Function full_build

code_review_graph/incremental.py:820–916

Full rebuild of the entire graph. Args: repo_root: Repository root directory. store: Graph database store. recurse_submodules: If True, include files from git submodules. When *None*, falls back to ``CRG_RECURSE_SUBMODULES`` env var.

(
    repo_root: Path,
    store: GraphStore,
    recurse_submodules: bool | None = None,
)

def full_build(
    repo_root: Path,
    store: GraphStore,
    recurse_submodules: bool | None = None,
) -> dict:
    """Full rebuild of the entire graph.

    Args:
        repo_root: Repository root directory.
        store: Graph database store.
        recurse_submodules: If True, include files from git submodules.
            When *None*, falls back to ``CRG_RECURSE_SUBMODULES`` env var.
    """
    parser = CodeParser(repo_root)
    files = collect_all_files(repo_root, recurse_submodules)

    # Purge stale data from files no longer on disk
    existing_files = set(store.get_all_files())
    current_abs = {str(repo_root / f) for f in files}
    stale_files = existing_files - current_abs
    for stale in stale_files:
        store.remove_file_data(stale)
    # Ensure deletions are persisted before store_file_nodes_edges()
    # starts its own explicit transaction via BEGIN IMMEDIATE.
    if stale_files:
        store.commit()

    total_nodes = 0
    total_edges = 0
    errors = []
    file_count = len(files)

    use_serial = os.environ.get("CRG_SERIAL_PARSE", "") == "1"

    if use_serial or file_count < 8:
        # Serial fallback (for debugging or tiny repos)
        for i, rel_path in enumerate(files, 1):
            full_path = repo_root / rel_path
            try:
                source = full_path.read_bytes()
                fhash = hashlib.sha256(source).hexdigest()
                nodes, edges = parser.parse_bytes(full_path, source)
                store.store_file_nodes_edges(str(full_path), nodes, edges, fhash)
                total_nodes += len(nodes)
                total_edges += len(edges)
            except (OSError, PermissionError) as e:
                errors.append({"file": rel_path, "error": str(e)})
            except Exception as e:
                logger.warning("Error parsing %s: %s", rel_path, e)
                errors.append({"file": rel_path, "error": str(e)})
            if i % 50 == 0 or i == file_count:
                logger.info("Progress: %d/%d files parsed", i, file_count)
    else:
        # Parallel parsing — store calls remain serial (SQLite single-writer).
        # Executor kind auto-selected: process on Linux/macOS/Windows-TTY,
        # thread on Windows-MCP-stdio to avoid pipe-handle inheritance
        # deadlock (issues #46, #136). Override via CRG_PARSE_EXECUTOR env.
        args_list = [(rel_path, str(repo_root)) for rel_path in files]