import pandas as pd from sqlalchemy import create_engine, text import logging import math import re import time # --- 配置 (与之前相同) --- EXCEL_FILE_PATH = 'Z:\\xiaohu\\web_data40000-50000.xlsx' SHEET_NAME = 0 DB_USER = 'zsjie' DB_PASSWORD = 'xRekX6Cc3RRK6mBe' DB_HOST = '111.180.203.166' DB_PORT = 25506 DB_NAME = 'zsjie' TABLE_NAME = 'resource' UNIQUE_KEY_COLUMNS = ['id'] DATE_COLUMN_TO_CONVERT = 'update_date' DEFAULT_FOR_STRING = '' DEFAULT_FOR_NUMERIC = 0 COMMIT_BATCH_SIZE = 300 # --- 新增：用于条件判断的列名 --- CONDITION_COLUMN = 'resource_url' # 基于此列在数据库中的值来决定是否更新 logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(levelname)s - %(message)s') def excel_to_mysql_upsert_conditional_on_url(): engine = None connection = None transaction = None rows_processed = 0 rows_since_last_commit = 0 total_inserted = 0 total_updated = 0 # 注意：此计数现在可能包含实际未更改但匹配了重复键的行 start_time = time.time() try: # 1. 创建引擎 (带 pool_recycle) engine_url = f"mysql+mysqlconnector://{DB_USER}:{DB_PASSWORD}@{DB_HOST}:{DB_PORT}/{DB_NAME}" engine = create_engine(engine_url, pool_recycle=3600) logging.info(f"数据库引擎创建成功: {DB_HOST}:{DB_PORT}/{DB_NAME}") # 2. 读取 Excel read_start_time = time.time() logging.info(f"开始读取 Excel: {EXCEL_FILE_PATH} (Sheet: {SHEET_NAME})") df = pd.read_excel(EXCEL_FILE_PATH, sheet_name=SHEET_NAME, dtype='object', keep_default_na=True) logging.info(f"Excel 读取完成，共 {len(df)} 行。耗时: {time.time() - read_start_time:.2f} 秒。") # --- 3. 数据清理和准备 --- transform_start_time = time.time() logging.info("开始清理和转换数据...") # a. 重命名列 rename_map = { 'URL': 'url', 'Title': 'title', 'Tags': 'tags', 'source': 'resource_url', 'password': 'resource_pd', } df.rename(columns=rename_map, inplace=True) # b. 删除 Unnamed 列 unnamed_cols = [col for col in df.columns if str(col).startswith('Unnamed:')] if unnamed_cols: df.drop(columns=unnamed_cols, inplace=True) # c. 清理列名空格 df.columns = [str(col).strip() for col in df.columns] # d. 检查 'id' 和 条件列 存在 if 'id' not in df.columns: raise ValueError("错误：唯一键列 'id' 在 DataFrame 中未找到。") if CONDITION_COLUMN not in df.columns: # 如果条件列必须存在，则报错 raise ValueError(f"错误：条件更新所需的列 '{CONDITION_COLUMN}' 在 DataFrame 中未找到。") # 如果条件列是可选的，可以只记录警告并构建不带条件的SQL # logging.warning(f"警告: 列 '{CONDITION_COLUMN}' 未找到，将执行无条件 Upsert。") # build_conditional_sql = False # 控制下方 SQL 构建逻辑 # e. update_date 转换 if DATE_COLUMN_TO_CONVERT in df.columns: date_series = df[DATE_COLUMN_TO_CONVERT].astype(str) date_series_numeric_str = date_series.str.replace(r'\D', '', regex=True) df[DATE_COLUMN_TO_CONVERT] = pd.to_numeric(date_series_numeric_str, errors='coerce') # f. 确定列类型 col_types = {} for col in df.columns: if pd.api.types.is_string_dtype(df[col]) or pd.api.types.is_object_dtype(df[col]): col_types[col] = 'string' elif pd.api.types.is_numeric_dtype(df[col]): col_types[col] = 'numeric' else: col_types[col] = 'other' logging.info(f"数据清理转换完成。耗时: {time.time() - transform_start_time:.2f} 秒。") # g. *** 修改：构建带条件的 SQL 模板 *** all_columns = df.columns.tolist() update_columns = [col for col in all_columns if col not in UNIQUE_KEY_COLUMNS] cols_str = ", ".join([f"`{col}`" for col in all_columns]) placeholders_str = ", ".join([f":{col}" for col in all_columns]) # 构建 ON DUPLICATE KEY UPDATE 部分 if update_columns: update_clause_list = [] for col in update_columns: # --- 核心条件逻辑 --- # 如果数据库现有的 resource_url 是 NULL 或空, 则更新为新值, 否则保持旧值 update_clause = f"`{col}` = IF(`{CONDITION_COLUMN}` IS NULL OR `{CONDITION_COLUMN}` = '', VALUES(`{col}`), `{col}`)" update_clause_list.append(update_clause) # --- 结束核心条件逻辑 --- update_str = ", ".join(update_clause_list) sql_template = f""" INSERT INTO `{TABLE_NAME}` ({cols_str}) VALUES ({placeholders_str}) ON DUPLICATE KEY UPDATE {update_str} """ logging.info(f"将使用带条件 (基于数据库 '{CONDITION_COLUMN}' 值) 的 INSERT ... ON DUPLICATE KEY UPDATE 模式。") else: # 如果只有 id 列，则使用 INSERT IGNORE sql_template = f"INSERT IGNORE INTO `{TABLE_NAME}` ({cols_str}) VALUES ({placeholders_str})" logging.info("仅配置了唯一键，将使用 INSERT IGNORE 模式。") # --- 结束 SQL 构建 --- # --- 4. 数据库交互与周期性提交 (与之前类似) --- db_interaction_start_time = time.time() connection = engine.connect() logging.info("数据库连接成功。") transaction = connection.begin() logging.info("已开始第一个事务。") logging.info(f"开始处理 {len(df)} 行数据 (每 {COMMIT_BATCH_SIZE} 行提交一次)...") for record_original in df.to_dict(orient='records'): rows_processed += 1 record_processed = {} # --- 应用默认值 (与之前相同) --- for col_name, value in record_original.items(): processed_value = value if pd.isna(processed_value): if col_name not in UNIQUE_KEY_COLUMNS: column_type = col_types.get(col_name, 'other') if column_type == 'string': processed_value = DEFAULT_FOR_STRING elif column_type == 'numeric': processed_value = DEFAULT_FOR_NUMERIC else: processed_value = None record_processed[col_name] = processed_value # --- 检查 'id' (与之前相同) --- if record_processed.get('id') is None: logging.warning(f"跳过第 {rows_processed} 行，因为 'id' 列为空或无效: {record_original}") continue # 跳过该行 # 增加有效行批次计数器 rows_since_last_commit += 1 # --- 执行 SQL (使用新的条件模板) --- try: result = connection.execute(text(sql_template), record_processed) # --- 解释 rowcount (可能有歧义) --- # 1: 插入了新行 # 2: 匹配了重复键，并执行了 UPDATE 子句（即使所有 IF 条件都为 false，导致无实际更改） # 0: 匹配了重复键，但某些 MySQL 版本/配置下，无实际更改的 UPDATE 可能报告 0 if result.rowcount == 1: total_inserted += 1 elif result.rowcount == 2: # 认为匹配了重复键并尝试了更新 total_updated += 1 # 注意：total_updated 不再精确代表“实际发生值改变的更新行数” except Exception as row_error: logging.error(f"处理行数据时出错 (行号约 {rows_processed}):\n 原始: {record_original}\n 处理后: {record_processed}\n 错误: {row_error}") if transaction: try: transaction.rollback() except Exception as rb_err: logging.error(f"回滚事务时也出错: {rb_err}") raise # --- 周期性提交 (与之前相同) --- if rows_since_last_commit >= COMMIT_BATCH_SIZE: try: commit_start = time.time() transaction.commit() commit_duration = time.time() - commit_start logging.info(f"已提交 {rows_since_last_commit} 行 (处理总数: {rows_processed})。本次提交耗时: {commit_duration:.2f} 秒。") transaction = connection.begin() rows_since_last_commit = 0 except Exception as commit_error: logging.error(f"提交事务时出错 (行号约 {rows_processed}): {commit_error}") raise # --- 循环结束后的最终提交 (与之前相同) --- if rows_since_last_commit > 0: try: final_commit_start = time.time() logging.info(f"准备提交最后 {rows_since_last_commit} 行...") transaction.commit() final_commit_duration = time.time() - final_commit_start logging.info(f"最后 {rows_since_last_commit} 行已成功提交。耗时: {final_commit_duration:.2f} 秒。") except Exception as final_commit_error: logging.error(f"提交最后批次事务时出错: {final_commit_error}") raise total_db_time = time.time() - db_interaction_start_time logging.info(f"数据库交互完成。总耗时: {total_db_time:.2f} 秒。") logging.info(f"处理完成。总处理行: {rows_processed}, 总插入: {total_inserted}, 总匹配重复键(尝试更新): {total_updated}.") except ValueError as ve: logging.error(f"配置或数据错误: {ve}") except Exception as e: logging.error(f"发生严重错误，脚本已停止: {e}", exc_info=False) finally: if connection: connection.close() logging.info("数据库连接已关闭。") total_script_time = time.time() - start_time logging.info(f"脚本总运行时间: {total_script_time:.2f} 秒。") if __name__ == "__main__": excel_to_mysql_upsert_conditional_on_url()