"""完整靶场对账：CSV 原始预期 vs 引擎实际输出。

逐样本打印：case_id / 文件名 / CSV 预期（裁决+关键规则）/ 引擎实际（裁决+规则）/ 差距分类
"""
import sys, json, csv, asyncio, time
from pathlib import Path

sys.path.insert(0, '/app')

async def run():
    from app.main import scanner
    loop = asyncio.get_running_loop()
    
    # 读 CSV
    csv_rows = {}
    with open('/tmp/yongli.csv', encoding='utf-8') as f:
        for row in csv.DictReader(f):
            csv_rows[row['用例ID'].strip()] = row
    
    # 读 index.json（用于归类 L7 旧样本等）
    idx = json.load(open('/tmp/regen_v2.json')) if Path('/tmp/regen_v2.json').exists() else {}
    # 先从宿主机拷
    print('=== 开始全量靶场对账 ===\n')
    
    # 扫描所有文件
    from pathlib import Path as P
    samples_dir = P('/app/tests/samples')
    results = []
    
    files = sorted(p for p in samples_dir.iterdir() 
                   if p.is_file() and p.name not in ('index.json',) and not p.name.startswith('_'))
    
    # 先用 regen_v2 的 index 拿 case_id
    old_idx = json.loads(samples_dir.joinpath('index.json').read_text())
    meta_map = {s['filename']: s for s in old_idx['samples']}
    
    for fpath in files:
        meta = meta_map.get(fpath.name, {})
        cid = meta.get('case_id', '')
        
        # CSV 原始预期
        csv_exp = csv_rows.get(cid, {})
        csv_verdict = csv_exp.get('预期结果', '')
        csv_name = csv_exp.get('用例名称', '')
        
        # 引擎实际
        try:
            outcome = await loop.run_in_executor(None, lambda fp=fpath: scanner.scan_file(
                file_path=str(fp), original_filename=fp.name,
                task_id=f"chk-{fp.name}", channel="audit", client_id="console",
            ))
        except Exception as e:
            results.append({
                'cid': cid, 'file': fpath.name, 'layer': meta.get('category', ''),
                'csv_verdict': csv_verdict, 'csv_name': csv_name,
                'eng_verdict': 'error', 'eng_rules': [], 'score': 0, 'reason': str(e),
            })
            continue
        
        eng_rules = sorted({f.rule for f in outcome.findings})
        results.append({
            'cid': cid, 'file': fpath.name, 'layer': meta.get('category', ''),
            'csv_verdict': csv_verdict, 'csv_name': csv_name,
            'eng_verdict': outcome.verdict, 'eng_rules': eng_rules,
            'score': outcome.score,
        })
    
    # 分类差距
    def classify(r):
        csv_v = r['csv_verdict']
        eng_v = r['eng_verdict']
        eng_reject = eng_v in ('reject', 'malicious')
        eng_pass = eng_v in ('pass', 'clean')
        
        if not csv_v:
            return 'NO_CSV'
        # CSV 预期 reject
        if any(k in csv_v for k in ['拦截', '告警', '高危', '拒绝', '命中']) and eng_pass:
            return 'CSV想拦→引擎放'
        # CSV 预期 pass
        if any(k in csv_v for k in ['无风险', '无告警', '无命中', '正常解析']) and eng_reject:
            return 'CSV想放→引擎拦'
        # verdict 类型不匹配
        if '拦截' in csv_v and eng_v == 'suspicious':
            return 'CSV要拦截→引擎仅可疑'
        if '低危' in csv_v and eng_reject:
            return 'CSV说低危→引擎拦截'
        if '低危' in csv_v and eng_pass:
            return 'CSV说低危→引擎放（合理，低危=放行）'
        return 'OK'
    
    # 打印汇总
    from collections import Counter, defaultdict
    gap_cnt = Counter()
    gap_by_layer = defaultdict(Counter)
    for r in results:
        gap = classify(r)
        gap_cnt[gap] += 1
        gap_by_layer[r['layer']][gap] += 1
    
    print('=== 总体分类 ===')
    for gap, cnt in gap_cnt.most_common():
        print(f'  {gap:20s}: {cnt}')
    print()
    
    print('=== 按层 × 分类 ===')
    for layer in sorted(set(r['layer'] for r in results)):
        total = sum(1 for r in results if r['layer'] == layer)
        ok = sum(1 for r in results if r['layer'] == layer and classify(r) == 'OK')
        print(f'  {layer:5s}: {ok}/{total} OK')
        for gap, cnt in gap_by_layer[layer].most_common():
            if gap != 'OK':
                print(f'    └ {gap}: {cnt}')
    print()
    
    # 逐类打印差距详情
    for gap_type in ['CSV想拦→引擎放', 'CSV想放→引擎拦', 'CSV要拦截→引擎仅可疑', 'CSV说低危→引擎拦截', 'NO_CSV']:
        items = [r for r in results if classify(r) == gap_type]
        if not items:
            continue
        print(f'=== {gap_type} ({len(items)} 条) ===')
        for r in items:
            print(f"  {r['cid']:10s} [{r['layer']:3s}] {r['file']:45s}")
            print(f"    CSV预期: {r['csv_verdict'][:80]}")
            print(f"    引擎输出: verdict={r['eng_verdict']:10s} rules={r['eng_rules'][:5]}")
            print()

asyncio.run(run())