"""Tests for the canonical naming module—R4 SHORT grammar (Phase 1 naming-reset). Covers BUILD_SPEC Phase 1: - single-shot, strategy grouping, multi-shot pass, happy-accident suffix, sidecar compound ext - round-trip build/parse for every shape - parse_filename dual-reads legacy _PASS_ / strategy-token filenames - LEGACY_FILENAME_PATTERN still parses R3-era verbose names (lives in the migration script post-Phase-11) - bad take_suffix shape rejected by ValueError """ from __future__ import annotations import pytest from recoil.core.naming import ( build_filename, next_take_number, parse_filename, parse_to_build_args, ) def test_next_take_number_increments_no_overwrite(tmp_path): """REC-102 regression: re-generation must advance the take, never reuse (overwrite) an existing one. The scan stem is derived from build_filename, so it tracks the strategy-token grammar (the bug was a stale glob that no longer matched ``_COV_...`` names → repeat gen kept choosing take1).""" args = dict( episode=1, strategy="coverage", ordinal=7, shot_ids=["EP001_SH16", "EP001_SH17", "EP001_SH18"], ) # empty dir → first take is 1 assert next_take_number(tmp_path, **args) == 1 # drop a REAL take1 file under the canonical strategy-token scheme take1 = tmp_path / build_filename(take=1, **args) take1.write_bytes(b"v1") assert take1.name == "EP001_COV_007_SH16_17_18_take1.mp4" # next must be 2 — returning 1 here is the overwrite bug assert next_take_number(tmp_path, **args) == 2 # add take2 → next is 3 (max+1) (tmp_path / build_filename(take=2, **args)).write_bytes(b"v2") assert next_take_number(tmp_path, **args) == 3 # the original take1 is never touched assert take1.exists() and take1.read_bytes() == b"v1" def test_next_take_number_scoped_to_identity(tmp_path): """A different ordinal / shot set must NOT count toward this pass's takes.""" a = dict(episode=1, strategy="coverage", ordinal=7, shot_ids=["EP001_SH16", "EP001_SH17"]) b = dict(episode=1, strategy="coverage", ordinal=8, shot_ids=["EP001_SH20"]) (tmp_path / build_filename(take=1, **b)).write_bytes(b"v") (tmp_path / build_filename(take=2, **b)).write_bytes(b"v") # pass A shares none of B's stem → starts fresh at 1 assert next_take_number(tmp_path, **a) == 1 # pass B correctly sees its own two takes → next is 3 assert next_take_number(tmp_path, **b) == 3 def test_single_shot_short_grammar(): name = build_filename(episode=1, shot_ids=["EP001_SH10"], take=16) assert name == "EP001_SH10_take16.mp4" def test_multi_shot_pass_segment(): name = build_filename( episode=1, pass_counter=7, shot_ids=["EP001_SH16", "EP001_SH17", "EP001_SH18"], take=2, ) assert name == "EP001_PASS_007_SH16_17_18_take2.mp4" @pytest.mark.parametrize( ("strategy", "ordinal", "token"), [ ("continuity", 2, "CONT"), ("coverage", 11, "COV"), ("oner", 3, "ONER"), ], ) def test_grouping_strategy_filename_builds_and_parses(strategy, ordinal, token): name = build_filename( episode=1, strategy=strategy, ordinal=ordinal, shot_ids=["EP001_SH16", "EP001_SH17", "EP001_SH18"], take=2, ) assert name == f"EP001_{token}_{ordinal:03d}_SH16_17_18_take2.mp4" parsed = parse_filename(name) assert parsed is not None assert parsed["strategy"] == strategy assert parsed["ordinal"] == ordinal assert parsed["grouping_token"] == token assert parsed["legacy_grouping"] is False def test_solo_strategy_is_tokenless_grouping_with_ordinal_zero(): name = build_filename( episode=1, strategy="solo", ordinal=0, shot_ids=["EP001_SH10"], take=16, ) assert name == "EP001_SH10_take16.mp4" parsed = parse_filename(name) assert parsed is not None assert parsed["strategy"] == "solo" assert parsed["ordinal"] == 0 assert parsed["grouping_token"] is None assert parsed["pass_counter"] is None def test_legacy_pass_parses_as_coverage_grouping_and_rebuilds_legacy(): legacy_name = "EP001_PASS_017_SH33_33a_34_35_take1.mp4" parsed = parse_filename(legacy_name) assert parsed is not None assert parsed["strategy"] == "coverage" assert parsed["ordinal"] == 17 assert parsed["pass_counter"] == 17 assert parsed["grouping_token"] == "PASS" assert parsed["legacy_grouping"] is True build_args = parse_to_build_args(parsed) assert build_args["pass_counter"] == 17 assert "strategy" not in build_args assert "ordinal" not in build_args assert build_filename(**build_args) == legacy_name def test_strategy_build_rejects_mixed_legacy_pass_counter(): with pytest.raises(ValueError): build_filename( episode=1, strategy="coverage", ordinal=1, pass_counter=1, shot_ids=["EP001_SH10"], take=1, ) with pytest.raises(ValueError): build_filename( episode=1, ordinal=1, pass_counter=1, shot_ids=["EP001_SH10"], take=1, ) def test_happy_accident_suffix(): name = build_filename( episode=1, shot_ids=["EP001_SH10"], take=16, take_suffix="~a", ) assert name == "EP001_SH10_take16~a.mp4" def test_sidecar_compound_ext(): name = build_filename( episode=1, shot_ids=["EP001_SH33"], take=7, ext=".mp4.json", ) assert name == "EP001_SH33_take7.mp4.json" def test_round_trip_all_shapes(): for args in [ dict(episode=1, shot_ids=["EP001_SH10"], take=16), dict(episode=1, strategy="solo", ordinal=0, shot_ids=["EP001_SH10"], take=16), dict(episode=1, strategy="continuity", ordinal=2, shot_ids=["EP001_SH16", "EP001_SH17"], take=2), dict(episode=1, strategy="coverage", ordinal=11, shot_ids=["EP001_SH16", "EP001_SH17", "EP001_SH18"], take=2), dict(episode=1, strategy="oner", ordinal=3, shot_ids=["EP001_SH16", "EP001_SH17", "EP001_SH18"], take=2), dict(episode=1, shot_ids=["EP001_SH10"], take=16, take_suffix="~a"), dict(episode=1, pass_counter=7, shot_ids=["EP001_SH16", "EP001_SH17", "EP001_SH18"], take=2), dict(episode=1, shot_ids=["EP001_SH33"], take=7, ext=".mp4.json"), dict(episode=2, pass_counter=21, shot_ids=["EP002_SH1", "EP002_SH2a"], take=1), ]: name = build_filename(**args) parsed = parse_filename(name) assert parsed is not None, f"parse_filename(None) for {name!r}" rebuilt = build_filename(**parse_to_build_args(parsed)) assert rebuilt == name, f"round-trip drift: {name!r} -> {rebuilt!r}" def test_parse_returns_none_on_no_match(): assert parse_filename("not_a_canonical_name.mp4") is None assert parse_filename( "TARTARUS_EP001_PASS_000_SH10_SOLO_ENV_seeddance-2-0_take16.mp4" ) is None # legacy grammar rejected by new regex def test_legacy_pattern_matches_old_grammar(): """R3-era verbose names still parse under the legacy regex. The migration script (recoil/pipeline/tools/migrate_pass_names_r4.py) that once exported LEGACY_FILENAME_PATTERN was deleted in commit 8395db8d once the v2->v3 migration completed. The pattern is inlined here so legacy-name coverage survives the script's retirement. """ import re as _re LEGACY_FILENAME_PATTERN = _re.compile( r"^([A-Z][A-Z0-9_]*?)" # 1: project r"_(EP\d{3})" # 2: episode r"_PASS_(\d{3})" # 3: pass counter r"_SH(\d{1,4}[a-z]?(?:_\d{1,4}[a-z]?)*)" # 4: shot list r"_([A-Z][A-Z0-9]*(?:_[A-Z][A-Z0-9]*)*)" # 5: semantic tag r"_([a-z][a-z0-9\-]*)" # 6: model filename_id r"_take(\d+)" # 7: take r"(\.[a-zA-Z0-9]+(?:\.[a-zA-Z0-9]+)?)$" # 8: extension ) m = LEGACY_FILENAME_PATTERN.match( "TARTARUS_EP001_PASS_000_SH10_SOLO_ENV_seeddance-2-0_take16.mp4" ) assert m is not None assert m.group(1) == "TARTARUS" assert m.group(6) == "seeddance-2-0" def test_bad_take_suffix_rejected(): with pytest.raises(ValueError): build_filename(episode=1, shot_ids=["EP001_SH10"], take=1, take_suffix="~AB") with pytest.raises(ValueError): build_filename(episode=1, shot_ids=["EP001_SH10"], take=1, take_suffix="a")