def persist_db_history_and_latest(table_name,
num_cols,
value_list,
is_many=True,
need_history=False,
need_flow=False):
tbl_latest = '%s_latest' % table_name
ps_list = []
for i in xrange(num_cols):
ps_list.append('%s')
values_str = ','.join(ps_list)
t1 = time.time()
tcur = t1
affected_rows = 99999
affected_rows3 = 99999
if need_history:
sql = 'replace into %s values(%s)' % (table_name, values_str)
affected_rows = dbhelper.executeSqlWriteMany(sql,
value_list,
is_dirty=True)
t2 = time.time()
tcur = t2
logging.debug('persist_db_history_and_latest, history using time: %s' %
(t2 - t1))
if need_flow:
sql = 'replace into %s values(%s)' % (table_name + '_flow', values_str)
affected_rows3 = dbhelper.executeSqlWriteMany(sql,
value_list,
is_dirty=True)
t21 = time.time()
logging.debug('persist_db_history_and_latest, flow using time: %s' %
(t21 - tcur))
tcur = t21
sql2 = 'replace into %s values(%s)' % (tbl_latest, values_str)
affected_rows2 = dbhelper.executeSqlWriteMany(sql2,
value_list,
is_dirty=True)
t3 = time.time()
logging.debug('persist_db_history_and_latest, latest using time: %s' %
(t3 - tcur))
status = -1
if affected_rows > 0 and affected_rows2 > 0 and affected_rows3 > 0:
status = 0
ret = {
'status': status,
'affected_rows_latest': affected_rows2,
'affected_rows_history': affected_rows,
'affected_rows_flow': affected_rows3,
}
return ret
def __record_task_complete__(self, task_list):
vlist = []
ut = timeHelper.getNowLong()
for item in task_list:
tp = (self.job_name, item, ut)
vlist.append(tp)
sql = 'insert into task_status(job_name,task_id,update_time) values(%s,%s,%s)'
affected_rows = dbhelper.executeSqlWriteMany(sql, vlist)
return affected_rows
def __record_task_complete__(self, task_list):
vlist = []
ut = timeHelper.getNowLong()
for item in task_list:
tp = (self.job_name,item,ut)
vlist.append(tp)
sql = 'insert into task_status(job_name,task_id,update_time) values(%s,%s,%s)'
affected_rows = dbhelper.executeSqlWriteMany(sql,vlist)
return affected_rows
def persist_db_history_and_latest(table_name, num_cols, value_list, is_many=True, need_history=False, need_flow=False):
tbl_latest = '%s_latest' %table_name
ps_list = []
for i in xrange(num_cols):
ps_list.append('%s')
values_str = ','.join(ps_list)
t1 = time.time()
tcur = t1
affected_rows = 99999
affected_rows3 = 99999
if need_history:
sql = 'replace into %s values(%s)' %(table_name,values_str)
affected_rows = dbhelper.executeSqlWriteMany(sql,value_list,is_dirty=True)
t2 = time.time()
tcur = t2
logging.debug('persist_db_history_and_latest, history using time: %s' %(t2-t1))
if need_flow:
sql = 'replace into %s values(%s)' %(table_name+'_flow',values_str)
affected_rows3 = dbhelper.executeSqlWriteMany(sql,value_list,is_dirty=True)
t21 = time.time()
logging.debug('persist_db_history_and_latest, flow using time: %s' %(t21-tcur))
tcur = t21
sql2 = 'replace into %s values(%s)' %(tbl_latest,values_str)
affected_rows2 = dbhelper.executeSqlWriteMany(sql2,value_list,is_dirty=True)
t3 = time.time()
logging.debug('persist_db_history_and_latest, latest using time: %s' %(t3-tcur))
status = -1
if affected_rows>0 and affected_rows2>0 and affected_rows3>0:
status = 0
ret = {
'status':status,
'affected_rows_latest': affected_rows2,
'affected_rows_history': affected_rows,
'affected_rows_flow': affected_rows3,
}
return ret
def crawl_detail_images(sku_id):
html = __get_detail_page_content__(sku_id)
img_list = jd_detail_resolver.resolve_Images(html)
# logging.debug(img_list)
if len(img_list)==0:
return {'status':-1}
vlist = []
update_time = timeHelper.getNow()
for img in img_list:
tp = (sku_id, update_time, img)
vlist.append(tp)
# sql = 'replace into jd_item_images values(%s,%s,%s)'
# affected_rows = dbhelper.executeSqlWriteMany(sql,vlist)
sql2 = 'replace into jd_item_images_latest values(%s,%s,%s)'
affected_rows2 = dbhelper.executeSqlWriteMany(sql2,vlist)
ret = {
'status': -1,
# 'affected_rows': affected_rows,
'affected_rows2': affected_rows2
}
if affected_rows2>0:
ret['status'] = 0
return ret
def crawl_detail_property(sku_id):
html = __get_detail_page_content__(sku_id)
prop_map = jd_detail_resolver.resolve_Properties(html)
update_time = timeHelper.getNow()
vlist = []
if len(prop_map) == 0:
return {'status':0}
for p_key in prop_map:
p_value = prop_map[p_key]
tp = (sku_id, update_time, p_key, p_value)
vlist.append(tp)
# sql = 'replace into jd_item_property values(%s,%s,%s,%s)'
# affected_rows = dbhelper.executeSqlWriteMany(sql,vlist)
sql2 = 'replace into jd_item_property_latest values(%s,%s,%s,%s)'
affected_rows2 = dbhelper.executeSqlWriteMany(sql2,vlist)
ret = {
'status': -1,
# 'affected_rows': affected_rows,
'affected_rows2': affected_rows2
}
if affected_rows2 > 0:
ret['status'] = 0
return ret
def loadCategoryList():
html = url_utils.getWebResponse(JD_CATEGORY_WEBSERVICE_URL, JD_ENC)
json_str = url_utils.removeJsonP(html)
obj = json.loads(json_str)
clist = __extractCategoryList_fromJson__(obj)
cat_list = []
for item in clist:
print item
vals = item.split('|')
if len(vals) < 4:
print 'error in length of category line'
print item
continue
cat_name = vals[1]
vals0 = vals[0]
cat_id = cat_url = cat_memo = ""
if '.com' in vals0:
cat_url = vals0
else:
cat_id = vals0
if len(vals[2]) > 0:
cat_memo = vals[2]
if len(cat_id) > 0:
tp = (cat_id, cat_name, timeHelper.getNow())
cat_list.append(tp)
# persist categories
sql = 'replace into jd_category values(%s,%s,%s)'
affected_rows = dbhelper.executeSqlWriteMany(sql, cat_list)
print 'rows affected : jd_category : %s' % affected_rows
return 0
def loadCategoryList():
html = url_utils.getWebResponse(JD_CATEGORY_WEBSERVICE_URL,JD_ENC)
json_str = url_utils.removeJsonP(html)
obj = json.loads(json_str)
clist = __extractCategoryList_fromJson__(obj)
cat_list = []
for item in clist:
print item
vals = item.split('|')
if len(vals)<4:
print 'error in length of category line'
print item
continue
cat_name = vals[1]
vals0 = vals[0]
cat_id = cat_url = cat_memo = ""
if '.com' in vals0:
cat_url = vals0
else:
cat_id = vals0
if len(vals[2]) > 0:
cat_memo = vals[2]
if len(cat_id) > 0:
tp = (cat_id,cat_name, timeHelper.getNow())
cat_list.append(tp)
# persist categories
sql = 'replace into jd_category values(%s,%s,%s)'
affected_rows = dbhelper.executeSqlWriteMany(sql,cat_list)
print 'rows affected : jd_category : %s' %affected_rows
return 0
def crawl_category(category_id):
logging.debug('category_id = %s -- page 1' %(category_id))
url = __get_category_page_url__(category_id,1)
# print url
html = url_utils.getWebResponse(url,'utf-8')
if html == "":
html = url_utils.getWebResponse(url,'gb18030')
if html == "":
html = url_utils.getWebResponse(url, 'gbk')
total_pages = jd_list_resolver.resolveTotalPageNum(html)
product_list = jd_list_resolver.resolveProductListFromPage(html)
while len(product_list) == 0 and category_id is not None:
category_id = __up_roll_category_id__(category_id)
return crawl_category(category_id)
if category_id is None or len(product_list)==0:
return {'status':-1, 'msg': 'No item in category product list'}
for page_iter in range(2,total_pages+1):
logging.debug('category_id = %s -- page %s' %(category_id,page_iter))
url = __get_category_page_url__(category_id,page_iter)
html = url_utils.getWebResponse(url,'utf-8')
product_list = product_list + jd_list_resolver.resolveProductListFromPage(html)
time.sleep(SLEEP_TIME)
sku_list = []
for product_tp in product_list:
sku_id = product_tp[0]
sku_list.append(sku_id)
# Get price of all products
#price_obj = jd_API.getPrices_JD(sku_list,sleep_time=SLEEP_PRICE_API)
ret_obj = {
'status': -1,
'affected_rows': -1,
'sku_count': -1
}
total_goods_num = len(product_list)
# for item in product_list:
# print item[0]
# print '='*80
# combine product list and price list, timestamp, category_id
for i in xrange(total_goods_num):
product_id = product_list[i][0]
pkey = '%s' %product_id
# if pkey in price_obj:
# product_list[i] = product_list[i] + (price_obj[pkey][0],price_obj[pkey][1],price_obj[pkey][2],) #nowdate,nowtime,)
# else:
# logging.error('Error: product_id=%s cannot get result' %(product_id,price_id))
# continue
product_list[i] = product_list[i] + (0,0,0,)
# persist in database
# (sku_id,sku_title,sku_url,sku_thumnail_url,sku_stock,comment_count,is_global,is_pay_on_delivery,is_free_gift,sku_icon_url, price, price_m, update_date,update_time, category_id)
# sql = '''
# replace into jd_item_dynamic (sku_id,title,url,thumbnail_url,stock_status,comment_count,is_global,is_pay_on_delivery,
# has_free_gift,icon_url,price,price_m,price_pcp,update_date,update_time) values (%s,%s,%s,%s,%s,%s,%s,%s,%s,%s,%s,%s,%s,%s,%s)
# '''
# affected_rows = dbhelper.executeSqlWriteMany(sql,product_list)
ret = crawler_helper.persist_db_history_and_latest(
table_name='jd_item_dynamic',
num_cols=len(product_list[0]),
value_list=product_list,
is_many=True,
need_history=False, # was True - changed 01/03
need_flow=False, # was True - changed 12/23
)
logging.debug('Saved to DB -- category_id = %s -- sku_count=%s' %(category_id,total_goods_num))
logging.debug('%s' %ret)
# HANDLE JD_ITEM_CATEGORY
item_cat_list = []
for prod in product_list:
item_cat_list.append((prod[0],category_id,))
sql2 = 'replace into jd_item_category values (%s,%s)'
affected_rows2 = dbhelper.executeSqlWriteMany(sql2,item_cat_list)
logging.debug('Saved to DB - item_category - affected rows = %s' %affected_rows2)
if affected_rows2<=0:
logging.error('Saving to item_category error, category_id = %s' %category_id)
# HANDLE JD_ITEM_FIRSTSEEN
nowtime = timeHelper.getNowLong()
nowdate = timeHelper.getNow()
sql3 = 'insert ignore into jd_item_firstseen values(%s,"%s","%s")'
ftlist = []
for item in product_list:
ftlist.append([item[0],nowtime,nowdate])
affected_rows3 = dbhelper.executeSqlWriteMany(sql3,ftlist)
ret_obj = {
'status': 0 if ret['status']==0 and affected_rows2>0 else -1,
'item_dynamic': ret,
'item_category': affected_rows2,
'item_first_seen': affected_rows3,
}
return ret_obj
def crawl_category(category_id):
logging.debug('category_id = %s -- page 1' % (category_id))
url = __get_category_page_url__(category_id, 1)
# print url
html = url_utils.getWebResponse(url, 'utf-8')
if html == "":
html = url_utils.getWebResponse(url, 'gb18030')
if html == "":
html = url_utils.getWebResponse(url, 'gbk')
total_pages = jd_list_resolver.resolveTotalPageNum(html)
product_list = jd_list_resolver.resolveProductListFromPage(html)
while len(product_list) == 0 and category_id is not None:
category_id = __up_roll_category_id__(category_id)
return crawl_category(category_id)
if category_id is None or len(product_list) == 0:
return {'status': -1, 'msg': 'No item in category product list'}
for page_iter in range(2, total_pages + 1):
logging.debug('category_id = %s -- page %s' % (category_id, page_iter))
url = __get_category_page_url__(category_id, page_iter)
html = url_utils.getWebResponse(url, 'utf-8')
product_list = product_list + jd_list_resolver.resolveProductListFromPage(
html)
time.sleep(SLEEP_TIME)
sku_list = []
for product_tp in product_list:
sku_id = product_tp[0]
sku_list.append(sku_id)
# Get price of all products
#price_obj = jd_API.getPrices_JD(sku_list,sleep_time=SLEEP_PRICE_API)
ret_obj = {'status': -1, 'affected_rows': -1, 'sku_count': -1}
total_goods_num = len(product_list)
# for item in product_list:
# print item[0]
# print '='*80
# combine product list and price list, timestamp, category_id
for i in xrange(total_goods_num):
product_id = product_list[i][0]
pkey = '%s' % product_id
# if pkey in price_obj:
# product_list[i] = product_list[i] + (price_obj[pkey][0],price_obj[pkey][1],price_obj[pkey][2],) #nowdate,nowtime,)
# else:
# logging.error('Error: product_id=%s cannot get result' %(product_id,price_id))
# continue
product_list[i] = product_list[i] + (
0,
0,
0,
)
# persist in database
# (sku_id,sku_title,sku_url,sku_thumnail_url,sku_stock,comment_count,is_global,is_pay_on_delivery,is_free_gift,sku_icon_url, price, price_m, update_date,update_time, category_id)
# sql = '''
# replace into jd_item_dynamic (sku_id,title,url,thumbnail_url,stock_status,comment_count,is_global,is_pay_on_delivery,
# has_free_gift,icon_url,price,price_m,price_pcp,update_date,update_time) values (%s,%s,%s,%s,%s,%s,%s,%s,%s,%s,%s,%s,%s,%s,%s)
# '''
# affected_rows = dbhelper.executeSqlWriteMany(sql,product_list)
ret = crawler_helper.persist_db_history_and_latest(
table_name='jd_item_dynamic',
num_cols=len(product_list[0]),
value_list=product_list,
is_many=True,
need_history=False, # was True - changed 01/03
need_flow=False, # was True - changed 12/23
)
logging.debug('Saved to DB -- category_id = %s -- sku_count=%s' %
(category_id, total_goods_num))
logging.debug('%s' % ret)
# HANDLE JD_ITEM_CATEGORY
item_cat_list = []
for prod in product_list:
item_cat_list.append((
prod[0],
category_id,
))
sql2 = 'replace into jd_item_category values (%s,%s)'
affected_rows2 = dbhelper.executeSqlWriteMany(sql2, item_cat_list)
logging.debug('Saved to DB - item_category - affected rows = %s' %
affected_rows2)
if affected_rows2 <= 0:
logging.error('Saving to item_category error, category_id = %s' %
category_id)
# HANDLE JD_ITEM_FIRSTSEEN
nowtime = timeHelper.getNowLong()
nowdate = timeHelper.getNow()
sql3 = 'insert ignore into jd_item_firstseen values(%s,"%s","%s")'
ftlist = []
for item in product_list:
ftlist.append([item[0], nowtime, nowdate])
affected_rows3 = dbhelper.executeSqlWriteMany(sql3, ftlist)
ret_obj = {
'status': 0 if ret['status'] == 0 and affected_rows2 > 0 else -1,
'item_dynamic': ret,
'item_category': affected_rows2,
'item_first_seen': affected_rows3,
}
return ret_obj