"""
分析特定原始特征的搜索执行情况
"""

import json
import sys
from typing import Dict, Any, List


def analyze_feature_searches(stage4_path: str, stage6_path: str, feature_name: str):
    """分析指定原始特征的搜索情况"""

    # 加载数据
    with open(stage4_path, 'r', encoding='utf-8') as f:
        stage4_data = json.load(f)

    with open(stage6_path, 'r', encoding='utf-8') as f:
        stage6_data = json.load(f)

    # 找到指定特征
    stage4_feature = None
    stage6_feature = None

    for item in stage4_data:
        if item['原始特征名称'] == feature_name:
            stage4_feature = item
            break

    for item in stage6_data:
        if item['原始特征名称'] == feature_name:
            stage6_feature = item
            break

    if not stage4_feature:
        print(f"❌ 在 Stage4 中未找到特征: {feature_name}")
        return

    if not stage6_feature:
        print(f"❌ 在 Stage6 中未找到特征: {feature_name}")
        return

    print("=" * 80)
    print(f"原始特征: {feature_name}")
    print("=" * 80)

    # 收集 Stage4 的所有搜索词
    stage4_search_words = []
    for association in stage4_feature.get('找到的关联', []):
        assoc_name = association.get('分类名称', '')
        for feature in association.get('特征列表', []):
            search_word = feature.get('search_word')
            llm_eval = feature.get('llm_evaluation', {})

            if search_word:
                stage4_search_words.append({
                    'search_word': search_word,
                    'association': assoc_name,
                    'feature_name': feature.get('特征名称', ''),
                    'llm_score': llm_eval.get('score'),
                    'llm_rank': llm_eval.get('rank'),
                    'reasoning': llm_eval.get('reasoning', '')
                })

    # 收集 Stage6 的所有搜索词及其执行状态
    stage6_search_words = []
    for association in stage6_feature.get('找到的关联', []):
        assoc_name = association.get('分类名称', '')
        for feature in association.get('特征列表', []):
            search_word = feature.get('search_word')
            search_result = feature.get('search_result')
            search_metadata = feature.get('search_metadata', {})
            llm_eval = feature.get('llm_evaluation', {})

            if search_word:
                stage6_search_words.append({
                    'search_word': search_word,
                    'association': assoc_name,
                    'feature_name': feature.get('特征名称', ''),
                    'llm_score': llm_eval.get('score'),
                    'llm_rank': llm_eval.get('rank'),
                    'has_result': search_result is not None,
                    'status': search_metadata.get('status', 'not_searched'),
                    'note_count': search_metadata.get('note_count', 0)
                })

    # 统计
    total_stage4 = len(stage4_search_words)
    total_stage6 = len(stage6_search_words)
    searched = sum(1 for w in stage6_search_words if w['has_result'])
    not_searched = total_stage6 - searched

    print(f"\n📊 统计信息:")
    print(f"  Stage4 生成的搜索词数: {total_stage4}")
    print(f"  Stage6 保留的搜索词数: {total_stage6}")
    print(f"  已执行搜索: {searched} 个")
    print(f"  未执行搜索: {not_searched} 个")
    print(f"  搜索执行率: {searched/total_stage6*100:.1f}%")

    # 按 rank 排序并展示
    stage6_sorted = sorted(stage6_search_words, key=lambda x: x['llm_rank'] if x['llm_rank'] else 999)

    print(f"\n📋 详细搜索词列表 (按 LLM Rank 排序):")
    print(f"{'Rank':<6} {'评分':<6} {'搜索状态':<12} {'帖子数':<8} 搜索词")
    print("-" * 80)

    for idx, word in enumerate(stage6_sorted, 1):
        rank = word['llm_rank'] if word['llm_rank'] else 'N/A'
        score = f"{word['llm_score']:.2f}" if word['llm_score'] else 'N/A'
        status = '✅ 已搜索' if word['has_result'] else '⏸️  未搜索'
        note_count = word['note_count'] if word['has_result'] else '-'

        print(f"{rank:<6} {score:<6} {status:<12} {note_count:<8} {word['search_word']}")

    # 展示已搜索的搜索词详情
    searched_words = [w for w in stage6_sorted if w['has_result']]
    if searched_words:
        print(f"\n✅ 已执行搜索的 {len(searched_words)} 个搜索词:")
        for idx, word in enumerate(searched_words, 1):
            print(f"\n  【{idx}】 {word['search_word']}")
            print(f"       关联: {word['association']}")
            print(f"       特征: {word['feature_name']}")
            print(f"       评分: {word['llm_score']:.2f}, 排名: #{word['llm_rank']}")
            print(f"       结果: {word['note_count']} 个帖子")

    # 展示未搜索的搜索词
    not_searched_words = [w for w in stage6_sorted if not w['has_result']]
    if not_searched_words:
        print(f"\n⏸️  未执行搜索的 {len(not_searched_words)} 个搜索词:")
        for idx, word in enumerate(not_searched_words, 1):
            print(f"\n  【{idx}】 {word['search_word']}")
            print(f"       关联: {word['association']}")
            print(f"       特征: {word['feature_name']}")
            print(f"       评分: {word['llm_score']:.2f}, 排名: #{word['llm_rank']}")

    # 分析为什么只搜索了部分
    print(f"\n🔍 搜索策略分析:")
    if searched == 10:
        print(f"  系统使用了 Top-10 策略")
        top_10_ranks = sorted([w['llm_rank'] for w in searched_words if w['llm_rank']])
        print(f"  实际搜索的 Rank 范围: {top_10_ranks}")

        # 检查是否严格按 rank 取的 top-10
        expected_top_10_ranks = sorted([w['llm_rank'] for w in stage6_sorted[:10] if w['llm_rank']])
        if top_10_ranks == expected_top_10_ranks:
            print(f"  ✓ 严格按照 LLM Rank 取了 Top-10")
        else:
            print(f"  ⚠️  不是严格的 Top-10 (期望: {expected_top_10_ranks})")
    elif searched > 0:
        print(f"  系统执行了 {searched} 个搜索")
    else:
        print(f"  该特征的搜索尚未执行")

    print("\n" + "=" * 80)


if __name__ == '__main__':
    stage4_path = 'output_v2/stage4_with_llm_scores.json'
    stage6_path = 'output_v2/stage6_with_evaluations.json'
    feature_name = '墨镜'

    if len(sys.argv) > 1:
        feature_name = sys.argv[1]
    if len(sys.argv) > 2:
        stage4_path = sys.argv[2]
    if len(sys.argv) > 3:
        stage6_path = sys.argv[3]

    analyze_feature_searches(stage4_path, stage6_path, feature_name)