Question

I’ve managed to code a very simple crawler with Scrapy, with these given constraints:

• Store all link info (e.g.: anchor text, page title), hence the 2 callbacks
• Use CrawlSpider to take advantage of rules, hence no BaseSpider

It runs well, except it doesn’t implement rules if I add a callback to the first request!

Here is my code: (works but not properly, with a live example)

from scrapy.contrib.spiders import CrawlSpider,Rule
from scrapy.selector import HtmlXPathSelector
from scrapy.http import Request
from scrapySpider.items import SPage
from scrapy.contrib.linkextractors.sgml import SgmlLinkExtractor

class TestSpider4(CrawlSpider):
    name = "spiderSO"
    allowed_domains = ["cumulodata.com"]
    start_urls = ["http://www.cumulodata.com"]
    extractor = SgmlLinkExtractor()

    def parse_start_url(self, response):
        #3
        print('----------manual call of',response)
        self.parse_links(response)
        print('----------manual call done')
        # 1 return Request(self.start_urls[0]) # does not call parse_links(example.com)
        # 2 return Request(self.start_urls[0],callback = self.parse_links) # does not call parse_links(example.com)

    rules = (
        Rule(extractor,callback='parse_links',follow=True),
        )

    def parse_links(self, response):
        hxs = HtmlXPathSelector(response)
        print('----------- manual parsing links of',response.url)
        links = hxs.select('//a')
        for link in links:
                title = link.select('@title')
                url = link.select('@href').extract()[0]
                meta={'title':title,}
                yield Request(url, callback = self.parse_page,meta=meta)

    def parse_page(self, response):
        print('----------- parsing page: ',response.url)
        hxs = HtmlXPathSelector(response)
        item=SPage()
        item['url'] = str(response.request.url)
        item['title']=response.meta['title']
        item['h1']=hxs.select('//h1/text()').extract()
        yield item

I’ve tried solving this issue in 3 ways:

• 1: To return a Request with the start url – rules are not executed
• 2: Same as above, but with a callback to parse_links – Same issue
• 3: Call parse_links after scraping the start url, by implementing parse_start_url, function does not get called

Here are the logs:

----------manual call of <200 http://www.cumulodata.com>)

----------manual call done

#No '----------- manual parsing links', so `parse_links` is never called!

Versions

• Python 2.7.2
• Scrapy 0.14.4

Answer 1

Here’s a scraper that works perfectly:

from scrapy.contrib.spiders import CrawlSpider,Rule
from scrapy.selector import HtmlXPathSelector
from scrapy.http import Request
from scrapySpider.items import SPage
from scrapy.contrib.linkextractors.sgml import SgmlLinkExtractor

class TestSpider4(CrawlSpider):
    name = "spiderSO"
    allowed_domains = ["cumulodata.com"]
    start_urls = ["http://www.cumulodata.com/"]

    extractor = SgmlLinkExtractor()

    rules = (
        Rule(extractor,callback='parse_links',follow=True),
        )

    def parse_start_url(self, response):
        list(self.parse_links(response))

    def parse_links(self, response):
        hxs = HtmlXPathSelector(response)
        links = hxs.select('//a')
        for link in links:
            title = ''.join(link.select('./@title').extract())
            url = ''.join(link.select('./@href').extract())
            meta={'title':title,}
            cleaned_url = "%s/?1" % url if not '/' in url.partition('//')[2] else "%s?1" % url
            yield Request(cleaned_url, callback = self.parse_page, meta=meta,)

    def parse_page(self, response):
        hxs = HtmlXPathSelector(response)
        item=SPage()
        item['url'] = response.url
        item['title']=response.meta['title']
        item['h1']=hxs.select('//h1/text()').extract()
        return item

Changes:

1. Implemented parse_start_url – Unfortunately, when you specify a callback for the first request, rules are not executed. This is inbuilt into Scrapy, and we can only manage this with a workaround. So we do a list(self.parse_links(response)) inside this function. Why the list()? Because parse_links is a generator, and generators are lazy. So we need to explicitly call it fully.

2. cleaned_url = "%s/?1" % url if not '/' in url.partition('//')[2] else "%s?1" % url – There are a couple of things going on here:

   a. We’re adding ‘/?1’ to the end of the URL – Since parse_links returns duplicate URLs, Scrapy filters them out. An easier way to avoid that is to pass dont_filter=True to Request(). However, all your pages are interlinked (back to index from pageAA, etc.) and a dont_filter here results in too many duplicate requests & items.

   b. if not '/' in url.partition('//')[2] – Again, this is because of the linking in your website. One of the internal links is to ‘www.cumulodata.com’ and another to ‘www.cumulodata.com/’. Since we’re explicitly adding a mechanism to allow duplicates, this was resulting in one extra item. Since we needed perfect, I implemented this hack.

3. title = ''.join(link.select('./@title').extract()) – You don’t want to return the node, but the data. Also: ”.join(list) is better than list[0] in case of an empty list.

Congrats on creating a test website which posed a curious problem – Duplicates are both necessary as well as unwanted!