Scrapy shell

The Scrapy shell is an interactive shell where you can try and debug your scraping code very quickly, without having to run the spider. It’s meant to be used for testing data extraction code, but you can actually use it for testing any kind of code as it is also a regular Python shell.

The shell is used for testing XPath or CSS expressions and see how they work and what data they extract from the web pages you’re trying to scrape. It allows you to interactively test your expressions while you’re writing your spider, without having to run the spider to test every change.

Once you get familiarized with the Scrapy shell, you’ll see that it’s an invaluable tool for developing and debugging your spiders.

If you have IPython installed, the Scrapy shell will use it (instead of the standard Python console). The IPython console is much more powerful and provides smart auto-completion and colorized output, among other things.

We highly recommend you install IPython, specially if you’re working on Unix systems (where IPython excels). See the IPython installation guide for more info.

Launch the shell

To launch the Scrapy shell you can use the shell command like this:

scrapy shell <url>

Where the <url> is the URL you want to scrape.

Using the shell

The Scrapy shell is just a regular Python console (or IPython console if you have it available) which provides some additional shortcut functions for convenience.

Available Shortcuts

  • shelp() - print a help with the list of available objects and shortcuts
  • fetch(request_or_url) - fetch a new response from the given request or URL and update all related objects accordingly.
  • view(response) - open the given response in your local web browser, for inspection. This will add a <base> tag to the response body in order for external links (such as images and style sheets) to display properly. Note, however,that this will create a temporary file in your computer, which won’t be removed automatically.

Available Scrapy objects

The Scrapy shell automatically creates some convenient objects from the downloaded page, like the Response object and the Selector objects (for both HTML and XML content).

Those objects are:

  • spider - the Spider which is known to handle the URL, or a BaseSpider object if there is no spider found for the current URL
  • request - a Request object of the last fetched page. You can modify this request using replace() or fetch a new request (without leaving the shell) using the fetch shortcut.
  • response - a Response object containing the last fetched page
  • sel - a Selector object constructed with the last response fetched
  • settings - the current Scrapy settings

Example of shell session

Here’s an example of a typical shell session where we start by scraping the page, and then proceed to scrape the page. Finally, we modify the (Slashdot) request method to POST and re-fetch it getting a HTTP 405 (method not allowed) error. We end the session by typing Ctrl-D (in Unix systems) or Ctrl-Z in Windows.

Keep in mind that the data extracted here may not be the same when you try it, as those pages are not static and could have changed by the time you test this. The only purpose of this example is to get you familiarized with how the Scrapy shell works.

First, we launch the shell:

scrapy shell --nolog

Then, the shell fetches the URL (using the Scrapy downloader) and prints the list of available objects and useful shortcuts (you’ll notice that these lines all start with the [s] prefix):

[s] Available objects
[s]   sel       <Selector ( xpath=None>
[s]   item      Item()
[s]   request   <>
[s]   response  <>
[s]   settings  <Settings 'mybot.settings'>
[s]   spider    <scrapy.spider.models.BaseSpider object at 0x2bed9d0>
[s] Useful shortcuts:
[s]   shelp()           Prints this help.
[s]   fetch(req_or_url) Fetch a new request or URL and update objects
[s]   view(response)    View response in a browser


After that, we can star playing with the objects:

>>> sel.xpath("//h2/text()").extract()[0]
u'Welcome to Scrapy'

>>> fetch("")
[s] Available Scrapy objects:
[s]   sel        <Selector ( xpath=None>
[s]   item       JobItem()
[s]   request    <GET>
[s]   response   <200>
[s]   settings   <Settings 'jobsbot.settings'>
[s]   spider     <BaseSpider 'default' at 0x3c44a10>
[s] Useful shortcuts:
[s]   shelp()           Shell help (print this help)
[s]   fetch(req_or_url) Fetch request (or URL) and update local objects
[s]   view(response)    View response in a browser

>>> sel.xpath("//h2/text()").extract()
[u'News for nerds, stuff that matters']

>>> request = request.replace(method="POST")

>>> fetch(request)
2009-04-03 00:57:39-0300 [default] ERROR: Downloading <> from <None>: 405 Method Not Allowed


Invoking the shell from spiders to inspect responses

Sometimes you want to inspect the responses that are being processed in a certain point of your spider, if only to check that response you expect is getting there.

This can be achieved by using the function.

Here’s an example of how you would call it from your spider:

class MySpider(BaseSpider):

    def parse(self, response):
        if response.url == '':
            from import inspect_response

        # ... your parsing code ..

When you run the spider, you will get something similar to this:

2009-08-27 19:15:25-0300 [] DEBUG: Crawled <> (referer: <None>)
2009-08-27 19:15:26-0300 [] DEBUG: Crawled <> (referer: <>)
[s] Available objects
[s]   sel       <Selector ( xpath=None>

>>> response.url

Then, you can check if the extraction code is working:

>>> sel.xpath('//h1')

Nope, it doesn’t. So you can open the response in your web browser and see if it’s the response you were expecting:

>>> view(response)

Finally you hit Ctrl-D (or Ctrl-Z in Windows) to exit the shell and resume the crawling:

>>> ^D
2009-08-27 19:15:25-0300 [] DEBUG: Crawled <> (referer: <None>)
2009-08-27 19:15:25-0300 [] DEBUG: Crawled <> (referer: <None>)
# ...

Note that you can’t use the fetch shortcut here since the Scrapy engine is blocked by the shell. However, after you leave the shell, the spider will continue crawling where it stopped, as shown above.