XPath Tester

Paste a document, type an expression, and see what it selects. The evaluation is the browser's own XPath engine, so what you get here is what document.evaluate will give your code — matched nodes listed one by one with their serialised form, or the string, number or boolean that a function like string(), count() or starts-with() returns instead of a node-set.

The engine is XPath 1.0, and that is the first thing to know. Every browser ships 1.0 and nothing more, so the 2.0 and 3.1 functions people reach for are simply not there: matches(), lower-case(), ends-with() and tokenize() are rejected as syntax errors rather than returning nothing. Their 1.0 equivalents are contains(), translate(), the substring() family and a substring(s, string-length(s) - n) idiom for suffixes, and a rejected expression here will be rejected by lxml and by libxml2-based tooling for the same reason.

Namespaces are where most real documents defeat a first attempt. XPath 1.0 has no concept of a default namespace: if a document opens with <feed xmlns="http://www.w3.org/2005/Atom">, then //entry matches nothing at all, because the unprefixed name in the expression means "no namespace" while every element in the document is in the Atom one. There are two ways out, and the tool supports both — bind a prefix in the namespace box and write //a:entry, or sidestep the whole question with //*[local-name()='entry']. When a query returns nothing from a document that declares a default namespace, this says so rather than leaving you to guess.

The parse mode matters for the same reason in reverse. XML parsing is strict, so a stray unclosed tag produces a parser error — which the tool reports, because the browser still hands back a partial document and XPath will run against the fragment that survived, giving answers that look plausible and are not. HTML parsing is forgiving and repairs the markup instead, which is what you want for scraping a real page; note that element names come back uppercase there, since that is what the DOM reports for HTML elements.

Everything runs locally; the document is never uploaded.