attributes.py
python
| 1 | """Muse attributes — ``.museattributes`` TOML parser and per-path strategy resolver. |
| 2 | |
| 3 | ``.museattributes`` lives in the repository root (next to ``.muse/`` and |
| 4 | ``muse-work/``) and declares merge strategies for specific paths and |
| 5 | dimensions. It uses TOML syntax with an optional ``[meta]`` section for |
| 6 | domain declaration and an ordered ``[[rules]]`` array. |
| 7 | |
| 8 | Format |
| 9 | ------ |
| 10 | |
| 11 | .. code-block:: toml |
| 12 | |
| 13 | # .museattributes |
| 14 | # Merge strategy overrides for this repository. |
| 15 | |
| 16 | [meta] |
| 17 | domain = "midi" # optional — validated against .muse/repo.json |
| 18 | |
| 19 | [[rules]] |
| 20 | path = "drums/*" # fnmatch glob against workspace-relative POSIX paths |
| 21 | dimension = "*" # domain axis name, or "*" to match any dimension |
| 22 | strategy = "ours" # resolution strategy (see below) |
| 23 | comment = "Drums are always authored by branch A — always prefer ours." |
| 24 | priority = 10 # optional; higher priority rules are tried first |
| 25 | |
| 26 | [[rules]] |
| 27 | path = "keys/*" |
| 28 | dimension = "pitch_bend" |
| 29 | strategy = "theirs" |
| 30 | comment = "Remote always has the better pitch-bend automation." |
| 31 | |
| 32 | [[rules]] |
| 33 | path = "*" |
| 34 | dimension = "*" |
| 35 | strategy = "auto" |
| 36 | |
| 37 | Strategies |
| 38 | ---------- |
| 39 | |
| 40 | ``ours`` |
| 41 | Take the left / current-branch version; the path is removed from the |
| 42 | conflict list. |
| 43 | |
| 44 | ``theirs`` |
| 45 | Take the right / incoming-branch version; the path is removed from the |
| 46 | conflict list. |
| 47 | |
| 48 | ``union`` |
| 49 | Include **all** additions from both sides. Deletions are honoured only |
| 50 | when **both** sides agree. For independent element sets (MIDI notes, |
| 51 | code symbol additions, import sets) this produces a combined result with |
| 52 | no conflicts. For opaque binary blobs where full unification is |
| 53 | impossible, the left / current-branch blob is preferred and the path is |
| 54 | removed from the conflict list. |
| 55 | |
| 56 | ``base`` |
| 57 | Revert to the common merge-base version — discard changes from *both* |
| 58 | branches. Useful for generated files, lock files, or any path that |
| 59 | should always stay at a known-good state during a merge. |
| 60 | |
| 61 | ``auto`` |
| 62 | Default behaviour. Defer to the engine's three-way algorithm. |
| 63 | |
| 64 | ``manual`` |
| 65 | Force the path into the conflict list even if the engine would |
| 66 | auto-resolve it. Use this to guarantee human review on safety-critical |
| 67 | paths. |
| 68 | |
| 69 | Rule fields |
| 70 | ----------- |
| 71 | |
| 72 | ``path`` (required) — ``fnmatch`` glob matched against workspace-relative |
| 73 | POSIX paths (e.g. ``"tracks/*.mid"``, ``"src/**/*.py"``). |
| 74 | |
| 75 | ``dimension`` (required) — domain axis name (e.g. ``"notes"``, |
| 76 | ``"pitch_bend"``, ``"symbols"``) or ``"*"`` to match any |
| 77 | dimension. |
| 78 | |
| 79 | ``strategy`` (required) — one of the six strategies listed above. |
| 80 | |
| 81 | ``comment`` (optional) — free-form documentation string; ignored at |
| 82 | runtime. Use it to explain *why* the rule exists. |
| 83 | |
| 84 | ``priority`` (optional, default 0) — integer used to order rules before |
| 85 | file order. Higher-priority rules are evaluated first. Rules |
| 86 | with equal priority preserve their declaration order. |
| 87 | |
| 88 | **First matching rule wins** after sorting by priority (descending) then |
| 89 | file order (ascending). |
| 90 | |
| 91 | ``[meta]`` is optional; its absence has no effect on merge correctness. |
| 92 | When both ``[meta] domain`` and a repo ``domain`` are known, a mismatch |
| 93 | logs a warning. |
| 94 | |
| 95 | Public API |
| 96 | ---------- |
| 97 | |
| 98 | - :class:`AttributesMeta` — TypedDict for the ``[meta]`` section. |
| 99 | - :class:`AttributesRuleDict` — TypedDict for a single ``[[rules]]`` entry. |
| 100 | - :class:`MuseAttributesFile` — TypedDict for the full parsed file. |
| 101 | - :class:`AttributeRule` — a single resolved rule (dataclass). |
| 102 | - :func:`read_attributes_meta` — read only the ``[meta]`` section. |
| 103 | - :func:`load_attributes` — read ``.museattributes`` from a repo root. |
| 104 | - :func:`resolve_strategy` — first-match strategy lookup. |
| 105 | """ |
| 106 | |
| 107 | from __future__ import annotations |
| 108 | |
| 109 | import fnmatch |
| 110 | import logging |
| 111 | import pathlib |
| 112 | import tomllib |
| 113 | from dataclasses import dataclass, field |
| 114 | from typing import TypedDict |
| 115 | |
| 116 | _logger = logging.getLogger(__name__) |
| 117 | |
| 118 | VALID_STRATEGIES: frozenset[str] = frozenset( |
| 119 | {"ours", "theirs", "union", "base", "auto", "manual"} |
| 120 | ) |
| 121 | |
| 122 | _FILENAME = ".museattributes" |
| 123 | |
| 124 | |
| 125 | class AttributesMeta(TypedDict, total=False): |
| 126 | """Typed representation of the ``[meta]`` section in ``.museattributes``.""" |
| 127 | |
| 128 | domain: str |
| 129 | |
| 130 | |
| 131 | class AttributesRuleDict(TypedDict, total=False): |
| 132 | """Typed representation of a single ``[[rules]]`` entry. |
| 133 | |
| 134 | ``path``, ``dimension``, and ``strategy`` are required at parse time. |
| 135 | ``comment`` and ``priority`` are optional. |
| 136 | """ |
| 137 | |
| 138 | path: str |
| 139 | dimension: str |
| 140 | strategy: str |
| 141 | comment: str |
| 142 | priority: int |
| 143 | |
| 144 | |
| 145 | class MuseAttributesFile(TypedDict, total=False): |
| 146 | """Typed representation of the complete ``.museattributes`` file.""" |
| 147 | |
| 148 | meta: AttributesMeta |
| 149 | rules: list[AttributesRuleDict] |
| 150 | |
| 151 | |
| 152 | @dataclass(frozen=True) |
| 153 | class AttributeRule: |
| 154 | """A single rule resolved from ``.museattributes``. |
| 155 | |
| 156 | Attributes: |
| 157 | path_pattern: ``fnmatch`` glob matched against workspace-relative paths. |
| 158 | dimension: Domain axis name (e.g. ``"notes"``) or ``"*"``. |
| 159 | strategy: Resolution strategy: one of ``ours | theirs | union | |
| 160 | base | auto | manual``. |
| 161 | comment: Human-readable annotation explaining the rule's purpose. |
| 162 | Ignored at runtime. |
| 163 | priority: Ordering weight. Higher values are evaluated before |
| 164 | lower values. Rules with equal priority preserve |
| 165 | declaration order. |
| 166 | source_index: 0-based index of the rule in the ``[[rules]]`` array. |
| 167 | """ |
| 168 | |
| 169 | path_pattern: str |
| 170 | dimension: str |
| 171 | strategy: str |
| 172 | comment: str = "" |
| 173 | priority: int = 0 |
| 174 | source_index: int = 0 |
| 175 | |
| 176 | |
| 177 | def _parse_raw(root: pathlib.Path) -> MuseAttributesFile: |
| 178 | """Read and TOML-parse ``.museattributes``, returning a typed file structure. |
| 179 | |
| 180 | Builds ``MuseAttributesFile`` from the raw TOML dict using explicit |
| 181 | ``isinstance`` checks — no ``Any`` propagated into the return value. |
| 182 | |
| 183 | Raises: |
| 184 | ValueError: On TOML syntax errors. |
| 185 | """ |
| 186 | attr_file = root / _FILENAME |
| 187 | raw_bytes = attr_file.read_bytes() |
| 188 | try: |
| 189 | raw = tomllib.loads(raw_bytes.decode("utf-8")) |
| 190 | except tomllib.TOMLDecodeError as exc: |
| 191 | raise ValueError(f"{_FILENAME}: TOML parse error — {exc}") from exc |
| 192 | |
| 193 | result: MuseAttributesFile = {} |
| 194 | |
| 195 | # [meta] section |
| 196 | meta_raw = raw.get("meta") |
| 197 | if isinstance(meta_raw, dict): |
| 198 | meta: AttributesMeta = {} |
| 199 | domain_val = meta_raw.get("domain") |
| 200 | if isinstance(domain_val, str): |
| 201 | meta["domain"] = domain_val |
| 202 | result["meta"] = meta |
| 203 | |
| 204 | # [[rules]] array |
| 205 | rules_raw = raw.get("rules") |
| 206 | if isinstance(rules_raw, list): |
| 207 | rules: list[AttributesRuleDict] = [] |
| 208 | for idx, entry in enumerate(rules_raw): |
| 209 | if not isinstance(entry, dict): |
| 210 | continue |
| 211 | path_val = entry.get("path") |
| 212 | dim_val = entry.get("dimension") |
| 213 | strat_val = entry.get("strategy") |
| 214 | if ( |
| 215 | isinstance(path_val, str) |
| 216 | and isinstance(dim_val, str) |
| 217 | and isinstance(strat_val, str) |
| 218 | ): |
| 219 | rule: AttributesRuleDict = { |
| 220 | "path": path_val, |
| 221 | "dimension": dim_val, |
| 222 | "strategy": strat_val, |
| 223 | } |
| 224 | comment_val = entry.get("comment") |
| 225 | if isinstance(comment_val, str): |
| 226 | rule["comment"] = comment_val |
| 227 | priority_val = entry.get("priority") |
| 228 | if isinstance(priority_val, int): |
| 229 | rule["priority"] = priority_val |
| 230 | rules.append(rule) |
| 231 | else: |
| 232 | missing = [ |
| 233 | f |
| 234 | for f, v in ( |
| 235 | ("path", path_val), |
| 236 | ("dimension", dim_val), |
| 237 | ("strategy", strat_val), |
| 238 | ) |
| 239 | if not isinstance(v, str) |
| 240 | ] |
| 241 | raise ValueError( |
| 242 | f"{_FILENAME}: rule[{idx}] is missing required field(s): " |
| 243 | + ", ".join(missing) |
| 244 | ) |
| 245 | result["rules"] = rules |
| 246 | |
| 247 | return result |
| 248 | |
| 249 | |
| 250 | def read_attributes_meta(root: pathlib.Path) -> AttributesMeta: |
| 251 | """Return the ``[meta]`` section of ``.museattributes``, or an empty dict. |
| 252 | |
| 253 | Does not validate or resolve rules — use this to inspect metadata only. |
| 254 | |
| 255 | Args: |
| 256 | root: Repository root directory. |
| 257 | |
| 258 | Returns: |
| 259 | The ``[meta]`` TypedDict, which may be empty if the section is absent |
| 260 | or the file does not exist. |
| 261 | """ |
| 262 | attr_file = root / _FILENAME |
| 263 | if not attr_file.exists(): |
| 264 | return {} |
| 265 | try: |
| 266 | parsed = _parse_raw(root) |
| 267 | except ValueError: |
| 268 | return {} |
| 269 | meta = parsed.get("meta") |
| 270 | if meta is None: |
| 271 | return {} |
| 272 | return meta |
| 273 | |
| 274 | |
| 275 | def load_attributes( |
| 276 | root: pathlib.Path, |
| 277 | *, |
| 278 | domain: str | None = None, |
| 279 | ) -> list[AttributeRule]: |
| 280 | """Parse ``.museattributes`` from *root* and return the ordered rule list. |
| 281 | |
| 282 | Rules are sorted by ``priority`` (descending) then by declaration order |
| 283 | (ascending), so higher-priority rules are evaluated first. |
| 284 | |
| 285 | Args: |
| 286 | root: Repository root directory (the directory that contains ``.muse/`` |
| 287 | and ``muse-work/``). |
| 288 | domain: Optional domain name from the active repository. When provided |
| 289 | and the file contains ``[meta] domain``, a mismatch logs a |
| 290 | warning. Pass ``None`` to skip domain validation. |
| 291 | |
| 292 | Returns: |
| 293 | A list of :class:`AttributeRule` sorted by priority then file order. |
| 294 | Returns an empty list when ``.museattributes`` is absent or contains |
| 295 | no valid rules. |
| 296 | |
| 297 | Raises: |
| 298 | ValueError: If a rule entry is missing required fields, or contains an |
| 299 | invalid strategy. |
| 300 | """ |
| 301 | attr_file = root / _FILENAME |
| 302 | if not attr_file.exists(): |
| 303 | return [] |
| 304 | |
| 305 | data = _parse_raw(root) |
| 306 | |
| 307 | # Domain validation |
| 308 | meta = data.get("meta", {}) |
| 309 | file_domain = meta.get("domain") if meta else None |
| 310 | if file_domain and domain and file_domain != domain: |
| 311 | _logger.warning( |
| 312 | "⚠️ %s: [meta] domain %r does not match active repo domain %r — " |
| 313 | "rules may target a different domain", |
| 314 | _FILENAME, |
| 315 | file_domain, |
| 316 | domain, |
| 317 | ) |
| 318 | |
| 319 | raw_rules = data.get("rules", []) |
| 320 | |
| 321 | rules: list[AttributeRule] = [] |
| 322 | for idx, entry in enumerate(raw_rules): |
| 323 | strategy = entry["strategy"] |
| 324 | if strategy not in VALID_STRATEGIES: |
| 325 | raise ValueError( |
| 326 | f"{_FILENAME}: rule[{idx}]: unknown strategy {strategy!r}. " |
| 327 | f"Valid strategies: {sorted(VALID_STRATEGIES)}" |
| 328 | ) |
| 329 | |
| 330 | rules.append( |
| 331 | AttributeRule( |
| 332 | path_pattern=entry["path"], |
| 333 | dimension=entry["dimension"], |
| 334 | strategy=strategy, |
| 335 | comment=entry.get("comment", ""), |
| 336 | priority=entry.get("priority", 0), |
| 337 | source_index=idx, |
| 338 | ) |
| 339 | ) |
| 340 | |
| 341 | # Stable sort: higher priority first, ties preserve declaration order. |
| 342 | rules.sort(key=lambda r: -r.priority) |
| 343 | return rules |
| 344 | |
| 345 | |
| 346 | def resolve_strategy( |
| 347 | rules: list[AttributeRule], |
| 348 | path: str, |
| 349 | dimension: str = "*", |
| 350 | ) -> str: |
| 351 | """Return the first matching strategy for *path* and *dimension*. |
| 352 | |
| 353 | Matching rules: |
| 354 | |
| 355 | - **path**: ``fnmatch.fnmatch(path, rule.path_pattern)`` must be ``True``. |
| 356 | - **dimension**: ``rule.dimension`` must be ``"*"`` (matches anything) **or** |
| 357 | equal *dimension*. |
| 358 | |
| 359 | First-match wins after priority ordering applied by :func:`load_attributes`. |
| 360 | Returns ``"auto"`` when no rule matches. |
| 361 | |
| 362 | Args: |
| 363 | rules: Rule list from :func:`load_attributes`. |
| 364 | path: Workspace-relative POSIX path (e.g. ``"tracks/drums.mid"``). |
| 365 | dimension: Domain axis name or ``"*"`` to match any rule dimension. |
| 366 | |
| 367 | Returns: |
| 368 | A strategy string: ``"ours"``, ``"theirs"``, ``"union"``, ``"base"``, |
| 369 | ``"auto"``, or ``"manual"``. |
| 370 | """ |
| 371 | for rule in rules: |
| 372 | path_match = fnmatch.fnmatch(path, rule.path_pattern) |
| 373 | dim_match = ( |
| 374 | rule.dimension == "*" |
| 375 | or rule.dimension == dimension |
| 376 | or dimension == "*" |
| 377 | ) |
| 378 | if path_match and dim_match: |
| 379 | return rule.strategy |
| 380 | return "auto" |