test_harmony_phase1.py
python
sha256:88ac91129873e6a496e9189515aa690eb893ae25d69c8f72af141a2be5068eb3
docs: docstring sprint contract→find-symbol — idiomatic run…
Sonnet 4.6
patch
140 days ago
| 1 | """Tests for muse/core/harmony.py — Phase 1: Core data model. |
| 2 | |
| 3 | Coverage tiers |
| 4 | -------------- |
| 5 | I Unit — fingerprints, validation, namespaces, dataclasses, condition matching |
| 6 | II Integration — all CRUD operations (patterns, resolutions, policies, audit, gc) |
| 7 | III End-to-end — full conflict lifecycle from record → resolve → replay → gc |
| 8 | IV Stress — 10 k pattern scan, concurrent writes under parallel threads |
| 9 | V Data integrity— atomic writes (no temp files left), JSON round-trip, field types |
| 10 | VI Security — path traversal, symlink guards, size caps, crafted IDs |
| 11 | VII Performance — per-operation timing assertions |
| 12 | """ |
| 13 | from __future__ import annotations |
| 14 | |
| 15 | import concurrent.futures |
| 16 | import datetime |
| 17 | import hashlib |
| 18 | import json |
| 19 | import os |
| 20 | import pathlib |
| 21 | import tempfile |
| 22 | import threading |
| 23 | import time |
| 24 | from dataclasses import FrozenInstanceError |
| 25 | from typing import Any |
| 26 | from unittest import mock |
| 27 | |
| 28 | import pytest |
| 29 | |
| 30 | import muse.core.harmony as h |
| 31 | from muse.core.harmony import ( |
| 32 | AgentProvenance, |
| 33 | AuditEvent, |
| 34 | AuditEventType, |
| 35 | ConflictPattern, |
| 36 | ConflictType, |
| 37 | Policy, |
| 38 | PolicyAction, |
| 39 | PolicyCondition, |
| 40 | PolicyScope, |
| 41 | Resolution, |
| 42 | ResolutionProposal, |
| 43 | ResolutionStrategy, |
| 44 | _MAX_AUDIT_BYTES, |
| 45 | _MAX_PATTERN_BYTES, |
| 46 | _MAX_POLICY_BYTES, |
| 47 | _MAX_RESOLUTION_BYTES, |
| 48 | _MAX_SCAN, |
| 49 | _condition_matches, |
| 50 | append_audit, |
| 51 | best_resolution, |
| 52 | blob_fingerprint, |
| 53 | clear_all, |
| 54 | compute_pattern_id, |
| 55 | compute_resolution_id, |
| 56 | forget_pattern, |
| 57 | gc_stale, |
| 58 | increment_applied_count, |
| 59 | list_audit, |
| 60 | list_patterns, |
| 61 | list_policies, |
| 62 | list_resolutions, |
| 63 | load_pattern, |
| 64 | load_policy, |
| 65 | load_resolution, |
| 66 | match_policy, |
| 67 | record_pattern, |
| 68 | remove_policy, |
| 69 | save_policy, |
| 70 | save_resolution, |
| 71 | ) |
| 72 | |
| 73 | |
| 74 | # --------------------------------------------------------------------------- |
| 75 | # Shared fixtures |
| 76 | # --------------------------------------------------------------------------- |
| 77 | |
| 78 | |
| 79 | @pytest.fixture() |
| 80 | def repo(tmp_path: pathlib.Path) -> pathlib.Path: |
| 81 | """Return a temporary directory acting as a bare repo root.""" |
| 82 | (tmp_path / ".muse").mkdir() |
| 83 | return tmp_path |
| 84 | |
| 85 | |
| 86 | def _fake_id(seed: str) -> str: |
| 87 | """Return a deterministic 64-char hex string from *seed*.""" |
| 88 | return hashlib.sha256(seed.encode()).hexdigest() |
| 89 | |
| 90 | |
| 91 | def _utc_now() -> datetime.datetime: |
| 92 | return datetime.datetime.now(datetime.timezone.utc) |
| 93 | |
| 94 | |
| 95 | def _make_pattern( |
| 96 | path: str = "track.mid", |
| 97 | domain: str = "midi", |
| 98 | conflict_type: str = ConflictType.CONTENT, |
| 99 | ours: str = "ours", |
| 100 | theirs: str = "theirs", |
| 101 | description: dict[str, Any] | None = None, |
| 102 | recorded_by: str = "claude-code", |
| 103 | ) -> ConflictPattern: |
| 104 | """Build a ConflictPattern with sensible defaults.""" |
| 105 | ours_id = _fake_id(ours) |
| 106 | theirs_id = _fake_id(theirs) |
| 107 | blob_fp = blob_fingerprint(ours_id, theirs_id) |
| 108 | semantic_fp = blob_fp |
| 109 | pattern_id = compute_pattern_id(path, blob_fp, semantic_fp) |
| 110 | return ConflictPattern( |
| 111 | pattern_id=pattern_id, |
| 112 | path=path, |
| 113 | domain=domain, |
| 114 | conflict_type=conflict_type, |
| 115 | blob_fingerprint=blob_fp, |
| 116 | semantic_fingerprint=semantic_fp, |
| 117 | ours_id=ours_id, |
| 118 | theirs_id=theirs_id, |
| 119 | description=description or {}, |
| 120 | recorded_at=_utc_now(), |
| 121 | recorded_by=recorded_by, |
| 122 | ) |
| 123 | |
| 124 | |
| 125 | def _make_resolution( |
| 126 | pattern: ConflictPattern, |
| 127 | strategy: str = ResolutionStrategy.MANUAL, |
| 128 | confidence: float = 0.9, |
| 129 | human_verified: bool = False, |
| 130 | provenance: AgentProvenance | None = None, |
| 131 | policy_id: str | None = None, |
| 132 | ) -> Resolution: |
| 133 | """Build a Resolution tied to *pattern* with sensible defaults.""" |
| 134 | outcome_blob = _fake_id(f"outcome-{pattern.pattern_id[:8]}") |
| 135 | prov = provenance or AgentProvenance.agent("claude-code", "claude-sonnet-4-6") |
| 136 | resolved_at = _utc_now() |
| 137 | resolution_id = compute_resolution_id( |
| 138 | pattern.pattern_id, outcome_blob, strategy, prov, resolved_at |
| 139 | ) |
| 140 | return Resolution( |
| 141 | resolution_id=resolution_id, |
| 142 | pattern_id=pattern.pattern_id, |
| 143 | strategy=strategy, |
| 144 | policy_id=policy_id, |
| 145 | outcome_blob=outcome_blob, |
| 146 | resolved_by=prov, |
| 147 | human_verified=human_verified, |
| 148 | confidence=confidence, |
| 149 | rationale="Test rationale", |
| 150 | resolved_at=resolved_at, |
| 151 | ) |
| 152 | |
| 153 | |
| 154 | def _make_policy( |
| 155 | policy_id: str = "always-prefer-ours", |
| 156 | scope: str = PolicyScope.REPO, |
| 157 | action: str = PolicyAction.PREFER_OURS, |
| 158 | confidence: float = 0.95, |
| 159 | conflict_type: str | None = None, |
| 160 | domain: str | None = None, |
| 161 | path_pattern: str | None = None, |
| 162 | ) -> Policy: |
| 163 | """Build a Policy with sensible defaults.""" |
| 164 | return Policy( |
| 165 | policy_id=policy_id, |
| 166 | description="Test policy", |
| 167 | when=PolicyCondition( |
| 168 | conflict_type=conflict_type, |
| 169 | domain=domain, |
| 170 | path_pattern=path_pattern, |
| 171 | ), |
| 172 | action=action, |
| 173 | confidence=confidence, |
| 174 | escalate_to=None, |
| 175 | delegate_to=None, |
| 176 | scope=scope, |
| 177 | created_at=_utc_now(), |
| 178 | created_by="claude-code", |
| 179 | ) |
| 180 | |
| 181 | |
| 182 | # =========================================================================== |
| 183 | # Tier I — Unit tests |
| 184 | # =========================================================================== |
| 185 | |
| 186 | |
| 187 | class TestBlobFingerprint: |
| 188 | """I: blob_fingerprint must be commutative and deterministic.""" |
| 189 | |
| 190 | def test_commutativity(self) -> None: |
| 191 | a, b = _fake_id("A"), _fake_id("B") |
| 192 | assert blob_fingerprint(a, b) == blob_fingerprint(b, a) |
| 193 | |
| 194 | def test_determinism(self) -> None: |
| 195 | a, b = _fake_id("X"), _fake_id("Y") |
| 196 | fp1 = blob_fingerprint(a, b) |
| 197 | fp2 = blob_fingerprint(a, b) |
| 198 | assert fp1 == fp2 |
| 199 | |
| 200 | def test_output_is_64_hex(self) -> None: |
| 201 | a, b = _fake_id("p"), _fake_id("q") |
| 202 | fp = blob_fingerprint(a, b) |
| 203 | assert len(fp) == 64 |
| 204 | assert all(c in "0123456789abcdef" for c in fp) |
| 205 | |
| 206 | def test_distinct_pairs_differ(self) -> None: |
| 207 | ab = blob_fingerprint(_fake_id("A"), _fake_id("B")) |
| 208 | cd = blob_fingerprint(_fake_id("C"), _fake_id("D")) |
| 209 | assert ab != cd |
| 210 | |
| 211 | def test_same_id_both_sides(self) -> None: |
| 212 | a = _fake_id("same") |
| 213 | # Should not crash; result is deterministic |
| 214 | fp = blob_fingerprint(a, a) |
| 215 | assert len(fp) == 64 |
| 216 | |
| 217 | |
| 218 | class TestComputePatternId: |
| 219 | """I: compute_pattern_id includes path, so same content → different IDs for different paths.""" |
| 220 | |
| 221 | def test_deterministic(self) -> None: |
| 222 | blob_fp = _fake_id("blob") |
| 223 | sem_fp = _fake_id("sem") |
| 224 | p1 = compute_pattern_id("track.mid", blob_fp, sem_fp) |
| 225 | p2 = compute_pattern_id("track.mid", blob_fp, sem_fp) |
| 226 | assert p1 == p2 |
| 227 | |
| 228 | def test_path_changes_id(self) -> None: |
| 229 | blob_fp = _fake_id("blob") |
| 230 | sem_fp = _fake_id("sem") |
| 231 | p1 = compute_pattern_id("track.mid", blob_fp, sem_fp) |
| 232 | p2 = compute_pattern_id("drums.mid", blob_fp, sem_fp) |
| 233 | assert p1 != p2 |
| 234 | |
| 235 | def test_blob_changes_id(self) -> None: |
| 236 | sem_fp = _fake_id("sem") |
| 237 | p1 = compute_pattern_id("track.mid", _fake_id("blobA"), sem_fp) |
| 238 | p2 = compute_pattern_id("track.mid", _fake_id("blobB"), sem_fp) |
| 239 | assert p1 != p2 |
| 240 | |
| 241 | def test_64_hex_output(self) -> None: |
| 242 | pid = compute_pattern_id("f.py", _fake_id("b"), _fake_id("s")) |
| 243 | assert len(pid) == 64 |
| 244 | assert all(c in "0123456789abcdef" for c in pid) |
| 245 | |
| 246 | |
| 247 | class TestComputeResolutionId: |
| 248 | """I: compute_resolution_id is deterministic and encodes actor.""" |
| 249 | |
| 250 | def test_deterministic(self) -> None: |
| 251 | prov = AgentProvenance.agent("claude-code", "claude-sonnet-4-6") |
| 252 | ts = datetime.datetime(2025, 1, 1, tzinfo=datetime.timezone.utc) |
| 253 | pid = _fake_id("pattern") |
| 254 | ob = _fake_id("outcome") |
| 255 | r1 = compute_resolution_id(pid, ob, ResolutionStrategy.MANUAL, prov, ts) |
| 256 | r2 = compute_resolution_id(pid, ob, ResolutionStrategy.MANUAL, prov, ts) |
| 257 | assert r1 == r2 |
| 258 | |
| 259 | def test_different_agents_differ(self) -> None: |
| 260 | ts = datetime.datetime(2025, 1, 1, tzinfo=datetime.timezone.utc) |
| 261 | pid = _fake_id("pattern") |
| 262 | ob = _fake_id("outcome") |
| 263 | p1 = AgentProvenance.agent("claude-code") |
| 264 | p2 = AgentProvenance.agent("codex") |
| 265 | r1 = compute_resolution_id(pid, ob, ResolutionStrategy.MANUAL, p1, ts) |
| 266 | r2 = compute_resolution_id(pid, ob, ResolutionStrategy.MANUAL, p2, ts) |
| 267 | assert r1 != r2 |
| 268 | |
| 269 | def test_human_provenance_encodes_as_human(self) -> None: |
| 270 | ts = datetime.datetime(2025, 1, 1, tzinfo=datetime.timezone.utc) |
| 271 | pid = _fake_id("pattern") |
| 272 | ob = _fake_id("outcome") |
| 273 | # Should not crash and should produce 64 hex chars |
| 274 | rid = compute_resolution_id(pid, ob, ResolutionStrategy.MANUAL, AgentProvenance.human(), ts) |
| 275 | assert len(rid) == 64 |
| 276 | |
| 277 | |
| 278 | class TestValidation: |
| 279 | """I: _validate_id and _validate_policy_id must reject bad inputs.""" |
| 280 | |
| 281 | def test_validate_id_accepts_64_hex(self) -> None: |
| 282 | h._validate_id("a" * 64) # no exception |
| 283 | |
| 284 | def test_validate_id_rejects_63_chars(self) -> None: |
| 285 | with pytest.raises(ValueError, match="64 lowercase hexadecimal"): |
| 286 | h._validate_id("a" * 63) |
| 287 | |
| 288 | def test_validate_id_rejects_65_chars(self) -> None: |
| 289 | with pytest.raises(ValueError, match="64 lowercase hexadecimal"): |
| 290 | h._validate_id("a" * 65) |
| 291 | |
| 292 | def test_validate_id_rejects_uppercase(self) -> None: |
| 293 | with pytest.raises(ValueError): |
| 294 | h._validate_id("A" * 64) |
| 295 | |
| 296 | def test_validate_id_rejects_path_traversal(self) -> None: |
| 297 | # Attempt to inject a path traversal via the ID |
| 298 | with pytest.raises(ValueError): |
| 299 | h._validate_id("../../../etc/passwd" + "a" * 45) |
| 300 | |
| 301 | def test_validate_id_rejects_empty(self) -> None: |
| 302 | with pytest.raises(ValueError): |
| 303 | h._validate_id("") |
| 304 | |
| 305 | def test_validate_policy_id_accepts_alphanumeric(self) -> None: |
| 306 | h._validate_policy_id("my-policy_123") # no exception |
| 307 | |
| 308 | def test_validate_policy_id_rejects_slash(self) -> None: |
| 309 | with pytest.raises(ValueError, match="alphanumeric"): |
| 310 | h._validate_policy_id("bad/policy") |
| 311 | |
| 312 | def test_validate_policy_id_rejects_dot(self) -> None: |
| 313 | with pytest.raises(ValueError): |
| 314 | h._validate_policy_id("bad.policy") |
| 315 | |
| 316 | def test_validate_policy_id_rejects_empty(self) -> None: |
| 317 | with pytest.raises(ValueError): |
| 318 | h._validate_policy_id("") |
| 319 | |
| 320 | def test_validate_policy_id_rejects_129_chars(self) -> None: |
| 321 | with pytest.raises(ValueError): |
| 322 | h._validate_policy_id("a" * 129) |
| 323 | |
| 324 | def test_validate_policy_id_accepts_128_chars(self) -> None: |
| 325 | h._validate_policy_id("a" * 128) # no exception |
| 326 | |
| 327 | |
| 328 | class TestNamespaces: |
| 329 | """I: Open string-constant namespaces are plain strings — plugin extensibility.""" |
| 330 | |
| 331 | def test_conflict_type_are_strings(self) -> None: |
| 332 | assert isinstance(ConflictType.CONTENT, str) |
| 333 | assert isinstance(ConflictType.STRUCTURAL, str) |
| 334 | assert isinstance(ConflictType.METADATA, str) |
| 335 | assert isinstance(ConflictType.RELATIONAL, str) |
| 336 | assert isinstance(ConflictType.UNKNOWN, str) |
| 337 | |
| 338 | def test_resolution_strategy_are_strings(self) -> None: |
| 339 | assert isinstance(ResolutionStrategy.POLICY, str) |
| 340 | assert isinstance(ResolutionStrategy.EXACT_REPLAY, str) |
| 341 | assert isinstance(ResolutionStrategy.SEMANTIC_PROPOSAL, str) |
| 342 | assert isinstance(ResolutionStrategy.MANUAL, str) |
| 343 | |
| 344 | def test_policy_action_are_strings(self) -> None: |
| 345 | for attr in ("PREFER_OURS", "PREFER_THEIRS", "ESCALATE", "REQUIRE_HUMAN", "DELEGATE"): |
| 346 | assert isinstance(getattr(PolicyAction, attr), str) |
| 347 | |
| 348 | def test_policy_scope_are_strings(self) -> None: |
| 349 | for attr in ("WORKSPACE", "REPO", "DOMAIN", "FILE"): |
| 350 | assert isinstance(getattr(PolicyScope, attr), str) |
| 351 | |
| 352 | def test_audit_event_type_are_strings(self) -> None: |
| 353 | for attr in ( |
| 354 | "PATTERN_RECORDED", "RESOLUTION_SAVED", "RESOLUTION_APPLIED", |
| 355 | "PATTERN_FORGOTTEN", "POLICY_SAVED", "POLICY_REMOVED", "GC_RUN", "CLEAR_RUN", |
| 356 | ): |
| 357 | assert isinstance(getattr(AuditEventType, attr), str) |
| 358 | |
| 359 | def test_plugin_can_use_custom_conflict_type(self) -> None: |
| 360 | # Plugins may add strings at runtime — no import needed |
| 361 | custom_type = "note_collision" |
| 362 | pattern = _make_pattern(conflict_type=custom_type) |
| 363 | assert pattern.conflict_type == custom_type |
| 364 | |
| 365 | |
| 366 | class TestDataclasses: |
| 367 | """I: Frozen dataclasses, AgentProvenance constructors, PolicyCondition wildcards.""" |
| 368 | |
| 369 | def test_agent_provenance_human(self) -> None: |
| 370 | p = AgentProvenance.human() |
| 371 | assert p.type == "human" |
| 372 | assert p.agent_id is None |
| 373 | assert p.model_id is None |
| 374 | |
| 375 | def test_agent_provenance_agent(self) -> None: |
| 376 | p = AgentProvenance.agent("claude-code", "claude-sonnet-4-6") |
| 377 | assert p.type == "agent" |
| 378 | assert p.agent_id == "claude-code" |
| 379 | assert p.model_id == "claude-sonnet-4-6" |
| 380 | |
| 381 | def test_agent_provenance_agent_no_model(self) -> None: |
| 382 | p = AgentProvenance.agent("codex") |
| 383 | assert p.model_id is None |
| 384 | |
| 385 | def test_agent_provenance_frozen(self) -> None: |
| 386 | p = AgentProvenance.human() |
| 387 | with pytest.raises(FrozenInstanceError): |
| 388 | p.type = "agent" # type: ignore[misc] |
| 389 | |
| 390 | def test_policy_condition_all_none_wildcard(self) -> None: |
| 391 | cond = PolicyCondition() |
| 392 | assert cond.conflict_type is None |
| 393 | assert cond.domain is None |
| 394 | assert cond.path_pattern is None |
| 395 | assert cond.min_confidence is None |
| 396 | |
| 397 | def test_policy_condition_frozen(self) -> None: |
| 398 | cond = PolicyCondition(conflict_type=ConflictType.CONTENT) |
| 399 | with pytest.raises(FrozenInstanceError): |
| 400 | cond.conflict_type = ConflictType.STRUCTURAL # type: ignore[misc] |
| 401 | |
| 402 | def test_conflict_pattern_frozen(self) -> None: |
| 403 | p = _make_pattern() |
| 404 | with pytest.raises(FrozenInstanceError): |
| 405 | p.path = "evil.mid" # type: ignore[misc] |
| 406 | |
| 407 | def test_resolution_frozen(self) -> None: |
| 408 | pattern = _make_pattern() |
| 409 | res = _make_resolution(pattern) |
| 410 | with pytest.raises(FrozenInstanceError): |
| 411 | res.confidence = 0.0 # type: ignore[misc] |
| 412 | |
| 413 | def test_resolution_proposal_defaults(self) -> None: |
| 414 | prop = ResolutionProposal( |
| 415 | pattern_id=_fake_id("p"), |
| 416 | strategy=ResolutionStrategy.SEMANTIC_PROPOSAL, |
| 417 | proposed_action=PolicyAction.PREFER_OURS, |
| 418 | confidence=0.7, |
| 419 | rationale="fuzzy match", |
| 420 | ) |
| 421 | assert prop.policy_id is None |
| 422 | assert prop.similar_pattern_id is None |
| 423 | assert prop.similarity is None |
| 424 | assert prop.requires_confirmation is False |
| 425 | |
| 426 | |
| 427 | class TestConditionMatching: |
| 428 | """I: _condition_matches and match_policy first-match-wins semantics.""" |
| 429 | |
| 430 | def _pattern(self, **kwargs: Any) -> ConflictPattern: |
| 431 | return _make_pattern(**kwargs) |
| 432 | |
| 433 | def test_all_none_matches_anything(self) -> None: |
| 434 | cond = PolicyCondition() |
| 435 | assert _condition_matches(cond, self._pattern()) is True |
| 436 | |
| 437 | def test_conflict_type_match(self) -> None: |
| 438 | cond = PolicyCondition(conflict_type=ConflictType.CONTENT) |
| 439 | assert _condition_matches(cond, self._pattern(conflict_type=ConflictType.CONTENT)) is True |
| 440 | assert _condition_matches(cond, self._pattern(conflict_type=ConflictType.METADATA)) is False |
| 441 | |
| 442 | def test_domain_match(self) -> None: |
| 443 | cond = PolicyCondition(domain="midi") |
| 444 | assert _condition_matches(cond, self._pattern(domain="midi")) is True |
| 445 | assert _condition_matches(cond, self._pattern(domain="code")) is False |
| 446 | |
| 447 | def test_path_pattern_glob(self) -> None: |
| 448 | cond = PolicyCondition(path_pattern="*.mid") |
| 449 | assert _condition_matches(cond, self._pattern(path="track.mid")) is True |
| 450 | assert _condition_matches(cond, self._pattern(path="src/main.py")) is False |
| 451 | |
| 452 | def test_path_pattern_prefix_glob(self) -> None: |
| 453 | cond = PolicyCondition(path_pattern="audio/*") |
| 454 | assert _condition_matches(cond, self._pattern(path="audio/kick.mid")) is True |
| 455 | assert _condition_matches(cond, self._pattern(path="video/clip.mp4")) is False |
| 456 | |
| 457 | def test_all_conditions_must_match(self) -> None: |
| 458 | cond = PolicyCondition(conflict_type=ConflictType.CONTENT, domain="midi") |
| 459 | matching = self._pattern(conflict_type=ConflictType.CONTENT, domain="midi") |
| 460 | wrong_domain = self._pattern(conflict_type=ConflictType.CONTENT, domain="code") |
| 461 | wrong_type = self._pattern(conflict_type=ConflictType.METADATA, domain="midi") |
| 462 | assert _condition_matches(cond, matching) is True |
| 463 | assert _condition_matches(cond, wrong_domain) is False |
| 464 | assert _condition_matches(cond, wrong_type) is False |
| 465 | |
| 466 | def test_match_policy_first_match_wins(self) -> None: |
| 467 | policy_a = _make_policy("policy-a", scope=PolicyScope.WORKSPACE, conflict_type=ConflictType.CONTENT) |
| 468 | policy_b = _make_policy("policy-b", scope=PolicyScope.REPO, conflict_type=ConflictType.CONTENT) |
| 469 | pattern = self._pattern(conflict_type=ConflictType.CONTENT) |
| 470 | result = match_policy([policy_a, policy_b], pattern) |
| 471 | assert result is not None |
| 472 | assert result.policy_id == "policy-a" |
| 473 | |
| 474 | def test_match_policy_no_match_returns_none(self) -> None: |
| 475 | policy = _make_policy("p", conflict_type=ConflictType.STRUCTURAL) |
| 476 | pattern = self._pattern(conflict_type=ConflictType.CONTENT) |
| 477 | assert match_policy([policy], pattern) is None |
| 478 | |
| 479 | def test_match_policy_empty_list(self) -> None: |
| 480 | pattern = self._pattern() |
| 481 | assert match_policy([], pattern) is None |
| 482 | |
| 483 | def test_min_confidence_not_evaluated_here(self) -> None: |
| 484 | # min_confidence is an engine-level filter, not evaluated by _condition_matches |
| 485 | cond = PolicyCondition(min_confidence=0.99) |
| 486 | # Should still match since no other fields constrain the pattern |
| 487 | assert _condition_matches(cond, self._pattern()) is True |
| 488 | |
| 489 | |
| 490 | # =========================================================================== |
| 491 | # Tier II — Integration tests |
| 492 | # =========================================================================== |
| 493 | |
| 494 | |
| 495 | class TestPatternCRUD: |
| 496 | """II: record_pattern, load_pattern, list_patterns, forget_pattern, clear_all.""" |
| 497 | |
| 498 | def test_record_and_load(self, repo: pathlib.Path) -> None: |
| 499 | pattern = _make_pattern() |
| 500 | record_pattern(repo, pattern) |
| 501 | loaded = load_pattern(repo, pattern.pattern_id) |
| 502 | assert loaded is not None |
| 503 | assert loaded.pattern_id == pattern.pattern_id |
| 504 | assert loaded.path == pattern.path |
| 505 | assert loaded.domain == pattern.domain |
| 506 | assert loaded.conflict_type == pattern.conflict_type |
| 507 | |
| 508 | def test_record_is_idempotent(self, repo: pathlib.Path) -> None: |
| 509 | pattern = _make_pattern() |
| 510 | record_pattern(repo, pattern) |
| 511 | # Second call should not raise and should return same ID |
| 512 | pid = record_pattern(repo, pattern) |
| 513 | assert pid == pattern.pattern_id |
| 514 | # Only one pattern.json should exist |
| 515 | entry_dir = h.patterns_dir(repo) / pattern.pattern_id |
| 516 | assert list(entry_dir.glob("pattern.json")) == [entry_dir / "pattern.json"] |
| 517 | |
| 518 | def test_load_nonexistent_returns_none(self, repo: pathlib.Path) -> None: |
| 519 | assert load_pattern(repo, "a" * 64) is None |
| 520 | |
| 521 | def test_load_invalid_id_returns_none(self, repo: pathlib.Path) -> None: |
| 522 | assert load_pattern(repo, "not-a-hex-id") is None |
| 523 | |
| 524 | def test_list_patterns_empty(self, repo: pathlib.Path) -> None: |
| 525 | assert list_patterns(repo) == [] |
| 526 | |
| 527 | def test_list_patterns_multiple(self, repo: pathlib.Path) -> None: |
| 528 | p1 = _make_pattern(path="a.mid", ours="oa", theirs="ta") |
| 529 | p2 = _make_pattern(path="b.mid", ours="ob", theirs="tb") |
| 530 | record_pattern(repo, p1) |
| 531 | record_pattern(repo, p2) |
| 532 | results = list_patterns(repo) |
| 533 | assert len(results) == 2 |
| 534 | pids = {r.pattern_id for r in results} |
| 535 | assert {p1.pattern_id, p2.pattern_id} == pids |
| 536 | |
| 537 | def test_list_patterns_sorted_newest_first(self, repo: pathlib.Path) -> None: |
| 538 | older = _make_pattern(path="old.mid", ours="oa", theirs="ta") |
| 539 | newer = _make_pattern(path="new.mid", ours="ob", theirs="tb") |
| 540 | # Force newer to be newer by manipulating recorded_at |
| 541 | import dataclasses |
| 542 | ts_old = datetime.datetime(2024, 1, 1, tzinfo=datetime.timezone.utc) |
| 543 | ts_new = datetime.datetime(2025, 1, 1, tzinfo=datetime.timezone.utc) |
| 544 | older = dataclasses.replace(older, recorded_at=ts_old) |
| 545 | newer = dataclasses.replace(newer, recorded_at=ts_new) |
| 546 | record_pattern(repo, older) |
| 547 | record_pattern(repo, newer) |
| 548 | results = list_patterns(repo) |
| 549 | assert results[0].pattern_id == newer.pattern_id |
| 550 | |
| 551 | def test_forget_pattern_removes_entry(self, repo: pathlib.Path) -> None: |
| 552 | pattern = _make_pattern() |
| 553 | record_pattern(repo, pattern) |
| 554 | assert forget_pattern(repo, pattern.pattern_id) is True |
| 555 | assert load_pattern(repo, pattern.pattern_id) is None |
| 556 | |
| 557 | def test_forget_nonexistent_returns_false(self, repo: pathlib.Path) -> None: |
| 558 | assert forget_pattern(repo, "a" * 64) is False |
| 559 | |
| 560 | def test_forget_invalid_id_returns_false(self, repo: pathlib.Path) -> None: |
| 561 | assert forget_pattern(repo, "../evil") is False |
| 562 | |
| 563 | def test_forget_also_removes_resolutions(self, repo: pathlib.Path) -> None: |
| 564 | pattern = _make_pattern() |
| 565 | record_pattern(repo, pattern) |
| 566 | res = _make_resolution(pattern) |
| 567 | save_resolution(repo, res) |
| 568 | forget_pattern(repo, pattern.pattern_id) |
| 569 | # Resolution directory should be gone |
| 570 | res_dir = h._resolutions_dir(repo, pattern.pattern_id) |
| 571 | assert not res_dir.exists() |
| 572 | |
| 573 | def test_clear_all_removes_all(self, repo: pathlib.Path) -> None: |
| 574 | for i in range(5): |
| 575 | record_pattern(repo, _make_pattern(path=f"f{i}.mid", ours=f"o{i}", theirs=f"t{i}")) |
| 576 | removed = clear_all(repo) |
| 577 | assert removed == 5 |
| 578 | assert list_patterns(repo) == [] |
| 579 | |
| 580 | def test_clear_all_empty_store(self, repo: pathlib.Path) -> None: |
| 581 | assert clear_all(repo) == 0 |
| 582 | |
| 583 | def test_record_pattern_invalid_id_raises(self, repo: pathlib.Path) -> None: |
| 584 | import dataclasses |
| 585 | pattern = _make_pattern() |
| 586 | bad = dataclasses.replace(pattern, pattern_id="bad-id") |
| 587 | with pytest.raises(ValueError): |
| 588 | record_pattern(repo, bad) |
| 589 | |
| 590 | |
| 591 | class TestResolutionCRUD: |
| 592 | """II: save_resolution, load_resolution, list_resolutions, increment_applied_count, best_resolution.""" |
| 593 | |
| 594 | def test_save_and_load(self, repo: pathlib.Path) -> None: |
| 595 | pattern = _make_pattern() |
| 596 | record_pattern(repo, pattern) |
| 597 | res = _make_resolution(pattern) |
| 598 | save_resolution(repo, res) |
| 599 | loaded = load_resolution(repo, pattern.pattern_id, res.resolution_id) |
| 600 | assert loaded is not None |
| 601 | assert loaded.resolution_id == res.resolution_id |
| 602 | assert loaded.pattern_id == pattern.pattern_id |
| 603 | assert loaded.strategy == res.strategy |
| 604 | assert loaded.confidence == res.confidence |
| 605 | |
| 606 | def test_save_is_idempotent(self, repo: pathlib.Path) -> None: |
| 607 | pattern = _make_pattern() |
| 608 | record_pattern(repo, pattern) |
| 609 | res = _make_resolution(pattern) |
| 610 | save_resolution(repo, res) |
| 611 | save_resolution(repo, res) # second call is no-op |
| 612 | assert len(list_resolutions(repo, pattern.pattern_id)) == 1 |
| 613 | |
| 614 | def test_save_requires_parent_pattern(self, repo: pathlib.Path) -> None: |
| 615 | pattern = _make_pattern() |
| 616 | res = _make_resolution(pattern) |
| 617 | with pytest.raises(FileNotFoundError, match="No harmony pattern"): |
| 618 | save_resolution(repo, res) |
| 619 | |
| 620 | def test_load_nonexistent_returns_none(self, repo: pathlib.Path) -> None: |
| 621 | assert load_resolution(repo, "a" * 64, "b" * 64) is None |
| 622 | |
| 623 | def test_load_invalid_ids_return_none(self, repo: pathlib.Path) -> None: |
| 624 | assert load_resolution(repo, "bad", "b" * 64) is None |
| 625 | assert load_resolution(repo, "a" * 64, "bad") is None |
| 626 | |
| 627 | def test_list_resolutions_empty(self, repo: pathlib.Path) -> None: |
| 628 | pattern = _make_pattern() |
| 629 | record_pattern(repo, pattern) |
| 630 | assert list_resolutions(repo, pattern.pattern_id) == [] |
| 631 | |
| 632 | def test_list_resolutions_sorted_by_quality(self, repo: pathlib.Path) -> None: |
| 633 | """human_verified > confidence > applied_count (desc).""" |
| 634 | pattern = _make_pattern() |
| 635 | record_pattern(repo, pattern) |
| 636 | |
| 637 | low_conf = _make_resolution(pattern, confidence=0.3) |
| 638 | high_conf = _make_resolution(pattern, confidence=0.9) |
| 639 | verified = _make_resolution(pattern, confidence=0.5, human_verified=True) |
| 640 | |
| 641 | # Build distinct resolutions (different outcomes) |
| 642 | import dataclasses |
| 643 | low_conf = dataclasses.replace( |
| 644 | low_conf, |
| 645 | outcome_blob=_fake_id("low_outcome"), |
| 646 | resolution_id=_fake_id("low_res"), |
| 647 | ) |
| 648 | high_conf = dataclasses.replace( |
| 649 | high_conf, |
| 650 | outcome_blob=_fake_id("high_outcome"), |
| 651 | resolution_id=_fake_id("high_res"), |
| 652 | ) |
| 653 | verified = dataclasses.replace( |
| 654 | verified, |
| 655 | outcome_blob=_fake_id("ver_outcome"), |
| 656 | resolution_id=_fake_id("ver_res"), |
| 657 | ) |
| 658 | |
| 659 | for r in (low_conf, high_conf, verified): |
| 660 | save_resolution(repo, r) |
| 661 | |
| 662 | results = list_resolutions(repo, pattern.pattern_id) |
| 663 | assert results[0].resolution_id == verified.resolution_id # human_verified first |
| 664 | assert results[-1].resolution_id == low_conf.resolution_id # lowest confidence last |
| 665 | |
| 666 | def test_increment_applied_count(self, repo: pathlib.Path) -> None: |
| 667 | pattern = _make_pattern() |
| 668 | record_pattern(repo, pattern) |
| 669 | res = _make_resolution(pattern) |
| 670 | save_resolution(repo, res) |
| 671 | assert increment_applied_count(repo, pattern.pattern_id, res.resolution_id) is True |
| 672 | loaded = load_resolution(repo, pattern.pattern_id, res.resolution_id) |
| 673 | assert loaded is not None |
| 674 | assert loaded.applied_count == 1 |
| 675 | |
| 676 | def test_increment_applied_count_multiple_times(self, repo: pathlib.Path) -> None: |
| 677 | pattern = _make_pattern() |
| 678 | record_pattern(repo, pattern) |
| 679 | res = _make_resolution(pattern) |
| 680 | save_resolution(repo, res) |
| 681 | for _ in range(5): |
| 682 | increment_applied_count(repo, pattern.pattern_id, res.resolution_id) |
| 683 | loaded = load_resolution(repo, pattern.pattern_id, res.resolution_id) |
| 684 | assert loaded is not None |
| 685 | assert loaded.applied_count == 5 |
| 686 | |
| 687 | def test_increment_nonexistent_returns_false(self, repo: pathlib.Path) -> None: |
| 688 | assert increment_applied_count(repo, "a" * 64, "b" * 64) is False |
| 689 | |
| 690 | def test_best_resolution_returns_highest_quality(self, repo: pathlib.Path) -> None: |
| 691 | pattern = _make_pattern() |
| 692 | record_pattern(repo, pattern) |
| 693 | |
| 694 | import dataclasses |
| 695 | r1 = _make_resolution(pattern, confidence=0.5) |
| 696 | r2 = _make_resolution(pattern, confidence=0.9) |
| 697 | r1 = dataclasses.replace(r1, outcome_blob=_fake_id("r1ob"), resolution_id=_fake_id("r1id")) |
| 698 | r2 = dataclasses.replace(r2, outcome_blob=_fake_id("r2ob"), resolution_id=_fake_id("r2id")) |
| 699 | save_resolution(repo, r1) |
| 700 | save_resolution(repo, r2) |
| 701 | |
| 702 | best = best_resolution(repo, pattern.pattern_id) |
| 703 | assert best is not None |
| 704 | assert best.resolution_id == r2.resolution_id |
| 705 | |
| 706 | def test_best_resolution_none_when_no_resolutions(self, repo: pathlib.Path) -> None: |
| 707 | pattern = _make_pattern() |
| 708 | record_pattern(repo, pattern) |
| 709 | assert best_resolution(repo, pattern.pattern_id) is None |
| 710 | |
| 711 | |
| 712 | class TestPolicyCRUD: |
| 713 | """II: save_policy, load_policy, list_policies scope-sorted, remove_policy.""" |
| 714 | |
| 715 | def test_save_and_load(self, repo: pathlib.Path) -> None: |
| 716 | policy = _make_policy() |
| 717 | save_policy(repo, policy) |
| 718 | loaded = load_policy(repo, policy.policy_id) |
| 719 | assert loaded is not None |
| 720 | assert loaded.policy_id == policy.policy_id |
| 721 | assert loaded.action == policy.action |
| 722 | assert loaded.scope == policy.scope |
| 723 | |
| 724 | def test_save_overwrites_existing(self, repo: pathlib.Path) -> None: |
| 725 | policy = _make_policy(action=PolicyAction.PREFER_OURS) |
| 726 | save_policy(repo, policy) |
| 727 | import dataclasses |
| 728 | updated = dataclasses.replace(policy, action=PolicyAction.PREFER_THEIRS) |
| 729 | save_policy(repo, updated) |
| 730 | loaded = load_policy(repo, policy.policy_id) |
| 731 | assert loaded is not None |
| 732 | assert loaded.action == PolicyAction.PREFER_THEIRS |
| 733 | |
| 734 | def test_load_nonexistent_returns_none(self, repo: pathlib.Path) -> None: |
| 735 | assert load_policy(repo, "missing-policy") is None |
| 736 | |
| 737 | def test_load_invalid_id_returns_none(self, repo: pathlib.Path) -> None: |
| 738 | assert load_policy(repo, "bad/policy/id") is None |
| 739 | |
| 740 | def test_list_policies_empty(self, repo: pathlib.Path) -> None: |
| 741 | assert list_policies(repo) == [] |
| 742 | |
| 743 | def test_list_policies_scope_order(self, repo: pathlib.Path) -> None: |
| 744 | """workspace → repo → domain → file regardless of insertion order.""" |
| 745 | file_p = _make_policy("file-p", scope=PolicyScope.FILE) |
| 746 | workspace_p = _make_policy("workspace-p", scope=PolicyScope.WORKSPACE) |
| 747 | domain_p = _make_policy("domain-p", scope=PolicyScope.DOMAIN) |
| 748 | repo_p = _make_policy("repo-p", scope=PolicyScope.REPO) |
| 749 | for p in (file_p, workspace_p, domain_p, repo_p): |
| 750 | save_policy(repo, p) |
| 751 | results = list_policies(repo) |
| 752 | scopes = [r.scope for r in results] |
| 753 | assert scopes.index(PolicyScope.WORKSPACE) < scopes.index(PolicyScope.REPO) |
| 754 | assert scopes.index(PolicyScope.REPO) < scopes.index(PolicyScope.DOMAIN) |
| 755 | assert scopes.index(PolicyScope.DOMAIN) < scopes.index(PolicyScope.FILE) |
| 756 | |
| 757 | def test_remove_policy_returns_true(self, repo: pathlib.Path) -> None: |
| 758 | policy = _make_policy() |
| 759 | save_policy(repo, policy) |
| 760 | assert remove_policy(repo, policy.policy_id) is True |
| 761 | assert load_policy(repo, policy.policy_id) is None |
| 762 | |
| 763 | def test_remove_nonexistent_returns_false(self, repo: pathlib.Path) -> None: |
| 764 | assert remove_policy(repo, "no-such-policy") is False |
| 765 | |
| 766 | def test_remove_invalid_id_returns_false(self, repo: pathlib.Path) -> None: |
| 767 | assert remove_policy(repo, "bad/id") is False |
| 768 | |
| 769 | def test_save_invalid_id_raises(self, repo: pathlib.Path) -> None: |
| 770 | import dataclasses |
| 771 | policy = _make_policy() |
| 772 | bad = dataclasses.replace(policy, policy_id="bad/id") |
| 773 | with pytest.raises(ValueError): |
| 774 | save_policy(repo, bad) |
| 775 | |
| 776 | def test_condition_round_trips(self, repo: pathlib.Path) -> None: |
| 777 | policy = _make_policy(conflict_type=ConflictType.CONTENT, domain="midi", path_pattern="*.mid") |
| 778 | save_policy(repo, policy) |
| 779 | loaded = load_policy(repo, policy.policy_id) |
| 780 | assert loaded is not None |
| 781 | assert loaded.when.conflict_type == ConflictType.CONTENT |
| 782 | assert loaded.when.domain == "midi" |
| 783 | assert loaded.when.path_pattern == "*.mid" |
| 784 | |
| 785 | |
| 786 | class TestAuditLog: |
| 787 | """II: append_audit, list_audit sorted newest-first.""" |
| 788 | |
| 789 | def test_append_and_list(self, repo: pathlib.Path) -> None: |
| 790 | actor = AgentProvenance.agent("claude-code") |
| 791 | append_audit(repo, AuditEventType.PATTERN_RECORDED, actor, pattern_id="a" * 64) |
| 792 | entries = list_audit(repo) |
| 793 | assert len(entries) == 1 |
| 794 | assert entries[0]["event_type"] == AuditEventType.PATTERN_RECORDED |
| 795 | |
| 796 | def test_entries_sorted_newest_first(self, repo: pathlib.Path) -> None: |
| 797 | actor = AgentProvenance.human() |
| 798 | for i in range(3): |
| 799 | append_audit(repo, AuditEventType.GC_RUN, actor, metadata={"i": i}) |
| 800 | time.sleep(0.01) # slight delay so filenames differ |
| 801 | entries = list_audit(repo) |
| 802 | assert len(entries) == 3 |
| 803 | # Filenames encode date+uuid — sorted descending means newest at [0] |
| 804 | names_in_dir = sorted( |
| 805 | (f.name for f in h.audit_dir(repo).iterdir()), |
| 806 | reverse=True, |
| 807 | ) |
| 808 | assert entries[0]["audit_id"] in names_in_dir[0] |
| 809 | |
| 810 | def test_list_audit_empty(self, repo: pathlib.Path) -> None: |
| 811 | assert list_audit(repo) == [] |
| 812 | |
| 813 | def test_limit_respected(self, repo: pathlib.Path) -> None: |
| 814 | actor = AgentProvenance.human() |
| 815 | for _ in range(10): |
| 816 | append_audit(repo, AuditEventType.GC_RUN, actor) |
| 817 | entries = list_audit(repo, limit=3) |
| 818 | assert len(entries) == 3 |
| 819 | |
| 820 | def test_audit_fields_present(self, repo: pathlib.Path) -> None: |
| 821 | actor = AgentProvenance.agent("claude-code", "claude-sonnet-4-6") |
| 822 | append_audit( |
| 823 | repo, AuditEventType.RESOLUTION_SAVED, actor, |
| 824 | pattern_id="a" * 64, resolution_id="b" * 64, |
| 825 | metadata={"extra": "data"}, |
| 826 | ) |
| 827 | entry = list_audit(repo)[0] |
| 828 | assert entry["event_type"] == AuditEventType.RESOLUTION_SAVED |
| 829 | assert entry["pattern_id"] == "a" * 64 |
| 830 | assert entry["resolution_id"] == "b" * 64 |
| 831 | assert entry["acted_by"]["agent_id"] == "claude-code" |
| 832 | assert entry["metadata"]["extra"] == "data" |
| 833 | assert "audit_id" in entry |
| 834 | assert "occurred_at" in entry |
| 835 | |
| 836 | |
| 837 | class TestGcStale: |
| 838 | """II: gc_stale keeps resolved patterns and removes old unresolved ones.""" |
| 839 | |
| 840 | def test_gc_removes_old_unresolved(self, repo: pathlib.Path) -> None: |
| 841 | import dataclasses |
| 842 | old_pattern = _make_pattern() |
| 843 | old_ts = datetime.datetime(2020, 1, 1, tzinfo=datetime.timezone.utc) |
| 844 | old_pattern = dataclasses.replace(old_pattern, recorded_at=old_ts) |
| 845 | record_pattern(repo, old_pattern) |
| 846 | removed = gc_stale(repo, age_days=1) |
| 847 | assert removed == 1 |
| 848 | assert load_pattern(repo, old_pattern.pattern_id) is None |
| 849 | |
| 850 | def test_gc_keeps_resolved_pattern(self, repo: pathlib.Path) -> None: |
| 851 | import dataclasses |
| 852 | pattern = _make_pattern() |
| 853 | old_ts = datetime.datetime(2020, 1, 1, tzinfo=datetime.timezone.utc) |
| 854 | pattern = dataclasses.replace(pattern, recorded_at=old_ts) |
| 855 | record_pattern(repo, pattern) |
| 856 | res = _make_resolution(pattern) |
| 857 | save_resolution(repo, res) |
| 858 | removed = gc_stale(repo, age_days=1) |
| 859 | assert removed == 0 |
| 860 | assert load_pattern(repo, pattern.pattern_id) is not None |
| 861 | |
| 862 | def test_gc_keeps_recent_unresolved(self, repo: pathlib.Path) -> None: |
| 863 | pattern = _make_pattern() # recorded_at = now |
| 864 | record_pattern(repo, pattern) |
| 865 | removed = gc_stale(repo, age_days=90) |
| 866 | assert removed == 0 |
| 867 | |
| 868 | def test_gc_empty_store(self, repo: pathlib.Path) -> None: |
| 869 | assert gc_stale(repo, age_days=1) == 0 |
| 870 | |
| 871 | |
| 872 | # =========================================================================== |
| 873 | # Tier III — End-to-end lifecycle |
| 874 | # =========================================================================== |
| 875 | |
| 876 | |
| 877 | class TestFullLifecycle: |
| 878 | """III: record → save_resolution → best_resolution → increment → gc won't touch it.""" |
| 879 | |
| 880 | def test_complete_lifecycle(self, repo: pathlib.Path) -> None: |
| 881 | # 1. Record the conflict pattern |
| 882 | pattern = _make_pattern( |
| 883 | path="tracks/lead.mid", |
| 884 | domain="midi", |
| 885 | conflict_type=ConflictType.CONTENT, |
| 886 | ) |
| 887 | pid = record_pattern(repo, pattern) |
| 888 | assert pid == pattern.pattern_id |
| 889 | assert load_pattern(repo, pid) is not None |
| 890 | |
| 891 | # 2. Save a resolution |
| 892 | res = _make_resolution(pattern, strategy=ResolutionStrategy.MANUAL, confidence=0.85) |
| 893 | save_resolution(repo, res) |
| 894 | |
| 895 | # 3. Retrieve best resolution |
| 896 | best = best_resolution(repo, pid) |
| 897 | assert best is not None |
| 898 | assert best.resolution_id == res.resolution_id |
| 899 | assert best.confidence == pytest.approx(0.85) |
| 900 | |
| 901 | # 4. Replay — increment applied count |
| 902 | increment_applied_count(repo, pid, res.resolution_id) |
| 903 | increment_applied_count(repo, pid, res.resolution_id) |
| 904 | reloaded = load_resolution(repo, pid, res.resolution_id) |
| 905 | assert reloaded is not None |
| 906 | assert reloaded.applied_count == 2 |
| 907 | |
| 908 | # 5. GC should NOT remove — has a resolution |
| 909 | import dataclasses |
| 910 | old_ts = datetime.datetime(2020, 1, 1, tzinfo=datetime.timezone.utc) |
| 911 | pattern_old = dataclasses.replace(pattern, recorded_at=old_ts) |
| 912 | # Re-record is idempotent so we can't update recorded_at via record_pattern. |
| 913 | # Write updated pattern.json directly for this edge-case test. |
| 914 | entry_p = h.patterns_dir(repo) / pid / "pattern.json" |
| 915 | entry_p.write_text( |
| 916 | json.dumps(h._pattern_to_dict(pattern_old), indent=2), encoding="utf-8" |
| 917 | ) |
| 918 | gc_count = gc_stale(repo, age_days=1) |
| 919 | assert gc_count == 0 # Protected because it has a resolution |
| 920 | assert load_pattern(repo, pid) is not None |
| 921 | |
| 922 | def test_policy_fires_on_matching_pattern(self, repo: pathlib.Path) -> None: |
| 923 | policy = _make_policy( |
| 924 | "midi-prefer-ours", |
| 925 | scope=PolicyScope.DOMAIN, |
| 926 | action=PolicyAction.PREFER_OURS, |
| 927 | domain="midi", |
| 928 | ) |
| 929 | save_policy(repo, policy) |
| 930 | |
| 931 | pattern = _make_pattern(domain="midi", conflict_type=ConflictType.CONTENT) |
| 932 | record_pattern(repo, pattern) |
| 933 | |
| 934 | policies = list_policies(repo) |
| 935 | matched = match_policy(policies, pattern) |
| 936 | assert matched is not None |
| 937 | assert matched.policy_id == "midi-prefer-ours" |
| 938 | assert matched.action == PolicyAction.PREFER_OURS |
| 939 | |
| 940 | def test_policy_does_not_fire_wrong_domain(self, repo: pathlib.Path) -> None: |
| 941 | policy = _make_policy("midi-only", domain="midi") |
| 942 | save_policy(repo, policy) |
| 943 | pattern = _make_pattern(domain="code") |
| 944 | record_pattern(repo, pattern) |
| 945 | matched = match_policy(list_policies(repo), pattern) |
| 946 | assert matched is None |
| 947 | |
| 948 | def test_audit_trail_through_lifecycle(self, repo: pathlib.Path) -> None: |
| 949 | actor = AgentProvenance.agent("claude-code", "claude-sonnet-4-6") |
| 950 | pattern = _make_pattern() |
| 951 | pid = record_pattern(repo, pattern) |
| 952 | append_audit(repo, AuditEventType.PATTERN_RECORDED, actor, pattern_id=pid) |
| 953 | |
| 954 | res = _make_resolution(pattern) |
| 955 | save_resolution(repo, res) |
| 956 | append_audit( |
| 957 | repo, AuditEventType.RESOLUTION_SAVED, actor, |
| 958 | pattern_id=pid, resolution_id=res.resolution_id, |
| 959 | ) |
| 960 | |
| 961 | entries = list_audit(repo) |
| 962 | event_types = [e["event_type"] for e in entries] |
| 963 | assert AuditEventType.RESOLUTION_SAVED in event_types |
| 964 | assert AuditEventType.PATTERN_RECORDED in event_types |
| 965 | |
| 966 | |
| 967 | # =========================================================================== |
| 968 | # Tier IV — Stress tests |
| 969 | # =========================================================================== |
| 970 | |
| 971 | |
| 972 | class TestStress: |
| 973 | """IV: 10k pattern scan, concurrent record_pattern, concurrent save_resolution.""" |
| 974 | |
| 975 | def test_100_patterns_scan(self, repo: pathlib.Path) -> None: |
| 976 | """Store 100 patterns and verify list_patterns returns all of them.""" |
| 977 | n = 100 |
| 978 | for i in range(n): |
| 979 | record_pattern(repo, _make_pattern(path=f"f{i}.mid", ours=f"o{i}", theirs=f"t{i}")) |
| 980 | results = list_patterns(repo) |
| 981 | assert len(results) == n |
| 982 | |
| 983 | def test_concurrent_record_pattern_no_corruption(self, repo: pathlib.Path) -> None: |
| 984 | """Concurrent record_pattern from 20 threads — all patterns must be loadable.""" |
| 985 | patterns = [ |
| 986 | _make_pattern(path=f"concurrent{i}.mid", ours=f"co{i}", theirs=f"ct{i}") |
| 987 | for i in range(20) |
| 988 | ] |
| 989 | errors: list[Exception] = [] |
| 990 | |
| 991 | def worker(p: ConflictPattern) -> None: |
| 992 | try: |
| 993 | record_pattern(repo, p) |
| 994 | except Exception as exc: |
| 995 | errors.append(exc) |
| 996 | |
| 997 | with concurrent.futures.ThreadPoolExecutor(max_workers=10) as executor: |
| 998 | futures = [executor.submit(worker, p) for p in patterns] |
| 999 | concurrent.futures.wait(futures) |
| 1000 | |
| 1001 | assert errors == [], f"Thread errors: {errors}" |
| 1002 | for p in patterns: |
| 1003 | assert load_pattern(repo, p.pattern_id) is not None |
| 1004 | |
| 1005 | def test_concurrent_increment_applied_count(self, repo: pathlib.Path) -> None: |
| 1006 | """20 concurrent increments — none must crash, final count must be ≥ 1. |
| 1007 | |
| 1008 | ``increment_applied_count`` is a read-modify-write cycle; ``os.replace`` |
| 1009 | makes each individual write atomic but does not serialise the full cycle. |
| 1010 | Under heavy concurrency, updates may be lost (last writer wins). The |
| 1011 | guarantee is: no exception, file always valid, count always ≥ 1. |
| 1012 | """ |
| 1013 | pattern = _make_pattern() |
| 1014 | record_pattern(repo, pattern) |
| 1015 | res = _make_resolution(pattern) |
| 1016 | save_resolution(repo, res) |
| 1017 | |
| 1018 | lock = threading.Lock() |
| 1019 | errors: list[Exception] = [] |
| 1020 | |
| 1021 | def worker() -> None: |
| 1022 | try: |
| 1023 | increment_applied_count(repo, pattern.pattern_id, res.resolution_id) |
| 1024 | except Exception as exc: |
| 1025 | with lock: |
| 1026 | errors.append(exc) |
| 1027 | |
| 1028 | threads = [threading.Thread(target=worker) for _ in range(20)] |
| 1029 | for t in threads: |
| 1030 | t.start() |
| 1031 | for t in threads: |
| 1032 | t.join() |
| 1033 | |
| 1034 | assert errors == [], f"Thread errors: {errors}" |
| 1035 | loaded = load_resolution(repo, pattern.pattern_id, res.resolution_id) |
| 1036 | assert loaded is not None |
| 1037 | # At least one increment must have landed; file must be valid JSON |
| 1038 | assert loaded.applied_count >= 1 |
| 1039 | |
| 1040 | def test_list_patterns_scan_cap_does_not_crash( |
| 1041 | self, repo: pathlib.Path, monkeypatch: pytest.MonkeyPatch |
| 1042 | ) -> None: |
| 1043 | """If _MAX_SCAN is set to 5, list_patterns truncates rather than crashing.""" |
| 1044 | monkeypatch.setattr(h, "_MAX_SCAN", 5) |
| 1045 | for i in range(10): |
| 1046 | record_pattern(repo, _make_pattern(path=f"s{i}.mid", ours=f"o{i}", theirs=f"t{i}")) |
| 1047 | results = list_patterns(repo) |
| 1048 | assert len(results) <= 5 |
| 1049 | |
| 1050 | |
| 1051 | # =========================================================================== |
| 1052 | # Tier V — Data integrity |
| 1053 | # =========================================================================== |
| 1054 | |
| 1055 | |
| 1056 | class TestDataIntegrity: |
| 1057 | """V: atomic writes (no temp files left), JSON round-trip, field type preservation.""" |
| 1058 | |
| 1059 | def test_no_temp_files_after_record_pattern(self, repo: pathlib.Path) -> None: |
| 1060 | pattern = _make_pattern() |
| 1061 | record_pattern(repo, pattern) |
| 1062 | entry_dir = h.patterns_dir(repo) / pattern.pattern_id |
| 1063 | tmp_files = list(entry_dir.glob(".harmony-tmp-*")) |
| 1064 | assert tmp_files == [] |
| 1065 | |
| 1066 | def test_no_temp_files_after_save_resolution(self, repo: pathlib.Path) -> None: |
| 1067 | pattern = _make_pattern() |
| 1068 | record_pattern(repo, pattern) |
| 1069 | res = _make_resolution(pattern) |
| 1070 | save_resolution(repo, res) |
| 1071 | res_dir = h._resolutions_dir(repo, pattern.pattern_id) |
| 1072 | tmp_files = list(res_dir.glob(".harmony-tmp-*")) |
| 1073 | assert tmp_files == [] |
| 1074 | |
| 1075 | def test_no_temp_files_after_save_policy(self, repo: pathlib.Path) -> None: |
| 1076 | policy = _make_policy() |
| 1077 | save_policy(repo, policy) |
| 1078 | tmp_files = list(h.policies_dir(repo).glob(".harmony-tmp-*")) |
| 1079 | assert tmp_files == [] |
| 1080 | |
| 1081 | def test_pattern_json_round_trip(self, repo: pathlib.Path) -> None: |
| 1082 | pattern = _make_pattern( |
| 1083 | path="round/trip.mid", |
| 1084 | domain="midi", |
| 1085 | conflict_type=ConflictType.STRUCTURAL, |
| 1086 | description={"beats": 4, "key": "Cmaj"}, |
| 1087 | ) |
| 1088 | record_pattern(repo, pattern) |
| 1089 | loaded = load_pattern(repo, pattern.pattern_id) |
| 1090 | assert loaded is not None |
| 1091 | assert loaded.path == "round/trip.mid" |
| 1092 | assert loaded.domain == "midi" |
| 1093 | assert loaded.conflict_type == ConflictType.STRUCTURAL |
| 1094 | assert loaded.description == {"beats": 4, "key": "Cmaj"} |
| 1095 | assert loaded.ours_id == pattern.ours_id |
| 1096 | assert loaded.theirs_id == pattern.theirs_id |
| 1097 | |
| 1098 | def test_resolution_json_round_trip(self, repo: pathlib.Path) -> None: |
| 1099 | pattern = _make_pattern() |
| 1100 | record_pattern(repo, pattern) |
| 1101 | prov = AgentProvenance.agent("claude-code", "claude-sonnet-4-6") |
| 1102 | res = _make_resolution(pattern, provenance=prov, confidence=0.77, human_verified=True) |
| 1103 | save_resolution(repo, res) |
| 1104 | loaded = load_resolution(repo, pattern.pattern_id, res.resolution_id) |
| 1105 | assert loaded is not None |
| 1106 | assert loaded.confidence == pytest.approx(0.77) |
| 1107 | assert loaded.human_verified is True |
| 1108 | assert loaded.resolved_by.type == "agent" |
| 1109 | assert loaded.resolved_by.agent_id == "claude-code" |
| 1110 | assert loaded.resolved_by.model_id == "claude-sonnet-4-6" |
| 1111 | |
| 1112 | def test_policy_json_round_trip(self, repo: pathlib.Path) -> None: |
| 1113 | policy = _make_policy( |
| 1114 | "round-trip-policy", |
| 1115 | scope=PolicyScope.WORKSPACE, |
| 1116 | action=PolicyAction.ESCALATE, |
| 1117 | confidence=0.6, |
| 1118 | conflict_type=ConflictType.RELATIONAL, |
| 1119 | domain="code", |
| 1120 | path_pattern="src/**", |
| 1121 | ) |
| 1122 | save_policy(repo, policy) |
| 1123 | loaded = load_policy(repo, policy.policy_id) |
| 1124 | assert loaded is not None |
| 1125 | assert loaded.scope == PolicyScope.WORKSPACE |
| 1126 | assert loaded.action == PolicyAction.ESCALATE |
| 1127 | assert loaded.confidence == pytest.approx(0.6) |
| 1128 | assert loaded.when.conflict_type == ConflictType.RELATIONAL |
| 1129 | assert loaded.when.domain == "code" |
| 1130 | assert loaded.when.path_pattern == "src/**" |
| 1131 | |
| 1132 | def test_agent_provenance_round_trip(self) -> None: |
| 1133 | p = AgentProvenance.agent("claude-code", "claude-sonnet-4-6") |
| 1134 | d = p.to_dict() |
| 1135 | restored = AgentProvenance.from_dict(d) |
| 1136 | assert restored == p |
| 1137 | |
| 1138 | def test_agent_provenance_human_round_trip(self) -> None: |
| 1139 | p = AgentProvenance.human() |
| 1140 | restored = AgentProvenance.from_dict(p.to_dict()) |
| 1141 | assert restored == p |
| 1142 | |
| 1143 | def test_recorded_at_is_utc_aware(self, repo: pathlib.Path) -> None: |
| 1144 | pattern = _make_pattern() |
| 1145 | record_pattern(repo, pattern) |
| 1146 | loaded = load_pattern(repo, pattern.pattern_id) |
| 1147 | assert loaded is not None |
| 1148 | assert loaded.recorded_at.tzinfo is not None |
| 1149 | |
| 1150 | def test_resolved_at_is_utc_aware(self, repo: pathlib.Path) -> None: |
| 1151 | pattern = _make_pattern() |
| 1152 | record_pattern(repo, pattern) |
| 1153 | res = _make_resolution(pattern) |
| 1154 | save_resolution(repo, res) |
| 1155 | loaded = load_resolution(repo, pattern.pattern_id, res.resolution_id) |
| 1156 | assert loaded is not None |
| 1157 | assert loaded.resolved_at.tzinfo is not None |
| 1158 | |
| 1159 | def test_applied_count_starts_at_zero(self, repo: pathlib.Path) -> None: |
| 1160 | pattern = _make_pattern() |
| 1161 | record_pattern(repo, pattern) |
| 1162 | res = _make_resolution(pattern) |
| 1163 | save_resolution(repo, res) |
| 1164 | loaded = load_resolution(repo, pattern.pattern_id, res.resolution_id) |
| 1165 | assert loaded is not None |
| 1166 | assert loaded.applied_count == 0 |
| 1167 | |
| 1168 | |
| 1169 | # =========================================================================== |
| 1170 | # Tier VI — Security |
| 1171 | # =========================================================================== |
| 1172 | |
| 1173 | |
| 1174 | class TestSecurity: |
| 1175 | """VI: path traversal, symlink guards, size caps, crafted policy_id.""" |
| 1176 | |
| 1177 | def test_path_traversal_in_pattern_id_rejected(self, repo: pathlib.Path) -> None: |
| 1178 | with pytest.raises(ValueError): |
| 1179 | h._validate_id("../../../etc/passwd") |
| 1180 | |
| 1181 | def test_path_traversal_in_load_pattern(self, repo: pathlib.Path) -> None: |
| 1182 | result = load_pattern(repo, "../evil") |
| 1183 | assert result is None |
| 1184 | |
| 1185 | def test_path_traversal_in_load_resolution(self, repo: pathlib.Path) -> None: |
| 1186 | result = load_resolution(repo, "a" * 64, "../evil") |
| 1187 | assert result is None |
| 1188 | |
| 1189 | def test_path_traversal_in_forget_pattern(self, repo: pathlib.Path) -> None: |
| 1190 | result = forget_pattern(repo, "../evil/../../../../../etc") |
| 1191 | assert result is False |
| 1192 | |
| 1193 | def test_crafted_policy_id_with_slash(self, repo: pathlib.Path) -> None: |
| 1194 | result = load_policy(repo, "../../etc/passwd") |
| 1195 | assert result is None |
| 1196 | |
| 1197 | def test_crafted_policy_id_with_null_byte(self, repo: pathlib.Path) -> None: |
| 1198 | with pytest.raises(ValueError): |
| 1199 | h._validate_policy_id("policy\x00id") |
| 1200 | |
| 1201 | def test_symlinks_in_patterns_dir_skipped(self, repo: pathlib.Path) -> None: |
| 1202 | pdir = h.patterns_dir(repo) |
| 1203 | pdir.mkdir(parents=True, exist_ok=True) |
| 1204 | # Create a symlink in the patterns dir |
| 1205 | link = pdir / ("s" * 64) |
| 1206 | target = repo / "other" |
| 1207 | target.mkdir() |
| 1208 | link.symlink_to(target) |
| 1209 | results = list_patterns(repo) |
| 1210 | assert results == [] # symlink skipped |
| 1211 | |
| 1212 | def test_symlinks_in_policies_dir_skipped(self, repo: pathlib.Path) -> None: |
| 1213 | poldir = h.policies_dir(repo) |
| 1214 | poldir.mkdir(parents=True, exist_ok=True) |
| 1215 | link = poldir / "linked-policy.json" |
| 1216 | target = repo / "victim.json" |
| 1217 | target.write_text('{"evil": true}') |
| 1218 | link.symlink_to(target) |
| 1219 | results = list_policies(repo) |
| 1220 | assert results == [] |
| 1221 | |
| 1222 | def test_oversized_pattern_file_rejected(self, repo: pathlib.Path) -> None: |
| 1223 | pattern = _make_pattern() |
| 1224 | record_pattern(repo, pattern) |
| 1225 | meta_p = h.patterns_dir(repo) / pattern.pattern_id / "pattern.json" |
| 1226 | # Overwrite with a file exceeding cap |
| 1227 | meta_p.write_bytes(b"x" * (_MAX_PATTERN_BYTES + 1)) |
| 1228 | result = load_pattern(repo, pattern.pattern_id) |
| 1229 | assert result is None |
| 1230 | |
| 1231 | def test_oversized_resolution_file_rejected(self, repo: pathlib.Path) -> None: |
| 1232 | pattern = _make_pattern() |
| 1233 | record_pattern(repo, pattern) |
| 1234 | res = _make_resolution(pattern) |
| 1235 | save_resolution(repo, res) |
| 1236 | dest = h._resolutions_dir(repo, pattern.pattern_id) / f"{res.resolution_id}.json" |
| 1237 | dest.write_bytes(b"y" * (_MAX_RESOLUTION_BYTES + 1)) |
| 1238 | result = load_resolution(repo, pattern.pattern_id, res.resolution_id) |
| 1239 | assert result is None |
| 1240 | |
| 1241 | def test_oversized_policy_file_rejected(self, repo: pathlib.Path) -> None: |
| 1242 | policy = _make_policy() |
| 1243 | save_policy(repo, policy) |
| 1244 | dest = h.policies_dir(repo) / f"{policy.policy_id}.json" |
| 1245 | dest.write_bytes(b"z" * (_MAX_POLICY_BYTES + 1)) |
| 1246 | result = load_policy(repo, policy.policy_id) |
| 1247 | assert result is None |
| 1248 | |
| 1249 | def test_malformed_json_pattern_returns_none(self, repo: pathlib.Path) -> None: |
| 1250 | pdir = h.patterns_dir(repo) |
| 1251 | pid = "a" * 64 |
| 1252 | entry = pdir / pid |
| 1253 | entry.mkdir(parents=True) |
| 1254 | (entry / "pattern.json").write_text("not json", encoding="utf-8") |
| 1255 | assert load_pattern(repo, pid) is None |
| 1256 | |
| 1257 | def test_malformed_json_resolution_returns_none(self, repo: pathlib.Path) -> None: |
| 1258 | pattern = _make_pattern() |
| 1259 | record_pattern(repo, pattern) |
| 1260 | res_dir = h._resolutions_dir(repo, pattern.pattern_id) |
| 1261 | res_dir.mkdir(parents=True, exist_ok=True) |
| 1262 | rid = "b" * 64 |
| 1263 | (res_dir / f"{rid}.json").write_text("{{bad", encoding="utf-8") |
| 1264 | assert load_resolution(repo, pattern.pattern_id, rid) is None |
| 1265 | |
| 1266 | def test_non_hex_dir_in_patterns_dir_skipped(self, repo: pathlib.Path) -> None: |
| 1267 | pdir = h.patterns_dir(repo) |
| 1268 | pdir.mkdir(parents=True, exist_ok=True) |
| 1269 | (pdir / "not-a-valid-id").mkdir() |
| 1270 | assert list_patterns(repo) == [] |
| 1271 | |
| 1272 | |
| 1273 | # =========================================================================== |
| 1274 | # Tier VII — Performance |
| 1275 | # =========================================================================== |
| 1276 | |
| 1277 | |
| 1278 | class TestPerformance: |
| 1279 | """VII: operation timing assertions — single operations must be fast.""" |
| 1280 | |
| 1281 | def test_record_pattern_under_50ms(self, repo: pathlib.Path) -> None: |
| 1282 | pattern = _make_pattern() |
| 1283 | start = time.monotonic() |
| 1284 | record_pattern(repo, pattern) |
| 1285 | elapsed = (time.monotonic() - start) * 1000 |
| 1286 | assert elapsed < 50, f"record_pattern took {elapsed:.1f}ms — expected <50ms" |
| 1287 | |
| 1288 | def test_load_pattern_under_10ms(self, repo: pathlib.Path) -> None: |
| 1289 | pattern = _make_pattern() |
| 1290 | record_pattern(repo, pattern) |
| 1291 | start = time.monotonic() |
| 1292 | load_pattern(repo, pattern.pattern_id) |
| 1293 | elapsed = (time.monotonic() - start) * 1000 |
| 1294 | assert elapsed < 10, f"load_pattern took {elapsed:.1f}ms — expected <10ms" |
| 1295 | |
| 1296 | def test_save_resolution_under_50ms(self, repo: pathlib.Path) -> None: |
| 1297 | pattern = _make_pattern() |
| 1298 | record_pattern(repo, pattern) |
| 1299 | res = _make_resolution(pattern) |
| 1300 | start = time.monotonic() |
| 1301 | save_resolution(repo, res) |
| 1302 | elapsed = (time.monotonic() - start) * 1000 |
| 1303 | assert elapsed < 50, f"save_resolution took {elapsed:.1f}ms — expected <50ms" |
| 1304 | |
| 1305 | def test_list_100_patterns_under_500ms(self, repo: pathlib.Path) -> None: |
| 1306 | n = 100 |
| 1307 | for i in range(n): |
| 1308 | record_pattern(repo, _make_pattern(path=f"perf{i}.mid", ours=f"o{i}", theirs=f"t{i}")) |
| 1309 | start = time.monotonic() |
| 1310 | results = list_patterns(repo) |
| 1311 | elapsed = (time.monotonic() - start) * 1000 |
| 1312 | assert len(results) == n |
| 1313 | assert elapsed < 500, f"list_patterns(100) took {elapsed:.1f}ms — expected <500ms" |
| 1314 | |
| 1315 | def test_save_and_load_policy_under_20ms(self, repo: pathlib.Path) -> None: |
| 1316 | policy = _make_policy() |
| 1317 | start = time.monotonic() |
| 1318 | save_policy(repo, policy) |
| 1319 | load_policy(repo, policy.policy_id) |
| 1320 | elapsed = (time.monotonic() - start) * 1000 |
| 1321 | assert elapsed < 20, f"save+load policy took {elapsed:.1f}ms — expected <20ms" |
| 1322 | |
| 1323 | def test_append_audit_under_20ms(self, repo: pathlib.Path) -> None: |
| 1324 | actor = AgentProvenance.human() |
| 1325 | start = time.monotonic() |
| 1326 | append_audit(repo, AuditEventType.GC_RUN, actor) |
| 1327 | elapsed = (time.monotonic() - start) * 1000 |
| 1328 | assert elapsed < 20, f"append_audit took {elapsed:.1f}ms — expected <20ms" |
| 1329 | |
| 1330 | def test_increment_applied_count_under_20ms(self, repo: pathlib.Path) -> None: |
| 1331 | pattern = _make_pattern() |
| 1332 | record_pattern(repo, pattern) |
| 1333 | res = _make_resolution(pattern) |
| 1334 | save_resolution(repo, res) |
| 1335 | start = time.monotonic() |
| 1336 | increment_applied_count(repo, pattern.pattern_id, res.resolution_id) |
| 1337 | elapsed = (time.monotonic() - start) * 1000 |
| 1338 | assert elapsed < 20, f"increment_applied_count took {elapsed:.1f}ms — expected <20ms" |
File History
1 commit
sha256:88ac91129873e6a496e9189515aa690eb893ae25d69c8f72af141a2be5068eb3
docs: docstring sprint contract→find-symbol — idiomatic run…
Sonnet 4.6
patch
140 days ago