraction (#645). Handles, in order:
 *   - `make_unique<T>()` / `make_shared<T>()`        → T
 *   - single-level member call `recv.method()`       → recv's type, then method's return
 *   - `Class::method()` / free `func()`              → the callee's recorded return type
 *   - direct construction `Type()` / `ns::Type()`    → Type
 * Returns null when undeterminable. Callers MUST still validate the outer method
 * exists on the result before creating an edge, so a wrong guess stays silent.
 */
function resolveCppCallResultType(
  inner: string,
  ref: UnresolvedRef,
  context: ResolutionContext,
  depth = 0,
): string | null {
  if (depth > 3) return null; // guard against pathological mutual recursion
  const expr = inner.trim();

  const make = expr.match(/(?:^|::)(?:make_unique|make_shared)\s*<\s*([A-Za-z_]\w*)/);
  if (make) return make[1] ?? null;

  // Single-level member call `recv.method` (the `manager.view().render()` shape).
  const dotIdx = expr.lastIndexOf('.');
  if (dotIdx > 0) {
    const recv = expr.slice(0, dotIdx);
    const method = expr.slice(dotIdx + 1);
    if (recv.includes('.') || recv.includes('(') || recv.includes('::')) return null; // single level only
    const recvType = inferCppReceiverType(recv, ref, context, depth + 1);
    if (!recvType) return null;
    return lookupCalleeReturnType(`${recvType}::${method}`, ref, context);
  }

  const ret = lookupCalleeReturnType(expr, ref, context);
  if (ret) return ret;

  // Direct construction — the callee itself names a class/struct.
  if (cppClassExists(expr, ref, context)) return cppLastSegment(expr);

  return null;
}

/**
 * Recover the type of an `auto`-declared local from its initializer on the
 * declaration line — `auto x = Foo::instance();`, `auto w = make_unique<W>();`,
 * `auto p = new W();`, `auto w = Widget();` (#645).
 */
function inferCppAutoInitializerType(
  line: string,
  receiverName: string,
  ref: UnresolvedRef,
  context: ResolutionContext,
  depth: number,
): string | null {
  const escaped = receiverName.replace(/[.*+?^${}()|[\]\\]/g, '\\$&');
  const m = line.match(new RegExp(`\\b${escaped}\\b\\s*=\\s*([^;]+)`));
  if (!m || !m[1]) return null;
  const init = m[1].trim();

  const neu = init.match(/^new\s+([A-Za-z_][\w:]*)/);
  if (neu && neu[1]) return cppLastSegment(neu[1]);

  // A call or construction: `Foo(...)`, `A::b(...)`, `make_unique<T>(...)`.
  const call = init.match(/^([A-Za-z_][\w:]*(?:\s*<[^>;]*>)?)\s*\(/);
  if (call && call[1]) return resolveCppCallResultType(call[1].replace(/\s+/g, ''), ref, context, depth + 1);

  return null;
}

/**
 * Resolve a C++ chained call whose receiver is itself a call — encoded by the
 * extractor as `<innerCallee>().<method>` (#645). The receiver's type is what
 * the inner call returns; the outer method is then resolved and VALIDATED on it
 * (resolveMethodOnType requires `cls::method` to exist), so a wrong inference
 * produces no edge rather than a wrong one.
 */
export function matchCppCallChain(
  ref: UnresolvedRef,
  context: ResolutionContext,
): ResolvedRef | null {
  const m = ref.referenceName.match(/^(.+)\(\)\.(\w+)$/);
  if (!m || !m[1] || !m[2]) return null;
  const cls = resolveCppCallResultType(m[1], ref, context);
  if (!cls) return null;
  return resolveMethodOnType(cls, m[2], ref, context, 0.85, 'instance-method');
}

/**
 * Resolve a `::`-scoped factory chain whose receiver is a scoped/static call —
 * PHP `Cls::for($x)->method()` (#608, the per-credential Laravel client idiom) or
 * Rust `Foo::new().bar()` (an associated-function call) — both encoded by the
 * extractor as `Cls::factory().method`. The receiver's type is what `Cls::factory`
 * returns: a `self` marker (PHP `: self`/`: static`, Rust `-> Self`) resolves to
 * the factory's own type, a concrete return type to that type. The outer method is
 * then resolved and VALIDATED on it (resolveMethodOnType requires the method to
 * exist on the type or a supertype it conforms to), so a wrong inference yields no
 * edge rather than a wrong one. Shared by the `::`-receiver languages (PHP, Rust).
 */
export function matchScopedCallChain(
  ref: UnresolvedRef,
  context: ResolutionContext,
): ResolvedRef | null {
  const m = ref.referenceName.match(/^(.+)\(\)\.(\w+)$/);
  if (!m || !m[1] || !m[2]) return null;
  const inner = m[1];
  const method = m[2];
  if (!inner.includes('::')) return null; // only static-factory (`Cls::method`) chains
  const factoryClass = inner.slice(0, inner.lastIndexOf('::'));
  const ret = lookupCalleeReturnType(inner, ref, context);
  if (!ret) return null;
  // `self` (the extractor's marker for self/static/$this) → the factory's class.
  const resolvedClass = ret === 'self' ? factoryClass : ret;
  return resolveMethodOnType(resolvedClass, method, ref, context, 0.85, 'instance-method');
}

/**
 * Languages where an unprefixed capitalized call `Foo(args)` constructs the
 * class (so a `Foo(args).method()` receiver's type is `Foo`). Java/C# need `new`,
 * so a bare `Foo()` there is a method call, not construction — excluded. Scala's
 * `Foo(args)` is a case-class / companion `apply`, which conventionally returns
 * `Foo` — and resolveMethodOnType validates, so a non-conventional `apply` that
 * returns another type simply yields no edge rather than a wrong one.
 */
const CONSTRUCTS_VIA_BARE_CALL = new Set(['kotlin', 'swift', 'scala', 'dart']);

/**
 * Resolve a dotted chained call whose receiver is a static factory / fluent call —
 * `Foo.getInstance().bar()`, encoded by the extractor as `Foo.getInstance().bar`
 * (#645/#608 mechanism). The receiver's type is what `Foo.getInstance` returns
 * (its declared return type); the outer method is then resolved and VALIDATED on
 * it (resolveMethodOnType requires `Type::method` to exist), so a wrong inference
 * yields no edge rather than a wrong one (e.g. a same-named `bar()` on an
 * unrelated class is never matched). Shared by the dot-notation languages
 * (Java, Kotlin, C#, Swift) — same receiver shape, same `Class::method` qualified names.
 */
export function matchDottedCallChain(
  ref: UnresolvedRef,
  context: ResolutionContext,
): ResolvedRef | null {
  const m = ref.referenceName.match(/^(.+)\(\)\.(\w+)$/);
  if (!m || !m[1] || !m[2]) return null;
  const inner = m[1]; // `Foo.getInstance`
  const method = m[2]; // `bar`
  const lastDot = inner.lastIndexOf('.');

  if (lastDot <= 0) {
    // Go: bare package-level factory FUNCTION `New().method()` — the receiver's
    // type is what `New` returns; resolve the method on that.
    if (ref.language === 'go') {
      const ret = lookupCalleeReturnType(inner, ref, context);
      if (ret) {
        return resolveMethodOnType(ret, method, ref, context, 0.85, 'instance-method', importedFqnOf(ret, ref, context));
      }
      // `inner` isn't a function with a captured return type — typically a
      // package-level VARIABLE holding a function value (e.g. gin's `engine()`),
      // whose type we can't recover. Fall back to bare-name resolution of the
      // method so we don't DROP an edge the un-re-encoded bare path would have
      // found. (When `inner` IS a real factory function but the method doesn't
      // exist on its return type, `ret` is truthy and we returned no edge above —
      // the absent-method safety guarantee is preserved.)
      //
      // CRITICAL: resolve the TARGET via a synthetic bare-name ref, but return the
      // match tied to the ORIGINAL `ref` (referenceName `inner().method`). The
      // batched resolver (resolveAndPersistBatched) reads unresolved rows from
      // offset 0 every pass and relies on deleteSpecificResolvedReferences —
      // keyed on referenceName — to clear each resolved row so the batch empties.
      // If we propagated the synthetic ref's bare `method` as `.original`, the
      // delete would never match the stored `inner().method` row, the batch would
      // never drain, and the loop would re-resolve + re-insert forever (a runaway
      // that grew gin's graph to 5M edges / 1.4 GB before this fix).
      const bareRef = { ...ref, referenceName: method };
      const bareMatch = matchByExactName(bareRef, context) ?? matchFuzzy(bareRef, context);
      return bareMatch ? { ...bareMatch, original: ref } : null;
    }
    // Constructor receiver `Foo(args).method()` (encoded `Foo().method`): a bare,
    // capitalized inner is a class construction, so the receiver's type is the
    // class itself — resolve the method on it. Only in languages where an
    // unprefixed capitalized call constructs the class (Kotlin, Swift); in Java/C#
    // a bare `Foo()` is a method call (constructors need `new`), so we must not
    // assume construction. A lowercase bare inner is a top-level `factory().method()`
    // whose type we can't recover — bail.
    if (!CONSTRUCTS_VIA_BARE_CALL.has(ref.language) || !/^[A-Z]/.test(inner)) return null;
    return resolveMethodOnType(inner, method, ref, context, 0.85, 'instance-method', importedFqnOf(inner, ref, context));
  }

  // Factory/fluent receiver `Receiver.factory(args).method()`: the receiver's
  // type is what `Receiver.factory` returns (its declared return type).
  const factoryClass = inner.slice(0, lastDot).split('.').pop(); // simple class name
  const factoryMethod = inner.slice(lastDot + 1);
  if (!factoryClass || !factoryMethod) return null;
  const ret = lookupCalleeReturnType(`${factoryClass}::${factoryMethod}`, ref, context);
  if (!ret) {
    // Objective-C: a class-message factory — `[X alloc]`, `[X new]`,
    // `[X sharedFoo]` — returns an instance of the RECEIVER class `X` by
    // convention (`instancetype`). So when the factory's own return type isn't
    // recoverable (its selector returns `instancetype`, or `alloc`/`new` aren't
    // user-defined nodes at all), the receiver's type is the class `X` itself.
    // This resolves the ubiquitous `[[X alloc] init]` and singleton chains.
    // resolveMethodOnType validates against X (and its supertypes), so a class
    // whose method actually lives elsewhere yields NO edge, not a wrong one — and
    // crucially this does NOT fire when a concrete return type WAS captured but
    // simply lacks the method (that already returned null above: absent-method
    // safety, so a same-named decoy is still never matched).
    if (ref.language === 'objc' && /^[A-Z]/.test(factoryClass)) {
      return resolveMethodOnType(factoryClass, method, ref, context, 0.8, 'instance-method', importedFqnOf(factoryClass, ref, context));
    }
    return null;
  }
  return resolveMethodOnType(ret, method, ref, context, 0.85, 'instance-method', importedFqnOf(ret, ref, context));
}

/**
 * When several classes share a simple type name, the caller file's import of
 * that type is the only signal that names WHICH one (#314). Returns the imported
 * FQN for `typeName` in the ref's file, or undefined.
 */
function importedFqnOf(
  typeName: string,
  ref: UnresolvedRef,
  context: ResolutionContext,
): string | undefined {
  const imports = context.getImportMappings(ref.filePath, ref.language);
  return imports.find((i) => i.localName === typeName)?.source;
}

/**
 * Java/Kotlin: infer a receiver's declared type by walking field declarations
 * in the class enclosing the call site. The field's `signature` is already in
 * the form "<TypeName> <fieldName>" (set by tree-sitter.ts extractField), so we
 * pull the type from there. Handles Spring `@Resource UserBO userbo;` /
 * `@Autowired private UserService userService;` where the receiver field name
 * doesn't match the class name by Java naming convention.
 *
 * Returns the bare type name (generics stripped, dotted package stripped) or
 * null when no matching field is in the enclosing class.
 */
function inferJavaFieldReceiverType(
  receiverName: string,
  ref: UnresolvedRef,
  context: ResolutionContext,
): string | null {
  const inFile = context.getNodesInFile(ref.filePath);
  if (inFile.length === 0) return null;

  // Find the class enclosing the call line (tightest match by latest start).
  let enclosing: Node | null = null;
  for (const n of inFile) {
    if (n.kind !== 'class' && n.kind !== 'interface') continue;
    if (n.language !== ref.language) continue;
    const end = n.endLine ?? n.startLine;
    if (n.startLine <= ref.line && end >= ref.line) {
      if (!enclosing || n.startLine >= enclosing.startLine) enclosing = n;
    }
  }
  if (!enclosing) return null;

  const enclosingEnd = enclosing.endLine ?? enclosing.startLine;
  const field = inFile.find(
    (n) =>
      n.kind === 'field' &&
      n.name === receiverName &&
      n.language === ref.language &&
      n.startLine >= enclosing.startLine &&
      (n.endLine ?? n.startLine) <= enclosingEnd,
  );
  if (!field || !field.signature) return null;

  // Signature shape: "<TypeName> <fieldName>" (extractField). Pull the type,
  // strip generics + dotted package, drop array/varargs markers.
  const beforeName = field.signature.slice(
    0,
    field.signature.lastIndexOf(field.name),
  );
  const typeRaw = beforeName.trim();
  if (!typeRaw) return null;

  const typeNoGenerics = typeRaw.replace(/<[^>]*>/g, '').trim();
  const typeNoArray = typeNoGenerics.replace(/\[\s*\]/g, '').replace(/\.\.\.$/, '').trim();
  const parts = typeNoArray.split(/[.\s]+/).filter(Boolean);
  const lastPart = parts[parts.length - 1];
  if (!lastPart) return null;
  if (!/^[A-Z]/.test(lastPart)) return null; // primitives / lowercase → skip
  return lastPart;
}

/**
 * Try to resolve by method name on a class/object
 */
export function matchMethodCall(
  ref: UnresolvedRef,
  context: ResolutionContext
): ResolvedRef | null {
  // Parse method call patterns like "obj.method" or "Class::method". The method
  // part allows trailing `:` keywords so Objective-C selectors resolve
  // (`SDImageCache.storeImage:`, `obj.setX:y:`); colons never appear in other
  // languages' method refs, so this is a no-op for them.
  // The receiver allows dots (`builder.Services.AddCoreServices`) so a CHAINED
  // call resolves by its last segment — Strategy 3 below name-matches the method
  // (with its existing single-candidate / receiver-overlap guards). Without this
  // a multi-dot extension-method call (C# DI `builder.Services.AddCoreServices()`,
  // `Guard.Against.X()`) matched no pattern and never resolved.
  const dotMatch = ref.referenceName.match(/^([\w.]+)\.(\w+:?(?:\w+:)*)$/);
  const colonMatch = ref.referenceName.match(/^(\w+)::(\w+)$/);

  const match = dotMatch || colonMatch;
  if (!match) {
    return null;
  }

  const [, objectOrClass, methodName] = match;

  if (ref.language === 'cpp' && dotMatch) {
    const inferredType = inferCppReceiverType(objectOrClass!, ref, context);
    if (inferredType) {
      const typedMatch = resolveMethodOnType(
        inferredType,
        methodName!,
        ref,
        context,
        0.9,
        'instance-method',
      );
      if (typedMatch) {
        return typedMatch;
      }
    }
  }

  // Java/Kotlin: receiver may be a field whose name doesn't match the type by
  // Java naming convention (`userbo` → class `UserBO`, abbreviated). Look up
  // the field in the enclosing class to get its declared type, then resolve
  // the method on that type. Covers Spring `@Resource`/`@Autowired` field
  // injection where the field type is the concrete bean class.
  if ((ref.language === 'java' || ref.language === 'kotlin') && dotMatch) {
    const inferredType = inferJavaFieldReceiverType(objectOrClass!, ref, context);
    if (inferredType) {
      // When two classes share the same simple name, the caller file's
      // import is the only signal that names WHICH one — pass the
      // imported FQN so resolveMethodOnType can disambiguate (#314).
      const imports = context.getImportMappings(ref.filePath, ref.language);
      const importedFqn = imports.find((i) => i.localName === inferredType)?.source;
      const typedMatch = resolveMethodOnType(
        inferredType,
        methodName!,
        ref,
        context,
        0.9,
        'instance-method',
        importedFqn,
      );
      if (typedMatch) {
        return typedMatch;
      }
    }
  }

  // Strategy 1: Direct class name match (existing logic)
  const classCandidates = context.getNodesByName(objectOrClass!);

  for (const classNode of classCandidates) {
    if (classNode.kind === 'class' || classNode.kind === 'struct' || classNode.kind === 'interface') {
      // Skip cross-language class matches
      if (classNode.language !== ref.language) continue;

      const nodesInFile = context.getNodesInFile(classNode.filePath);
      const methodNode = nodesInFile.find(
        (n) =>
          n.kind === 'method' &&
          n.name === methodName &&
          n.qualifiedName.includes(classNode.name)
      );

      if (methodNode) {
        return {
          original: ref,
          targetNodeId: methodNode.id,
          confidence: 0.85,
          resolvedBy: 'qualified-name',
        };
      }
    }
  }

  // Strategy 2: Instance variable receiver - try capitalized form to find class
  // e.g., "permissionEngine" → look for classes containing "PermissionEngine"
  const capitalizedReceiver = objectOrClass!.charAt(0).toUpperCase() + objectOrClass!.slice(1);
  if (capitalizedReceiver !== objectOrClass) {
    const fuzzyClassCandidates = context.getNodesByName(capitalizedReceiver);
    for (const classNode of fuzzyClassCandidates) {
      if (classNode.kind === 'class' || classNode.kind === 'struct' || classNode.kind === 'interface') {
        // Skip cross-language class matches
        if (classNode.language !== ref.language) continue;

        const nodesInFile = context.getNodesInFile(classNode.filePath);
        const methodNode = nodesInFile.find(
          (n) =>
            n.kind === 'method' &&
            n.name === methodName &&
            n.qualifiedName.includes(classNode.name)
        );

        if (methodNode) {
          return {
            original: ref,
            targetNodeId: methodNode.id,
            confidence: 0.8,
            resolvedBy: 'instance-method',
          };
        }
      }
    }
  }

  // Strategy 3: Find methods by name across the codebase, match by receiver
  // name similarity with the containing class. Handles abbreviated variable
  // names like permissionEngine → PermissionRuleEngine.
  if (methodName) {
    const methodCandidates = context.getNodesByName(methodName!);
    const methods = methodCandidates.filter(
      (n) => n.kind === 'method' && n.name === methodName
    );

    // Filter to same-language candidates first
    const sameLanguageMethods = methods.filter(m => m.language === ref.language);
    const targetMethods = sameLanguageMethods.length > 0 ? sameLanguageMethods : methods;

    // If only one same-language method with this name exists, use it
    if (targetMethods.length === 1 && targetMethods[0]!.language === ref.language) {
      return {
        original: ref,
        targetNodeId: targetMethods[0]!.id,
        confidence: 0.7,
        resolvedBy: 'instance-method',
      };
    }

    // Multiple methods: score by receiver name word overlap with class name
    if (targetMethods.length > 1) {
      const receiverWords = splitCamelCase(objectOrClass!);
      let bestMatch: typeof targetMethods[0] | undefined;
      let bestScore = 0;

      for (const method of targetMethods) {
        const classWords = splitCamelCase(method.qualifiedName);
        let score = receiverWords.filter(w =>
          classWords.some(cw => cw.toLowerCase() === w.toLowerCase())
        ).length;
        // Bonus for same language
        if (method.language === ref.language) score += 1;
        if (score > bestScore) {
          bestScore = score;
          bestMatch = method;
        }
      }

      if (bestMatch && bestScore >= 2) {
        return {
          original: ref,
          targetNodeId: bestMatch.id,
          confidence: 0.65,
          resolvedBy: 'instance-method',
        };
      }
    }
  }

  return null;
}

/**
 * Split a camelCase or PascalCase string into words.
 */
function splitCamelCase(str: string): string[] {
  return str.replace(/([a-z])([A-Z])/g, '$1 $2')
    .replace(/([A-Z]+)([A-Z][a-z])/g, '$1 $2')
    .split(/[\s._:\/\\]+/)
    .filter(w => w.length > 1);
}

/**
 * Compute directory proximity between two file paths.
 * Returns a score based on the number of shared directory segments.
 * Higher score = closer in directory tree.
 */
function computePathProximity(filePath1: string, filePath2: string): number {
  const dir1 = filePath1.split('/').slice(0, -1);
  const dir2 = filePath2.split('/').slice(0, -1);

  let shared = 0;
  for (let i = 0; i < Math.min(dir1.length, dir2.length); i++) {
    if (dir1[i] === dir2[i]) {
      shared++;
    } else {
      break;
    }
  }

  // Each shared directory segment contributes 15 points, capped at 80
  return Math.min(shared * 15, 80);
}

/**
 * Find the best matching node when there are multiple candidates
 */
function findBestMatch(
  ref: UnresolvedRef,
  candidates: Node[],
  _context: ResolutionContext
): Node | null {
  // Prioritization rules:
  // 1. Same file > different file
  // 2. Directory proximity (same module/package > different module)
  // 3. Same language > different language
  // 4. Functions/methods > classes/types (for call references)
  // 5. Exported > non-exported

  let bestScore = -1;
  let bestNode: Node | null = null;

  for (const candidate of candidates) {
    let score = 0;

    // Same file bonus
    if (candidate.filePath === ref.filePath) {
      score += 100;
    }

    // Directory proximity bonus — strongly prefer same module/package
    score += computePathProximity(ref.filePath, candidate.filePath);

    // Language matching: strongly prefer same language, penalize cross-language
    if (candidate.language === ref.language) {
      score += 50;
    } else {
      score -= 80;
    }

    // For call references, prefer functions/methods
    if (ref.referenceKind === 'calls') {
      if (candidate.kind === 'function' || candidate.kind === 'method') {
        score += 25;
      }
    }

    // For instantiation references (`new Foo()`), prefer class-like
    // targets — without this, a function named `Foo` in another module
    // could outscore the actual class.
    if (ref.referenceKind === 'instantiates') {
      if (
        candidate.kind === 'class' ||
        candidate.kind === 'struct' ||
        candidate.kind === 'interface'
      ) {
        score += 25;
      }
    }

    // For decorator references (`@Foo`), prefer functions. Class
    // decorators (Python `@SomeClass`, Java annotation interfaces)
    // also resolve here, hence the smaller class bonus.
    if (ref.referenceKind === 'decorates') {
      if (candidate.kind === 'function' || candidate.kind === 'method') {
        score += 25;
      } else if (candidate.kind === 'class' || candidate.kind === 'interface') {
        score += 15;
      }
    }

    // Exported bonus
    if (candidate.isExported) {
      score += 10;
    }

    // Closer line number (within same file)
    if (candidate.filePath === ref.filePath && candidate.startLine) {
      const distance = Math.abs(candidate.startLine - ref.line);
      score += Math.max(0, 20 - distance / 10);
    }

    if (score > bestScore) {
      bestScore = score;
      bestNode = candidate;
    }
  }

  return bestNode;
}

/**
 * Fuzzy match - last resort with lower confidence
 */
export function matchFuzzy(
  ref: UnresolvedRef,
  context: ResolutionContext
): ResolvedRef | null {
  const lowerName = ref.referenceName.toLowerCase();

  // Use pre-built lowercase index for O(1) lookup instead of scanning all nodes
  const candidates = context.getNodesByLowerName(lowerName);

  // Filter to callable kinds only (function, method, class)
  const callableKinds = new Set(['function', 'method', 'class']);
  const callableCandidates = applyLanguageGate(candidates.filter((n) => callableKinds.has(n.kind)), ref);

  // Prefer same-language matches
  const sameLanguageCandidates = callableCandidates.filter(n => n.language === ref.language);
  const finalCandidates = sameLanguageCandidates.length > 0 ? sameLanguageCandidates : callableCandidates;

  if (finalCandidates.length === 1) {
    const isCrossLanguage = finalCandidates[0]!.language !== ref.language;
    return {
      original: ref,
      targetNodeId: finalCandidates[0]!.id,
      confidence: isCrossLanguage ? 0.3 : 0.5,
      resolvedBy: 'fuzzy',
    };
  }

  return null;
}

/**
 * Match all strategies in order of confidence
 */
export function matchReference(
  ref: UnresolvedRef,
  context: ResolutionContext
): ResolvedRef | null {
  // Try strategies in order of confidence
  let result: ResolvedRef | null;

  // 0. File path match (e.g., "snippets/drawer-menu.liquid" → file node)
  result = matchByFilePath(ref, context);
  if (result) return result;

  // 1. Qualified name match (highest confidence)
  result = matchByQualifiedName(ref, context);
  if (result) return result;

  // 1b. C++ chained call whose receiver is another call — `Foo::instance().bar()`
  // encoded as `Foo::instance().bar` by the extractor (#645). Resolve the
  // receiver's type from what the inner call returns, then the method on it.
  if (ref.language === 'cpp' || ref.language === 'c') {
    result = matchCppCallChain(ref, context);
    if (result) return result;
  }

  // 1c. `::`-scoped factory chain — PHP `Cls::for($x)->method()` (#608) or Rust
  // `Foo::new().bar()`, both encoded as `Cls::factory().method`. The receiver's
  // type is the factory's `self` (PHP `: self`/`: static`, Rust `-> Self`) or
  // concrete return type.
  if (ref.language === 'php' || ref.language === 'rust') {
    result = matchScopedCallChain(ref, context);
    if (result) return result;
  }

  // 1d. Dotted chained static-factory / fluent call (Java / Kotlin / C# / Swift /
  // Go / Scala / Dart / Objective-C) — `Foo.getInstance().bar()` encoded as
  // `Foo.getInstance().bar`, Go's bare-factory `New().Method()` as `New().Method`,
  // Scala's companion factory, Dart's static factory / factory-constructor, or
  // ObjC's chained message send `[[Foo create] doIt]` encoded as `Foo.create().doIt`
  // (#645/#608 mechanism). Resolve the method's class from the inner call's
  // declared return type, then validate it.
  if (
    ref.language === 'java' ||
    ref.language === 'kotlin' ||
    ref.language === 'csharp' ||
    ref.language === 'swift' ||
    ref.language === 'go' ||
    ref.language === 'scala' ||
    ref.language === 'dart' ||
    ref.language === 'objc'
  ) {
    result = matchDottedCallChain(ref, context);
    if (result) return result;
  }

  // 2. Method call pattern
  result = matchMethodCall(ref, context);
  if (result) return result;

  // 3. Exact name match
  result = matchByExactName(ref, context);
  if (result) return result;

  // 4. Fuzzy match (lowest confidence)
  result = matchFuzzy(ref, context);
  if (result) return result;

  return null;
}