Using beautiful soup to clean up scraped HTML from scrapy

Question

I'm using scrapy to try and scrape some data that I need off Google Scholar. Consider, as an example the following link: http://scholar.google.com/scholar?q=intitle%3Apython+xpath

Now, I'd like to scrape all the titles off this page. The process that I am following is as follows:

scrapy shell "http://scholar.google.com/scholar?q=intitle%3Apython+xpath"

which gives me the scrapy shell, inside which I do:

>>> sel.xpath('//h3[@class="gs_rt"]/a').extract()

[
 u'<a href="http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.122.4438&amp;rep=rep1&amp;type=pdf"><b>Python </b>Paradigms for XML</a>', 
 u'<a href="https://svn.eecs.jacobs-university.de/svn/eecs/archive/bsc-2009/sbhushan.pdf">NCClient: A <b>Python </b>Library for NETCONF Clients</a>', 
 u'<a href="http://hal.archives-ouvertes.fr/hal-00759589/">PALSE: <b>Python </b>Analysis of Large Scale (Computer) Experiments</a>', 
 u'<a href="http://i.iinfo.cz/r2/kd/xmlprague2007.pdf#page=53"><b>Python </b>and XML</a>', 
 u'<a href="http://www.loadaveragezero.com/app/drx/Programming/Languages/Python/">drx: <b>Python </b>Programming Language [Computers: Programming: Languages: <b>Python</b>]-loadaverageZero</a>', 
 u'<a href="http://www.worldcolleges.info/sites/default/files/py10.pdf">XML and <b>Python </b>Tutorial</a>', 
 u'<a href="http://dl.acm.org/citation.cfm?id=2555791">Zato\u2014agile ESB, SOA, REST and cloud integrations in <b>Python</b></a>', 
 u'<a href="ftp://ftp.sybex.com/4021/4021index.pdf">XML Processing with Perl, <b>Python</b>, and PHP</a>', 
 u'<a href="http://books.google.com/books?hl=en&amp;lr=&amp;id=El4TAgAAQBAJ&amp;oi=fnd&amp;pg=PT8&amp;dq=python+xpath&amp;ots=RrFv0f_Y6V&amp;sig=tSXzPJXbDi6KYnuuXEDnZCI7rDA"><b>Python </b>&amp; XML</a>', 
 u'<a href="https://code.grnet.gr/projects/ncclient/repository/revisions/efed7d4cd5ac60cbb7c1c38646a6d6dfb711acc9/raw/docs/proposal.pdf">A <b>Python </b>Module for NETCONF Clients</a>'
]

As you can see, this output is raw HTML that needs cleaning. I now have a good sense of how to clean this HTML up. The simplest way is probably to just BeautifulSoup and try something like:

t = sel.xpath('//h3[@class="gs_rt"]/a').extract()
soup = BeautifulSoup(t)
text_parts = soup.findAll(text=True)
text = ''.join(text_parts)

This is based off an earlier SO question. The regexp version has been suggested, but I am guessing that BeautifulSoup will be more robust.

I'm a scrapy n00b and can't figure out how to embed this in my spider. I tried

from scrapy.spider import Spider
from scrapy.selector import Selector
from bs4 import BeautifulSoup

from scholarscrape.items import ScholarscrapeItem

class ScholarSpider(Spider):
    name = "scholar"
    allowed_domains = ["scholar.google.com"]
    start_urls = [
        "http://scholar.google.com/scholar?q=intitle%3Apython+xpath"
    ]

    def parse(self, response):
        sel = Selector(response)
        item = ScholarscrapeItem()        
        t = sel.xpath('//h3[@class="gs_rt"]/a').extract()
        soup = BeautifulSoup(t)
        text_parts = soup.findAll(text=True)
        text = ''.join(text_parts)
        item['title'] = text
        return(item)

But that didn't quite work. Any suggestions would be helpful.

---

Edit 3: Based on suggestions, I have modified my spider file to:

from scrapy.spider import Spider
from scrapy.selector import Selector
from bs4 import BeautifulSoup

from scholarscrape.items import ScholarscrapeItem

class ScholarSpider(Spider):
    name = "dmoz"
    allowed_domains = ["sholar.google.com"]
    start_urls = [
        "http://scholar.google.com/scholar?q=intitle%3Anine+facts+about+top+journals+in+economics"
    ]

    def parse(self, response):
        sel = Selector(response)
        item = ScholarscrapeItem()        
        titles = sel.xpath('//h3[@class="gs_rt"]/a')

        for title in titles:
            title = item.xpath('.//text()').extract()
            print "".join(title)

However, I get the following output:

2014-02-17 15:11:12-0800 [scrapy] INFO: Scrapy 0.22.2 started (bot: scholarscrape)
2014-02-17 15:11:12-0800 [scrapy] INFO: Optional features available: ssl, http11
2014-02-17 15:11:12-0800 [scrapy] INFO: Overridden settings: {'NEWSPIDER_MODULE': 'scholarscrape.spiders', 'SPIDER_MODULES': ['scholarscrape.spiders'], 'BOT_NAME': 'scholarscrape'}
2014-02-17 15:11:12-0800 [scrapy] INFO: Enabled extensions: LogStats, TelnetConsole, CloseSpider, WebService, CoreStats, SpiderState
2014-02-17 15:11:13-0800 [scrapy] INFO: Enabled downloader middlewares: HttpAuthMiddleware, DownloadTimeoutMiddleware, UserAgentMiddleware, RetryMiddleware, DefaultHeadersMiddleware, MetaRefreshMiddleware, HttpCompressionMiddleware, RedirectMiddleware, CookiesMiddleware, ChunkedTransferMiddleware, DownloaderStats
2014-02-17 15:11:13-0800 [scrapy] INFO: Enabled spider middlewares: HttpErrorMiddleware, OffsiteMiddleware, RefererMiddleware, UrlLengthMiddleware, DepthMiddleware
2014-02-17 15:11:13-0800 [scrapy] INFO: Enabled item pipelines:
2014-02-17 15:11:13-0800 [dmoz] INFO: Spider opened
2014-02-17 15:11:13-0800 [dmoz] INFO: Crawled 0 pages (at 0 pages/min), scraped 0 items (at 0 items/min)
2014-02-17 15:11:13-0800 [scrapy] DEBUG: Telnet console listening on 0.0.0.0:6023
2014-02-17 15:11:13-0800 [scrapy] DEBUG: Web service listening on 0.0.0.0:6080
2014-02-17 15:11:13-0800 [dmoz] DEBUG: Crawled (200) <GET http://scholar.google.com/scholar?q=intitle%3Apython+xml> (referer: None)
2014-02-17 15:11:13-0800 [dmoz] ERROR: Spider error processing <GET http://scholar.google.com/scholar?q=intitle%3Apython+xml>
 Traceback (most recent call last):
   File "/System/Library/Frameworks/Python.framework/Versions/2.7/Extras/lib/python/twisted/internet/base.py", line 1178, in mainLoop
     self.runUntilCurrent()
   File "/System/Library/Frameworks/Python.framework/Versions/2.7/Extras/lib/python/twisted/internet/base.py", line 800, in runUntilCurrent
     call.func(*call.args, **call.kw)
   File "/System/Library/Frameworks/Python.framework/Versions/2.7/Extras/lib/python/twisted/internet/defer.py", line 368, in callback
     self._startRunCallbacks(result)
   File "/System/Library/Frameworks/Python.framework/Versions/2.7/Extras/lib/python/twisted/internet/defer.py", line 464, in _startRunCallbacks
     self._runCallbacks()
 --- <exception caught here> ---
   File "/System/Library/Frameworks/Python.framework/Versions/2.7/Extras/lib/python/twisted/internet/defer.py", line 551, in _runCallbacks
     current.result = callback(current.result, *args, **kw)
   File "/Users/krishnan/work/research/journals/code/scholarscrape/scholarscrape/spiders/scholar_spider.py", line 20, in parse
     title = item.xpath('.//text()').extract()
   File "/Library/Python/2.7/site-packages/scrapy/item.py", line 65, in __getattr__
     raise AttributeError(name)
 exceptions.AttributeError: xpath

2014-02-17 15:11:13-0800 [dmoz] INFO: Closing spider (finished)
2014-02-17 15:11:13-0800 [dmoz] INFO: Dumping Scrapy stats:
 {'downloader/request_bytes': 247,
  'downloader/request_count': 1,
  'downloader/request_method_count/GET': 1,
  'downloader/response_bytes': 108851,
  'downloader/response_count': 1,
  'downloader/response_status_count/200': 1,
  'finish_reason': 'finished',
  'finish_time': datetime.datetime(2014, 2, 17, 23, 11, 13, 196648),
  'log_count/DEBUG': 3,
  'log_count/ERROR': 1,
  'log_count/INFO': 7,
  'response_received_count': 1,
  'scheduler/dequeued': 1,
  'scheduler/dequeued/memory': 1,
  'scheduler/enqueued': 1,
  'scheduler/enqueued/memory': 1,
  'spider_exceptions/AttributeError': 1,
  'start_time': datetime.datetime(2014, 2, 17, 23, 11, 13, 21701)}
2014-02-17 15:11:13-0800 [dmoz] INFO: Spider closed (finished)

---

Edit 2: My original question was quite different, but I am now convinced that this is the right way to proceed. Original question (and first edit below):

I'm using scrapy to try and scrape some data that I need off Google Scholar. Consider, as an example the following link:

http://scholar.google.com/scholar?q=intitle%3Apython+xpath

Now, I'd like to scrape all the titles off this page. The process that I am following is as follows:

scrapy shell "http://scholar.google.com/scholar?q=intitle%3Apython+xpath"

which gives me the scrapy shell, inside which I do:

>>> sel.xpath('string(//h3[@class="gs_rt"]/a)').extract()
[u'Python Paradigms for XML']

As you can see, this only selects the first title, and none of the others on the page. I can't figure out what I should modify my XPath to, so that I select all such elements on the page. Any help is greatly appreciated.

---

Edit 1: My first approach was to try

>>> sel.xpath('//h3[@class="gs_rt"]/a/text()').extract()
[u'Paradigms for XML', u'NCClient: A ', u'Library for NETCONF Clients', 
 u'PALSE: ', u'Analysis of Large Scale (Computer) Experiments', u'and XML', 
 u'drx: ', u'Programming Language [Computers: Programming: Languages: ',
 u']-loadaverageZero', u'XML and ', u'Tutorial', 
 u'Zato\u2014agile ESB, SOA, REST and cloud integrations in ', 
 u'XML Processing with Perl, ', u', and PHP', u'& XML', u'A ', 
 u'Module for NETCONF Clients']

The problem with this is approach is that if you look at the actual Google Scholar page, you will see that the first entry is actually 'Python Paradigms for XML' and not 'Paradigms for XML' as scrapy returns. My guess for this behaviour is that 'Python' is trapped inside tags which is why text() is not doing what we want him to do.

Answer 1

The first entry is actually 'Python Paradigms for XML' and not 'Paradigms for XML' as Scrapy returns.

You need to use normalize-space() that will return blank nodes as well because text() will ignore blank text nodes. So your initial XPath would look like this:

sel.xpath('//h3[@class="gs_rt"]/a').xpath("normalize-space()").extract()

Example:

# HTML has been simplified

from parsel import Selector

html = '''
<a href="https://erdincuzun.com/wp-content/uploads/download/plovdiv_2018_01.pdf"><span class=gs_ctg2>[PDF]</span> erdincuzun.com</a>
<div class="gs_ri">
    <h3 class="gs_rt">
        <span class="gs_ctc"><span class="gs_ct1">[PDF]</span><span class="gs_ct2">[PDF]</span></span>
        <a href="https://erdincuzun.com/wp-content/uploads/download/plovdiv_2018_01.pdf">Comparison of <b>Python </b>libraries used for Web
            data extraction</a>
'''

selector = Selector(HTML)

# get() to get textual data
print("Without normalize-space:\n", selector.xpath('//*[@class="gs_rt"]/a/text()').get())
print("\nWith normalize-space:\n", selector.xpath('//*[@class="gs_rt"]/a').xpath("normalize-space()").get())


"""
Without normalize-space:
 Comparison
        of 

With normalize-space:
 Comparison of Python libraries used for Web data extraction
"""

---

Actual code and example in the online IDE to get titles from Google Scholar Organic results:

import scrapy

class ScholarSpider(scrapy.Spider):
    name = "scholar_titles"
    allowed_domains = ["scholar.google.com"]
    start_urls = ["https://scholar.google.com/scholar?q=intitle%3Apython+xpath"]

    def parse(self, response):
        for quote in response.xpath('//*[@class="gs_rt"]/a'):
            yield {
                "title": quote.xpath("normalize-space()").get()
            }

Run it:

$ scrapy runspider -O <file_name>.jl <file_name>.py 

• -O stands for override. -o for appending.
• jl is a JSON line file format.

Output with normalize-space():

{"title": "Comparison of Python libraries used for Web data extraction"}
{"title": "Approaching the largest 'API': extracting information from the internet with python"}
{"title": "News crawling based on Python crawler"}
{"title": "A survey on python libraries used for social media content scraping"}
{"title": "Design and Implementation of Crawler Program Based on Python"}
{"title": "DECEPTIVE SECURITY USING PYTHON"}
{"title": "Hands-On Web Scraping with Python: Perform advanced scraping operations using various Python libraries and tools such as Selenium, Regex, and others"}
{"title": "Python Paradigms for XML"}
{"title": "Using Web Scraping In A Knowledge Environment To Build Ontologies Using Python And Scrapy"}
{"title": "XML processing with Python"}

Output without normalize-space():

{"title": "Comparison of "}
{"title": "libraries used for Web data extraction"}
{"title": "Approaching the largest 'API': extracting information from the internet with "}
{"title": "News crawling based on "}
{"title": "crawler"}
{"title": "A survey on "}
{"title": "libraries used for social media content scraping"}
{"title": "Design and Implementation of Crawler Program Based on "}
{"title": "DECEPTIVE SECURITY USING "}
{"title": "Hands-On Web Scraping with "}
{"title": ": Perform advanced scraping operations using various "}
{"title": "libraries and tools such as Selenium, Regex, and others"}
{"title": "Paradigms for XML"}
{"title": "Using Web Scraping In A Knowledge Environment To Build Ontologies Using "}
{"title": "And Scrapy"}
{"title": "XML processing with "}

---

Alternatively, you can achieve it with Google Scholar Organic Results API from SerpApi.

It's a paid API with a free plan. You don't have to figure out the extraction part and maintain it, how to scale it and how to bypass blocks from search engines since it's already done for the end-user.

Example code to integrate:

from serpapi import GoogleScholarSearch

params = {
  "api_key": "Your SerpApi API key",  # API Key
  "engine": "google_scholar",         # parsing engine
  "q": "intitle:python XPath",        # search query
  "hl": "en"                          # language
}

search = GoogleScholarSearch(params)  # where extraction happens on the SerpApi back-end
results = search.get_dict()           # JSON -> Python dictionary

for result in results["organic_results"]:
    title = result["title"]
    print(title)

Output:

Comparison of Python libraries used for Web data extraction
Approaching the largest 'API': extracting information from the internet with python
News crawling based on Python crawler
A survey on python libraries used for social media content scraping
Design and Implementation of Crawler Program Based on Python
DECEPTIVE SECURITY USING PYTHON
Hands-On Web Scraping with Python: Perform advanced scraping operations using various Python libraries and tools such as Selenium, Regex, and others
Python Paradigms for XML
Using Web Scraping In A Knowledge Environment To Build Ontologies Using Python And Scrapy
XML processing with Python

Disclaimer, I work for SerpApi.

Answer 2

This is a really interesting and rather difficult question. The problem you're facing concerns the fact that "Python" in the title is in bold, and it is treated as node, while the rest of the title is simply a text, therefore text() extracts only textual content and not content of <b> node.

Here's my solution. First get all the links:

titles = sel.xpath('//h3[@class="gs_rt"]/a')

then iterate over them and select all textual content of each node, in other words join <b> node with text node for each children of this link

for item in titles:
    title = item.xpath('.//text()').extract()
    print "".join(title)

This works because in a for loop you will be dealing with textual content of children of each link and thus you will be able to join matching elements. Title in the loop will be equal for instance :[u'Python ', u'Paradigms for XML'] or [u'NCClient: A ', u'Python ', u'Library for NETCONF Clients']

Answer 3

The XPath string() function only returns the string representation of the first node you pass to it.

Just extract nodes normally, don't use string().

sel.xpath('//h3[@class="gs_rt"]/a').extract()

or

sel.xpath('//h3[@class="gs_rt"]/a/text()').extract()