feat(normalizer): шапка документа из data_ocr — ИНН, единицы, дата отчёта (§11.1 п. 8)

Co-Authored-By: Claude Haiku 4.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_012vJkmBGKKqMoEMzvsXG3y2

api/src/normalizer/ocr/header.py

@@ -0,0 +1,62 @@
+"""Шапка документа из полей `data_ocr` (§11.1 п. 8): ИНН — только цифры, единицы — очищенный текст, дата отчёта —
+`parse_date` после замены двойников; «распознано» — `confidence == 100` и сырой текст без мусора."""
+
+from collections.abc import Mapping
+from dataclasses import dataclass
+from datetime import date
+
+from src.normalizer.ocr.dates import parse_date
+from src.normalizer.ocr.text import clean_value, collapse_whitespace, digits_only, fix_lookalikes
+
+__all__ = ["FileHeader", "parse_file_header"]
+
+FULL_CONFIDENCE = 100.0
+CAPPED_CONFIDENCE = 99.0
+
+
+@dataclass(frozen=True)
+class FileHeader:
+    organization_name: str | None
+    organization_confidence: float | None
+    inn: str | None
+    inn_confidence: float | None
+    units: str | None
+    report_date_text: str | None
+    report_date: date | None
+    report_date_recognized: bool
+
+
+def _field(data_ocr: Mapping[str, object], name: str) -> tuple[str, float | None]:
+    raw = data_ocr.get(name)
+    if not isinstance(raw, Mapping):
+        return "", None
+    text = raw.get("form_name")
+    confidence = raw.get("confidence")
+    text_str = str(text) if isinstance(text, str) else ""
+    conf_float = float(confidence) if isinstance(confidence, (int, float)) else None
+    return (text_str, conf_float)
+
+
+def parse_file_header(data_ocr: Mapping[str, object], lookalikes: Mapping[str, str]) -> FileHeader:
+    organization, organization_confidence = _field(data_ocr, "organization")
+    inn_raw, inn_confidence = _field(data_ocr, "inn")
+    units_raw, _ = _field(data_ocr, "units")
+    date_raw, date_confidence = _field(data_ocr, "date")
+
+    organization_name = collapse_whitespace(organization) or None
+    inn = digits_only(inn_raw) or None
+    if inn is None:
+        inn_confidence = None
+    elif inn_confidence is not None and inn_raw.strip() != inn:
+        inn_confidence = min(inn_confidence, CAPPED_CONFIDENCE)
+    report_date = parse_date(fix_lookalikes(date_raw, lookalikes)) if date_raw.strip() else None
+    return FileHeader(
+        organization_name=organization_name,
+        organization_confidence=organization_confidence if organization_name is not None else None,
+        inn=inn,
+        inn_confidence=inn_confidence,
+        units=clean_value(units_raw) or None,
+        report_date_text=date_raw.strip() or None,
+        report_date=report_date,
+        report_date_recognized=date_confidence == FULL_CONFIDENCE and report_date is not None,
+    )

api/tests/unit/normalizer/test_header.py

@@ -0,0 +1,75 @@
+"""Шапка документа из `data_ocr` (§11.1 п. 8) на реальных `aep_3.json` (мусор) и `aep_6.json` (чисто)."""
+
+import json
+from datetime import date
+from pathlib import Path
+
+import pytest
+
+from src.normalizer.ocr.header import FileHeader, parse_file_header
+
+API_DIR = Path(__file__).resolve().parents[3]
+REAL_FIXTURES = API_DIR.parent.parent / "fixtures" / "ocr"
+LOOKALIKES = {"О": "0", "о": "0", "З": "3", "l": "1", "I": "1", "|": "1", "б": "6", "S": "5", "Ч": "4"}
+
+
+def _real(name: str) -> dict:
+    if not REAL_FIXTURES.is_dir():
+        pytest.skip("fixtures недоступен: положите репозиторий fixtures рядом с workspace")
+    return dict(json.loads((REAL_FIXTURES / name).read_text(encoding="utf-8"))["data_ocr"])
+
+
+def _field(text: str, confidence: float) -> dict:
+    return {"form_name": text, "confidence": confidence}
+
+
+def test_clean_header() -> None:
+    header = parse_file_header(
+        {
+            "organization": _field("ООО Ромашка", 97.5),
+            "inn": _field("7700000000", 100),
+            "units": _field("тыс. руб.", 100),
+            "date": _field("на 31 декабря 2024 г.", 100),
+        },
+        LOOKALIKES,
+    )
+    assert header == FileHeader(
+        "ООО Ромашка",
+        97.5,
+        "7700000000",
+        100.0,
+        "тыс. руб.",
+        "на 31 декабря 2024 г.",
+        date(2024, 12, 31),
+        True,
+    )
+
+
+def test_noisy_inn_caps_confidence() -> None:
+    header = parse_file_header({"inn": _field("3|1 21 2| о! О' 0 0; 3: 51", 100)}, LOOKALIKES)
+    assert header.inn == "3121200351"
+    assert header.inn_confidence == 99.0
+    assert header.organization_name is None and header.organization_confidence is None
+
+
+def test_missing_fields() -> None:
+    header = parse_file_header({}, LOOKALIKES)
+    assert header == FileHeader(None, None, None, None, None, None, None, False)
+
+
+@pytest.mark.fixtures
+def test_aep6_header() -> None:
+    header = parse_file_header(_real("aep_6.json"), LOOKALIKES)
+    assert header.inn is None  # текст «+»
+    assert header.units == "тыс. руб."
+    assert header.report_date is None  # в поле date — название организации
+    assert header.report_date_text is not None and header.report_date_recognized is False
+    assert header.organization_name is not None and "Дорстроймонтаж" in header.organization_name
+
+
+@pytest.mark.fixtures
+def test_aep3_header() -> None:
+    header = parse_file_header(_real("aep_3.json"), LOOKALIKES)
+    assert header.inn == "3121200351" and header.inn_confidence is not None and header.inn_confidence < 100
+    assert header.report_date == date(2024, 12, 31)
+    assert header.report_date_recognized is False  # confidence 75