codemap.py
python
sha256:be3641f35bdbcc094677776a77b9aa6a5dab891f8fab201dc162d03c2bab5aea
fix(read): strip position:null from structured_delta ops in…
Sonnet 4.6
patch
23 days ago
| 1 | """muse code codemap — repository semantic topology. |
| 2 | |
| 3 | Generates a structural map of the codebase from committed snapshot data: |
| 4 | |
| 5 | * **Modules ranked by size** — symbol count and lines of code per file |
| 6 | * **Import in-degree** — how many other files import each module |
| 7 | * **Import cycles** — circular dependency chains detected via iterative DFS |
| 8 | * **High-centrality symbols** — functions called from the most callers |
| 9 | * **Boundary files** — high fan-out (imports many) but low fan-in (few import it) |
| 10 | * **Agent-safe zones** — completely isolated files with no import coupling |
| 11 | |
| 12 | This is a semantic topology view, not a file-system listing. It reveals the |
| 13 | actual shape of a codebase — where the load-bearing columns are, where the |
| 14 | cycles hide, and where parallel agents can safely work without collision. |
| 15 | |
| 16 | Usage:: |
| 17 | |
| 18 | muse code codemap |
| 19 | muse code codemap --commit HEAD~10 |
| 20 | muse code codemap --language Python |
| 21 | muse code codemap --top 20 |
| 22 | muse code codemap --min-importers 2 |
| 23 | muse code codemap --json |
| 24 | |
| 25 | Output:: |
| 26 | |
| 27 | Semantic codemap — commit a1b2c3d4 |
| 28 | |
| 29 | Top modules by size: |
| 30 | src/billing.py 42 symbols (12 importers) ⬛ HIGH CENTRALITY |
| 31 | src/models.py 31 symbols (8 importers) |
| 32 | src/auth.py 18 symbols (5 importers) |
| 33 | |
| 34 | Import cycles (2): |
| 35 | src/billing.py → src/utils.py → src/billing.py |
| 36 | src/api.py → src/auth.py → src/api.py |
| 37 | |
| 38 | High-centrality symbols (most callers): |
| 39 | src/billing.py::compute_total 14 callers |
| 40 | src/auth.py::validate_token 9 callers |
| 41 | |
| 42 | Boundary files (high fan-out, low fan-in): |
| 43 | src/cli.py imports 8 modules ← imported by 0 |
| 44 | |
| 45 | Agent-safe zones (no import coupling — safe for parallel work): |
| 46 | src/utils.py |
| 47 | src/constants.py |
| 48 | |
| 49 | Flags: |
| 50 | |
| 51 | ``--commit, -c REF`` |
| 52 | Analyse a historical snapshot instead of HEAD. |
| 53 | |
| 54 | ``--language LANG`` |
| 55 | Restrict analysis to files of this language. |
| 56 | |
| 57 | ``--top N`` |
| 58 | Show top N entries in each section (default: 15, must be ≥ 1). |
| 59 | |
| 60 | ``--min-importers N`` |
| 61 | Only include modules imported by at least N other files in the ranked |
| 62 | module list (default: 0 = show all). |
| 63 | |
| 64 | ``--json`` |
| 65 | Emit the full codemap as JSON. |
| 66 | """ |
| 67 | |
| 68 | import argparse |
| 69 | import json |
| 70 | import logging |
| 71 | import pathlib |
| 72 | from typing import TypedDict |
| 73 | |
| 74 | from muse.core.envelope import EnvelopeJson, make_envelope |
| 75 | from muse.core.errors import ExitCode |
| 76 | from muse.core.repo import require_repo |
| 77 | from muse.core.types import Manifest |
| 78 | from muse.core.refs import read_current_branch |
| 79 | from muse.core.commits import resolve_commit_ref |
| 80 | from muse.core.snapshots import get_commit_snapshot_manifest |
| 81 | from muse.core.symbol_cache import load_symbol_cache |
| 82 | from muse.core.timing import start_timer |
| 83 | from muse.plugins.code._callgraph import build_reverse_graph |
| 84 | from muse.plugins.code._query import symbols_for_snapshot |
| 85 | from muse.plugins.code.ast_parser import SymbolTree |
| 86 | from muse.core.validation import clamp_int, sanitize_display |
| 87 | |
| 88 | type _SymbolTreeMap = dict[str, SymbolTree] |
| 89 | type _ImportOut = dict[str, list[str]] |
| 90 | type _CounterMap = dict[str, int] |
| 91 | |
| 92 | logger = logging.getLogger(__name__) |
| 93 | |
| 94 | # --------------------------------------------------------------------------- |
| 95 | # Typed output shape |
| 96 | # --------------------------------------------------------------------------- |
| 97 | |
| 98 | class _ModuleEntry(TypedDict): |
| 99 | file: str |
| 100 | symbol_count: int |
| 101 | importers: int |
| 102 | imports: int |
| 103 | |
| 104 | class _CentralityEntry(TypedDict): |
| 105 | name: str |
| 106 | callers: int |
| 107 | |
| 108 | class _BoundaryEntry(TypedDict): |
| 109 | file: str |
| 110 | fan_out: int |
| 111 | fan_in: int |
| 112 | |
| 113 | class _CodemapOutputJson(EnvelopeJson): |
| 114 | """Top-level JSON envelope emitted by ``muse code codemap --json``. |
| 115 | |
| 116 | Fields |
| 117 | ------ |
| 118 | schema_version Muse version string. |
| 119 | commit Short commit ID that was analysed. |
| 120 | branch Current branch name. |
| 121 | language_filter Language filter applied, or None. |
| 122 | modules Files ranked by symbol count. |
| 123 | import_cycles Detected circular dependency chains. |
| 124 | high_centrality Symbols with the most callers. |
| 125 | boundary_files Files with high fan-out and zero fan-in. |
| 126 | agent_safe_zones Files with no import coupling (safe for parallel work). |
| 127 | exit_code Always 0 — errors raise SystemExit before JSON emits. |
| 128 | duration_ms Wall-clock time for the full analysis in milliseconds. |
| 129 | """ |
| 130 | |
| 131 | commit: str |
| 132 | branch: str |
| 133 | language_filter: str | None |
| 134 | modules: list[_ModuleEntry] |
| 135 | import_cycles: list[list[str]] |
| 136 | high_centrality: list[_CentralityEntry] |
| 137 | boundary_files: list[_BoundaryEntry] |
| 138 | agent_safe_zones: list[str] |
| 139 | |
| 140 | def _build_import_graph( |
| 141 | sym_map: _SymbolTreeMap, |
| 142 | ) -> tuple[_ImportOut, _CounterMap]: |
| 143 | """Return ``(imports_out, in_degree)`` for the files in *sym_map*. |
| 144 | |
| 145 | Builds a stem-based heuristic import graph: for each import symbol in |
| 146 | each file, check whether the imported module stem matches a known file in |
| 147 | the map. Edges are deduplicated so a file importing the same module via |
| 148 | multiple ``from X import a, b`` statements counts as one edge. |
| 149 | |
| 150 | Args: |
| 151 | sym_map: Pre-parsed symbol trees, keyed by file path. Already |
| 152 | filtered by language when the caller applies a filter. |
| 153 | |
| 154 | Returns: |
| 155 | ``imports_out`` — adjacency list: file → list of files it imports. |
| 156 | ``in_degree`` — import fan-in count per file. |
| 157 | """ |
| 158 | stem_to_file: Manifest = { |
| 159 | pathlib.PurePosixPath(fp).stem: fp for fp in sym_map |
| 160 | } |
| 161 | |
| 162 | imports_out: _ImportOut = {fp: [] for fp in sym_map} |
| 163 | in_degree: _CounterMap = {fp: 0 for fp in sym_map} |
| 164 | |
| 165 | for file_path, tree in sym_map.items(): |
| 166 | seen_targets: set[str] = set() |
| 167 | for rec in tree.values(): |
| 168 | if rec["kind"] != "import": |
| 169 | continue |
| 170 | # rec["name"] is the bare module name (e.g. "utils" for both |
| 171 | # `import utils` and `from utils import X`). |
| 172 | target = stem_to_file.get(rec["name"]) |
| 173 | if target and target != file_path and target not in seen_targets: |
| 174 | seen_targets.add(target) |
| 175 | imports_out[file_path].append(target) |
| 176 | in_degree[target] += 1 |
| 177 | |
| 178 | return imports_out, in_degree |
| 179 | |
| 180 | def _find_cycles(imports_out: _ImportOut) -> list[list[str]]: |
| 181 | """Detect import cycles via iterative DFS. Returns cycle paths. |
| 182 | |
| 183 | Uses an explicit stack instead of recursion so that deeply nested import |
| 184 | graphs (thousands of files in a chain) cannot exhaust Python's call stack. |
| 185 | O(V+E) — every node is visited at most once globally. |
| 186 | |
| 187 | The ``in_stack`` dict maps each node on the current DFS path to its index |
| 188 | in ``path``, giving O(1) lookups for both cycle detection and cycle |
| 189 | extraction (replacing the previous O(N) ``path.index()`` call). |
| 190 | """ |
| 191 | cycles: list[list[str]] = [] |
| 192 | visited: set[str] = set() |
| 193 | |
| 194 | for start in imports_out: |
| 195 | if start in visited: |
| 196 | continue |
| 197 | # Stack frame: (node, path-so-far, in_stack: node→index-in-path) |
| 198 | stack: list[tuple[str, list[str], dict[str, int]]] = [(start, [], {})] |
| 199 | while stack: |
| 200 | node, path, in_stack = stack.pop() |
| 201 | if node in in_stack: |
| 202 | idx = in_stack[node] |
| 203 | cycles.append(path[idx:] + [node]) |
| 204 | continue |
| 205 | if node in visited: |
| 206 | continue |
| 207 | visited.add(node) |
| 208 | new_in_stack = {**in_stack, node: len(path)} |
| 209 | new_path = path + [node] |
| 210 | for neighbour in imports_out.get(node, []): |
| 211 | stack.append((neighbour, new_path, new_in_stack)) |
| 212 | |
| 213 | return cycles |
| 214 | |
| 215 | def register(subparsers: "argparse._SubParsersAction[argparse.ArgumentParser]") -> None: |
| 216 | """Register the codemap subcommand.""" |
| 217 | parser = subparsers.add_parser( |
| 218 | "codemap", |
| 219 | help="Generate a semantic topology map of the repository.", |
| 220 | description=__doc__, |
| 221 | formatter_class=argparse.RawDescriptionHelpFormatter, |
| 222 | ) |
| 223 | parser.add_argument( |
| 224 | "--commit", "-c", default=None, metavar="REF", dest="ref", |
| 225 | help="Analyse this commit instead of HEAD.", |
| 226 | ) |
| 227 | parser.add_argument( |
| 228 | "--language", "-l", default=None, metavar="LANG", dest="language", |
| 229 | help="Restrict analysis to this language.", |
| 230 | ) |
| 231 | parser.add_argument( |
| 232 | "--top", "-n", type=int, default=15, metavar="N", dest="top", |
| 233 | help="Number of entries to show in each ranked section (must be ≥ 1).", |
| 234 | ) |
| 235 | parser.add_argument( |
| 236 | "--min-importers", type=int, default=0, metavar="N", dest="min_importers", |
| 237 | help="Only show modules imported by at least N other files (default: 0 = all).", |
| 238 | ) |
| 239 | parser.add_argument( |
| 240 | "--json", "-j", action="store_true", dest="json_out", |
| 241 | help="Emit results as JSON.", |
| 242 | ) |
| 243 | parser.set_defaults(func=run) |
| 244 | |
| 245 | def run(args: argparse.Namespace) -> None: |
| 246 | """Generate a semantic topology map of the repository. |
| 247 | |
| 248 | Ranks modules by size, detects import cycles, finds high-centrality symbols, |
| 249 | identifies boundary files (high fan-out, low fan-in), and surfaces |
| 250 | agent-safe zones — files with no import coupling safe for parallel work. |
| 251 | All analysis runs from the committed snapshot; the working tree is never read. |
| 252 | |
| 253 | Agent quickstart |
| 254 | ---------------- |
| 255 | :: |
| 256 | |
| 257 | muse code codemap --json |
| 258 | muse code codemap --language Python --top 20 --json |
| 259 | muse code codemap --ref dev --json |
| 260 | |
| 261 | JSON fields |
| 262 | ----------- |
| 263 | ref Commit ref analysed. |
| 264 | language_filter Language filter applied (``null`` if none). |
| 265 | modules List of module objects ranked by size: ``file``, |
| 266 | ``symbol_count``, ``importers``, ``imports``. |
| 267 | cycles List of import cycle groups (each a list of file paths). |
| 268 | high_centrality List of symbol addresses with the highest centrality. |
| 269 | boundary_files Files with high fan-out and low fan-in. |
| 270 | agent_safe Files with no import coupling (safe for parallel work). |
| 271 | exit_code Integer exit code (see below). |
| 272 | |
| 273 | Exit codes |
| 274 | ---------- |
| 275 | 0 Analysis complete. |
| 276 | 1 Invalid arguments or commit not found. |
| 277 | 2 Not inside a Muse repository. |
| 278 | """ |
| 279 | elapsed = start_timer() |
| 280 | ref: str | None = args.ref |
| 281 | language: str | None = args.language |
| 282 | top: int = clamp_int(args.top, 1, 10_000, 'top') |
| 283 | min_importers: int = clamp_int(args.min_importers, 0, 10000, 'min_importers') |
| 284 | json_out: bool = args.json_out |
| 285 | |
| 286 | if top < 1: |
| 287 | logger.error("--top must be at least 1, got %d", top) |
| 288 | raise SystemExit(ExitCode.USER_ERROR) |
| 289 | if min_importers < 0: |
| 290 | logger.error("--min-importers must be non-negative, got %d", min_importers) |
| 291 | raise SystemExit(ExitCode.USER_ERROR) |
| 292 | |
| 293 | root = require_repo() |
| 294 | |
| 295 | branch = read_current_branch(root) |
| 296 | commit = resolve_commit_ref(root, branch, ref) |
| 297 | if commit is None: |
| 298 | logger.error("Commit %r not found.", ref or "HEAD") |
| 299 | raise SystemExit(ExitCode.USER_ERROR) |
| 300 | |
| 301 | manifest = get_commit_snapshot_manifest(root, commit.commit_id) or {} |
| 302 | |
| 303 | # Single shared cache — each blob is parsed at most once across all passes. |
| 304 | cache = load_symbol_cache(root) |
| 305 | |
| 306 | sym_map = symbols_for_snapshot(root, manifest, language_filter=language, cache=cache) |
| 307 | |
| 308 | if language and not sym_map: |
| 309 | logger.warning("No files matched language filter %r — output will be empty.", language) |
| 310 | |
| 311 | file_sym_counts: _CounterMap = {fp: len(tree) for fp, tree in sym_map.items()} |
| 312 | |
| 313 | # Import graph built from the pre-parsed sym_map (no second parse pass). |
| 314 | imports_out, in_degree = _build_import_graph(sym_map) |
| 315 | |
| 316 | cycles = _find_cycles(imports_out) |
| 317 | |
| 318 | # Call graph uses the shared cache (no re-parse of Python blobs). |
| 319 | reverse = build_reverse_graph(root, manifest, cache=cache) |
| 320 | centrality: list[tuple[str, int]] = sorted( |
| 321 | ((name, len(callers)) for name, callers in reverse.items()), |
| 322 | key=lambda t: t[1], |
| 323 | reverse=True, |
| 324 | )[:top] |
| 325 | |
| 326 | # Boundary files: imports many but is imported by few. |
| 327 | fan_out = {fp: len(targets) for fp, targets in imports_out.items() if targets} |
| 328 | boundaries: list[tuple[str, int, int]] = sorted( |
| 329 | [ |
| 330 | (fp, fan_out.get(fp, 0), in_degree.get(fp, 0)) |
| 331 | for fp in sym_map |
| 332 | if fan_out.get(fp, 0) >= 3 and in_degree.get(fp, 0) == 0 |
| 333 | ], |
| 334 | key=lambda t: t[1], |
| 335 | reverse=True, |
| 336 | )[:top] |
| 337 | |
| 338 | # Agent-safe zones: completely decoupled files (no imports in or out). |
| 339 | agent_safe: list[str] = sorted( |
| 340 | fp for fp in sym_map |
| 341 | if in_degree.get(fp, 0) == 0 and not imports_out.get(fp) |
| 342 | )[:top] |
| 343 | |
| 344 | # Ranked modules — optionally filtered by --min-importers. |
| 345 | ranked: list[tuple[str, int]] = sorted( |
| 346 | ( |
| 347 | (fp, cnt) for fp, cnt in file_sym_counts.items() |
| 348 | if in_degree.get(fp, 0) >= min_importers |
| 349 | ), |
| 350 | key=lambda t: t[1], |
| 351 | reverse=True, |
| 352 | )[:top] |
| 353 | |
| 354 | if json_out: |
| 355 | print(json.dumps(_CodemapOutputJson( |
| 356 | **make_envelope(elapsed), |
| 357 | commit=commit.commit_id, |
| 358 | branch=branch, |
| 359 | language_filter=language, |
| 360 | modules=[ |
| 361 | _ModuleEntry( |
| 362 | file=fp, |
| 363 | symbol_count=cnt, |
| 364 | importers=in_degree.get(fp, 0), |
| 365 | imports=len(imports_out.get(fp, [])), |
| 366 | ) |
| 367 | for fp, cnt in ranked |
| 368 | ], |
| 369 | import_cycles=cycles, |
| 370 | high_centrality=[ |
| 371 | _CentralityEntry(name=name, callers=cnt) |
| 372 | for name, cnt in centrality |
| 373 | ], |
| 374 | boundary_files=[ |
| 375 | _BoundaryEntry(file=fp, fan_out=fo, fan_in=fi) |
| 376 | for fp, fo, fi in boundaries |
| 377 | ], |
| 378 | agent_safe_zones=agent_safe, |
| 379 | ))) |
| 380 | return |
| 381 | |
| 382 | print(f"\nSemantic codemap — commit {commit.commit_id}") |
| 383 | if language: |
| 384 | print(f" (language: {language})") |
| 385 | if min_importers: |
| 386 | print(f" (min-importers: {min_importers})") |
| 387 | print("─" * 62) |
| 388 | |
| 389 | print(f"\nTop modules by size (top {min(top, len(ranked))}):") |
| 390 | if ranked: |
| 391 | max_fp = max(len(fp) for fp, _ in ranked) |
| 392 | for fp, cnt in ranked: |
| 393 | imp = in_degree.get(fp, 0) |
| 394 | imp_label = f"({imp} importers)" if imp else "(not imported)" |
| 395 | print(f" {sanitize_display(fp):<{max_fp}} {cnt:>3} symbols {imp_label}") |
| 396 | else: |
| 397 | print(" (no files match the current filters)") |
| 398 | |
| 399 | print(f"\nImport cycles ({len(cycles)}):") |
| 400 | if cycles: |
| 401 | for cycle in cycles[:top]: |
| 402 | print(f" {' → '.join(cycle)}") |
| 403 | else: |
| 404 | print(" ✅ No import cycles detected") |
| 405 | |
| 406 | print(f"\nHigh-centrality symbols — most callers (Python):") |
| 407 | if centrality: |
| 408 | for name, cnt in centrality: |
| 409 | print(f" {sanitize_display(name):<40} {cnt} caller(s)") |
| 410 | else: |
| 411 | print(" (no Python call graph available)") |
| 412 | |
| 413 | print(f"\nBoundary files — high fan-out, zero fan-in:") |
| 414 | if boundaries: |
| 415 | for fp, fo, fi in boundaries: |
| 416 | print(f" {sanitize_display(fp)} imports {fo} ← imported by {fi}") |
| 417 | else: |
| 418 | print(" (none detected)") |
| 419 | |
| 420 | print(f"\nAgent-safe zones — no import coupling ({len(agent_safe)}):") |
| 421 | if agent_safe: |
| 422 | for fp in agent_safe: |
| 423 | print(f" {sanitize_display(fp)}") |
| 424 | else: |
| 425 | print(" (all files are coupled via imports)") |
| 426 |
File History
3 commits
sha256:36c3cb3e76619d4c30a6d9bf81b5ec4ff148e30dcfed913e3114ca7b43b81c7e
fix: rename objects→blobs in push client and all stale test…
Sonnet 4.6
patch
22 days ago
sha256:c06a9b9b9fee26c68ea725b44d54b2c0a171301ce9de746d5b656617b4463a9a
fix: repair four test failures from post-migration audit
Sonnet 4.6
patch
29 days ago
sha256:1900655993c83c4107067375548a7be823e471d2515830842f1a12cba4bd3cdf
fix: unified object store migration — idempotent writes, JS…
Sonnet 4.6
minor
⚠
29 days ago