Back to the room

Pass 3 of 3 · Deep dive of PHAROSPass 1 · OverviewPass 2 · How it works

Code Observatory / Source

Resolve calls conservatively

Same file, then import, then a name unique in the repository. Anything ambiguous stays unlinked.

PINNED SOURCE Copied from the repository at this commit. Source, not an execution.

PHAROS at 975a38dMIT license

build_edgesbackend/app/codeintel/graph.py, lines 144–244Open source
def build_edges(nodes: list[ChunkNode]) -> tuple[list[Edge], GraphStats]:    stats = GraphStats()    edges: list[Edge] = []    seen: set[tuple[str, str, str]] = set()    def add(src: ChunkNode, dst: ChunkNode, name: str, etype: str, resolution: str) -> bool:        if src.id == dst.id:            return False        key = (src.id, dst.id, etype)        if key in seen:            return False        seen.add(key)        edges.append(Edge(src.id, dst.id, name, etype, resolution))        return True    by_qualified: dict[str, ChunkNode] = {}    by_file: dict[str, list[ChunkNode]] = defaultdict(list)    modules: dict[str, ChunkNode] = {}    by_name: dict[str, list[ChunkNode]] = defaultdict(list)    for n in nodes:        by_qualified.setdefault(n.qualified_name, n)        by_file[n.path].append(n)        if n.kind == "module":            modules[n.path] = n        elif n.kind != "block":            by_name[n.name].append(n)    by_basename: dict[str, list[ChunkNode]] = defaultdict(list)    for path, mod in modules.items():        by_basename[posixpath.basename(path)].append(mod)    py_index = _python_module_index([m for m in modules.values() if m.language == "python"])    # contains    for file_nodes in by_file.values():        local = {n.qualified_name: n for n in file_nodes}        for n in file_nodes:            if n.parent and n.parent in local:                if add(local[n.parent], n, n.name, "contains", "same_file"):                    stats.contains += 1    # imports (module -> module)    imported_files: dict[str, set[str]] = defaultdict(set)    for path, mod in modules.items():        targets: list[ChunkNode] = []        for spec in mod.imports:            if mod.language == "python":                hit = _resolve_python_import(spec, py_index)                targets.extend([hit] if hit else [])            else:                targets.extend(_resolve_path_import(path, spec, mod.language, modules, by_basename))        for t in targets:            if t.path != path:                imported_files[path].add(t.path)                if add(mod, t, t.qualified_name, "imports", "imported_file"):                    stats.imports += 1    # calls    for n in nodes:        if not n.calls or n.kind == "block":            continue        file_nodes = by_file[n.path]        module = modules.get(n.path)        import_map = module.import_map if module else {}        for name in n.calls:            if name in _IGNORED_CALLS or len(name) < 2:                continue            target, resolution = None, None            same = [c for c in file_nodes if c.name == name and c.kind != "module"]            if len(same) == 1:                target, resolution = same[0], "same_file"            elif len(same) > 1:                stats.calls_ambiguous += 1                continue            if target is None and name in import_map:                qual = import_map[name]                hit = by_qualified.get(qual)                if hit is None:                    # Imported module path may carry a src/ prefix in the repo.                    cands = [c for q, c in by_qualified.items() if q.endswith("." + qual)]                    hit = cands[0] if len(cands) == 1 else None                if hit is not None and hit.kind != "module":                    target, resolution = hit, "imported_file"            # Python has an exact import map (above); this file-level            # heuristic is for languages without one (JS, Go, C, ...).            if target is None and n.language != "python" and imported_files.get(n.path):                cands = [c for c in by_name.get(name, []) if c.path in imported_files[n.path]]                if len(cands) == 1:                    target, resolution = cands[0], "imported_file"            if target is None:                cands = by_name.get(name, [])                if len(cands) == 1:                    target, resolution = cands[0], "unique_name"                elif len(cands) > 1:                    stats.calls_ambiguous += 1                    continue            if target is None:                stats.calls_unresolved += 1                continue            if add(n, target, name, "calls", resolution):                stats.calls_resolved += 1                stats.by_resolution[resolution] += 1    return edges, stats

Where this comes from

Commit
975a38d5d17839c26ee219295a5be6ac66b6d910
File
backend/app/codeintel/graph.py
Lines
144–244
Git blob
3791a0283d18722675047af567570cb2c1911c63
License
MIT, © 2024–2026 Pharos Contributors. Notice

Read it on GitHub at this commit

Check it yourself

From a clone of the repository, the first command prints the blob above and the second prints these lines.

git rev-parse 975a38d:backend/app/codeintel/graph.py
git show 975a38d:backend/app/codeintel/graph.py | sed -n '144,244p'