lxml supports a number of interesting languages for tree traversal and element selection. The most important is obviously XPath, but there is also ObjectPath in the lxml.objectify module. The newest child of this family is CSS selection, which is implemented in the new lxml.cssselect module.


The CSSSelector class

The most important class in the cssselect module is CSSSelector. It provides the same interface as the XPath class, but accepts a CSS selector expression as input:

>>> from lxml.cssselect import CSSSelector
>>> sel = CSSSelector('div.content')
>>> sel  #doctest: +ELLIPSIS
<CSSSelector ... for 'div.content'>
>>> sel.css

The selector actually compiles to XPath, and you can see the expression by inspecting the object:

>>> sel.path
"descendant-or-self::div[contains(concat(' ', normalize-space(@class), ' '), ' content ')]"

To use the selector, simply call it with a document or element object:

>>> from lxml.etree import fromstring
>>> h = fromstring('''<div id="outer">
...   <div id="inner" class="content body">
...       text
...   </div></div>''')
>>> [e.get('id') for e in sel(h)]

CSS Selectors

This libraries attempts to implement CSS selectors as described in the w3c specification. Many of the pseudo-classes do not apply in this context, including all dynamic pseudo-classes. In particular these will not be available:

Also, none of the psuedo-elements apply, because the selector only returns elements and psuedo-elements select portions of text, like ::first-line.


In CSS you can use namespace-prefix|element, similar to namespace-prefix:element in an XPath expression. In fact, it maps one-to-one, and the same rules are used to map namespace prefixes to namespace URIs.


These applicable pseudoclasses are not yet implemented:

Unlike XPath you cannot provide parameters in your expressions -- all expressions are completely static.

XPath has underspecified string quoting rules (there seems to be no string quoting at all), so if you use expressions that contain characters that requiring quoting you might have problems with the translation from CSS to XPath.