What does preceding-sibling do in XPath?

preceding-sibling:: is an XPath axis that selects sibling nodes appearing before the context node under the same parent. The syntax is context/preceding-sibling::tag[predicate], so //input[@id='email']/preceding-sibling::label[1] returns the closest <label> before the email input. Its forward counterpart is following-sibling::, and preceding-sibling::* matches every prior sibling regardless of tag.

The common gotcha is index direction. On the preceding-sibling axis, position counts backward from the context node, so [1] is the immediately previous sibling, [2] is the one before that, and so on. This trips up developers who expect document order. If you want the first sibling in document order under the parent, do not write preceding-sibling::li[1]. Use preceding-sibling::li[last()] instead.

Nearest previous td:

//td[@class='value']/preceding-sibling::td[1]

First td under the parent (in document order):

//td[@class='value']/preceding-sibling::td[last()]

Do not confuse preceding-sibling with preceding. The preceding axis selects every node appearing before the context node anywhere in the document, except ancestors, attributes, and namespace nodes. preceding-sibling is restricted to nodes that share the same parent. //h2[@id='intro']/preceding::p walks back across the whole document, while //h2[@id='intro']/preceding-sibling::p stays inside the parent of that <h2>.