XPath

XPath is a query language for XML documents. Given a document tree, an XPath expression selects one or more nodes from it — elements, attributes, text, or any combination. It was designed alongside XSLT and is standardised by the W3C. Current versions are XPath 2.0, 3.0, and 3.1, each adding features; XPath 1.0 is still widely deployed in simpler tools.

Example

Given:

<library>
  <book id="1">
    <title>The C Programming Language</title>
    <author>Kernighan</author>
    <author>Ritchie</author>
  </book>
  <book id="2">
    <title>Code</title>
    <author>Petzold</author>
  </book>
</library>

• /library/book — both <book> elements.
• //book[@id="2"] — the book with id="2".
• //book[1] — the first book (XPath is 1-indexed, not 0-indexed).
• //book[author="Petzold"]/title/text() — the text "Code".
• //author[1] — first <author> child of each book (careful: this is per-parent, not globally).
• //book[count(author) > 1]/title — titles of books with multiple authors.

Axes and predicates

XPath is built around two concepts:

• Axes — directions in the document tree: child::, parent::, descendant::, ancestor::, following-sibling::, and more. The / shorthand is child::.
• Predicates — square-bracket filters that constrain which nodes a step selects: book[@id="1"], *[position() = last()].

Combining them gives a lot of expressive power in one line.

Where you see XPath

• XSLT transformations — XPath is the addressing language XSLT uses to find what to transform.
• Selenium and browser automation — driver.findElement(By.xpath("...")) to locate DOM nodes.
• Python’s lxml, JavaScript’s document.evaluate, Java’s javax.xml.xpath — generic XML traversal.
• XSD uses XPath expressions in <xs:selector> / <xs:field> for identity constraints.
• YAML / JSON alternatives — XPath is the inspiration for JSONPath and JMESPath.

Why it matters

Before XPath, extracting “the email of every customer with a pending order” from an XML document required writing a DOM traversal by hand. XPath turns that into one expression. The same idea drove JSONPath and later jq for JSON, and (centrally) CSS selectors in web scraping.

Simple mental model

If you know CSS selectors, XPath feels like “CSS selectors with more power and more punctuation”. .foo .bar in CSS is roughly //foo//bar in XPath. Beyond that, XPath can walk up the tree, match by position, compute with strings, and evaluate boolean predicates — things CSS selectors cannot do.