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

base_dir = "resources/knowledge/original/원가계산/건설공사_표준품셈"
merged_file = os.path.join(base_dir, "2026년_건설공사_표준품셈.md")

with open(merged_file, 'r', encoding='utf-8') as f:
    merged_text = f.read()

all_split_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_split_files.append(os.path.join(root, f))

# Function to extract all tables with their unique text fingerprints
def extract_table_blocks(text, label):
    lines = text.split('\n')
    tables = []
    curr_lines = []
    start_line = 0
    
    for idx, line in enumerate(lines):
        line_num = idx + 1
        l_strip = line.strip()
        if l_strip.startswith('|') and l_strip.endswith('|'):
            if not curr_lines:
                start_line = line_num
            curr_lines.append(l_strip)
        else:
            if curr_lines:
                # check divider
                if any(re.match(r'^\|(?:\s*:?-+:?\s*\|)+$', r) for r in curr_lines):
                    # take first 2 rows as key
                    header_str = " // ".join(curr_lines[:2])
                    tables.append({
                        "file": label,
                        "start_line": start_line,
                        "line_count": len(curr_lines),
                        "header_key": header_str,
                        "first_row": curr_lines[0],
                        "sample_cell": curr_lines[min(2, len(curr_lines)-1)]
                    })
                curr_lines = []
    if curr_lines and any(re.match(r'^\|(?:\s*:?-+:?\s*\|)+$', r) for r in curr_lines):
        tables.append({
            "file": label,
            "start_line": start_line,
            "line_count": len(curr_lines),
            "header_key": " // ".join(curr_lines[:2]),
            "first_row": curr_lines[0],
            "sample_cell": curr_lines[min(2, len(curr_lines)-1)]
        })
    return tables

merged_tbls = extract_table_blocks(merged_text, "합본")
split_tbls = []
for p in sorted(all_split_files):
    rel = os.path.relpath(p, base_dir)
    with open(p, 'r', encoding='utf-8') as f:
        t = f.read()
    split_tbls.extend(extract_table_blocks(t, rel))

print(f"Merged tables: {len(merged_tbls)}")
print(f"Split tables: {len(split_tbls)}")

# Match each split table in merged_text
# If first_row and sample_cell not found together in merged_text
truly_missing_in_merged = []
for st in split_tbls:
    # check if first_row exists in merged_text
    r0 = st['first_row']
    # remove spacing differences
    norm_r0 = re.sub(r'\s+', '', r0)
    # Check if this row exists in any merged table
    found = False
    for mt in merged_tbls:
        norm_mt = re.sub(r'\s+', '', mt['first_row'])
        if norm_r0 == norm_mt:
            # check sample_cell too
            norm_sc = re.sub(r'\s+', '', st['sample_cell'])
            norm_msc = re.sub(r'\s+', '', mt['sample_cell'])
            if norm_sc == norm_msc or norm_sc[:15] in norm_msc or norm_msc[:15] in norm_sc:
                found = True
                break
    if not found:
        truly_missing_in_merged.append(st)

print(f"\nTruly missing tables in 합본: {len(truly_missing_in_merged)}")
for t in truly_missing_in_merged:
    print(f" - {t['file']} (L{t['start_line']}): {t['first_row'][:60]} | sample={t['sample_cell'][:40]}")

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