normalization.py

from __future__ import annotations

import hashlib
import re
from collections import defaultdict
from decimal import Decimal, InvalidOperation
from typing import Any

from .schema import Category, Characteristic, ReviewStatus

TECHNICAL_LINE = re.compile(
    r"(?:[⌀Ø∅⌓⌖⏥⌯∥⊥±°]|\bM\s*\d|\bR\s*\d|\bRa\s*\d|"
    r"\d+(?:\.\d+)?\s*[±+\-/]|H\d\b|6H\b|THRU|TYP|MAX|MIN)",
    re.IGNORECASE,
)


def enrich_characteristic(item: Characteristic) -> Characteristic:
    text = " ".join(part for part in (item.requirement, item.source_text) if part).strip()
    if not item.requirement:
        item.requirement = item.source_text
    if item.category == Category.OTHER:
        item.category = infer_category(text)
    if item.quantity is None:
        quantity = re.search(r"(?<!\d)(\d{1,3})\s*[Xx](?!\w)", text)
        if quantity:
            item.quantity = int(quantity.group(1))
    if not item.thread:
        thread = re.search(
            r"\bM\s*\d+(?:\.\d+)?\s*[×xX]\s*\d+(?:\.\d+)?(?:\s*[-–]\s*[0-9A-Za-z]+)?",
            text,
        )
        if thread:
            item.thread = re.sub(r"\s+", "", thread.group(0)).replace("x", "×").replace("X", "×")
    if not item.surface_finish:
        finish = re.search(r"\bR[azq]\s*\d+(?:\.\d+)?(?:\s*µ?m)?", text, re.I)
        if finish:
            item.surface_finish = finish.group(0)
    if not item.nominal:
        nominal = _extract_nominal(text)
        if nominal is not None:
            item.nominal = nominal
    _extract_tolerances(item, text)
    _calculate_limits(item)
    if not item.record_id:
        seed = f"{item.page}|{item.balloon_id}|{item.requirement}|{item.source_text}"
        item.record_id = hashlib.sha1(seed.encode("utf-8")).hexdigest()[:12]
    return item


def infer_category(text: str) -> Category:
    normalized = text.strip()
    if re.search(r"\bM\s*\d+(?:\.\d+)?\s*[×xX]", normalized):
        return Category.THREAD
    if re.search(r"\bR[azq]\s*\d", normalized, re.I):
        return Category.SURFACE_FINISH
    if any(symbol in normalized for symbol in ("⌖", "⏥", "⌯", "∥", "⊥", "◎", "○")):
        return Category.GD_AND_T
    if re.search(r"[⌀Ø∅]\s*\d", normalized):
        return Category.DIAMETER
    if re.search(r"(?:^|\s)R\s*\d", normalized, re.I):
        return Category.RADIUS
    if "°" in normalized:
        return Category.ANGLE
    if re.search(r"\d", normalized) and any(
        token in normalized for token in ("±", "+", "-", "/", "MAX", "MIN")
    ):
        return Category.LINEAR_DIMENSION
    if re.search(r"MATERIAL|GRADE|HARDNESS|HBW|HRC", normalized, re.I):
        return Category.MATERIAL
    if normalized:
        return Category.NOTE
    return Category.OTHER


def records_from_ocr_text(text: str, page: int = 1) -> list[Characteristic]:
    """Conservative fallback for OCR output; unknown associations remain unassigned."""
    records: list[Characteristic] = []
    counter = 1
    for raw_line in text.splitlines():
        line = re.sub(r"\s+", " ", raw_line).strip(" |\t")
        if len(line) < 2 or not TECHNICAL_LINE.search(line):
            continue
        balloon = ""
        requirement = line
        matched = re.match(
            r"^(?:BALLOON\s*)?[#(]?\s*(\d{1,4})\s*[)\]:.\-]\s*(.+)$",
            line,
            re.I,
        )
        if matched:
            balloon, requirement = matched.group(1), matched.group(2)
        else:
            balloon = f"UNASSIGNED-{counter:03d}"
            counter += 1
        record = Characteristic(
            balloon_id=balloon,
            page=page,
            requirement=requirement,
            source_text=line,
            confidence=0.35 if balloon.startswith("UNASSIGNED") else 0.5,
            status=(
                ReviewStatus.NEEDS_MAPPING
                if balloon.startswith("UNASSIGNED")
                else ReviewStatus.NEEDS_REVIEW
            ),
        )
        records.append(enrich_characteristic(record))
    return records


def validate_characteristics(records: list[Characteristic]) -> list[Characteristic]:
    groups: dict[tuple[int, str], list[Characteristic]] = defaultdict(list)
    for record in records:
        enrich_characteristic(record)
        record.validation_issues = list(dict.fromkeys(record.validation_issues))
        if not record.balloon_id:
            record.validation_issues.append("Balloon identifier was not read.")
            record.status = ReviewStatus.NEEDS_MAPPING
        if record.balloon_id.upper().startswith("UNASSIGNED"):
            record.validation_issues.append("Requirement is not mapped to a balloon.")
            record.status = ReviewStatus.NEEDS_MAPPING
        if not record.requirement:
            record.validation_issues.append("Requirement text is empty or unreadable.")
        if record.category in {
            Category.LINEAR_DIMENSION,
            Category.DIAMETER,
            Category.RADIUS,
            Category.ANGLE,
        } and not record.nominal:
            record.validation_issues.append("Nominal value requires confirmation.")
        if not record.source_text:
            record.validation_issues.append("No verbatim source text was preserved.")
        # AI results never become approved merely because confidence is high.
        if record.status == ReviewStatus.VERIFIED:
            record.status = ReviewStatus.NEEDS_REVIEW
        groups[(record.page, record.balloon_id)].append(record)

    for grouped in groups.values():
        unique_requirements = {
            _key(record.requirement) for record in grouped if record.requirement.strip()
        }
        if len(unique_requirements) > 1:
            for record in grouped:
                record.validation_issues.append(
                    "Conflicting requirements were extracted for the same balloon."
                )
    return records


def _extract_nominal(text: str) -> str | None:
    patterns = [
        r"[⌀Ø∅]\s*([+-]?\d+(?:\.\d+)?)",
        r"(?:^|\s)R\s*([+-]?\d+(?:\.\d+)?)",
        r"(?<![A-Za-z\d])([+-]?\d+(?:\.\d+)?)(?=\s*(?:±|\+|\-|°|$))",
    ]
    for pattern in patterns:
        match = re.search(pattern, text, re.I)
        if match:
            return match.group(1)
    return None


def _extract_tolerances(item: Characteristic, text: str) -> None:
    if item.upper_tolerance or item.lower_tolerance:
        return
    bilateral = re.search(r"±\s*(\d+(?:\.\d+)?)", text)
    if bilateral:
        item.upper_tolerance = bilateral.group(1)
        item.lower_tolerance = f"-{bilateral.group(1)}"
        return
    unilateral = re.search(
        r"\+\s*(\d+(?:\.\d+)?)\s*(?:/|\s)\s*-\s*(\d+(?:\.\d+)?)",
        text,
    )
    if unilateral:
        item.upper_tolerance = unilateral.group(1)
        item.lower_tolerance = f"-{unilateral.group(2)}"


def _calculate_limits(item: Characteristic) -> None:
    if item.upper_limit or item.lower_limit:
        return
    nominal = _decimal(item.nominal)
    upper = _decimal(item.upper_tolerance)
    lower = _decimal(item.lower_tolerance)
    if nominal is None or upper is None or lower is None:
        return
    item.upper_limit = _decimal_text(nominal + upper)
    item.lower_limit = _decimal_text(nominal + lower)


def _decimal(value: Any) -> Decimal | None:
    if value in (None, ""):
        return None
    text = str(value).strip().replace("+", "")
    if not re.fullmatch(r"-?\d+(?:\.\d+)?", text):
        return None
    try:
        return Decimal(text)
    except InvalidOperation:
        return None


def _decimal_text(value: Decimal) -> str:
    return format(value.normalize(), "f")


def _key(value: str) -> str:
    return re.sub(r"\W+", "", value.casefold())