sorter: deepest-matching-dir re-homing + weak 0.50 confidence tier

Inner directory names now compete with the pack/file name: a subtree
under a dir whose name matches the catalog better than the pack-level
resolution is re-homed there (fixes multi-unit sculptor packs like
'Retro Elvez' dumping everything into one keyword class). Junk dir
names (files/images/HEADS/2/...) never match. Hits between 0.50 and the
threshold are accepted as low-confidence: [w] in the plan, confidence
in the audit, listed in sorter_low_confidence.txt. Works in both
provenance and manual-tree mode.

tests/test_sorter_units.py

@@ -0,0 +1,149 @@
+"""Offline sorter tests: dir re-homing, weak tier, junk dirs (plain python3)."""
+import json
+import shutil
+import sys
+import tempfile
+from pathlib import Path
+
+sys.path.insert(0, str(Path(__file__).resolve().parent.parent))
+
+from dlphelper.sorter import (  # noqa: E402
+    apply_plan,
+    load_catalog,
+    sort_plan,
+)
+
+CATALOG = {
+    "entries": [
+        {"name": "Warp Spiders", "system": "Warhammer 40,000",
+         "group": "Xenos Armies", "faction": "Aeldari",
+         "class": "Fly", "aliases": []},
+        {"name": "Harlequin Troupe", "system": "Warhammer 40,000",
+         "group": "Xenos Armies", "faction": "Harlequins",
+         "class": "Battleline", "aliases": []},
+        {"name": "Guardian Squad", "system": "Warhammer 40,000",
+         "group": "Xenos Armies", "faction": "Aeldari",
+         "class": "Battleline", "aliases": []},
+    ]
+}
+
+
+def make_lib():
+    return Path(tempfile.mkdtemp(prefix="sorter_test_lib_"))
+
+
+def write(p: Path):
+    p.parent.mkdir(parents=True, exist_ok=True)
+    p.write_bytes(b"x")
+
+
+_LOADED = None
+
+
+def get_catalog():
+    global _LOADED
+    if _LOADED is None:
+        _LOADED = load_catalog_path()
+    return _LOADED
+
+
+def prov_items(root: Path, library):
+    (root / "extract_manifest.jsonl").write_text(json.dumps({
+        "archive": str(root / "Multi Pack.zip"),
+        "dest": "Multi Pack", "unit": "Multi Pack.zip",
+    }), encoding="utf-8")
+    files = sorted((root / "Multi Pack").rglob("*"))
+    return sort_plan(files, library, get_catalog(), provenance={
+        "Multi Pack": "Multi Pack.zip"}, root=root)
+
+
+def test_prov_dir_rehome():
+    """A matching inner dir name claims its subtree; junk dirs don't."""
+    root = Path(tempfile.mkdtemp(prefix="sorter_test_src_"))
+    pack = root / "Multi Pack"
+    write(pack / "14 Warp spiders" / "spider_body.stl")
+    write(pack / "14 Warp spiders" / "files" / "spider_arm.stl")
+    write(pack / "HEADS" / "head_a.stl")          # junk dir: stays with pack
+    write(pack / "Retro GrimDark Thing" / "t.stl")  # no catalog match
+    library = make_lib()
+    items = prov_items(root, library)
+    tgt = {i.name: i.target for i in items}
+    lib = str(library)
+    # warp spiders re-homed, incl. nested junk dir structure, class from catalog
+    assert "Warp Spiders" in str(tgt["spider_body.stl"]), tgt["spider_body.stl"]
+    assert "Fly" in str(tgt["spider_body.stl"]), tgt["spider_body.stl"]
+    assert "Hero" not in str(tgt["spider_body.stl"])
+    assert "files" in str(tgt["spider_arm.stl"])
+    # junk-dir file stays under the pack unit folder
+    assert "_unsorted/Multi Pack/HEADS" in str(tgt["head_a.stl"]).replace("\\", "/")
+    # unmatched subtree keeps the pack unit
+    assert "Multi Pack" in str(tgt["t.stl"])
+    shutil.rmtree(root); shutil.rmtree(library)
+
+
+def test_prov_weak_tier():
+    """Below-threshold dir hit is accepted as low confidence."""
+    root = Path(tempfile.mkdtemp(prefix="sorter_test_src_"))
+    pack = root / "Multi Pack"
+    # 'guardian squad extras' scores below 1.0 but high vs 'Guardian Squad'
+    write(pack / "guardian squad extras" / "g.stl")
+    library = make_lib()
+    items = prov_items(root, library)
+    it = next(i for i in items if i.name == "g.stl")
+    assert it.match_kind in ("weak", "catalog"), it.match_kind
+    if it.match_kind == "weak":
+        assert it.confidence == "low"
+    assert "Guardian" in str(it.target)
+    shutil.rmtree(root); shutil.rmtree(library)
+
+
+def test_nonprov_rehome():
+    """Loose noise file inside a named dir re-homes to the dir's unit."""
+    root = Path(tempfile.mkdtemp(prefix="sorter_test_src_"))
+    write(root / "Warp Spiders" / "body.stl")
+    library = make_lib()
+    files = sorted(root.rglob("*"))
+    items = sort_plan(files, library, get_catalog(), root=root)
+    it = next(i for i in items if i.name == "body.stl")
+    assert it.match_kind == "catalog", (it.match_kind, it.target)
+    assert "Warp Spiders" in str(it.target) and "Fly" in str(it.target)
+    shutil.rmtree(root); shutil.rmtree(library)
+
+
+def test_apply_weak_doc():
+    """apply_plan writes the low-confidence review document."""
+    root = Path(tempfile.mkdtemp(prefix="sorter_test_src_"))
+    pack = root / "Multi Pack"
+    write(pack / "guardian squad extras" / "g.stl")
+    library = make_lib()
+    items = prov_items(root, library)
+    linked, copied, skipped = apply_plan(items, library / "sorter_applied.jsonl")
+    doc = library / "sorter_low_confidence.txt"
+    kinds = [i.match_kind for i in items if i.match_kind == "weak"]
+    if kinds:
+        assert doc.exists(), "weak doc missing"
+        assert "low-confidence" in doc.read_text(encoding="utf-8")
+    else:
+        assert not doc.exists()
+    assert (library / "sorter_applied.jsonl").exists()
+    shutil.rmtree(root); shutil.rmtree(library)
+
+
+def load_catalog_path():
+    import tempfile as tf
+    fh = tf.NamedTemporaryFile(mode="w", suffix=".json", delete=False)
+    json.dump(CATALOG, fh)
+    fh.close()
+    return load_catalog(Path(fh.name))
+
+
+def main():
+    catalog = load_catalog_path()
+    for fn in (test_prov_dir_rehome, test_prov_weak_tier, test_nonprov_rehome,
+               test_apply_weak_doc):
+        fn()
+        print("OK", fn.__name__)
+
+
+if __name__ == "__main__":
+    main()