Changes: Format your xml document using xmllint

Revision as of 12:28, 29 March 2014

Tip 349 Printable Monobook Previous Next

created 2002 · complexity basic · author Daniel Allen · version 6.0

If you open an xml document that is either totally or partially unindented, you can use the GNU libxml2 libary's xmllint to reformat and align your document. This is especially good if you want to save your xml documents using as little space as possible (which would be totally unindented). Just add this under the autocmd section of your vimrc:

au FileType xml exe ":silent %!xmllint --format --recover - 2>/dev/null"

This instructs vim to take the entire contents of a *.xml file and pass it through xmllint, using the --format and --recover flags and silencing any errors that may occur. This is generally a very effective process and will only mess up on very poorly typed (a large amount of incorrect syntax) xml documents. Please note that xmllint only adds and removes structural space. It does not remove space from regular text nodes (as doing so would be incorrect).

Comments

xmllint as a filter

Q: How can make it as a filter?

I tried sth like

let&l:equalprg='xmllint --format --recover'

but it didn't work since xmllint doesn't take STDIN?

A: Notice the '-' given as the file argument to xmllint. This usually means stdin. So, the full command should be:

let &l:equalprg='xmllint --format --recover -'

A: Yes, xmllint can read from stdin, please note the '-' that occurs at the end of the options. This tells it to dump the file contents to the stdin pipe for xmllint. As far as I know, xmllint can be run in windows if you have gcc, but I am just projecting...look into it. If not, I am sure you can find something to take its place...the tip is just that, a tip, build on it.

A: Regarding the filter question:

Start playing with

" one or more lines:
vmap ,px !xmllint --format -<CR>
" pretty-print current line
nmap ,px !!xmllint --format -<CR>

Fixing indenting of comment lines

"xmllint --format" indents comment lines to 0 which screws up the formatting for vim. A simple fix for single line comments can be made in indent/xml.vim. Copy indent/xml.vim to ~/.vim/xml.vim (Unix systems) and change the line that says:

let lnum = prevnonblank(a:lnum - 1)

To this:

let lnum = a:lnum
while lnum > 0
    let lnum = prevnonblank(lnum - 1)
    let line = getline(lnum)
    if line =~ '-->'
        let [lnum,watever] = searchpos('<!--','b')
    else
        break
    endif
endwhile

Alone-on-line end-of-comment closing-tags may get badly indented though.

xmllint on Windows

Download xmllint for Windows Use the following command, assuming xml file is open in Vim and xmllint.exe is in the same directory as the file:

   :% !xmllint.exe % --format

The first % ensures the content of the open file is replaced with the results of passing through xmllint. The order of the second % and the flag --format is important as otherwise xmllint will try to open a file with the name "--format".

Altenative xmllint Windows binaries and instructions

Binaries

At previous wiki edit, the most recent version was http://www.zlatkovic.com/projects/libxml/libxml2-2.5.8.win32.zip

Instructions

For instructions see the Windows section under:

http://www.pinkjuice.com/howto/vimxml/setup.xml#xmllint

Alternatives to xmllint

Built-in Search and Replace

Seriously consider using Vim's built-in search and replace functionality. See this Pretty-formatting_XML and the third comment (:%s/></>\r</g , etc).

(HTML) Tidy

I was dissatisfied with what xmllint was doing to my XML documents. Its --format option seems to arbitrarily expand or remove blank lines. An alternative is to use the tidy program - formerly HTML Tidy. available at http://tidy.sourceforge.net/ I altered my .vimrc as follows:

au FileType xml exe ":silent 1,$!tidy --input-xml true --indent yes 2>/dev/null"

Here's how you use tidy with xml on Windows:

au FileType xml exe ":silent 1,$!tidy -q -i -xml"

I have this in my .vimrc

" select xml text to format and hit ,x
vmap ,x :!tidy -q -i -xml<CR>

UTF-8 format

Q: One problem with "xmllint --format" is that it turns non-ASCII UTF-8 characters into numeric references. This isn't a problem if you're taking the input of xmllint and editing it with an XML editor that internally converts everything to UTF-8 (like XMetal), but if what you want to do is use Vim to edit the file as a native UTF-8 (with :set encoding=utf8), then you *don't* want a bunch of numeric references instead of Unicode.

The xmllint developer knows about this issue but feels that adding an option to keep/not keep UTF-8 would complicate the code too much. So... is there another XML reformatter more suitable for producing output that Vim can use?

A: Try http://www.pinkjuice.com/howto/vimxml/tasks.xml#prettyprinting

A: I will change the solution described at the above URL since there is a very simple solution to the problem:

:%!xmllint --format --encode UTF-8 -

Changing tab/ident width

Q:. Any ideas on how to use xmllint, but change the tab/indent width? It seems to filter, on windows, with 2 spaces for a tab.

A:. In Unix/Linux, if you don't like the default indentation (2 spaces), set the $XMLLINT_INDENT environment variable, or prefix the xmllint command in your .vimrc file:

au FileType xml exe ":silent 1,$!XMLLINT_INDENT='    ' xmllint --format --recover - 2>/dev/null"