XPath 1.0 Functions Reference

Browsers, Selenium, Playwright and Appium all evaluate XPath 1.0. That gives you exactly 27 functions. This page lists all of them with the situations where each one earns its place in a locator.

If a function is not on this page, it will throw an “invalid expression” error in your tests. See the XPath versions guide for what lives in 2.0 and 3.1 and how to work around it.

Node-set functions

Function	Returns	Test automation use
last()	number of nodes in the context	(//tr)[last()] last row
position()	index of the context node	//li[position() < 4] first three
count(node-set)	how many nodes	//ul[count(li) > 5] lists with more than five items
id(string)	element with that ID	rarely used; //*[@id='x'] is clearer
local-name(node-set?)	name without namespace prefix	//*[local-name()='svg']
namespace-uri(node-set?)	the namespace URI	//*[namespace-uri()='http://www.w3.org/2000/svg']
name(node-set?)	full qualified name	//*[name()='svg:path'] in XML

String functions

Function	Returns	Test automation use
string(object?)	string value	string(//h1) for assertions
concat(s1, s2, ...)	joined string	quotes: concat("It", "'", "s")
starts-with(s, prefix)	boolean	//input[starts-with(@id, 'user-')] dynamic ids
contains(s, sub)	boolean	//*[contains(@class, 'btn')]
substring-before(s, sep)	text before separator	substring-before(@href, '?')
substring-after(s, sep)	text after separator	substring-after(@id, 'row-')
substring(s, start, len?)	slice, 1-based	substring(@id, 1, 4)='user'
string-length(s?)	length	//input[string-length(@value) > 0] non-empty
normalize-space(s?)	trimmed, single-spaced	//button[normalize-space()='Save']
translate(s, from, to)	character mapping	case folding, see below

substring counts from 1, not 0. substring('hello', 1, 2) is he.

translate maps each character in from to the character at the same position in to. Characters in from with no partner in to are deleted. Two idioms:

translate(text(), 'ABCDEFGHIJKLMNOPQRSTUVWXYZ', 'abcdefghijklmnopqrstuvwxyz')   lower-case
translate(@data-price, '$,', '')                                               strip $ and commas

Boolean functions

Function	Returns	Test automation use
boolean(object)	boolean conversion	boolean(//div[@role='alert']) “is there an alert”
not(boolean)	negation	//button[not(@disabled)]
true()	true	placeholders and generated expressions
false()	false	same
lang(string)	matches xml:lang	//p[lang('de')] localisation tests

not() is a function, not an operator. not @disabled is a syntax error.

Number functions

Function	Returns	Test automation use
number(object?)	numeric conversion	number(//span[@class='qty']) = 3
sum(node-set)	total	sum(//td[@class='price']) cart totals
floor(n)	round down	maths on positions
ceiling(n)	round up	same
round(n)	nearest integer	same

sum() converts each node’s string value to a number. Any non-numeric value makes the whole result NaN, so strip currency symbols with translate() first or use it only on clean data cells.

Functions that do not exist in XPath 1.0

These are the ones people reach for after reading generic XPath tutorials. None of them work in a browser:

Missing	Version	XPath 1.0 workaround
ends-with(s, suffix)	2.0	substring(s, string-length(s) - string-length('suffix') + 1) = 'suffix'
lower-case(s), upper-case(s)	2.0	translate() with the alphabet
matches(s, regex)	2.0	combine contains, starts-with, translate
replace(s, regex, with)	2.0	translate() for single characters only
string-join(), tokenize()	2.0	not possible; restructure the locator
exists(), empty()	2.0	boolean(), not()
compare(), codepoints-to-string()	2.0	not needed for locators

The ends-with workaround in full, for an id that ends in -email:

//input[substring(@id, string-length(@id) - string-length('-email') + 1) = '-email']

It is ugly. In practice contains(@id, '-email') is usually good enough.

Where each function goes

Functions can appear in predicates, as the whole expression, or nested:

//tr[normalize-space(td[1]) = 'Widget']              inside a predicate
count(//tr[td[3] > 0])                              as the whole expression, returns a number
//td[count(preceding-sibling::td) = 2]              nested in arithmetic
concat(normalize-space(//h1), ' | ', string(//title))  building a string

Selenium and Playwright locators must return elements, so an expression that returns a number, string or boolean will error when used as a locator. Use those forms in DevTools or scraping code instead.

Try It Yourself

Open in Playground →

Next Steps

1. Position and Indexing - position(), last() and reverse axes
2. Case-Insensitive Matching - translate() in depth
3. XPath 1.0 vs 2.0 vs 3.1 - Where the other 100+ functions live