XPath Playground

What you can do

• Test an XPath expression against your own XML before using it in code.
• Write and check an XPath selector for a web page or a Selenium test.
• Find the full path to a node in an XML document and copy it.
• Query XML that uses namespaces, with a prefix table you can edit.
• Extract attribute values, text nodes or whole elements from a document.
• Count or sum values with count(), sum() and string().
• Learn XPath from clickable examples, axes, predicates and functions.

How to use the XPath Playground

1. 1Paste your XML or HTML into the document editor, or load one of the samples.
2. 2Pick XML or HTML. XML parsing is strict and names the first error, HTML parsing accepts the usual unclosed tags.
3. 3Type an XPath expression, or click an example to load one.
4. 4Read the matches below. Each one shows its kind, its serialized value and its location path.
5. 5Copy the values, the paths or the matched XML, download the result, or send it to the next tool.

The same engine as your devtools

Expressions run through the XPath engine that is already part of the browser, the one behind $x() in the devtools console and behind document.evaluate() in code. What matches here is exactly what matches there, which is the point of testing an expression before you ship it. The Copy as $x() button gives you the expression ready to paste into a console.

That engine speaks XPath 1.0, the version every browser implements. XPath 2.0 and 3.1 additions like matches(), tokenize(), for loops or sequence types are not available anywhere in a browser, so an expression written for Saxon or a Java library may need trimming. Everything in the cheat sheet works.

Namespaces and the unprefixed name trap

XPath 1.0 has no default namespace. If a document starts with <feed xmlns="http://www.w3.org/2005/Atom">, every element in it belongs to that namespace, and //entry matches nothing at all, however right it looks. The name has to be written with a prefix that is bound to the URI, like //atom:entry.

The prefix table collects what the document declares and invents a readable prefix for the default namespace, so a namespaced feed is queryable the moment you paste it. The prefixes are yours to rename, and they do not have to match the ones in the source. When an expression comes back empty because of this, the result panel says so and offers to add the prefix for you. The other way out is //*[local-name(.)='entry'], which ignores namespaces entirely.

Node sets, strings, numbers and booleans

An XPath expression does not always return nodes. count(//book) gives a number, string(//title) a string, and a comparison or boolean(…) gives true or false. All four come back here, labelled with their type, instead of an empty list that looks like a mistake.

For a node set, each match is listed with its kind (element, attribute, text, comment or processing instruction), the node serialized back to XML and its location path, for example /catalog/book[2]/title. Steps are numbered only where a name repeats, so paths stay readable, and Number every path step switches to the fully numbered form that some tools expect.

XPath on HTML pages

HTML mode uses the forgiving HTML parser, so a page saved from a site works even with unclosed tags, stray attributes and no XML declaration. Unprefixed names match HTML elements directly, which is why selectors like //a[@class='external']/@href behave the same as in a scraper or a browser test.

If a paste fails to parse as XML and looks like a web page, the error offers to switch modes with one click, instead of leaving you with a mismatched-tag message you cannot fix.

XPath, JSONPath or jq

XPath queries markup. For JSON, the same job is done by JSONPath in the JSONPath Playground, and heavier reshaping of JSON belongs in the jq Playground. If the document you have is XML but the tool you feed it to wants JSON, convert it first with the Config Converter and query the result there.