Scrapy can't find XPath content - xpath

Question

I'm writing a web crawler with Scrapy to download the text of talk-backs on a certain webpage.
Here is the relevant part of the code behind the webpage, for a specific talkback:
<div id="site_comment_71339" class="site_comment site_comment-even large high-rank">
<div class="talkback-topic">
<a class="show-comment" data-ajax-url="/comments/71339.js?counter=97&num=57" href="/comments/71339?counter=97&num=57">57. talk back title here </a>
</div>
<div class="talkback-message"> blah blah blah talk-back message here </div>
....etc etc etc ......
While writing an XPath to get the the message:
titles = hxs.xpath("//div[#class='site_comment site_comment-even large high-rank']")
and later on:
item["title"] = titles.xpath("div[#class='talkback-message']text()").extract()
There's no bug, but it doesn't work. Any ideas why? I suppose I'm not writing the path correctly, but I can't find the error.
Thank you :)
The whole code:
from scrapy.spider import BaseSpider
from scrapy.selector import Selector
from craigslist_sample.items import CraigslistSampleItem
class MySpider(BaseSpider):
name = "craig"
allowed_domains = ["tbk.co.il"]
start_urls = ["http://www.tbk.co.il/tag/%D7%91%D7%A0%D7%99%D7%9E%D7%99%D7%9F_%D7%A0%D7%AA%D7%A0%D7%99%D7%94%D7%95/talkbacks"]
def parse(self, response):
hxs = Selector(response)
titles = hxs.xpath("//div[#class='site_comment site_comment-even large high-rank']")
items=[]
for titles in titles:
item = CraigslistSampleItem()
item["title"] = titles.xpath("div[#class='talkback-message']text()").extract()
items.append(item)
return items

Answer 1

Here's a snippet of the HTML page for #site_comment_74240
<div class="site_comment site_comment-even small normal-rank" id="site_comment_74240">
<div class="talkback-topic">
144. מדיניות
</div>
<div class="talkback-username">
<table><tr>
<td>קייזרמן פרדי </td>
<td>(01.11.2013)</td>
</tr></table>
</div>
The "talkback-message" div is not in the HTML page when you first fetch it, but rather is fetched asynchronously via some AJAX query when you click on a comment title, so you'll have to fetch it for each comment.
Comment blocks, titles in you code snipper, can be grabbed using an XPath like this: //div[starts-with(#id, "site_comment_"]), i.e. all divs that have an "id" attribute beginning with string ""site_comment_"
You can also use CSS selectors with Selector.css(). In your case, you can grab comment blocks using either the "id" approach (as I've done above using XPath), so:
titles = sel.css("div[id^=site_comment_]")
or using the "site_comment" class without the other "site_comment-even", "site_comment-odd", "small", "normal-rank" or "high-rank" that vary:
titles = sel.css("div.site_comment")
Then you would issue a new Request using the URL that's in ./div[#class="talkback-topic"]/a[#class="show-comment"]/#data-ajax-url inside that comment div. Or using CSS selectors, div.talkback-topic > a.show-comment::attr(data-ajax-url) (by the way, the ::attr(...) is not standard, but is a Scrapy extension to CSS selectors using pseudo elements functions)
What you get from the AJAX call is some Javascript code, and you want to grab the content inside old.after(...)
var old = $("#site_comment_72765");
old.attr('id', old.attr('id') + '_small');
old.hide();
old.after("\n<div class=\"site_comment site_comment-odd large high-rank\" id=\"site_comment_72765\">\n <div class=\"talkback-topic\">\n <a href=\"/comments/72765?counter=42&num=109\" class=\"show-comment\" data-ajax-url=\"/comments/72765.js?counter=42&num=109\">109. ביבי - האדם הנכון בראש ממשלת ישראל(לת)<\/a>\n <\/div>\n \n <div class=\"talkback-message\">\n \n <\/div>\n \n <div class=\"talkback-username\">\n <table><tr>\n <td>ישראל <\/td>\n <td>(11.03.2012)<\/td>\n <\/tr><\/table>\n <\/div>\n <div class=\"rank-controllers\">\n <table><tr>\n \n <td class=\"rabk-link\"><a href=\"#\" data-thumb=\"/comments/72765/thumb?type=up\"><img alt=\"\" src=\"/images/elements/thumbU.png?1376839523\" /><\/a><\/td>\n <td> | <\/td>\n <td class=\"rabk-link\"><a href=\"#\" data-thumb=\"/comments/72765/thumb?type=down\"><img alt=\"\" src=\"/images/elements/thumbD.png?1376839523\" /><\/a><\/td>\n \n <td> | <\/td>\n <td>11<\/td>\n \n <\/tr><\/table>\n <\/div>\n \n <div class=\"talkback-links\">\n <a href=\"/comments/new?add_to_root=true&html_id=site_comment_72765&sibling_id=72765\">תגובה חדשה<\/a>\n \n <a href=\"/comments/72765/comments/new?html_id=site_comment_72765\">הגיבו לתגובה<\/a>\n \n <a href=\"/i/offensive?comment_id=72765\" data-noajax=\"true\">דיווח תוכן פוגעני<\/a>\n <\/div>\n \n<\/div>");
var new_comment = $("#site_comment_72765");
This is HTML data that you'll need to parse again using something Selector(text=this_ajax_html_data) and a .//div[#class="talkback-message"]//text() XPath or div.talkback-message ::text CSS selector
Here's a skeleton spider to get you going with these ideas:
from scrapy.spider import BaseSpider
from scrapy.selector import Selector
from scrapy.http import Request
from craigslist_sample.items import CraigslistSampleItem
import urlparse
import re
class MySpider(BaseSpider):
name = "craig"
allowed_domains = ["tbk.co.il"]
start_urls = ["http://www.tbk.co.il/tag/%D7%91%D7%A0%D7%99%D7%9E%D7%99%D7%9F_%D7%A0%D7%AA%D7%A0%D7%99%D7%94%D7%95/talkbacks"]
def parse(self, response):
sel = Selector(response)
comments = sel.css("div.site_comment")
for comment in comments:
item = CraigslistSampleItem()
# this probably has to be fixed
#item["title"] = comment.xpath("div[#class='talkback-message']text()").extract()
# issue an additional request to fetch the Javascript
# data containing the comment text
# and pass the incomplete item via meta dict
for url in comment.css('div.talkback-topic > a.show-comment::attr(data-ajax-url)').extract():
yield Request(url=urlparse.urljoin(response.url, url),
callback=self.parse_javascript_comment,
meta={"item": item})
break
# the line we are looking for begins with "old.after"
# and we want everythin inside the parentheses
_re_comment_html = re.compile(r'^old\.after\((?P<html>.+)\);$')
def parse_javascript_comment(self, response):
item = response.meta["item"]
# loop on Javascript content lines
for line in response.body.split("\n"):
matching = self._re_comment_html.search(line.strip())
if matching:
# what's inside the parentheses is a Javascript strings
# with escaped double-quotes
# a simple way to decode that into a Python string
# is to use eval()
# then there are these "<\/tag>" we want to remove
html = eval(matching.group("html")).replace(r"<\/", "</")
# once we have the HTML snippet, decode it using Selector()
decoded = Selector(text=html, type="html")
# and save the message text in the item
item["message"] = u''.join(decoded.css('div.talkback-message ::text').extract()).strip()
# and return it
return item
You can try it out using scrapy runspider tbkspider.py.