def __op_new(self, data_loaded, connection):
item_id = int(data_loaded['id'])
item = UrlItem.load_with_content(
id=item_id,file_path=config.path_extractor_inbox)
count, maps = db.get_url_with_same_layout_hash(item['layout_hash'])
log.info(str(maps))
log.info(count)
if len(maps) > 0:
import operator
tar_ext = max(maps.iteritems(), key=operator.itemgetter(1))
log.info(float(tar_ext[1]) / len(maps))
if tar_ext[1] > config.extractor_same_layout_number:
extractor = tool.str2extractor(tar_ext[0])
self.__extract(item,extractor)
return
extractor = config.const_RULE_UNKNOW
self.__ext_queue[item_id] = {
"title": item['title'],
"url": item['url'],
"filename": item.filename(),
"decision": item['is_target'],
"extractor": extractor
}
log.info("[%s]: # %s " % (item_id, extractor))
pass
def __op_new(self, data_loaded, connection):
item_id = int(data_loaded['id'])
item = UrlItem.load_with_content(id=item_id,
file_path=config.path_extractor_inbox)
count, maps = db.get_url_with_same_layout_hash(item['layout_hash'])
log.info(str(maps))
log.info(count)
if len(maps) > 0:
import operator
tar_ext = max(maps.iteritems(), key=operator.itemgetter(1))
log.info(float(tar_ext[1]) / len(maps))
if tar_ext[1] > config.extractor_same_layout_number:
extractor = tool.str2extractor(tar_ext[0])
self.__extract(item, extractor)
return
extractor = config.const_RULE_UNKNOW
self.__ext_queue[item_id] = {
"title": item['title'],
"url": item['url'],
"filename": item.filename(),
"decision": item['is_target'],
"extractor": extractor
}
log.info("[%s]: # %s " % (item_id, extractor))
pass
def __op_refresh(self, data_loaded, connection):
delete_ids = []
for key, ent in self.__ext_queue.iteritems():
item_id = int(key)
item = UrlItem.load_with_content(
id=item_id, file_path=config.path_extractor_inbox)
count, maps = db.get_url_with_same_layout_hash(item['layout_hash'])
log.info(str(maps))
log.info(count)
if len(maps) > 0:
import operator
tar_ext = max(maps.iteritems(), key=operator.itemgetter(1))
log.info(float(tar_ext[1]) / len(maps))
if tar_ext[1] > config.extractor_same_layout_number:
extractor = tool.str2extractor(tar_ext[0])
self.__extract(item, extractor)
delete_ids.append(item_id)
# clear delete_ids
for ent_id in delete_ids:
del self.__ext_queue[ent_id]
def load(id=None, url=None):
r = UrlItem()
if id is not None:
res = db.get_url_by_id(id)
elif url is not None:
res = db.get_url_by_url(url)
else:
raise Exception("must provide id or url")
if res is None:
return None
r['id'] = res['id']
r['url'] = res['url']
r['is_target'] = res['is_target']
r['content_hash'] = res['content_hash']
r['layout_hash'] = res['layout_hash']
r['extractor'] = tool.str2extractor(res['extractor'])
r['last_access_ts'] = res['last_access_ts']
r['last_extract_ts'] = res['last_extract_ts']
r['title'] = res['title']
r['content_type'] = res['content_type']
return r
def __op_refresh(self,data_loaded,connection):
delete_ids = []
for key, ent in self.__ext_queue.iteritems():
item_id = int(key)
if not os.path.isfile(
config.path_extractor_inbox+"/"+str(item_id)+".html"):
delete_ids.append(item_id)
continue
item = UrlItem.load_with_content(
id=item_id,file_path=config.path_extractor_inbox)
count, maps = db.get_url_with_same_layout_hash(item['layout_hash'])
log.info(str(maps))
log.info(count)
if len(maps) > 0:
import operator
tar_ext = max(maps.iteritems(), key=operator.itemgetter(1))
log.info(float(tar_ext[1]) / len(maps))
if tar_ext[1] > config.extractor_same_layout_number:
extractor = tool.str2extractor(tar_ext[0])
self.__extract(item,extractor)
delete_ids.append(item_id)
for ent_id in delete_ids:
del self.__ext_queue[ent_id]
