java xml library that preserves attribute order

Do it twice:

Read the document in using a DOM parser so you have references, a repository, if you will.

Then read it again using SAX. At the point where you need to make the transformation, reference the DOM version to determine what you need, then output what you need in the middle of the SAX stream.

You might also want to try DecentXML, as it can preserve the attribute order, comments and even indentation.

It is very nice if you need to programmatically update an XML file that's also supposed to be human-editable. We use it for one of our configuration tools.

-- edit --

It seems it is no longer available on its original location; try these ones:

• https://github.com/cartermckinnon/decentxml
• https://github.com/haroldo-ok/decentxml (unnoficial and unmaintained fork; kept here just in case the other forks disappear, too)
• https://directory.fsf.org/wiki/DecentXML

Saxon these days offers a serialization option^[1] to control the order in which attributes are output. It doesn't retain the input order (because Saxon doesn't know the input order), but it does allow you to control, for example, that the ID attribute always appears first.

And this can be very useful if the XML is going to be hand-edited; XML in which the attributes appear in the "wrong" order can be very disorienting to a human reader or editor.

If you're using this as part of a diff process then you would want to put both files through a process that normalizes the attribute order before comparing them. However, for comparing files my preferred approach is to parse them both and use the XPath deep-equal() function; or to use a specialized tool like DeltaXML.

_{[1] saxon:attribute-order - see http://www.saxonica.com/documentation/index.html#!extensions/output-extras/serialization-parameters}