import sys, os, re, json
import fitz # PyMuPDF
sys.stdout.reconfigure(encoding='utf-8')

# Load target items
with open('scratch/located_const_sections.json', 'r', encoding='utf-8') as f:
    located_items = json.load(f)

md_map = {
    "공통_제1장": "resources/knowledge/original/원가계산/건설공사_표준품셈/01_공통부문/제1장_적용기준.md",
    "공통_제3장": "resources/knowledge/original/원가계산/건설공사_표준품셈/01_공통부문/제3장_토공사.md",
    "공통_제4장": "resources/knowledge/original/원가계산/건설공사_표준품셈/01_공통부문/제4장_조경공사.md",
    "공통_제6장": "resources/knowledge/original/원가계산/건설공사_표준품셈/01_공통부문/제6장_철근콘크리트공사.md",
    "토목_제1장": "resources/knowledge/original/원가계산/건설공사_표준품셈/02_토목부문/제1장_도로포장공사.md",
    "토목_제6장": "resources/knowledge/original/원가계산/건설공사_표준품셈/02_토목부문/제6장_관부설및접합공사.md",
    "건축_제9장": "resources/knowledge/original/원가계산/건설공사_표준품셈/03_건축부문/제9장_미장공사.md",
    "유지관리_제1장": "resources/knowledge/original/원가계산/건설공사_표준품셈/05_유지관리부문/제1장_공통.md",
    "유지관리_제3장": "resources/knowledge/original/원가계산/건설공사_표준품셈/05_유지관리부문/제3장_건축.md",
}

md_texts = {}
for k, p in md_map.items():
    if os.path.exists(p):
        with open(p, 'r', encoding='utf-8') as f:
            md_texts[k] = f.read()

pdf_path = "resources/knowledge/original/원가계산/건설공사_표준품셈/2026년_건설공사_표준품셈.pdf"
doc = fitz.open(pdf_path)

audit_details = []

for item in located_items:
    k = item['key']
    num = item['number']
    name = item['name']
    pn = item['path_name']
    pdf_pages = item.get('pdf_pages', [])
    
    # Exclude chapter 8
    if "공통부문 › 건설기계" in pn:
        continue

    # Determine chapter
    md_key = None
    if "공통부문 › 적용기준" in pn: md_key = "공통_제1장"
    elif "공통부문 › 토공사" in pn: md_key = "공통_제3장"
    elif "공통부문 › 조경공사" in pn: md_key = "공통_제4장"
    elif "공통부문 › 철근콘크리트공사" in pn: md_key = "공통_제6장"
    elif "토목부문 › 도로포장공사" in pn: md_key = "토목_제1장"
    elif "토목부문 › 관부설" in pn: md_key = "토목_제6장"
    elif "건축부문 › 미장공사" in pn: md_key = "건축_제9장"
    elif "유지관리부문 › 공 통" in pn: md_key = "유지관리_제1장"
    elif "유지관리부문 › 건 축" in pn: md_key = "유지관리_제3장"

    raw_md = md_texts.get(md_key, "")
    
    # Find how this section starts in raw_md
    # Look for num in raw_md
    escaped_num = re.escape(num)
    # Search with regex
    m = re.search(rf'([^\n]{{0,50}})({escaped_num}\s+[^\n]+)', raw_md)
    header_status = "정상 분리"
    preceding_context = ""
    if m:
        preceding = m.group(1)
        header_line = m.group(2)
        if preceding.strip():
            header_status = "⚠ 앞 문장과 경계 붙음 (줄바꿈 누락)"
            preceding_context = preceding.strip()[-30:]
    else:
        # Check if number appears at all
        if num in raw_md:
            header_status = "본문 중 인라인 출현"
        else:
            header_status = "❌ MD 본문 완전 부재"

    # Find PDF text for relevant pages
    actual_pages = [p for p in pdf_pages if p > 50]
    pdf_text = ""
    for p in actual_pages:
        pdf_text += doc[p - 1].get_text()

    # Check table collapse in this section's neighborhood
    # Grab 50 lines around the match
    section_flaws = []
    if m:
        start_pos = m.start()
        # look forward 3000 characters
        chunk = raw_md[start_pos:start_pos + 4000]
        # check collapsed table rows in chunk
        for line in chunk.split('\n'):
            if line.strip().startswith('|') and line.strip().endswith('|'):
                tokens = line.split()
                if len(tokens) >= 15:
                    section_flaws.append(f"표 한 줄 뭉침 ({len(tokens)}개 단어): {line.strip()[:60]}...")
            if '→' in line or '제1장' in line or '공통부문' in line:
                if re.search(r'→\d+|\d+제\d+장', line):
                    section_flaws.append(f"페이지 전이 찌꺼기 삽입: {line.strip()[:60]}")

    audit_entry = {
        "key": k,
        "section": num,
        "name": name,
        "chapter": md_key,
        "pdf_pages": actual_pages,
        "header_separation": header_status,
        "preceding_context": preceding_context,
        "flaws_detected": section_flaws
    }
    audit_details.append(audit_entry)
    print(f"[{k}] {num} {name}: {header_status} | flaws={len(section_flaws)}")

with open('scratch/audit_42_detailed_breakdown.json', 'w', encoding='utf-8') as f:
    json.dump(audit_details, f, ensure_ascii=False, indent=2)