test(pfs): lock in NPRM Addendum B data-quality handling; deterministic discovery (refs #592)

src/pfs/nprm.py

@@ -65,7 +65,8 @@ def _discover_addendum_b() -> Path:
             "JOIN items i ON i.id = a.item_id "
             "JOIN item_tags it ON it.item_id = i.id "
             "JOIN tags t ON t.id = it.tag_id "
-            "WHERE t.name = ?",
+            "WHERE t.name = ? "
+            "ORDER BY a.id",
             (_ADDENDUM_B_TAG,),
         )
         .fetchall()

tests/pfs/test_nprm.py

@@ -30,3 +30,63 @@ def test_load_rvu_proposed_loads_and_reloads(tmp_path, monkeypatch):
     assert row == ("99213", 1.3, "CMS-1832-P")
     nprm.load_rvu_proposed(con)  # idempotent delete-and-reload
     assert con.execute("SELECT count(*) FROM pfs.rvu_proposed").fetchone()[0] == 2
+
+
+def _dirty_fixture(tmp_path):
+    """Real-file quirks: padded cells, "NA" markers, a footnote row.
+
+    Mirrors what the actual CY2026 Addendum B workbook contains — this
+    combination broke silently on the real file (see task-5-report.md)
+    and was fixed via manual smoke-testing rather than a fixture test,
+    so these rows lock the fix in.
+    """
+    p = tmp_path / "CY2026_NPRM_Addendum_B_dirty.csv"
+    p.write_text(
+        "HCPCS,MOD,DESCRIPTION,STATUS CODE,WORK RVU,"
+        "NON-FAC PE RVU,FAC PE RVU,MP RVU\n"
+        "  99213  ,  ,Office visit est,A,  1.30  ,  1.50  ,  0.55  ,  0.10  \n"
+        "99214,,Office visit NA test,A,NA,NA,NA,NA\n"
+        "1 CPT codes and descriptors only are copyright 2025 American "
+        "Medical Association. All Rights Reserved.,,,,,,,\n"
+    )
+    return p
+
+
+def test_load_rvu_proposed_strips_padding(tmp_path, monkeypatch):
+    """Whitespace-padded HCPCS/numeric cells load clean, not null."""
+    path = _dirty_fixture(tmp_path)
+    monkeypatch.setattr(nprm, "_discover_addendum_b", lambda: path)
+    con = duckdb.connect()
+    nprm.load_rvu_proposed(con)
+    row = con.execute(
+        "SELECT hcpcs, description, work_rvu, nonfac_pe_rvu, fac_pe_rvu, mp_rvu "
+        "FROM pfs.rvu_proposed WHERE hcpcs='99213'"
+    ).fetchone()
+    assert row == ("99213", "Office visit est", 1.3, 1.5, 0.55, 0.1)
+
+
+def test_load_rvu_proposed_na_marker_becomes_null(tmp_path, monkeypatch):
+    """ "NA" in a numeric cell loads as SQL NULL, not a cast error."""
+    path = _dirty_fixture(tmp_path)
+    monkeypatch.setattr(nprm, "_discover_addendum_b", lambda: path)
+    con = duckdb.connect()
+    nprm.load_rvu_proposed(con)
+    row = con.execute(
+        "SELECT work_rvu, nonfac_pe_rvu, fac_pe_rvu, mp_rvu "
+        "FROM pfs.rvu_proposed WHERE hcpcs='99214'"
+    ).fetchone()
+    assert row == (None, None, None, None)
+
+
+def test_load_rvu_proposed_drops_footnote_rows(tmp_path, monkeypatch):
+    """A footnote row (first cell = long sentence) is not a HCPCS code."""
+    path = _dirty_fixture(tmp_path)
+    monkeypatch.setattr(nprm, "_discover_addendum_b", lambda: path)
+    con = duckdb.connect()
+    out = nprm.load_rvu_proposed(con)
+    assert out["rows"] == 2  # only 99213 and 99214 — footnote row excluded
+    assert con.execute("SELECT count(*) FROM pfs.rvu_proposed").fetchone()[0] == 2
+    footnote = con.execute(
+        "SELECT count(*) FROM pfs.rvu_proposed WHERE hcpcs LIKE '1 CPT%'"
+    ).fetchone()[0]
+    assert footnote == 0