XPath

XPath is a query language for selecting nodes in an XML document.

Selectors

Getting started

Test in Firefox or Chrome console:

$x('/html/body')
$x('//h1')
$x('//h1')[0].innerText
$x('//a[text()="XPath"]')[0].click()

Descendant selectors

Xpath CSS
//h1 h1
//div//p div p
//ul/li ul > li
//ul/li/a ul > li > a
//div/* div > *
/ :root
/html/body :root > body

Order selectors

Xpath CSS
//ul/li[1] ul > li:first-child
//ul/li[2] ul > li:nth-child(2)
//ul/li[last()] ul > li:last-child
//li[@id="id"][1] li#id:first-child
//a[1] a:first-child
//a[last()] a:last-child

Attribute selectors

Xpath CSS
//*[@id="id"] #id
//*[@class="class"] .class
//input[@type="submit"] input[type=”submit”]
//a[@id="abc"][@for="xyz"] a#abc[for=”xyz”]
//a[@rel] a[rel]
//a[starts-with(@href, '/')] a[href^=’/’]
//a[ends-with(@href, '.pdf')] a[href$=’pdf’]
//a[contains(@href, '://')] a[href*=’://’]
//a[contains(@rel, 'help')] a[rel~=’help’]

Siblings

Xpath CSS
//h1/following-sibling::ul h1 ~ ul
//h1/following-sibling::ul[1] h1 + ul
//h1/following-sibling::[@id="id"] h1 ~ #id

jQuery

Xpath CSS
//ul/li/.. $(‘ul > li’).parent()
//li/ancestor-or-self::section $(‘li’).closest(‘section’)
//a/@href $(‘a’).attr(‘href’)
//span/text() $(‘span’).text()

Misc selectors

Xpath CSS  
//h1[not(@id)] h1:not([id])  
//button[text()="Submit"] Text match  
//button[contains(text(),"Go")] Text contains (substring)  
//product[@price > 2.50] Arithmetic  
//ul[*] Has children  
//ul[li] Has children (specific)  
//a[@name or @href] Or logic  
`//a | //div` Union (joins results)  

Expressions

Steps and axes

// ul / a[@id='link']
Axis Step Axis Step

Prefixes

Prefix Example Means
// //hr[@class='edge'] Anywhere
/ /html/body/div Root
./ ./div/p Relative

Axes

Axis Example Means
/ //ul/li/a Child
// //[@id="list"]//a Descendant

Predicates

Predicates 

//div[true()]
//div[@class="head"]
//div[@class="head"][@id="top"]

Restricts a nodeset only if some condition is true. They can be chained.

Operators 

# Comparison
//a[@id = "xyz"]
//a[@id != "xyz"]
//a[@price > 25]

# Logic (and/or)
//div[@id="head" and position()=2]
//div[(x and y) or not(z)]

Using nodes 

# Use them inside functions
//ul[count(li) > 2]
//ul[count(li[@class='hide']) > 0]

# Returns `<ul>` that has a `<li>` child
//ul[li]

You can use nodes inside predicates.

Indexing 

//a[1]                # first <a>
//a[last()]           # last <a>
//ol/li[2]            # second <li>
//ol/li[position()=2] # same as above
//ol/li[position()>1] #:not(:first-child)

Use [] with a number, or last() or position().

Chaining order 

a[1][@href='/']
a[@href='/'][1]

Order is significant, these two are different.

Nesting predicates 

//section[.//h1[@id='hi']]

This returns <section> if it has an <h1> descendant with id='hi'.

Functions

Node functions 

name()            # //[starts-with(name(), 'h')]
text()            # //button[text()="Submit"]
                  # //button/text()
lang(str)
namespace-uri()

count()           # //table[count(tr)=1]
position()        # //ol/li[position()=2]

String functions 

contains()        # font[contains(@class,"head")]
starts-with()     # font[starts-with(@class,"head")]
ends-with()       # font[ends-with(@class,"head")]

concat(x,y)
substring(str, start, len)
substring-before("01/02", "/")  #=> 01
substring-after("01/02", "/")   #=> 02
translate()
normalize-space()
string-length()

Boolean functions 

not(expr)         # button[not(starts-with(text(),"Submit"))]

Type conversion 

string()
number()
boolean()

Axes

Using axes 

//ul/li                       # ul > li
//ul/child::li                # ul > li (same)
//ul/following-sibling::li    # ul ~ li
//ul/descendant-or-self::li   # ul li
//ul/ancestor-or-self::li     # $('ul').closest('li')
       
// ul /child:: li
Axis Step Axis Step

Steps of an expression are separated by /, usually used to pick child nodes. That’s not always true: you can specify a different “axis” with ::.

Child axis 

# both the same
//ul/li/a
//child::ul/child::li/child::a

child:: is the default axis. This makes //a/b/c work.

# both the same
# this works because `child::li` is truthy 
//ul[li]
//ul[child::li]

# both the same
//ul[count(li) > 2]
//ul[count(child::li) > 2]

Descendant-or-self axis 

# both the same
//div//h4
//div/descendant-or-self::h4

// is short for the descendant-or-self:: axis.

# both the same
//ul//[last()]
//ul/descendant-or-self::[last()]

Other axes

Axis Abbrev Notes
ancestor    
ancestor-or-self    
attribute @ @href is short for attribute::href
child   div is short for child::div
descendant    
descendant-or-self // // is short for /descendant-or-self::node()/
namespace    
self . . is short for self::node()
parent .. .. is short for parent::node()
following    
following-sibling    
preceding    
preceding-sibling    

There are other axes you can use.

Unions 

//a | //span

Use | to join two expressions.

More examples

Examples 

//*                 # all elements
count(//*)          # count all elements
(//h1)[1]/text()    # text of the first h1 heading
//li[span]          # find a <li> with an <span> inside it
                    # ...expands to //li[child::span]
//ul/li/..          # use .. to select a parent

Find a parent 

//section[h1[@id='section-name']]

Finds a <section> that directly contains h1#section-name

//section[//h1[@id='section-name']]

Finds a <section> that contains h1#section-name. (Same as above, but uses descendant-or-self instead of child)

Closest 

./ancestor-or-self::[@class="box"]

Works like jQuery’s $().closest('.box').

Attributes 

//item[@price > 2*@discount]

Finds <item> and check its attributes

See also