#!/usr/bin/env python3
"""
渠道4-5月线索 - 退款相关性分析

分析 Sheet1 退款状态与各维度的相关性：
- 渠道、课包、电话接通
- 用户注册时间、角色创建时间
- U0行课行为（L1/L2各课完成情况）
"""

import pandas as pd
import numpy as np
from openpyxl import Workbook
from openpyxl.styles import Font, Alignment, PatternFill, Border, Side
from openpyxl.utils.dataframe import dataframe_to_rows
import glob, sys
from datetime import datetime

def load_data():
    files = glob.glob('/root/.openclaw/media/inbound/*d58af024*')
    if not files:
        print("未找到输入文件")
        sys.exit(1)
    input_file = files[0]
    print(f"读取: {input_file}")
    df1 = pd.read_excel(input_file, sheet_name='Sheet1', dtype=str)
    df2 = pd.read_excel(input_file, sheet_name='Sheet2', dtype=str)
    return df1, df2

def prepare_data(df1, df2):
    """准备匹配后的分析数据集"""
    # Sheet1 清洗
    df1['账号ID_clean'] = pd.to_numeric(df1['账号ID'], errors='coerce')
    df1['is_refund'] = df1['订单状态'].apply(lambda x: '是' if x == '已退款' else '否')
    
    # Sheet2 清洗 - 聚合到用户级别
    df2['用户ID_num'] = pd.to_numeric(df2['用户ID'], errors='coerce')
    
    # U0 10课的列
    u0_cols = ['L1_U00_L01_完成时间', 'L1_U00_L02_完成时间', 'L1_U00_L03_完成时间', 
               'L1_U00_L04_完成时间', 'L1_U00_L05_完成时间',
               'L2_U00_L01_完成时间', 'L2_U00_L02_完成时间', 'L2_U00_L03_完成时间', 
               'L2_U00_L04_完成时间', 'L2_U00_L05_完成时间']
    l1_cols = u0_cols[:5]
    l2_cols = u0_cols[5:]
    
    # 按用户聚合 Sheet2
    def agg_user(group):
        row = {}
        row['注册时间'] = group['用户注册时间'].iloc[0]  # 同一用户注册时间相同
        # 最早角色创建时间
        create_times = pd.to_datetime(group['角色创建时间'], errors='coerce').dropna()
        row['最早角色创建时间'] = create_times.min() if len(create_times) > 0 else pd.NaT
        
        # U0行课：只要任一角色完成就算完成
        for c in u0_cols:
            if c in group.columns:
                vals = group[c].dropna()
                vals = vals[vals != '']
                vals = vals[vals != 'nan']
                row[c] = vals.iloc[0] if len(vals) > 0 else ''
            else:
                row[c] = ''
        
        # L1完成课数
        l1_done = 0
        for c in l1_cols:
            if c in group.columns:
                vals = group[c].dropna()
                vals = vals[vals != '']
                vals = vals[vals != 'nan']
                if len(vals) > 0:
                    l1_done += 1
        row['L1完成课数'] = l1_done
        
        # L2完成课数
        l2_done = 0
        for c in l2_cols:
            if c in group.columns:
                vals = group[c].dropna()
                vals = vals[vals != '']
                vals = vals[vals != 'nan']
                if len(vals) > 0:
                    l2_done += 1
        row['L2完成课数'] = l2_done
        
        row['U0总完成课数'] = l1_done + l2_done
        
        # 行课类型
        if l1_done > 0 and l2_done > 0:
            row['行课类型'] = 'L1+L2都有'
        elif l1_done > 0:
            row['行课类型'] = '仅L1'
        elif l2_done > 0:
            row['行课类型'] = '仅L2'
        else:
            row['行课类型'] = '无行课'
        
        # 是否有任何行课
        row['是否有行课'] = '有' if (l1_done + l2_done > 0) else '无'
        
        return pd.Series(row)
    
    df2_agg = df2.groupby('用户ID_num', as_index=False).apply(agg_user).reset_index()
    
    # 匹配
    merged = df1.merge(df2_agg, left_on='账号ID_clean', right_on='用户ID_num', how='left')
    merged['匹配状态'] = merged['用户ID_num'].notna().map({True: '已匹配', False: '未匹配'})
    
    return merged

def calc_stats(df, col, refund_col='is_refund'):
    """计算某维度的退款统计"""
    total = df.groupby(col).size().reset_index(name='总数')
    refund = df[df[refund_col] == '是'].groupby(col).size().reset_index(name='退款数')
    result = total.merge(refund, on=col, how='left')
    result['退款数'] = result['退款数'].fillna(0).astype(int)
    result['退费率'] = (result['退款数'] / result['总数'] * 100).round(1)
    result['退费率'] = result['退费率'].apply(lambda x: f'{x}%')
    result = result.sort_values('总数', ascending=False)
    return result

def analyze(df):
    """全面分析"""
    results = {}
    
    # 1. 渠道 vs 退款
    results['1-渠道'] = calc_stats(df, '渠道')
    
    # 2. 课包 vs 退款
    results['2-课包'] = calc_stats(df, '课包')
    
    # 3. 电话接通 vs 退款
    results['3-电话接通'] = calc_stats(df, '电话已接通')
    
    # 4. 渠道+课包 交叉 vs 退款
    cross = df.groupby(['渠道', '课包']).agg(
        总数=('is_refund', 'count'),
        退款数=('is_refund', lambda x: (x == '是').sum())
    ).reset_index()
    cross['退费率'] = (cross['退款数'] / cross['总数'] * 100).round(1)
    cross['退费率'] = cross['退费率'].apply(lambda x: f'{x}%')
    results['4-渠道×课包交叉'] = cross.sort_values('总数', ascending=False)
    
    # 5. 电话接通+课包 交叉
    cross2 = df.groupby(['电话已接通', '课包']).agg(
        总数=('is_refund', 'count'),
        退款数=('is_refund', lambda x: (x == '是').sum())
    ).reset_index()
    cross2['退费率'] = (cross2['退款数'] / cross2['总数'] * 100).round(1)
    cross2['退费率'] = cross2['退费率'].apply(lambda x: f'{x}%')
    results['5-电话接通×课包交叉'] = cross2.sort_values('总数', ascending=False)
    
    # 6. 用户注册时间（按月） vs 退款
    df['注册月份'] = pd.to_datetime(df['注册时间'], errors='coerce').dt.to_period('M').astype(str)
    df['注册月份'] = df['注册月份'].fillna('未知')
    results['6-注册月份'] = calc_stats(df, '注册月份')
    
    # 7. 角色创建时间（按月）
    df['角色创建月份'] = pd.to_datetime(df['最早角色创建时间'], errors='coerce').dt.to_period('M').astype(str)
    df['角色创建月份'] = df['角色创建月份'].fillna('未知')
    results['7-角色创建月份'] = calc_stats(df, '角色创建月份')
    
    # 8. 是否有行课 vs 退款
    results['8-是否有行课'] = calc_stats(df, '是否有行课')
    
    # 9. 行课类型 vs 退款
    results['9-行课类型'] = calc_stats(df, '行课类型')
    
    # 10. U0完成课数 vs 退款
    results['10-U0完成课数'] = calc_stats(df, 'U0总完成课数')
    
    # 11. L1完成课数 vs 退款
    results['11-L1完成课数'] = calc_stats(df, 'L1完成课数')
    
    # 12. L2完成课数 vs 退款
    results['12-L2完成课数'] = calc_stats(df, 'L2完成课数')
    
    # 13. 注册到创建时间差 vs 退款（天）
    df['注册时间_dt'] = pd.to_datetime(df['注册时间'], errors='coerce')
    df['角色创建时间_dt'] = pd.to_datetime(df['最早角色创建时间'], errors='coerce')
    df['注册到创建天数'] = (df['角色创建时间_dt'] - df['注册时间_dt']).dt.total_seconds() / 86400
    df['注册到创建天数'] = df['注册到创建天数'].fillna(-1)
    
    def day_bucket(d):
        if d < 0:
            return '未知'
        elif d <= 1:
            return '≤1天'
        elif d <= 7:
            return '2-7天'
        elif d <= 30:
            return '8-30天'
        elif d <= 90:
            return '31-90天'
        else:
            return '>90天'
    df['注册到创建间隔'] = df['注册到创建天数'].apply(day_bucket)
    results['13-注册到角色创建间隔'] = calc_stats(df, '注册到创建间隔')
    
    # 14. 匹配状态 vs 退款（未匹配的在Sheet2中找不到数据）
    results['14-匹配状态'] = calc_stats(df, '匹配状态')
    
    # 整体
    total_count = len(df)
    refund_count = (df['is_refund'] == '是').sum()
    results['_整体'] = {
        '总订单数': total_count,
        '退款订单数': refund_count,
        '退费率': f'{refund_count/total_count*100:.1f}%',
        '已匹配数': df['匹配状态'].value_counts().get('已匹配', 0),
        '未匹配数': df['匹配状态'].value_counts().get('未匹配', 0),
    }
    
    return results, df

def write_excel(results, df, output_path):
    """写入多Sheet Excel"""
    wb = Workbook()
    # 删除默认Sheet
    wb.remove(wb.active)
    
    header_font = Font(bold=True, size=11)
    header_fill = PatternFill(start_color='4472C4', end_color='4472C4', fill_type='solid')
    header_font_white = Font(bold=True, size=11, color='FFFFFF')
    thin_border = Border(
        left=Side(style='thin'), right=Side(style='thin'),
        top=Side(style='thin'), bottom=Side(style='thin')
    )
    red_fill = PatternFill(start_color='FFC7CE', end_color='FFC7CE', fill_type='solid')
    
    # 汇总页
    ws = wb.create_sheet('汇总')
    ws.append(['指标', '值'])
    for k, v in results['_整体'].items():
        ws.append([k, v])
    ws.append([])
    ws.append(['分析维度摘要'])
    # 各个维度的摘要
    for sheet_name, stat_df in results.items():
        if sheet_name.startswith('_'):
            continue
        if isinstance(stat_df, pd.DataFrame) and len(stat_df) > 0:
            # 找到退费率最高的
            if '退费率' in stat_df.columns:
                max_rate_row = stat_df.iloc[0]
                first_col = stat_df.columns[0]
                ws.append([sheet_name, f'{first_col}={max_rate_row[first_col]}: 退费率{max_rate_row["退费率"]} (退款{max_rate_row["退款数"]}/{max_rate_row["总数"]})'])
    
    for col in ws.columns:
        col[0].font = header_font
    ws.column_dimensions['A'].width = 30
    ws.column_dimensions['B'].width = 60
    
    # 各维度Sheet
    for sheet_name, stat_df in results.items():
        if sheet_name.startswith('_') or not isinstance(stat_df, pd.DataFrame):
            continue
        ws = wb.create_sheet(sheet_name[:31])  # Excel sheet name max 31 chars
        
        # 写入表头
        for col_idx, col_name in enumerate(stat_df.columns, 1):
            cell = ws.cell(row=1, column=col_idx, value=col_name)
            cell.font = header_font_white
            cell.fill = header_fill
            cell.border = thin_border
            cell.alignment = Alignment(horizontal='center')
        
        # 写入数据
        for row_idx, row in enumerate(stat_df.itertuples(index=False), 2):
            for col_idx, val in enumerate(row, 1):
                cell = ws.cell(row=row_idx, column=col_idx, value=val)
                cell.border = thin_border
                cell.alignment = Alignment(horizontal='center')
                # 高退费率标红 (>整体退费率)
                if '退费率' in stat_df.columns and col_idx == len(row):
                    try:
                        rate = float(str(val).replace('%', ''))
                        if rate > 44.0:  # 整体退费率约44%
                            cell.fill = red_fill
                            cell.font = Font(color='9C0006', bold=True)
                    except:
                        pass
        
        # 自动列宽
        for col_idx in range(1, len(stat_df.columns) + 1):
            max_len = max(
                len(str(stat_df.columns[col_idx-1])),
                stat_df.iloc[:, col_idx-1].astype(str).str.len().max()
            )
            ws.column_dimensions[chr(64+col_idx) if col_idx <= 26 else 'A'].width = min(max_len + 4, 30)
    
    # 明细页：所有匹配后的数据
    ws_detail = wb.create_sheet('匹配明细')
    detail_cols = ['下单日期', '手机号', '渠道', '课包', '订单状态', '电话已接通', '账号ID',
                   '匹配状态', '注册时间', '最早角色创建时间', '行课类型', 'U0总完成课数',
                   'L1完成课数', 'L2完成课数', '注册月份', '角色创建月份', '注册到创建间隔']
    detail_df = df[detail_cols].copy()
    
    for col_idx, col_name in enumerate(detail_df.columns, 1):
        cell = ws_detail.cell(row=1, column=col_idx, value=col_name)
        cell.font = header_font_white
        cell.fill = header_fill
        cell.border = thin_border
        cell.alignment = Alignment(horizontal='center')
    
    for row_idx, row in enumerate(detail_df.itertuples(index=False), 2):
        for col_idx, val in enumerate(row, 1):
            cell = ws_detail.cell(row=row_idx, column=col_idx, value=val if pd.notna(val) else '')
            cell.border = thin_border
            # 退款行标红背景
            if row[4] == '已退款':
                cell.fill = PatternFill(start_color='FFF2CC', end_color='FFF2CC', fill_type='solid')
    
    wb.save(output_path)
    print(f"报表已保存: {output_path}")

def main():
    df1, df2 = load_data()
    merged = prepare_data(df1, df2)
    
    print(f"\n整体: {len(merged)}条, 已匹配{(merged['匹配状态']=='已匹配').sum()}, 未匹配{(merged['匹配状态']=='未匹配').sum()}")
    print(f"退款: {(merged['is_refund']=='是').sum()}, 未退款: {(merged['is_refund']=='否').sum()}")
    
    results, merged = analyze(merged)
    
    print(f"\n=== 整体 ===")
    for k, v in results['_整体'].items():
        print(f"  {k}: {v}")
    
    print(f"\n=== 各维度退费率 ===")
    for name, stat in results.items():
        if name.startswith('_') or not isinstance(stat, pd.DataFrame):
            continue
        print(f"\n--- {name} ---")
        print(stat.to_string(index=False))
    
    ts = datetime.now().strftime('%Y%m%d_%H%M%S')
    output_path = f'/root/.openclaw/workspace/output/渠道4-5月线索_退款相关性分析_{ts}.xlsx'
    write_excel(results, merged, output_path)
    print(f"\n输出文件: {output_path}")
    return output_path

if __name__ == '__main__':
    main()