When you’re scraping web pages, the most common task you need to perform is to extract data from the HTML source. There are several libraries available to achieve this:

  • BeautifulSoup is a very popular web scraping library among Python programmers which constructs a Python object based on the structure of the HTML code and also deals with bad markup reasonably well, but it has one drawback: it’s slow.
  • lxml是一个XML解析库(也可以解析HTML),具有基于ElementTree的Python化的API。(lxml is not part of the Python standard library.)

Scrapy comes with its own mechanism for extracting data. 它们被称作选择器(seletors),因为他们通过特定的 XPath或者 CSS表达式来“选择” HTML文件中的某个部分。

XPath是一门用来在XML文件中选择节点的语言,也可以用在HTML上。CSS是一门将HTML文档样式化的语言。It defines selectors to associate those styles with specific HTML elements.

Scrapy selectors are built over the lxml library, which means they’re very similar in speed and parsing accuracy.

本页面解释了选择器如何工作,并描述了相应的API,不同于 lxml API的臃肿,该API短小而简洁,这是因为 lxml库除了用来选择标记化文档外,还可以用到许多任务上。

For a complete reference of the selectors API see Selector reference




>>> from scrapy.selector import Selector
>>> from scrapy.http import HtmlResponse


>>> body = '<html><body><span>good</span></body></html>'
>>> Selector(text=body).xpath('//span/text()').extract()


>>> response = HtmlResponse(url='', body=body)
>>> Selector(response=response).xpath('//span/text()').extract()

For convenience, response objects expose a selector on .selector attribute, it’s totally OK to use this shortcut when possible:

>>> response.selector.xpath('//span/text()').extract()

Using selectors

To explain how to use the selectors we’ll use the Scrapy shell (which provides interactive testing) and an example page located in the Scrapy documentation server:

Here’s its HTML code:

  <base href='' />
  <title>Example website</title>
  <div id='images'>
   <a href='image1.html'>Name: My image 1 <br /><img src='image1_thumb.jpg' /></a>
   <a href='image2.html'>Name: My image 2 <br /><img src='image2_thumb.jpg' /></a>
   <a href='image3.html'>Name: My image 3 <br /><img src='image3_thumb.jpg' /></a>
   <a href='image4.html'>Name: My image 4 <br /><img src='image4_thumb.jpg' /></a>
   <a href='image5.html'>Name: My image 5 <br /><img src='image5_thumb.jpg' /></a>

First, let’s open the shell:

scrapy shell

Then, after the shell loads, you’ll have the response available as response shell variable, and its attached selector in response.selector attribute.

Since we’re dealing with HTML, the selector will automatically use an HTML parser.

So, by looking at the HTML code of that page, let’s construct an XPath for selecting the text inside the title tag:

>>> response.selector.xpath('//title/text()')
[<Selector (text) xpath=//title/text()>]

Querying responses using XPath and CSS is so common that responses include two convenience shortcuts: response.xpath() and response.css():

>>> response.xpath('//title/text()')
[<Selector (text) xpath=//title/text()>]
>>> response.css('title::text')
[<Selector (text) xpath=//title/text()>]

As you can see, .xpath() and .css() methods return a SelectorList instance, which is a list of new selectors. This API can be used for quickly selecting nested data:

>>> response.css('img').xpath('@src').extract()

To actually extract the textual data, you must call the selector .extract() method, as follows:

>>> response.xpath('//title/text()').extract()
[u'Example website']

如果你只想要提取第一个匹配的元素,你可以调用选择器 .extract_first()

>>> response.xpath('//div[@id="images"]/a/text()').extract_first()
u'Name: My image 1 '

如果找不到元素,它返回 None:

>>> response.xpath('//div[@id="not-exists"]/text()').extract_first() is None


>>> response.xpath('//div[@id="not-exists"]/text()').extract_first(default='not-found')

Notice that CSS selectors can select text or attribute nodes using CSS3 pseudo-elements:

>>> response.css('title::text').extract()
[u'Example website']

Now we’re going to get the base URL and some image links:

>>> response.xpath('//base/@href').extract()

>>> response.css('base::attr(href)').extract()

>>> response.xpath('//a[contains(@href, "image")]/@href').extract()

>>> response.css('a[href*=image]::attr(href)').extract()

>>> response.xpath('//a[contains(@href, "image")]/img/@src').extract()

>>> response.css('a[href*=image] img::attr(src)').extract()


The selection methods (.xpath() or .css()) return a list of selectors of the same type, so you can call the selection methods for those selectors too. Here’s an example:

>>> links = response.xpath('//a[contains(@href, "image")]')
>>> links.extract()
[u'<a href="image1.html">Name: My image 1 <br><img src="image1_thumb.jpg"></a>',
 u'<a href="image2.html">Name: My image 2 <br><img src="image2_thumb.jpg"></a>',
 u'<a href="image3.html">Name: My image 3 <br><img src="image3_thumb.jpg"></a>',
 u'<a href="image4.html">Name: My image 4 <br><img src="image4_thumb.jpg"></a>',
 u'<a href="image5.html">Name: My image 5 <br><img src="image5_thumb.jpg"></a>']

>>> for index, link in enumerate(links):
...     args = (index, link.xpath('@href').extract(), link.xpath('img/@src').extract())
...     print 'Link number %d points to url %s and image %s' % args

Link number 0 points to url [u'image1.html'] and image [u'image1_thumb.jpg']
Link number 1 points to url [u'image2.html'] and image [u'image2_thumb.jpg']
Link number 2 points to url [u'image3.html'] and image [u'image3_thumb.jpg']
Link number 3 points to url [u'image4.html'] and image [u'image4_thumb.jpg']
Link number 4 points to url [u'image5.html'] and image [u'image5_thumb.jpg']


Selector also has a .re() method for extracting data using regular expressions. However, unlike using .xpath() or .css() methods, .re() returns a list of unicode strings. So you can’t construct nested .re() calls.

Here’s an example used to extract image names from the HTML code above:

>>> response.xpath('//a[contains(@href, "image")]/text()').re(r'Name:\s*(.*)')
[u'My image 1',
 u'My image 2',
 u'My image 3',
 u'My image 4',
 u'My image 5']

There’s an additional helper reciprocating .extract_first() for .re(), named .re_first(). Use it to extract just the first matching string:

>>> response.xpath('//a[contains(@href, "image")]/text()').re_first(r'Name:\s*(.*)')
u'My image 1'


Keep in mind that if you are nesting selectors and use an XPath that starts with /, that XPath will be absolute to the document and not relative to the Selector you’re calling it from.

For example, suppose you want to extract all <p> elements inside <div> elements. First, you would get all <div> elements:

>>> divs = response.xpath('//div')

At first, you may be tempted to use the following approach, which is wrong, as it actually extracts all <p> elements from the document, not only those inside <div> elements:

>>> for p in divs.xpath('//p'):  # this is wrong - gets all <p> from the whole document
...     print p.extract()

This is the proper way to do it (note the dot prefixing the .//p XPath):

>>> for p in divs.xpath('.//p'):  # extracts all <p> inside
...     print p.extract()

Another common case would be to extract all direct <p> children:

>>> for p in divs.xpath('p'):
...     print p.extract()

更多关于相对XPaths的细节详见XPath说明中的Location Paths部分。


因建于lxml之上, Scrapy选择器也支持一些 EXSLT扩展,可以在XPath表达式中使用这些预先制定的命名空间:



The test() function, for example, can prove quite useful when XPath’s starts-with() or contains() are not sufficient.

Example selecting links in list item with a “class” attribute ending with a digit:

>>> from scrapy import Selector
>>> doc = """
... <div>
...     <ul>
...         <li class="item-0"><a href="link1.html">first item</a></li>
...         <li class="item-1"><a href="link2.html">second item</a></li>
...         <li class="item-inactive"><a href="link3.html">third item</a></li>
...         <li class="item-1"><a href="link4.html">fourth item</a></li>
...         <li class="item-0"><a href="link5.html">fifth item</a></li>
...     </ul>
... </div>
... """
>>> sel = Selector(text=doc, type="html")
>>> sel.xpath('//li//@href').extract()
[u'link1.html', u'link2.html', u'link3.html', u'link4.html', u'link5.html']
>>> sel.xpath('//li[re:test(@class, "item-\d$")]//@href').extract()
[u'link1.html', u'link2.html', u'link4.html', u'link5.html']


C library libxslt doesn’t natively support EXSLT regular expressions so lxml‘s implementation uses hooks to Python’s re module. Thus, using regexp functions in your XPath expressions may add a small performance penalty.


These can be handy for excluding parts of a document tree before extracting text elements for example.


>>> doc = """
... <div itemscope itemtype="">
...   <span itemprop="name">Kenmore White 17" Microwave</span>
...   <img src="kenmore-microwave-17in.jpg" alt='Kenmore 17" Microwave' />
...   <div itemprop="aggregateRating"
...     itemscope itemtype="">
...    Rated <span itemprop="ratingValue">3.5</span>/5
...    based on <span itemprop="reviewCount">11</span> customer reviews
...   </div>
...   <div itemprop="offers" itemscope itemtype="">
...     <span itemprop="price">$55.00</span>
...     <link itemprop="availability" href="" />In stock
...   </div>
...   Product description:
...   <span itemprop="description">0.7 cubic feet countertop microwave.
...   Has six preset cooking categories and convenience features like
...   Add-A-Minute and Child Lock.</span>
...   Customer reviews:
...   <div itemprop="review" itemscope itemtype="">
...     <span itemprop="name">Not a happy camper</span> -
...     by <span itemprop="author">Ellie</span>,
...     <meta itemprop="datePublished" content="2011-04-01">April 1, 2011
...     <div itemprop="reviewRating" itemscope itemtype="">
...       <meta itemprop="worstRating" content = "1">
...       <span itemprop="ratingValue">1</span>/
...       <span itemprop="bestRating">5</span>stars
...     </div>
...     <span itemprop="description">The lamp burned out and now I have to replace
...     it. </span>
...   </div>
...   <div itemprop="review" itemscope itemtype="">
...     <span itemprop="name">Value purchase</span> -
...     by <span itemprop="author">Lucas</span>,
...     <meta itemprop="datePublished" content="2011-03-25">March 25, 2011
...     <div itemprop="reviewRating" itemscope itemtype="">
...       <meta itemprop="worstRating" content = "1"/>
...       <span itemprop="ratingValue">4</span>/
...       <span itemprop="bestRating">5</span>stars
...     </div>
...     <span itemprop="description">Great microwave for the price. It is small and
...     fits in my apartment.</span>
...   </div>
...   ...
... </div>
... """
>>> sel = Selector(text=doc, type="html")
>>> for scope in sel.xpath('//div[@itemscope]'):
...     print "current scope:", scope.xpath('@itemtype').extract()
...     props = scope.xpath('''
...                 set:difference(./descendant::*/@itemprop,
...                                .//*[@itemscope]/*/@itemprop)''')
...     print "    properties:", props.extract()
...     print

current scope: [u'']
    properties: [u'name', u'aggregateRating', u'offers', u'description', u'review', u'review']

current scope: [u'']
    properties: [u'ratingValue', u'reviewCount']

current scope: [u'']
    properties: [u'price', u'availability']

current scope: [u'']
    properties: [u'name', u'author', u'datePublished', u'reviewRating', u'description']

current scope: [u'']
    properties: [u'worstRating', u'ratingValue', u'bestRating']

current scope: [u'']
    properties: [u'name', u'author', u'datePublished', u'reviewRating', u'description']

current scope: [u'']
    properties: [u'worstRating', u'ratingValue', u'bestRating']


Here we first iterate over itemscope elements, and for each one, we look for all itemprops elements and exclude those that are themselves inside another itemscope.


根据 ScrapingHub的博客,您可能会发现使用XPath与Scrapy选择器时有用的一些提示。如果您对XPath不太熟悉,可以先看看这个XPath教程

Using text nodes in a condition

When you need to use the text content as argument to an XPath string function, avoid using .//text() and use just . instead.

This is because the expression .//text() yields a collection of text elements – a node-set. And when a node-set is converted to a string, which happens when it is passed as argument to a string function like contains() or starts-with(), it results in the text for the first element only.


>>> from scrapy import Selector
>>> sel = Selector(text='<a href="#">Click here to go to the <strong>Next Page</strong></a>')

Converting a node-set to string:

>>> sel.xpath('//a//text()').extract() # take a peek at the node-set
[u'Click here to go to the ', u'Next Page']
>>> sel.xpath("string(//a[1]//text())").extract() # convert it to string
[u'Click here to go to the ']

A node converted to a string, however, puts together the text of itself plus of all its descendants:

>>> sel.xpath("//a[1]").extract() # select the first node
[u'<a href="#">Click here to go to the <strong>Next Page</strong></a>']
>>> sel.xpath("string(//a[1])").extract() # convert it to string
[u'Click here to go to the Next Page']

So, using the .//text() node-set won’t select anything in this case:

>>> sel.xpath("//a[contains(.//text(), 'Next Page')]").extract()

But using the . to mean the node, works:

>>> sel.xpath("//a[contains(., 'Next Page')]").extract()
[u'<a href="#">Click here to go to the <strong>Next Page</strong></a>']

Beware of the difference between //node[1] and (//node)[1]

//node[1] selects all the nodes occurring first under their respective parents.

(//node)[1] selects all the nodes in the document, and then gets only the first of them.


>>> from scrapy import Selector
>>> sel = Selector(text="""
....:     <ul class="list">
....:         <li>1</li>
....:         <li>2</li>
....:         <li>3</li>
....:     </ul>
....:     <ul class="list">
....:         <li>4</li>
....:         <li>5</li>
....:         <li>6</li>
....:     </ul>""")
>>> xp = lambda x: sel.xpath(x).extract()

This gets all first <li> elements under whatever it is its parent:

>>> xp("//li[1]")
[u'<li>1</li>', u'<li>4</li>']

And this gets the first <li> element in the whole document:

>>> xp("(//li)[1]")

This gets all first <li> elements under an <ul> parent:

>>> xp("//ul/li[1]")
[u'<li>1</li>', u'<li>4</li>']

And this gets the first <li> element under an <ul> parent in the whole document:

>>> xp("(//ul/li)[1]")

When querying by class, consider using CSS

因为元素可以包含多个 CSS 类,由类选择元素的 XPath 方法是相当冗长的︰

*[contains(concat(' ', normalize-space(@class), ' '), ' someclass ')]

If you use @class='someclass' you may end up missing elements that have other classes, and if you just use contains(@class, 'someclass') to make up for that you may end up with more elements that you want, if they have a different class name that shares the string someclass.

As it turns out, Scrapy selectors allow you to chain selectors, so most of the time you can just select by class using CSS and then switch to XPath when needed:

>>> from scrapy import Selector
>>> sel = Selector(text='<div class="hero shout"><time datetime="2014-07-23 19:00">Special date</time></div>')
>>> sel.css('.shout').xpath('./time/@datetime').extract()
[u'2014-07-23 19:00']

这是比使用如上所示的详细的 XPath 伎俩干净。Just remember to use the . in the XPath expressions that will follow.

Built-in Selectors reference

class scrapy.selector.Selector(response=None, text=None, type=None)

An instance of Selector is a wrapper over response to select certain parts of its content.

response is an HtmlResponse or an XmlResponse object that will be used for selecting and extracting data.

text is a unicode string or utf-8 encoded text for cases when a response isn’t available. Using text and response together is undefined behavior.

type defines the selector type, it can be "html", "xml" or None (default).

If type is None, the selector automatically chooses the best type based on response type (see below), or defaults to "html" in case it is used together with text.

If type is None and a response is passed, the selector type is inferred from the response type as follows:

Otherwise, if type is set, the selector type will be forced and no detection will occur.


Find nodes matching the xpath query and return the result as a SelectorList instance with all elements flattened. List elements implement Selector interface too.

query is a string containing the XPATH query to apply.


For convenience, this method can be called as response.xpath()


Apply the given CSS selector and return a SelectorList instance.

query is a string containing the CSS selector to apply.

在后台,通过cssselect库和运行.xpath() 方法,CSS查询会被转换为XPath查询。


For convenience this method can be called as response.css()


Serialize and return the matched nodes as a list of unicode strings. Percent encoded content is unquoted.


Apply the given regex and return a list of unicode strings with the matches.

regex can be either a compiled regular expression or a string which will be compiled to a regular expression using re.compile(regex)

register_namespace(prefix, uri)

Register the given namespace to be used in this Selector. Without registering namespaces you can’t select or extract data from non-standard namespaces. See examples below.


Remove all namespaces, allowing to traverse the document using namespace-less xpaths. See example below.


Returns True if there is any real content selected or False otherwise. In other words, the boolean value of a Selector is given by the contents it selects.

SelectorList objects

class scrapy.selector.SelectorList

The SelectorList class is a subclass of the builtin list class, which provides a few additional methods.


Call the .xpath() method for each element in this list and return their results flattened as another SelectorList.

query is the same argument as the one in Selector.xpath()


Call the .css() method for each element in this list and return their results flattened as another SelectorList.

query is the same argument as the one in Selector.css()


Call the .extract() method for each element in this list and return their results flattened, as a list of unicode strings.


Call the .re() method for each element in this list and return their results flattened, as a list of unicode strings.


returns True if the list is not empty, False otherwise.

Selector examples on HTML response

Here’s a couple of Selector examples to illustrate several concepts. In all cases, we assume there is already a Selector instantiated with a HtmlResponse object like this:

sel = Selector(html_response)
  1. 从HTML响应主体中提取所有的<h1>元素,返回Selector对象(即SelectorList对象):

  2. Extract the text of all <h1> elements from an HTML response body, returning a list of unicode strings:

    sel.xpath("//h1").extract()         # this includes the h1 tag
    sel.xpath("//h1/text()").extract()  # this excludes the h1 tag
  3. Iterate over all <p> tags and print their class attribute:

    for node in sel.xpath("//p"):
        print node.xpath("@class").extract()


Here’s a couple of examples to illustrate several concepts. In both cases we assume there is already a Selector instantiated with an XmlResponse object like this:

sel = Selector(xml_response)
  1. 从XML响应主体中选择所有的<product>元素,返回 Selector对象(即SelectorList对象):

  2. Extract all prices from a Google Base XML feed which requires registering a namespace:

    sel.register_namespace("g", "")

Removing namespaces

When dealing with scraping projects, it is often quite convenient to get rid of namespaces altogether and just work with element names, to write more simple/convenient XPaths. You can use the Selector.remove_namespaces() method for that.

Let’s show an example that illustrates this with GitHub blog atom feed.

First, we open the shell with the url we want to scrape:

$ scrapy shell

Once in the shell we can try selecting all <link> objects and see that it doesn’t work (because the Atom XML namespace is obfuscating those nodes):

>>> response.xpath("//link")

But once we call the Selector.remove_namespaces() method, all nodes can be accessed directly by their names:

>>> response.selector.remove_namespaces()
>>> response.xpath("//link")
[<Selector xpath='//link' data=u'<link xmlns="'>,
 <Selector xpath='//link' data=u'<link xmlns="'>,

If you wonder why the namespace removal procedure isn’t always called by default instead of having to call it manually, this is because of two reasons, which, in order of relevance, are:

  1. Removing namespaces requires to iterate and modify all nodes in the document, which is a reasonably expensive operation to perform for all documents crawled by Scrapy
  2. There could be some cases where using namespaces is actually required, in case some element names clash between namespaces. These cases are very rare though.