Aislo/resources/master_data/scripts/check_master_body.py

# -*- coding: utf-8 -*-
"""마스터 데이터 검사 (2) 본문 — `check_master.py` 가 부름.

표형 요소의 출처 절을 본문 md 에서 찾아 수를 맞댐 · 결손(본문 표에 있고 요소에 없음)·허구(요소에 있고 본문 절에 없음).
"""

from __future__ import annotations

import re
import sys
from decimal import Decimal
from pathlib import Path

HERE = Path(__file__).resolve().parent
ROOT = HERE.parents[2]
COST = ROOT / "resources/knowledge/original/원가계산"
sys.path.insert(0, str(HERE))
sys.path.insert(0, str(ROOT / "resources/knowledge/original/_pipeline"))

import master_formula as mf  # noqa: E402
from pum_md_tool import _NUM, md_lines  # noqa: E402

DIVISIONS = {
    "공통": "01_공통부문",
    "토목": "02_토목부문",
    "건축": "03_건축부문",
    "기계설비": "04_기계설비부문",
    "유지관리": "05_유지관리부문",
}
_IDENT = re.compile(r"\d+-\d+(?:-\d+)*")


def _key(parts: str) -> tuple[int, ...]:
    return tuple(int(p) for p in parts.split("-"))


def section_of(source: str) -> tuple[Path | None, str, str]:
    """(본문 md, 절 번호, 절 글) — 못 찾으면 md 나 글이 빔."""
    ident = _IDENT.search(source)
    if not ident:
        return None, "", ""
    ident = ident.group()
    parts = _key(ident)
    if source.startswith("산림품셈"):
        chapters = COST / "산림_표준품셈/본문"
    elif source.startswith("건설품셈"):
        chapters = COST / "건설공사_표준품셈/본문" / DIVISIONS.get(source.split()[1], "")
    else:
        return None, ident, ""
    best = None
    for md in chapters.glob(f"제{parts[0]:02d}장_*/*.md"):
        head = md.name.split("_", 1)[0]
        if not _IDENT.fullmatch(head):
            continue
        own = _key(head)
        if (
            own[0]
            and own == parts[: len(own)]
            and (best is None or len(own) > len(_key(best.name.split("_", 1)[0])))
        ):
            best = md
    if best is None:
        return None, ident, ""
    lines = best.read_text(encoding="utf-8").split("\n")
    heading = re.compile(rf"^(#+)\s*{re.escape(ident)}\.?(\s|$)")
    for at, line in enumerate(lines):
        m = heading.match(line)
        if m:
            level, end = len(m.group(1)), len(lines)
            for later in range(at + 1, len(lines)):
                h = re.match(r"^(#+)\s", lines[later])
                if h and len(h.group(1)) <= level:
                    end = later
                    break
            return best, ident, "\n".join(lines[at:end])
    return best, ident, ""


def _canon(token: str) -> str:
    return format(Decimal(token.replace(",", "")).normalize(), "f")


def numbers_in(text: str) -> set[str]:
    # 칸 안에서 줄바뀐 천 단위 수(「5,1<br>50」)는 이어 붙임
    text = re.sub(r"(?<=\d,\d)<br>(?=\d{2})", "", text)
    return {_canon(t) for _, line in md_lines(text) for t in _NUM.findall(line)}


def _flat(value) -> set[str]:
    if isinstance(value, Decimal):
        return {_canon(str(abs(value)))}
    if isinstance(value, str):
        return {_canon(t) for t in _NUM.findall(value)}
    if isinstance(value, list):
        return set().union(*map(_flat, value)) if value else set()
    if isinstance(value, dict):
        return set().union(*map(_flat, value.values())) if value else set()
    return set()


def check_body(files: dict[str, dict]) -> list[str]:
    """표마다 허구 · 절마다 결손."""
    out, by_section = [], {}
    for name, data in files.items():
        for table in (data.get("표") or []) if data.get("그룹") in mf.TABLE_GROUPS else []:
            where = f"{name} · {table.get('키')} {table.get('원문번호')}"
            source = str(table.get("출처", ""))
            if not source.startswith(("산림품셈", "건설품셈")):
                continue  # 품셈 본문이 아닌 출처(질의회신 등) — 절 대조 대상 아님
            md, ident, text = section_of(source)
            if not text:
                out.append(f"{where} · 본문 절 못 찾음 「{table.get('출처')}」")
                continue
            mine = (
                _flat(table.get("줄"))
                | _flat(table.get("기준"))
                | _flat(table.get("주"))
                | _flat(table.get("값칸"))
            )
            fake = mine - numbers_in(text)
            if fake:
                out.append(f"{where} · 허구 {len(fake)} {sorted(fake, key=Decimal)}")
            slot = by_section.setdefault((md, ident), [text, set(), []])
            slot[1] |= mine
            slot[2].append(where)
    _elements_into(files, by_section)
    for (md, ident), (text, mine, tables) in by_section.items():
        for (md2, ident2), (_, mine2, tables2) in by_section.items():
            if md2 == md and ident2.startswith(ident + "-"):
                mine = mine | mine2
                tables = tables + tables2
        table_text = "\n".join(line for line in text.split("\n") if line.lstrip().startswith("|"))
        lack = numbers_in(table_text) - mine
        if lack:
            out.append(
                f"{md.name} {ident} · 결손 {len(lack)} {sorted(lack, key=Decimal)} ← {' · '.join(tables)}"
            )
    return out


def _elements_into(files: dict[str, dict], by_section: dict) -> None:
    """요소 줄(기계 운전경비처럼 표 대신 요소로 올린 절)의 수도 그 절 결손 대조에 넣음."""
    seen: dict[str, tuple] = {}
    for data in files.values():
        if data.get("그룹") in (*mf.TABLE_GROUPS, "로직"):
            continue
        for row in data.get("줄") or []:
            source = row.get("출처")
            for text in source.values() if isinstance(source, dict) else [source]:
                if not isinstance(text, str) or not text.startswith(("건설품셈", "산림품셈")):
                    continue
                if text not in seen:
                    seen[text] = section_of(text)[:2]
                slot = by_section.get(seen[text])
                if slot:
                    slot[1] |= _flat(row.get("값")) | _flat(
                        [row.get(k) for k in ("이름", "규격", "원문번호")]
                    )