Add explainable matching engine

ROADMAP.md

@@ -13,6 +13,7 @@
 - [ ] Database V2 read-only parser
 - [x] Configuration
 - [x] Logging
+- [x] Matching engine
 
 ## v0.2 — Diagnostics
 

docs/design/matching-engine.md

@@ -0,0 +1,29 @@
+# Explainable Matching Engine
+
+## Problem
+
+Missing references need ranked candidate files, but a filename-only yes/no check
+cannot explain ambiguity or cloud-provider conflict names.
+
+## Architecture
+
+The read-only matching engine indexes normalized filenames and scores only related
+candidates. Every score contains evidence for filename, extension, and parent
+folder. Exact filenames earn 60 points, normalized names 55, numeric conflict-name
+matches 50, extensions 10, and parent folders 20.
+
+The displayed percentage is an evidence score, not a statistical probability.
+Metadata, duration, hashes, and fingerprints can add stronger evidence later.
+
+## Edge Cases
+
+- Unicode and case differences.
+- OneDrive-style names such as `Track 2.mp3`.
+- Duplicate candidates in different folders.
+- Legitimate numbered song titles, which remain candidates but are never repaired.
+- Unrelated names, which are not emitted as candidates.
+
+## Verification
+
+Tests cover exact, normalized, conflict-suffix, ambiguous, and unrelated filenames.
+Candidate ordering is deterministic. The engine never changes a track or reference.

serato_doctor/matching.py

@@ -0,0 +1,86 @@
+import re
+import unicodedata
+from collections import defaultdict
+from pathlib import Path
+from typing import DefaultDict, Iterable, List, Tuple
+
+from serato_doctor.models.match import MatchEvidence, TrackMatch
+from serato_doctor.models.reference import TrackReference
+from serato_doctor.models.track import DiskTrack
+
+
+def normalize(value: str) -> str:
+    return unicodedata.normalize("NFKC", value).casefold()
+
+
+def cloud_conflict_name(filename: str) -> str:
+    """Remove a trailing numeric cloud-conflict suffix from a filename stem."""
+
+    path = Path(filename)
+    stem = re.sub(r" \d+$", "", path.stem)
+    return normalize(stem + path.suffix)
+
+
+def score_candidate(reference: TrackReference, track: DiskTrack) -> TrackMatch:
+    reference_name = reference.filename
+    track_name = track.filename
+
+    if reference_name == track_name:
+        filename_points = 60
+        filename_reason = "Filename is identical"
+    elif normalize(reference_name) == normalize(track_name):
+        filename_points = 55
+        filename_reason = "Filename matches after case and Unicode normalization"
+    elif cloud_conflict_name(reference_name) == cloud_conflict_name(track_name):
+        filename_points = 50
+        filename_reason = "Filename matches after removing a numeric conflict suffix"
+    else:
+        filename_points = 0
+        filename_reason = "Filename does not match"
+
+    same_extension = normalize(reference.path.suffix) == normalize(track.suffix)
+    same_parent = normalize(reference.path.parent.name) == normalize(
+        track.path.parent.name
+    )
+    evidence = (
+        MatchEvidence(
+            "filename",
+            filename_points > 0,
+            filename_points,
+            60,
+            filename_reason,
+        ),
+        MatchEvidence(
+            "extension",
+            same_extension,
+            10 if same_extension else 0,
+            10,
+            "File extension matches" if same_extension else "File extension differs",
+        ),
+        MatchEvidence(
+            "parent_folder",
+            same_parent,
+            20 if same_parent else 0,
+            20,
+            "Parent folder matches" if same_parent else "Parent folder differs",
+        ),
+    )
+    return TrackMatch(reference, track, evidence)
+
+
+class MatchingEngine:
+    """Find and rank filename-related disk candidates without modifying files."""
+
+    def __init__(self, tracks: Iterable[DiskTrack]):
+        self._by_conflict_name: DefaultDict[str, List[DiskTrack]] = defaultdict(list)
+        for track in tracks:
+            self._by_conflict_name[cloud_conflict_name(track.filename)].append(track)
+
+    def candidates_for(self, reference: TrackReference) -> Tuple[TrackMatch, ...]:
+        candidates = self._by_conflict_name.get(
+            cloud_conflict_name(reference.filename), []
+        )
+        matches = [score_candidate(reference, track) for track in candidates]
+        return tuple(
+            sorted(matches, key=lambda match: (-match.score, str(match.track.path)))
+        )

serato_doctor/models/__init__.py

@@ -1,6 +1,15 @@
 from serato_doctor.models.crate import Crate
 from serato_doctor.models.library import Library
+from serato_doctor.models.match import MatchEvidence, TrackMatch
 from serato_doctor.models.reference import ReferenceResult, TrackReference
 from serato_doctor.models.track import DiskTrack
 
-__all__ = ["Crate", "DiskTrack", "Library", "ReferenceResult", "TrackReference"]
+__all__ = [
+    "Crate",
+    "DiskTrack",
+    "Library",
+    "MatchEvidence",
+    "ReferenceResult",
+    "TrackMatch",
+    "TrackReference",
+]

serato_doctor/models/match.py

@@ -0,0 +1,39 @@
+from dataclasses import dataclass
+from typing import Tuple
+
+from serato_doctor.models.reference import TrackReference
+from serato_doctor.models.track import DiskTrack
+
+
+@dataclass(frozen=True)
+class MatchEvidence:
+    """One explainable scoring decision for a candidate track."""
+
+    field: str
+    matched: bool
+    points: int
+    max_points: int
+    explanation: str
+
+
+@dataclass(frozen=True)
+class TrackMatch:
+    """A ranked candidate backed by explicit, inspectable evidence."""
+
+    reference: TrackReference
+    track: DiskTrack
+    evidence: Tuple[MatchEvidence, ...]
+
+    @property
+    def score(self) -> int:
+        return sum(item.points for item in self.evidence)
+
+    @property
+    def max_score(self) -> int:
+        return sum(item.max_points for item in self.evidence)
+
+    @property
+    def score_percent(self) -> float:
+        if not self.max_score:
+            return 0.0
+        return round(self.score / self.max_score * 100, 1)

tests/test_matcher.py

@@ -0,0 +1,62 @@
+from pathlib import Path
+
+from serato_doctor.matching import MatchingEngine, score_candidate
+from serato_doctor.models.reference import TrackReference
+from serato_doctor.models.track import DiskTrack
+
+
+def reference(filename, folder="House"):
+    return TrackReference(
+        Path("Test.crate"), Path("/old") / folder / filename, filename
+    )
+
+
+def track(filename, folder="House"):
+    path = Path("/new") / folder / filename
+    return DiskTrack(path, filename, 100, path.suffix.lower())
+
+
+def test_exact_candidate_has_full_evidence_score():
+    match = score_candidate(reference("Track.mp3"), track("Track.mp3"))
+
+    assert match.score == 90
+    assert match.max_score == 90
+    assert match.score_percent == 100.0
+    assert all(item.matched for item in match.evidence)
+
+
+def test_cloud_conflict_suffix_is_explained():
+    match = score_candidate(reference("Track.mp3"), track("Track 2.mp3"))
+
+    assert match.score == 80
+    assert match.score_percent == 88.9
+    assert match.evidence[0].explanation == (
+        "Filename matches after removing a numeric conflict suffix"
+    )
+
+
+def test_case_normalized_match_scores_below_exact():
+    match = score_candidate(reference("TRACK.MP3"), track("track.mp3"))
+
+    assert match.score == 85
+    assert match.evidence[0].points == 55
+
+
+def test_engine_omits_unrelated_filenames():
+    engine = MatchingEngine([track("Different.mp3")])
+
+    assert engine.candidates_for(reference("Missing.mp3")) == ()
+
+
+def test_ambiguous_candidates_are_ranked_deterministically():
+    engine = MatchingEngine(
+        [track("Track 2.mp3", "Other"), track("Track 3.mp3", "House")]
+    )
+
+    matches = engine.candidates_for(reference("Track.mp3"))
+
+    assert [match.track.filename for match in matches] == [
+        "Track 3.mp3",
+        "Track 2.mp3",
+    ]
+    assert [match.score for match in matches] == [80, 60]