Take the 2-minute tour ×
Code Review Stack Exchange is a question and answer site for peer programmer code reviews. It's 100% free, no registration required.

I am using scrapy framework for data scraping and dumping item in MySQL database.

Here is my pipeline that is inserting output to MySQL, but its taking so much time. Any suggestions on how to optimize this?

class MysqlOutputPipeline(object):
  def __init__(self):
    dispatcher.connect(self.spider_opened, signals.spider_opened)
    dispatcher.connect(self.spider_closed, signals.spider_closed)

  def connect(self):
    try:
      self.conn = MySQLdb.connect(
        host='some_host',
        user='user',
        passwd='pwd',
        db='my_db',
        port=22)
    except (AttributeError, MySQLdb.OperationalError), e:
      raise e

  def query(self, sql, params=()):
    try:
      cursor = self.conn.cursor()
      cursor.execute(sql, params)
    except (AttributeError, MySQLdb.OperationalError) as e:
      print 'exception generated during sql connection: ', e
      self.connect()
      cursor = self.conn.cursor()
      cursor.execute(sql, params)
    return cursor

  def spider_opened(self, spider):
    self.connect()

  def process_item(self, item, spider):
    # clean_name
    clean_name = ''.join(e for e in item['store'] if e.isalnum()).lower()

    # conditional insertion in store_meta
    sql = """SELECT * FROM store_meta WHERE clean_name = %s"""
    curr = self.query(sql, clean_name)
    if not curr.fetchone():
      sql = """INSERT INTO store_meta (clean_name) VALUES (%s)"""
      self.query(sql, clean_name)
      self.conn.commit()

    # getting clean_id
    sql = """SELECT clean_id FROM store_meta WHERE clean_name = %s"""
    curr = self.query(sql, clean_name)
    clean_id = curr.fetchone()

    # conditional insertion in all_stores
    sql = """SELECT * FROM all_stores WHERE store_name = %s"""
    curr = self.query(sql, item['store'])
    if not curr.fetchone():
      sql = """INSERT INTO all_stores (store_name,clean_id) VALUES (%s,%s)"""
      self.query(sql, (item['store'], clean_id[0]))
      self.conn.commit()

    # getting store_id
    sql = """SELECT store_id FROM all_stores WHERE store_name =%s"""
    curr = self.query(sql, item['store'])
    store_id = curr.fetchone()

    if item and not item['is_coupon'] 
            and (item['store'] in ['null', ''] 
            or item['bonus'] in ['null', '']):
      raise DropItem(item)
    if item and  not item['is_coupon']:# conditional insertion in discounts table
      sql = """SELECT * 
               FROM discounts 
               WHERE mall=%s 
               AND store_id=%s 
               AND bonus=%s 
               AND deal_url=%s"""
      curr = self.query(
               sql, 
               (
                 item['mall'], 
                 store_id[0], 
                 item['bonus'], 
                 item['deal_url']
               )
             )
      if not curr.fetchone():
        self.query(
          "INSERT INTO discounts 
           (mall,store_id,bonus,per_action,more_than,up_to,deal_url) 
           VALUES (%s,%s,%s,%s,%s,%s,%s)",
           (item['mall'],
           store_id[0],
           item['bonus'],
           item['per_action'],
           item['more_than'],
           item['up_to'],
           item['deal_url']),
        )
        self.conn.commit()

    # conditional insertion in crawl_coupons table
    elif spider.name not in COUPONS_LESS_STORE: 
      if item['expiration'] is not 'null':
        item['expiration']=datetime.strptime(item['expiration'],'%m/%d/%Y').date()
        sql = """SELECT * 
               FROM crawl_coupons 
               WHERE mall=%s 
               AND clean_id=(SELECT clean_id FROM store_meta WHERE clean_name = %s) 
               AND coupon_code=%s 
               AND coupon_text=%s 
               AND expiry_date=%s"""
      curr = self.query(
               sql,
               (
                 item['mall'], 
                 clean_name, 
                 item['code'], 
                 self.conn.escape_string(item['discount']), 
                 item['expiration']
               )
             )
      if not curr.fetchone():
          sql = """INSERT INTO crawl_coupons 
                   (mall,clean_id,coupon_code,coupon_text,expiry_date) 
                   VALUES (
                     %s,
                     (SELECT clean_id FROM store_meta WHERE clean_name = %s),
                     %s,
                     %s,
                     %s
                   )"""
          self.query(
            sql, 
            (
              item['mall'], 
              clean_name, 
              item['code'], 
              self.conn.escape_string(item['discount']), 
              item['expiration']
            )
          )
          self.conn.commit()
    return item
share|improve this question
    
What makes you think this code is the bottleneck? –  Gareth Rees Sep 25 '12 at 16:31
    
on mysql queries part , other then that should i use adbapi ? –  akhter wahab Sep 25 '12 at 16:34
add comment

1 Answer

up vote 3 down vote accepted

It's hard to analyze the bottle neck without performance data on where the bottleneck is occurring. You should consider running the program several ways and comparing. Ie

  • run program as is
  • run program spider only with no db

This should tell you the problem is in your code as you suspect. If not you are crawling something that is slow to crawl. Your program can't control that, although you can look into why it's slow. Maybe you are hitting a tar pit (ie websites put in controls to allow normal users regular access but to impede crawlers) which would affect how you crawl.

If it is your program

  • run program as is
  • run program spider with just all your conditional code but no writes

My suspicion is that the writes are not the issue. You can then can analyze your conditions and try to optimize.

You are reading the db alot before writing. Whether any of those reads could be cached is data dependent on what you are crawling. If that has any predictability (eg it appears you are crawling malls and stores so I would assume alot of predictability) then maybe save at least the most recent q/a for each condition in ram so you don't need to go out to the db if you already have the answer. My guess is just caching most recent will be sufficient since crawler most likely walks 'near' items next. But if that isn't enough, analyze the db you've already gotten from your slow runs. If 90% of the db results are for Macy's then just cache Macy's condition results. Or you could put in more effort and put in a full caching class

share|improve this answer
add comment

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.