import copy import json import os import random import re import ast import sys import threading import time import logging import traceback import zlib import pandas as pd import numpy as np import redis from datetime import datetime sys.path.append("/opt/module/spark-3.2.0-bin-hadoop3.2/demo/py_demo/") sys.path.append(os.path.dirname(sys.path[0])) # 上级目录 from sqlalchemy import create_engine from utils.templates import Templates # from ..utils.templates import Templates from utils.templates_mysql import TemplatesMysql # from ..utils.templates_mysql import TemplatesMysql from pyspark.sql.types import IntegerType from pyspark.sql import functions as F from pyspark.sql.types import * from psycopg2.errors import NumericValueOutOfRange from sqlalchemy.exc import OperationalError, DataError, PendingRollbackError from utils.mysql_db import sql_connect, sql_update_many, sql_delete, get_country_engine from pyspark.sql import SparkSession logging.basicConfig(format='%(asctime)s %(name)s %(levelname)s %(message)s', level=logging.INFO) # from ..utils.DolphinschedulerHelper import DolphinschedulerHelper from utils.DolphinschedulerHelper import DolphinschedulerHelper from utils.db_util import DbTypes, DBUtil from queue import Queue class SpiderAsinDetail(Templates): def __init__(self, site_name='us', date_type="day", date_info='2022-10-01', consumer_type='lastest', topic_name="us_asin_detail", batch_size_history=100000, processing_time=900): super(SpiderAsinDetail, self).__init__() self.site_name = site_name self.date_type = date_type self.date_info = date_info self.consumer_type = consumer_type # 消费实时还是消费历史 self.topic_name = topic_name self.batch_size_history = batch_size_history self.processing_time = processing_time self.db_save = f'spider_asin_detail' # self.spark = self.create_spark_object(app_name=f"{self.db_save}: {self.site_name}, {self.date_type}, {self.date_info}, {self.consumer_type}") self.app_name = self.get_app_name() self.spark = self.create_spark_object(app_name=f"{self.app_name}") # 通过date_type 获取 topic self.get_topic_name() # 获取日期变量 self.get_year_week_tuple() # 连接数据库 # self.engine_mysql = DBUtil.get_db_engine(db_type=DbTypes.mysql.name, site_name=self.site_name) self.engine_pg14 = DBUtil.get_db_engine(db_type=DbTypes.postgresql_14.name, site_name=self.site_name) # 获取数据库表名 self.db_detail_name = str() self.db_variation_name = str() self.db_image_name = str() self.get_db_name() self.columns_detail_list = self.get_db_detail_columns() # self.get_db_columns() # 通过date_type 获取 schema self.init_schema() # self.topic_name = topic_name # 主题名字 # self.schema = self.init_schema() self.pdf_type_list = ["asin_vartion_list", "img_list", "asin_detail"] self.chunk_size = 1000 # 创建分区表队列 self.part_name_queue_image = Queue() self.part_name_queue_variation = Queue() # self.beginning_offsets = 2993_0000 if self.site_name == 'uk' else 0 # self.beginning_offsets = 3777_0000 if self.site_name == 'de' else 0 # self.beginning_offsets = 219_9000 if self.site_name == 'us' else 0 def init_schema(self): self.schema = StructType([ StructField("asin", StringType(), True), StructField("week", StringType(), True), StructField("month", StringType(), True), StructField("asin_vartion_list", StringType(), True), StructField("img_list", StringType(), True), StructField("title", StringType(), True), StructField("img_url", StringType(), True), StructField("rating", StringType(), True), StructField("total_comments", StringType(), True), StructField("price", FloatType(), True), StructField("rank", StringType(), True), StructField("category", StringType(), True), StructField("launch_time", StringType(), True), StructField("volume", StringType(), True), StructField("weight", StringType(), True), StructField("page_inventory", IntegerType(), True), StructField("buy_box_seller_type", IntegerType(), True), StructField("title_len", IntegerType(), True), StructField("img_num", IntegerType(), True), StructField("img_type", StringType(), True), StructField("activity_type", StringType(), True), StructField("one_two_val", StringType(), True), StructField("three_four_val", StringType(), True), StructField("eight_val", StringType(), True), StructField("qa_num", IntegerType(), True), StructField("five_star", IntegerType(), True), StructField("four_star", IntegerType(), True), StructField("three_star", IntegerType(), True), StructField("two_star", IntegerType(), True), StructField("one_star", IntegerType(), True), StructField("low_star", IntegerType(), True), StructField("together_asin", StringType(), True), StructField("brand", StringType(), True), StructField("ac_name", StringType(), True), StructField("material", StringType(), True), StructField("node_id", StringType(), True), StructField("data_type", IntegerType(), True), StructField("sp_num", StringType(), True), StructField("describe", StringType(), True), StructField("date_info", StringType(), True), StructField("weight_str", StringType(), True), StructField("package_quantity", StringType(), True), StructField("pattern_name", StringType(), True), StructField("seller_id", StringType(), True), StructField("variat_num", IntegerType(), True), StructField("site_name", StringType(), True), StructField("best_sellers_rank", StringType(), True), StructField("best_sellers_herf", StringType(), True), StructField("account_url", StringType(), True), StructField("account_name", StringType(), True), StructField("parentAsin", StringType(), True), StructField("asinUpdateTime", StringType(), True), StructField("spider_int", StringType(), True), StructField("follow_sellers", StringType(), True), StructField("buy_sales", StringType(), True), StructField("product_description", StringType(), True), StructField("category_state", IntegerType(), True), StructField("five_six_val", IntegerType(), True), StructField("image_view", IntegerType(), True), StructField("review_label_json", StringType(), True), StructField("product_json", StringType(), True), StructField("review_ai_text", StringType(), True), StructField("product_detail_json", StringType(), True), StructField("lob_asin_json", StringType(), True), StructField("sp_initial_seen_asins_json", StringType(), True), StructField("compare_similar_asin_json", StringType(), True), StructField("sp_4stars_initial_seen_asins_json", StringType(), True), StructField("customer_reviews_json", StringType(), True), StructField("sp_delivery_initial_seen_asins_json", StringType(), True), StructField("together_asin_json", StringType(), True), StructField("min_match_asin_json", StringType(), True), StructField("seller_json", StringType(), True), StructField("variat_num", IntegerType(), True), StructField("current_asin", StringType(), True), ]) # ['', '', '', '', '', ''] def get_topic_name(self): if self.site_name in ["us", "uk", "de"] and self.date_type == "month": self.topic_name = f"{site_name}_asin_detail_{self.date_type}_{self.date_info.replace('-', '_')}" else: self.topic_name = f"{site_name}_asin_detail" def get_db_name(self): self.db_detail_name = f"{self.site_name}_asin_detail_{self.date_info.split('-')[0]}_{self.date_info.split('-')[1]}" self.db_detail_name = self.db_detail_name.replace("_detail", "_detail_month") if self.date_type=='month' else self.db_detail_name self.db_variation_name = f"{self.site_name}_variat" self.db_image_name = f"{self.site_name}_asin_image" logging.info(f"db_detail_name:{self.db_detail_name}, db_variation_name:{self.db_variation_name}, db_image_name:{self.db_image_name}") def get_db_detail_columns(self): sql = f"select * from {self.db_detail_name} limit 0;" df = pd.read_sql(sql, con=self.engine_pg14) columns_list = list(set(df.columns)) columns_list.remove("id") columns_list.remove("updated_time") return columns_list def field_length_dispose(self, pdf): pdf.price = pdf.price.apply(lambda x: round(x, 2) if x is not None else None) # 截取字符 pdf.ac_name = pdf.ac_name.apply(lambda x: str(x)[:100] if x is not None else None) # 截取字符 pdf.brand = pdf.brand.apply(lambda x: str(x)[:100] if x is not None else None) # 截取字符 pdf.title = pdf.title.apply(lambda x: str(x)[:400] if x is not None else None) # 截取字符 pdf.category = pdf.category.apply(lambda x: str(x)[:400] if x is not None else None) # 截取字符 # pdf.img_url = pdf.img_url.apply(lambda x: str(x)[:400] if x is not None else None) # 截取字符 pdf.img_url = pdf.img_url.apply(lambda x: str(x)[:390] if x is not None else None) # 截取字符 pdf.material = pdf.material.apply(lambda x: str(x)[:150] if x is not None else None) # 截取字符 pdf.volume = pdf.volume.apply(lambda x: str(x)[:50] if x is not None else None) # 截取字符 if self.date_type in ["month", "week"]: pdf.package_quantity = pdf.package_quantity.apply(lambda x: str(x)[:50] if x is not None else None) # 截取字符 pdf.pattern_name = pdf.pattern_name.apply(lambda x: str(x)[:50] if x is not None else None) # 截取字符 pdf.weight_str = pdf.weight_str.apply(lambda x: str(x)[:250] if x is not None else None) # 截取字符 return pdf def start_process_instance(self): if site_name == 'us': # 最后一周走月流程 # year, week = self.year_week_tuple[-1].split("-") # sql = f"select count(*) as st_count from {self.site_name}_brand_analytics_{year} where week={week} ;" # print("sql:", sql) # year, month = self.date_info.split("-") # sql = f"select count(*) from {self.site_name}_brand_analytics_month_{year} where year={year} and month={month} ;" # df = pd.read_sql(sql, con=self.engine_mysql) # if list(df.st_count)[0] >= 100_0000: # process_df_name = f"{site_name}-月流程-ABA+反查(旧版)+流量选品(旧版)-api" # else: # self.date_type = "month_week" # process_df_name = f"{site_name}-30day+反查(旧版)+流量选品(旧版)-api" process_df_name = f"ALL站点-图片+变体表清洗" # 先走变体清洗 else: # process_df_name = f"{site_name}-ABA+反查(旧版)+流量选品(旧版)-api" process_df_name = f"ALL站点-图片+变体表清洗" # 先走变体清洗 print(f"process_df_name:{process_df_name}") DolphinschedulerHelper.start_process_instance( project_name="big_data_selection", process_df_name=process_df_name, startParams={ "site_name": self.site_name, "date_type": self.date_type, "date_info": self.date_info }, warning_Type="ALL" ) @staticmethod # 将asin转换成数值--从而可以划分指定分区表 def asin_to_number(asin): """ Convert a 10-character ASIN string to a unique number. This function assumes that ASIN consists of uppercase letters and digits. """ def char_to_number(char): if char.isdigit(): return int(char) else: return ord(char) - 55 # 'A' -> 10, 'B' -> 11, ..., 'Z' -> 35 if len(asin) != 10: raise ValueError(f"ASIN must be 10 characters long --{asin}--") base = 36 asin_number = 0 for i, char in enumerate(reversed(asin)): asin_number += char_to_number(char) * (base ** i) # The final number is taken modulo 1 billion to fit the range 1-10 billion return asin_number % 1000000000 @staticmethod # 列表均匀拆分成多个列表 def divide_list_into_equal_parts(lst, n): """ Divide a list into n equal parts. :param lst: List to be divided. :param n: Number of parts to divide into. :return: List of n lists. """ # Calculate the size of each part part_size = len(lst) // n return [lst[i * part_size:(i + 1) * part_size] for i in range(n)] @staticmethod # 将df对象的一行裂变成多行 def handle_data_df_explode(pdf, pdf_type='asin_vartion_list', columns=[]): # 根据不同表类型解析df对象 pdf[pdf_type] = pdf[pdf_type].apply(json.loads) # 对对应数据进行处理,将df_type内列表展开 exploded_list = pdf[pdf_type].explode() # 展开后转换为一个大列表 df_type_list = [i for i in exploded_list.tolist() if not isinstance(i, float)] df_type_list = [i for i in df_type_list if isinstance(i, list)] if df_type_list: df = pd.DataFrame(df_type_list, columns=columns) return df else: return None def save_data_image(self, pdf): logging.info("img处理") # 获取对应表字段 pdf['mapped_asin'] = pdf['asin'].apply(self.asin_to_number) while True: try: pdf.to_sql(name=f"{self.site_name}_asin_image", con=self.engine_pg14, if_exists='append', index=False, chunksize=100_0000) logging.info(f"入库{self.site_name}_asin_image, 图片数量为:{pdf.shape} 成功 {pdf.head(2)}") break except Exception as e: print(e, traceback.format_exc()) time.sleep(random.randint(3, 10)) self.engine_pg14 = DBUtil.get_db_engine(db_type=DbTypes.postgresql_14.name, site_name=self.site_name) continue def save_data_variation(self, pdf): pdf.drop_duplicates(subset=["asin", "parent_asin"], inplace=True) # 处理字段长度问题 pdf['color'] = pdf['color'].apply(lambda x: x.encode('utf-8', 'ignore').decode('utf-8')[:180] if x else None) pdf['size'] = pdf['size'].apply(lambda x: x.encode('utf-8', 'ignore').decode('utf-8')[:180] if x else None) pdf['style'] = pdf['style'].apply(lambda x: x.encode('utf-8', 'ignore').decode('utf-8')[:180] if x else None) pdf['column_2'] = pdf['column_2'].apply( lambda x: x.encode('utf-8', 'ignore').decode('utf-8')[:180] if x else None) logging.info(f"变体数量为:{pdf.shape}") # 处理分区名称问题 pdf['mapped_asin'] = pdf['parent_asin'].apply(self.asin_to_number) while True: try: pdf.to_sql(name=f"{self.site_name}_asin_variation", con=self.engine_pg14, if_exists='append', index=False, chunksize=100_0000) logging.info(f"入库{self.site_name}_asin_variation, 变体数量为:{pdf.shape} 成功 {pdf.head(2)}") break except Exception as e: print(e, traceback.format_exc()) time.sleep(random.randint(3, 10)) self.engine_pg14 = DBUtil.get_db_engine(db_type=DbTypes.postgresql_14.name, site_name=self.site_name) continue def save_data_asin_detail(self, pdf): print(f"{self.db_detail_name}: {pdf.columns}") # pdf.rename(columns={"asinUpdateTime": "created_time"}, inplace=True) pdf = pdf.loc[:, self.columns_detail_list] pdf = self.field_length_dispose(pdf) while True: try: # 分批次删除 # asin_tuple_all = tuple(pdf.asin) # for i in range(0, len(asin_tuple_all)+1, self.chunk_size): # asin_tuple = asin_tuple_all[i: i+self.chunk_size] # if asin_tuple: # asin_tuple = asin_tuple if len(asin_tuple) > 1 else f"('{asin_tuple[0]}')" # with self.engine_pg14.begin() as conn: # sql_del = f"delete from {self.db_detail_name} where asin in {asin_tuple};" # print("sql_del:", sql_del[:100]) # conn.execute(sql_del) # 存储 pdf.to_sql(name=self.db_detail_name, con=self.engine_pg14, if_exists='append', index=False, chunksize=100_0000) break except Exception as e: logging.info(f"error: {e}") time.sleep(random.randint(5, 20)) self.engine_pg14 = DBUtil.get_db_engine(db_type=DbTypes.postgresql_14.name, site_name=self.site_name) continue def save_data_common(self, pdf, pdf_type): logging.info(f"{pdf_type} 处理") start_time = time.time() pdf.rename(columns={"asinUpdateTime": "created_time"}, inplace=True) if pdf_type == 'asin_vartion_list': columns_list = ['asin', 'color', 'parent_asin', 'size', 'state', 'style', 'column_2'] # , 'created_time' pdf = self.handle_data_df_explode(pdf, pdf_type=pdf_type, columns=columns_list) pdf['asin'] = pdf['asin'].apply(lambda x: str(x).replace('/', '')) self.save_data_variation(pdf=pdf) elif pdf_type == "img_list": columns_list = ['asin', 'img_url', 'img_order_by', 'data_type'] # , 'created_time' pdf = self.handle_data_df_explode(pdf, pdf_type=pdf_type, columns=columns_list) pdf['asin'] = pdf['asin'].apply(lambda x: str(x).replace('/', '')) self.save_data_image(pdf=pdf) elif pdf_type == "asin_detail": # if self.site_name != 'us' and self.date_type != 'month': self.save_data_asin_detail(pdf=pdf) logging.info(f"{pdf_type}: 耗时 -- {time.time() - start_time}") def save_data(self, pdf): threads = [] for pdf_type in self.pdf_type_list: thread = threading.Thread(target=self.save_data_common, args=(pdf, pdf_type)) threads.append(thread) thread.start() for thread in threads: thread.join() logging.info("所有线程处理完成") def handle_kafka_df(self, kafka_df): kafka_df.show(20) # pyspark的kafka_df对象转换成pandas的df对象 pdf = kafka_df.toPandas() # pdf['asin'] = pdf['asin'].str.replace('/', '', regex=True) # 去掉脏数据 pdf['asin'] = pdf['asin'].apply(lambda x: str(x).replace('/', '')) # print(111111111111, pdf.loc[pdf.asin.str.contains("/")], pdf.loc[pdf.asin.str.contains("/")].shape) # 去重 pdf = pdf.drop_duplicates(['asin']) # 过滤--不符合当前周期的数据 pdf = pdf.loc[(~pdf.date_info.isna()) & (pdf.date_info == self.date_info)] if pdf.shape[0]: logging.info(f"{pdf.keys()}") logging.info(f"----------------------------") if self.date_type == "day": logging.info(f"天数据处理") img_columns = ['asin', 'img_url', 'img_order_by', 'data_type', 'created_time'] img_df = self.handle_data_df(pdf, df_type='asin_vartion_list', columns=img_columns) if img_df.shape[0]: self.save_data_image(df=img_df) vartion_columns = ['asin', 'color', 'parent_asin', 'size', 'state', 'style', 'column_2', 'created_time'] vartion_df = self.handle_data_df(pdf, df_type='asin_vartion_list', columns=vartion_columns) if vartion_df.shape[0]: self.save_data_variation(df=vartion_df) df = pdf[self.detail_col] df['site'] = df['site'].fillna(self.site_name) df.drop_duplicates(['asin', 'site'], inplace=True) now_date = time.strftime("%Y-%m-%d", time.gmtime(time.time())) detail_table_data_info = f"{self.site_name}_self_asin_detail" for name, group in df.groupby(['site']): asins = list(group["asin"]) # 详情入库表名 if asins: if len(asins) == 1: sql_del = f"delete from `{detail_table_data_info}` where `asin`= '{asins[0]}' and `site`='{name[0]}' and created_at>='{now_date}';" else: sql_del = f"delete from `{detail_table_data_info}` where `asin` in {tuple(asins)} and `site`='{name[0]}' and created_at>='{now_date}';" logging.info(f"{name}, {sql_del}") sql_delete(sql_del) logging.info(f"清理 {detail_table_data_info} 表中 {asins[0:10]} 数据") df.to_sql(name=f'{detail_table_data_info}', con=self.engine_pg14, if_exists='append', index=False) logging.info(f"入库 {detail_table_data_info} 成功 {df.head(10)}") else: self.save_data(pdf=pdf) else: logging.info(f"{pdf.shape}") def handle_kafka_history(self, kafka_df): # kafka_df = kafka_df.withColumn("asin", F.regexp_replace("asin", "/", "")) # kafka_df = kafka_df.withColumn("asin", F.translate("asin", "/", "")) self.handle_kafka_df(kafka_df) def handle_kafka_stream(self, kafka_df, epoch_id): # kafka_df = kafka_df.withColumn("asin", F.regexp_replace("asin", "/", "")) # kafka_df = kafka_df.withColumn("asin", F.translate("asin", "/", "")) self.handle_kafka_df(kafka_df) if __name__ == '__main__': site_name = sys.argv[1] # 参数1:站点 # batch_size_history = 15000 if site_name == 'us' else 10000 batch_size_history = 50000 date_type = sys.argv[2] # 参数2:类型:week/4_week/month/quarter/day date_info = sys.argv[3] # 参数3:年-周/年-月/年-季/年-月-日, 比如: 2022-1 consumer_type = sys.argv[4] # 参数4:实时 latest 历史 history # us day date_info 2023-11-07 handle_obj = SpiderAsinDetail(site_name=site_name, date_type=date_type, date_info=date_info, consumer_type=consumer_type, batch_size_history=batch_size_history) handle_obj.run_kafka()