Curl GET XML Example

The command

curl -H "Accept: application/xml" https://httpbin.org/xml

Run this request in the API Tester

Opens the request console with this request filled in. It is sent from api-tester.com's server rather than from your machine, so it shows what the endpoint returns, not how your own network or curl build behaves.

What the flags do

Formatting and querying the response

Servers send XML on a single line or with whatever indentation their serialiser produced. xmllint --format re-indents it. The same tool evaluates XPath, which is the dependable way to take one value out of a document: wrap the path in string() for text, or count() for a number of elements. The trailing dash tells xmllint to read from the pipe.

curl -s -H "Accept: application/xml" https://httpbin.org/xml | xmllint --format -
curl -s https://httpbin.org/xml | xmllint --xpath "string(//slide[1]/title)" -
curl -s https://httpbin.org/xml | xmllint --xpath "count(//slide)" -

Sitemaps, feeds and namespaces

The XML most people fetch is not from an API at all: it is a sitemap, an RSS or Atom feed, or a configuration document. Many of these declare a default namespace on the root element, and an XPath such as //loc then matches nothing, because the element's full name includes that namespace. Matching on local-name() sidesteps it without having to register a prefix.

curl -s --compressed https://example.com/sitemap.xml \
  | xmllint --xpath "//*[local-name()='loc']/text()" -

curl -s -H "Accept: application/rss+xml" https://example.com/feed \
  | xmllint --xpath "//item/title/text()" -

Checking what actually came back

Accept is a preference, and a server is free to ignore it. Before parsing, look at the status code and the response's own Content-Type; both are available as write-out variables while the body goes to a file. Saving to a file also helps with character sets: curl does not convert encodings, so a document declared as ISO-8859-1 looks garbled in a UTF-8 terminal yet is read correctly by a parser that honours the XML declaration.

curl -s -o feed.xml -w "%{http_code} %{content_type}\n" \
  -H "Accept: application/xml" https://httpbin.org/xml

Common mistakes

• An XPath that returns nothing usually means a namespace. Elements in a default namespace are not matched by their bare names.
• An HTML error page is not XML. A 404 or a login redirect fed to a parser produces a parse error far from the real cause.
• Avoid grep and sed for anything beyond a quick look. Attribute order, line breaks and entity encoding all vary between equivalent documents.

Related curl examples

More under Request bodies

• Curl POST JSON Example
• Curl POST Form Example
• Curl POST File Example
• Curl -d Flag Example
• Curl POST Body Example
• Curl JSON Request Example

References

• curl man page: -H — Adding a request header such as Accept.
• MDN: XPath — The query language used to pull values out of an XML document.

These links leave api-tester.com, which is independent of the projects they point to.

FAQ