Fix every real lint finding and drop degenerate splice fragments

ingestion/tests/test_segment_atc.py

@@ -0,0 +1,144 @@
+from ingestion.segment.atc import extract_atc_codes, is_stated_absent, normalize_atc_candidate
+
+
+def test_stray_whitespace_split_j04a_c01_recovered():
+    assert normalize_atc_candidate("J04A C01") == "J04AC01"
+
+
+def test_stray_whitespace_split_n05b_a06_recovered():
+    assert normalize_atc_candidate("N05B A06") == "N05BA06"
+
+
+def test_stray_whitespace_split_l01x_x02_recovered():
+    assert normalize_atc_candidate("L01X X02") == "L01XX02"
+
+
+def test_digit_letter_confusion_no3ax12_recovered():
+    assert normalize_atc_candidate("NO3AX12") == "N03AX12"
+
+
+def test_digit_letter_confusion_jo1dc07_recovered():
+    assert normalize_atc_candidate("JO1DC07") == "J01DC07"
+
+
+def test_clean_code_passes_through():
+    assert normalize_atc_candidate("N03AX12") == "N03AX12"
+
+
+def test_garbage_not_recovered():
+    assert normalize_atc_candidate("NOT AN ATC CODE") is None
+    assert normalize_atc_candidate("") is None
+
+
+def test_stated_absent_chua_co():
+    assert is_stated_absent("Mã ATC: Chưa có.") is True
+
+
+def test_stated_absent_khong_co():
+    assert is_stated_absent("Không có.") is True
+
+
+def test_stated_present_not_flagged_absent():
+    assert is_stated_absent("N03AX12") is False
+
+
+def test_extract_single_code():
+    result = extract_atc_codes("N03AX12")
+    assert result.codes == ["N03AX12"]
+    assert result.stated_absent is False
+
+
+def test_extract_multi_code_insulin_style():
+    result = extract_atc_codes("A10AB01, A10AC01, A10AD01")
+    assert result.codes == ["A10AB01", "A10AC01", "A10AD01"]
+
+
+def test_extract_multi_code_with_noise_mixed_in():
+    # one clean code, one noisy code recovered, matching the real corpus
+    # pattern where a monograph has some clean and some noisy ATC entries
+    result = extract_atc_codes("N03AX12, J04A C01")
+    assert result.codes == ["N03AX12", "J04AC01"]
+
+
+def test_extract_stated_absent_returns_no_codes():
+    result = extract_atc_codes("Mã ATC: Chưa có.")
+    assert result.codes == []
+    assert result.stated_absent is True
+
+
+def test_trailing_period_recovered_confirmed_real_abacavir_case():
+    # real field text is "J05AF06." — a sentence-ending period, not part of
+    # the code; an earlier version silently produced zero codes here.
+    assert normalize_atc_candidate("J05AF06.") == "J05AF06"
+    result = extract_atc_codes("J05AF06.")
+    assert result.codes == ["J05AF06"]
+
+
+def test_species_annotation_stripped_confirmed_real_insulin_case():
+    # annotation-stripping is extract_atc_codes's job (must run before the
+    # comma/semicolon split, see below) — normalize_atc_candidate itself
+    # only normalizes an already-isolated code token.
+    result = extract_atc_codes("A10AB01 (người); A10AB02 (bò)")
+    assert result.codes == ["A10AB01", "A10AB02"]
+
+
+def test_leading_colon_from_value_span_stripped_confirmed_real_alcuronium_case():
+    # real field text for ALCURONIUM CLORID (physical page 152): the bold
+    # label span is "Mã ATC" with no colon, and the plain value span is
+    # ": M03AA01." — the colon belongs to the value side here, not the
+    # label side (Abacavir's equivalent has it on the label side instead:
+    # "Mã ATC: " + "J05AF06."). See atc.py module docstring, defect 5.
+    assert normalize_atc_candidate(": M03AA01.") == "M03AA01"
+    result = extract_atc_codes(": M03AA01.")
+    assert result.codes == ["M03AA01"]
+
+
+def test_name_prefixed_code_stripped_confirmed_real_arginin_case():
+    # real field text for ARGININ (physical page 204): two salt forms, each
+    # its own "Name: CODE" line, not a bare code — see atc.py module
+    # docstring, defect 6.
+    assert normalize_atc_candidate("Arginin glutamat: A05BA01") == "A05BA01"
+    result = extract_atc_codes("Arginin glutamat: A05BA01\nArginin hydroclorid: B05XB01")
+    assert result.codes == ["A05BA01", "B05XB01"]
+
+
+def test_plain_code_with_no_colon_still_normalizes():
+    assert normalize_atc_candidate("N03AX12") == "N03AX12"
+
+
+def test_reversed_code_first_shape_confirmed_real_hmg_coa_case():
+    # real field text for CÁC CHẤT ỨC CHẾ HMG-CoA REDUCTASE (physical page
+    # 284): each statin is "CODE: Name", the opposite order from the
+    # "Name: CODE" shape above — see atc.py module docstring, defect 7.
+    # "C10A A01" also has the already-fixed stray-whitespace split.
+    assert normalize_atc_candidate("C10A A01: Simvastatin") == "C10AA01"
+    result = extract_atc_codes("C10A A01: Simvastatin\nC10A A02: Lovastatin")
+    assert result.codes == ["C10AA01", "C10AA02"]
+
+
+def test_annotation_containing_a_comma_does_not_break_the_split_confirmed_vaccine_case():
+    # real field text for VẮC XIN SỞI (physical page 1437): the English
+    # annotation "(Measles, live attenuated)" contains its own comma. An
+    # earlier version split on "," *before* stripping the annotation,
+    # breaking "J07BD01 (Measles, live attenuated)." into two unrecoverable
+    # fragments and silently returning zero codes — see atc.py module
+    # docstring, defect 4.
+    result = extract_atc_codes("J07BD01 (Measles, live attenuated).")
+    assert result.codes == ["J07BD01"]
+
+
+def test_extract_all_20_insulin_codes_from_real_field_text():
+    # exact real field text for INSULIN (physical page 809) — see atc.py
+    # module docstring; confirms the fix recovers all 20, not just 2.
+    field_text = (
+        "A10AB01 (người); A10AB02 (bò); A10AB03 (lợn);\n"
+        "A10AB04 (lispro); A10AB05 (aspart); A10AB06 (glulisin);\n"
+        "A10AC01 (người); A10AC02 (bò); A10AC03 (lợn); A10AC04\n"
+        "(lispro); A10AD01 (người), A10AD02 (bò), A10AD03 (lợn),\n"
+        "A10AD04 (lispro), A10AE01 (người); A10AE02 (bò); A10AE03\n"
+        "(lợn); A10AE04 (glargin); A10AE05 (detemir), A10AF01 (người)."
+    )
+    result = extract_atc_codes(field_text)
+    assert len(result.codes) == 20
+    assert "A10AB01" in result.codes
+    assert "A10AF01" in result.codes