feat: pfad-metadata-parser mit semester/fach/typ

app/ingest/metadata.py

@@ -0,0 +1,44 @@
+import re
+from dataclasses import dataclass
+from pathlib import PurePosixPath
+
+
+SEMESTER_RE = re.compile(r"^\d+\.Semester$")
+
+
+@dataclass(frozen=True)
+class PathMetadata:
+    semester: str
+    fach: str
+    typ: str | None
+
+
+def parse_path(file_path: str, ingest_root: str) -> PathMetadata | None:
+    """Parse a Nextcloud file path into structured metadata.
+
+    Returns None when the path is outside the ingest root or does not match
+    the expected `<root>/<N>.Semester/<Fach>/[<typ>/...]/<file>` pattern.
+    """
+    norm_path = file_path.lstrip("/")
+    norm_root = ingest_root.strip("/")
+
+    if not norm_path.startswith(norm_root + "/"):
+        return None
+
+    relative = norm_path[len(norm_root) + 1:]
+    parts = PurePosixPath(relative).parts
+
+    # Need at least: semester / fach / file.ext  → 3 parts
+    if len(parts) < 3:
+        return None
+
+    semester, fach = parts[0], parts[1]
+
+    if not SEMESTER_RE.match(semester):
+        return None
+
+    # parts[-1] is the filename. Anything between fach and filename is "deeper".
+    # The first deeper segment becomes `typ`. None if file lives directly in fach.
+    typ = parts[2] if len(parts) > 3 else None
+
+    return PathMetadata(semester=semester, fach=fach, typ=typ)

tests/test_metadata.py

@@ -0,0 +1,46 @@
+from app.ingest.metadata import parse_path, PathMetadata
+
+
+ROOT = "Documents/THB/Studium"
+
+
+def test_parse_path_with_typ():
+    md = parse_path("Documents/THB/Studium/2.Semester/Databases/Vorlesungen/DBS1.pdf", ROOT)
+    assert md == PathMetadata(semester="2.Semester", fach="Databases", typ="Vorlesungen")
+
+
+def test_parse_path_without_typ():
+    md = parse_path("Documents/THB/Studium/2.Semester/Databases/DBS1.pdf", ROOT)
+    assert md == PathMetadata(semester="2.Semester", fach="Databases", typ=None)
+
+
+def test_parse_path_deep_nested_keeps_first_subdir_as_typ():
+    md = parse_path("Documents/THB/Studium/2.Semester/Databases/Uebungen/01/sheet.pdf", ROOT)
+    assert md == PathMetadata(semester="2.Semester", fach="Databases", typ="Uebungen")
+
+
+def test_parse_path_outside_root_returns_none():
+    assert parse_path("Documents/Other/file.pdf", ROOT) is None
+
+
+def test_parse_path_loose_file_in_thb_returns_none():
+    assert parse_path("Documents/THB/Studienbescheinigung.pdf", ROOT) is None
+
+
+def test_parse_path_loose_file_under_root_returns_none():
+    # File directly under Studium with no semester folder
+    assert parse_path("Documents/THB/Studium/readme.txt", ROOT) is None
+
+
+def test_parse_path_invalid_semester_pattern_returns_none():
+    assert parse_path("Documents/THB/Studium/Sommersemester/Databases/x.pdf", ROOT) is None
+
+
+def test_parse_path_no_fach_returns_none():
+    # File directly in Semester folder with no fach subdir
+    assert parse_path("Documents/THB/Studium/2.Semester/loose.pdf", ROOT) is None
+
+
+def test_parse_path_with_leading_slash_normalizes():
+    md = parse_path("/Documents/THB/Studium/2.Semester/Databases/x.pdf", ROOT)
+    assert md == PathMetadata(semester="2.Semester", fach="Databases", typ=None)