update

crawler_sys/framework/search_page_single_process.py

@@ -7,6 +7,9 @@ Created on Tue Dec  4 14:00:03 2018
 
 import argparse
 import configparser
+import random
+from concurrent.futures.process import ProcessPoolExecutor
+
 from elasticsearch.helpers import scan
 from elasticsearch import Elasticsearch
 from crawler.crawler_sys.framework.platform_crawler_register import get_crawler
@@ -98,20 +101,28 @@ for platform in PLATFORM_LIST:
     initialize_crawler = get_crawler(platform)
     crawler = initialize_crawler()
     KEYWORD_dic = func_search_keywordlist(platform)
+    KEYWORD_dic = random.shuffle(KEYWORD_dic)
 
+    executor = ProcessPoolExecutor(max_workers=3)
+    futures = []
     for keyword in KEYWORD_dic:
         print("search keyword '%s' on platform %s" % (keyword, platform))
         search_pages = int(KEYWORD_dic[keyword])
-        try:
-            crawler.search_page(keyword=keyword,
-                                search_pages_max=search_pages,
-                                output_to_es_raw=OUTPUT_TO_ES_RAW,
-                                output_to_es_register=OUTPUT_TO_ES_REGISTER,
-                                es_index=ES_INDEX,proxies_num=proxies_num)
-
-        except Exception as e:
-            print(e)
-            continue
-
+        # try:
+        #     crawler.search_page(keyword=keyword,
+        #                         search_pages_max=search_pages,
+        #                         output_to_es_raw=OUTPUT_TO_ES_RAW,
+        #                         output_to_es_register=OUTPUT_TO_ES_REGISTER,
+        #                         es_index=ES_INDEX,proxies_num=proxies_num)
+        #
+        # except Exception as e:
+        #     print(e)
+        #     continue
+        future = executor.submit(crawler.search_page, keyword,search_pages_max=search_pages,
+                             output_to_es_raw=OUTPUT_TO_ES_RAW,
+                           output_to_es_register=OUTPUT_TO_ES_REGISTER,
+                           es_index=ES_INDEX,proxies_num=proxies_num)
+        futures.append(future)
+    executor.shutdown(True)
 # config file absolute path in serve
 # '/home/hanye/crawlers/crawler_sys/framework/config/search_keywords.ini'