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

base_dir = "resources/knowledge/original/원가계산/건설공사_표준품셈"

all_md_files = []
for root, dirs, files in os.walk(base_dir):
    for f in files:
        if f.endswith('.md') and not f.startswith('_') and '개정사항' not in f and '2026년_건설공사_표준품셈.md' not in f:
            all_md_files.append(os.path.join(root, f))

total_candidates = []

# Section pattern: digit-digit(-digit)? followed by Korean letters (section title)
# We want to detect when it is NOT preceded by newline/start of line
# e.g. "한다.1-2-2 단위표준" or "계산한다.1-2-3 토질"
sec_regex = re.compile(r'([^\n#\-\*\|\s])(\d+-\d+(?:-\d+)?\s+[가-힣\w\(\)·\,\s]+?(?:\(\'\d+[^)]*년\s*(?:보완|제정|신설)[^)]*\)|(?=\n|\([일|㎥|㎡|m|인|개|km|ton|hr|대|본|개소|소|개련|조|ha|회]))?)')

for fpath in sorted(all_md_files):
    rel_path = os.path.relpath(fpath, base_dir)
    with open(fpath, 'r', encoding='utf-8') as f:
        text = f.read()
        
    lines = text.split('\n')
    file_matches = []
    for line_idx, line in enumerate(lines):
        line_num = line_idx + 1
        # find patterns in line
        # check if line contains a section header that didn't start at beginning
        # skip markdown table rows (| ... |)
        if line.strip().startswith('|') and line.strip().endswith('|'):
            continue
            
        # Regex to find attached headers
        # Look for e.g. text followed immediately by \d+-\d+
        for m in re.finditer(r'([^\s#\-\*\>\|])(\d+-\d+(?:-\d+)?)\s+([가-힣][가-힣\s\w\(\)\/·]+)', line):
            prev_char = m.group(1)
            sec_num = m.group(2)
            sec_title = m.group(3).strip()
            
            # Filter out citations: e.g. '[공통부문] 5-3-1' or '‘5-3-1'
            if prev_char in ["'", '"', '‘', '’', '[', '(', '제', '표', '·', ':', '―', '-', '/']:
                continue
            if '참고' in sec_title[:10] or '따른다' in sec_title[:10]:
                continue
                
            file_matches.append({
                "file": rel_path,
                "line": line_num,
                "prev_char": prev_char,
                "sec_num": sec_num,
                "sec_title": sec_title[:30],
                "full_match": m.group(0)[:60],
                "line_snippet": line[:120]
            })
            
    if file_matches:
        print(f"{rel_path}: {len(file_matches)} candidates")
        total_candidates.extend(file_matches)

print(f"\nTotal attached section header candidates across all 45 files: {len(total_candidates)}")

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