`xmllint --format in.xml > out.xml`
`xmlstarlet format in.xml > out.xml`

Each of these two programs has switches for canonicalization, too.
But these are not XSLT solutions, so are bit off-topic.


> I have a some very large XML source files. Each is are currently
> formatted as one long string (ie no line breaks).
> 
> I need to add some processing instructions at various points in the
> file but I don't want to introduce any significant whitespace in
> the process of doing so.
> 
> I have XMLSpy but it does not seem to be capable of doing a
> schema-aware "pretty-print".
> 
> A web search seems to indicate that OXygen is capable of doing an
> intelligent pretty-print (the information I found is in the context
> of XHTML but I assume it's more generalized than that??), however,
> I can't really justify buying it just for that.
> 
> If needs be, I can write an XSLT to parse a DTD, classify each
> element as "inline", "mixed", "block" or "empty" and then do an
> identity transform against my XML source files inserting
> line-breaks and leading tabbing as required.
> 
> I am wondering if anyone else has any other suggestions for me?