arm-tabular-parser/api/src/parser/statements/header.py

"""Шапка формы по меткам словаря (design D-5): значение — правее метки в той же строке, иначе в следующей строке.

Особые правила: ИНН ниже метки «аудитор» не берётся; организация — после «Организация», а при пустом значении —
ближайшая строка выше с признаком ОПФ (НПЗ); дата — из подзаголовка формы («на <дата>», «за <период>») либо из тройки
«число | месяц | год»; единицы — текст как в файле. Найденные поля получают `*_confidence = 100`.
"""

import datetime
import re
from collections.abc import Iterator

from src.contract.result import StatementHeader
from src.parser.grid import Sheet
from src.parser.numbers import inn_digits
from src.parser.rules import Rules, StatementLabel
from src.parser.statements.columns import cell_label
from src.parser.statements.detect import FormBlock

__all__ = ["parse_header"]

_FULL_CONFIDENCE = 100
_DIGITS = re.compile(r"\d+")
NO_ORGANIZATION_WARNING = "в шапке не найдена организация"
NO_UNITS_WARNING = "в шапке не найдена единица измерения"


def _cells(sheet: Sheet, first_row: int, last_row: int) -> Iterator[tuple[int, int, str]]:
    for row in range(first_row, last_row + 1):
        for col, _ in sheet.row_cells(row):
            yield row, col, cell_label(sheet, row, col)


def _right(sheet: Sheet, row: int, col: int) -> list[tuple[int, str]]:
    return [(c, cell_label(sheet, row, c)) for c, _ in sheet.row_cells(row) if c > col]


def _value_right_or_below(sheet: Sheet, row: int, col: int, last_row: int) -> str | None:
    right = _right(sheet, row, col)
    if right:
        return right[0][1]
    if row + 1 <= last_row:
        below = [(c, cell_label(sheet, row + 1, c)) for c, _ in sheet.row_cells(row + 1)]
        if below:
            return below[0][1]
    return None


def _label_rows(
    sheet: Sheet, rules: Rules, key: StatementLabel, first_row: int, last_row: int
) -> list[tuple[int, int, str]]:
    return [
        (row, col, text)
        for row, col, text in _cells(sheet, first_row, last_row)
        if rules.statement_headers.label(key, text)
    ]


def _auditor_from(sheet: Sheet, rules: Rules, first_row: int, last_row: int) -> int:
    """Первая строка блока аудитора; ниже неё ИНН/названия — не организации."""
    rows = [
        row for row, _, text in _cells(sheet, first_row, last_row) if rules.statement_headers.is_auditor_label(text)
    ]
    return min(rows) if rows else last_row + 1


def _organization(sheet: Sheet, rules: Rules, first_row: int, last_row: int, auditor_from: int) -> str | None:
    for row, col, _ in _label_rows(sheet, rules, "organization", first_row, min(last_row, auditor_from - 1)):
        value = _value_right_or_below(sheet, row, col, last_row)
        if value and not rules.statement_headers.label("okpo", value) and not value.isdigit():
            return value
        for above in range(row - 1, first_row - 1, -1):
            texts = [text for _, _, text in _cells(sheet, above, above)]
            if any(rules.statement_headers.has_legal_form(text) for text in texts):
                return next(text for text in texts if rules.statement_headers.has_legal_form(text))
    return None


def _inn(sheet: Sheet, rules: Rules, first_row: int, last_row: int, auditor_from: int) -> str | None:
    for row, col, _ in _label_rows(sheet, rules, "inn", first_row, min(last_row, auditor_from - 1)):
        for c, _ in _right(sheet, row, col):
            digits = inn_digits(sheet.cell(row, c))
            if digits is not None:
                return digits
    return None


def _labelled(sheet: Sheet, rules: Rules, key: StatementLabel, first_row: int, last_row: int) -> str | None:
    for row, col, _ in _label_rows(sheet, rules, key, first_row, last_row):
        value = _value_right_or_below(sheet, row, col, last_row)
        if value:
            return value
    return None


def _okopf_okfs(sheet: Sheet, rules: Rules, first_row: int, last_row: int) -> tuple[str | None, str | None]:
    """«по ОКОПФ / ОКФС» с двумя числами правее (в той же или предыдущей строке — у 1С коды стоят строкой выше)."""
    for row, col, _ in _label_rows(sheet, rules, "okopf_okfs", first_row, last_row):
        for candidate in (row, row - 1):
            numbers = [text for _, text in _right(sheet, candidate, col) if _DIGITS.fullmatch(text)]
            if len(numbers) >= 2:  # noqa: PLR2004 - пара кодов
                return numbers[0], numbers[1]
            if len(numbers) == 1:
                return numbers[0], None
    return _labelled(sheet, rules, "okopf", first_row, last_row), _labelled(sheet, rules, "okfs", first_row, last_row)


def _units(sheet: Sheet, rules: Rules, first_row: int, last_row: int) -> str | None:
    for row, col, _ in _label_rows(sheet, rules, "units", first_row, last_row):
        for _, text in _right(sheet, row, col):
            if not rules.statement_headers.label("okei", text) and not text.isdigit():
                return text
    return None


def _report_date(
    sheet: Sheet, rules: Rules, block: FormBlock, first_row: int, last_row: int
) -> tuple[str | None, datetime.date | None]:
    periods = rules.statement_periods
    if block.title_row is not None and block.title_row + 1 <= last_row:
        # Подзаголовок — только следующая строка за заголовком формы
        # («на 31 декабря 2024 г.», «за 9 месяцев 2025 г.»);
        # глубже лежат «Форма по ОКУД» и тройка даты, где голый «2025» принялся бы за период
        for _, _, text in _cells(sheet, block.title_row + 1, block.title_row + 1):
            end = periods.period_end(text)
            if end is not None:
                return text, end
    for row, col, _ in _label_rows(sheet, rules, "date", first_row, last_row):
        parts = [text for _, text in _right(sheet, row, col)]
        if len(parts) >= 3:  # noqa: PLR2004 - число, месяц, год
            found = periods.date_triple(parts[0], parts[1], parts[2])
            if found is not None:
                return None, found
    return None, None


def parse_header(
    sheet: Sheet, rules: Rules, first_row: int, last_row: int, block: FormBlock | None
) -> tuple[StatementHeader, list[str]]:
    if block is None:
        marker = next(
            (
                text
                for _, _, text in _cells(sheet, first_row, last_row)
                if rules.statement_headers.is_continuation(text)
            ),
            None,
        )
        return StatementHeader(form_name=marker), []
    auditor_from = _auditor_from(sheet, rules, first_row, last_row)
    organization = _organization(sheet, rules, first_row, last_row, auditor_from)
    inn = _inn(sheet, rules, first_row, last_row, auditor_from)
    okopf, okfs = _okopf_okfs(sheet, rules, first_row, last_row)
    units = _units(sheet, rules, first_row, last_row)
    date_text, report_date = _report_date(sheet, rules, block, first_row, last_row)
    warnings: list[str] = []
    if organization is None:
        warnings.append(NO_ORGANIZATION_WARNING)
    if units is None:
        warnings.append(NO_UNITS_WARNING)
    header = StatementHeader(
        form_name=block.form_name,
        organization_name=organization,
        organization_name_confidence=_FULL_CONFIDENCE if organization else None,
        inn=inn,
        inn_confidence=_FULL_CONFIDENCE if inn else None,
        report_date_text=date_text,
        report_date=report_date,
        units=units,
        okud=block.okud,
        okpo=_labelled(sheet, rules, "okpo", first_row, last_row),
        okopf=okopf,
        okfs=okfs,
        address=_labelled(sheet, rules, "address", first_row, min(last_row, auditor_from - 1)),
    )
    return header, warnings