arm-statements/arm_statements/tests/test_pipeline.py

from decimal import Decimal as D

from arm_statements import document_status, normalize, normalize_candidate, status_input
from arm_statements.issues import make_warning
from arm_statements.loader import Rules
from arm_statements.model import DocStatus, DocumentType, FormVariant, Severity
from arm_statements.tests.builders import excel, hdr, ocr

HEAD = ["Наименование показателя", "Код", "На 30.09.2025"]
FULL_CONFIDENCE = 100.0
FULL_HEADER = hdr(organization_name="ООО «Пример»", inn="7700000000", report_date="30.09.2025", units="тыс. руб.")


def bal_rows() -> list[list[str]]:
    return [
        HEAD,
        ["Запасы", "1210", "60"],
        ["Дебиторы", "1230", "40"],
        ["Итого II", "1200", "100"],
        ["НМА", "1110", "50"],
        ["Итого I", "1100", "50"],
        ["БАЛАНС", "1600", "150"],
        ["Капитал", "1310", "150"],
        ["Итого III", "1300", "150"],
        ["БАЛАНС", "1700", "150"],
    ]


def test_clean_ocr_balance_done(rules: Rules) -> None:
    cand = ocr(bal_rows(), header=FULL_HEADER, type_hint=DocumentType.BALANCE, title="Бухгалтерский баланс")
    result = normalize_candidate(cand, rules)
    doc = result.document
    assert doc is not None and result.file_warnings == ()
    assert doc.header.form_name == "Бухгалтерский баланс"
    assert document_status(status_input(doc)) is doc.status
    assert (doc.document_type, doc.form_variant, doc.status) == (DocumentType.BALANCE, FormVariant.FULL, DocStatus.DONE)
    assert doc.checks is not None and doc.checks.relations_failed == ()
    assert doc.header.inn.value == "7700000000"


def test_disclosure_hint_with_balance_codes(rules: Rules) -> None:
    doc = normalize_candidate(ocr(bal_rows(), header=FULL_HEADER, type_hint=DocumentType.DISCLOSURE), rules).document
    assert doc is not None and doc.document_type is DocumentType.BALANCE
    assert [(w.code, w.severity) for w in doc.warnings if w.code == "TYPE_CONFLICT"] == [
        ("TYPE_CONFLICT", Severity.ERROR)
    ]


def test_balance_hint_with_income_codes(rules: Rules) -> None:
    rows = [
        HEAD,
        ["Выручка", "2110", "100"],
        ["Себестоимость", "2120", "(60)"],
        ["Валовая", "2100", "40"],
        ["Коммерческие", "2210", "(10)"],
        ["Прибыль от продаж", "2200", "30"],
    ]
    doc = normalize_candidate(ocr(rows, header=FULL_HEADER, type_hint=DocumentType.BALANCE), rules).document
    assert doc is not None and doc.document_type is DocumentType.INCOME
    assert "TYPE_CONFLICT" in [w.code for w in doc.warnings]


def test_excel_other_sheet(rules: Rules) -> None:
    result = normalize_candidate(excel([("Примечания", [["Пояснительный текст без цифр"]])]), rules)
    assert result.document is None
    assert [(w.code, w.severity, w.locator) for w in result.file_warnings] == [
        ("SHEET_OTHER", Severity.INFO, "sheet:Примечания")
    ]


def test_ocr_lost_and_skipped_tables(rules: Rules) -> None:
    lost = normalize_candidate(ocr([["Итого", "", "15 000"]], locator="file:a.json;table:1"), rules)
    assert [(w.code, w.locator) for w in lost.file_warnings] == [("TABLE_LOST", "file:a.json;table:1")]
    skipped = normalize_candidate(ocr([["Руководитель", "Иванов"]]), rules)
    assert [w.code for w in skipped.file_warnings] == ["TABLE_SKIPPED"] and skipped.document is None


def test_hint_without_code_column(rules: Rules) -> None:
    doc = normalize_candidate(
        ocr([["Итого", "15 000"]], type_hint=DocumentType.BALANCE, header=FULL_HEADER), rules
    ).document
    assert doc is not None and (doc.lines, doc.columns, doc.checks) == ((), (), None)
    assert doc.form_variant is FormVariant.FULL
    assert doc.status is DocStatus.NEEDS_CHECK and "CODE_COLUMN_NOT_FOUND" in [w.code for w in doc.warnings]
    disc = normalize_candidate(ocr([["Текст"]], type_hint=DocumentType.DISCLOSURE), rules).document
    assert disc is not None and (disc.status, disc.form_variant, disc.lines) == (DocStatus.DONE, None, ())
    assert disc.header.form_name is None


def test_disclosure_by_codes(rules: Rules) -> None:
    rows = [HEAD, ["Пояснение", "5101", "1"], ["Пояснение", "5102", "2"], ["Пояснение", "5100", "3"]]
    doc = normalize_candidate(ocr(rows, header=FULL_HEADER), rules).document
    assert doc is not None and (doc.document_type, doc.lines, doc.status) == (
        DocumentType.DISCLOSURE,
        (),
        DocStatus.DONE,
    )


def test_simplified_balance(rules: Rules) -> None:
    rows = [
        HEAD,
        ["Матактивы", "1150", "50"],
        ["Запасы", "1210", "50"],
        ["БАЛАНС", "1600", "100"],
        ["Капитал", "1300", "100"],
        ["БАЛАНС", "1700", "100"],
    ]
    doc = normalize_candidate(ocr(rows, header=FULL_HEADER), rules).document
    assert doc is not None and doc.form_variant is FormVariant.SIMPLIFIED
    assert doc.checks is not None and doc.checks.relations_failed == ()


def test_income_without_inn_still_done(rules: Rules) -> None:
    rows: list[list[str | D]] = [
        ["Наименование", "Код", "За 9 месяцев 2025 г."],
        ["Выручка", "2110", D("100")],
        ["Себестоимость", "2120", D("-60")],
        ["Валовая прибыль", "2100", D("40")],
        ["Прибыль от продаж", "2200", D("40")],
        ["Прибыль до налогообложения", "2300", D("40")],
        ["Чистая прибыль", "2400", D("40")],
    ]
    head = hdr(organization_name="ООО «Пример»", units="тыс. руб.")
    doc = normalize_candidate(excel([("ОФР", rows)], header=head), rules).document
    assert doc is not None and doc.status is DocStatus.DONE
    assert [(w.code, w.severity) for w in doc.warnings] == [("INN_NOT_FOUND", Severity.INFO)]


def test_normalize_collects(rules: Rules) -> None:
    result = normalize([ocr(bal_rows(), header=FULL_HEADER), ocr([["Руководитель", "Иванов"]])], rules)
    assert len(result.documents) == 1 and [w.code for w in result.file_warnings] == ["TABLE_SKIPPED"]


# --- P1-6: у DISCLOSURE предупреждений шапки нет, статус DONE (требование ревью T15.1) ---


def test_disclosure_by_hint_empty_header_has_no_warnings(rules: Rules) -> None:
    doc = normalize_candidate(ocr([["Текст"]], type_hint=DocumentType.DISCLOSURE), rules).document
    assert doc is not None
    assert (doc.warnings, doc.status) == ((), DocStatus.DONE)


def test_disclosure_by_codes_empty_header_has_no_warnings(rules: Rules) -> None:
    rows = [HEAD, ["Пояснение", "5101", "1"], ["Пояснение", "5102", "2"], ["Пояснение", "5100", "3"]]
    doc = normalize_candidate(ocr(rows), rules).document
    assert doc is not None and doc.document_type is DocumentType.DISCLOSURE
    assert (doc.warnings, doc.status) == ((), DocStatus.DONE)


# --- ROW_SPLIT: предупреждение документа без locator (место — locator документа, спека «Шаг 8 — предупреждения») ---


def test_row_split_has_no_locator(rules: Rules) -> None:
    rows = bal_rows()
    rows[1] = ["Запасы и дебиторы", "1210\n1230", "60\n40"]
    doc = normalize_candidate(ocr(rows, header=FULL_HEADER, type_hint=DocumentType.BALANCE), rules).document
    assert doc is not None
    split = [w for w in doc.warnings if w.code == "ROW_SPLIT"]
    assert len(split) == 1 and split[0].locator is None and split[0].severity is Severity.INFO


# --- строка «в тысячах рублей» над таблицей и уверенность даты отчёта у Excel ---


def test_units_row_above_table_is_not_header_and_excel_date_is_confident(rules: Rules) -> None:
    rows = [["Бухгалтерский баланс"], ["в тысячах рублей"], *bal_rows()]
    cand = excel([("Баланс", rows)], header=FULL_HEADER, type_hint=DocumentType.BALANCE)
    doc = normalize_candidate(cand, rules).document
    assert doc is not None
    assert doc.header.report_date.confidence == FULL_CONFIDENCE and doc.header.report_date.value is not None
    assert [c.column_index for c in doc.columns] == [2]
    assert (doc.status, [w.code for w in doc.warnings]) == (DocStatus.DONE, [])


# --- порядок предупреждений документа: адаптера, раскладки, строк, единиц, TYPE_CONFLICT, шапки ---


def test_document_warnings_order(rules: Rules) -> None:
    adapter = make_warning("NUMBER_FORMAT_UNPARSED", "формат", rules)
    rows = [["Наименование", "Код", "Итого"], *bal_rows()[1:]]
    rows[1] = ["Запасы и дебиторы", "1210\n1230", "60,5\n40,5"]
    cand = excel([("Баланс", rows)], header=hdr(units="тыс. руб."), type_hint=DocumentType.INCOME, warnings=[adapter])
    doc = normalize_candidate(cand, rules).document
    assert doc is not None
    codes = [w.code for w in doc.warnings]
    expected = [
        "NUMBER_FORMAT_UNPARSED",
        "COLUMN_WITHOUT_DATE",
        "ROW_SPLIT",
        "FRACTIONAL_IN_THOUSANDS",
        "TYPE_CONFLICT",
        "NO_ORGANIZATION_NO_INN",
    ]
    assert [c for c in codes if c in expected] == [c for c in expected if c in codes]
    assert codes[0] == "NUMBER_FORMAT_UNPARSED" and codes.index("TYPE_CONFLICT") < codes.index("NO_ORGANIZATION_NO_INN")