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

md_files = [
    ("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"),
]

report = {}

for label, path in md_files:
    if not os.path.exists(path):
        continue
    with open(path, 'r', encoding='utf-8') as f:
        lines = f.readlines()
        
    print(f"\n=======================================================")
    print(f"Scanning {label} ({os.path.basename(path)}) - {len(lines)} lines")
    
    file_issues = {
        "attached_section_headers": [],
        "collapsed_table_rows": [],
        "attached_notes_tail": [],
        "page_transition_artifacts": []
    }
    
    for idx, line in enumerate(lines):
        line_num = idx + 1
        line_clean = line.strip()
        
        # 1. Attached section headers: e.g. text ending immediately with 1-2-2 or 3-4-1 without newline
        m_sec = re.findall(r'([^\s#\-\*\>\|])(\d+-\d+(?:-\d+)?)', line)
        for prev_char, sec_num in m_sec:
            # check if it looks like a real section header
            if prev_char not in ['제', '표', '·', '(', '/', '-', ',', ':']:
                file_issues["attached_section_headers"].append({
                    "line": line_num,
                    "prev_char": prev_char,
                    "sec_num": sec_num,
                    "snippet": line_clean[:120]
                })
                
        # 2. Collapsed table rows: table row with extreme length or multiple repeated patterns in single cell
        if line_clean.startswith('|') and line_clean.endswith('|'):
            cells = [c.strip() for c in line_clean.split('|')[1:-1]]
            for c in cells:
                # if cell has multiple words separated by space that look like collapsed table rows (e.g. 5 or more items)
                tokens = c.split()
                if len(tokens) >= 15:
                    file_issues["collapsed_table_rows"].append({
                        "line": line_num,
                        "token_count": len(tokens),
                        "snippet": c[:100] + "..."
                    })
                    break
                    
        # 3. Page transition artifacts: e.g. "→4공통부문" or "3제1장 적용기준"
        m_page = re.findall(r'(?:→\d+|\d+제\d+장|\d+공통부문|\d+토목부문|\d+건축부문|\d+유지관리부문)', line)
        if m_page:
            file_issues["page_transition_artifacts"].append({
                "line": line_num,
                "artifacts": m_page,
                "snippet": line_clean[:100]
            })
            
        # 4. Attached notes tail: [주] ending with next section immediately
        if '[주]' in line or '주]' in line:
            m_tail = re.findall(r'(\[주\].*?)(\d+-\d+)', line)
            if m_tail:
                file_issues["attached_notes_tail"].append({
                    "line": line_num,
                    "snippet": line_clean[:100]
                })
                
    print(f" - Attached section headers: {len(file_issues['attached_section_headers'])}")
    print(f" - Collapsed table rows: {len(file_issues['collapsed_table_rows'])}")
    print(f" - Attached notes tails: {len(file_issues['attached_notes_tail'])}")
    print(f" - Page transition artifacts: {len(file_issues['page_transition_artifacts'])}")
    
    report[label] = file_issues

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