XSLT extensions for beautification

[spodzone]

Hi,
This follows from the thread over at Glavin001/atom-beautify#480 (comment)

Desirable features for XSLT-specific beautification over and above XML:

0. generally retain the XML-style 2-space indent (or 4-space, or whatever)

1. blank lines between xsl-specific nodes, especially xsl:template, xsl:output

2. blank lines before comments, especially comments immediately preceding an xsl:template

3. indent attributes, especially when there's a long list of xmlns decls on an xsl:transform.

4. double-check that it never breaks XML markup, especially regarding inserting bogus spaces in <xsl : text>.

Something like what this little shell-script achieves, except it doesn't get point 2 quite right:

#!/bin/sh

export PATH=/bin:/usr/bin:/sbin:/usr/sbin:/usr/local/bin

basic() {
  xmlstarlet fo -s 2
}

nltemplate() {
    sed -E 's#<xsl:template#\n  <xsl:template#g'
}

stylesheet() {
    sed -E '/<xsl:stylesheet/s# #\n    #g'
}

stylespace() {
    sed -E 's#<xsl:stylesheet#\n<xsl:stylesheet#'
}

paramspace() {
    sed -E '1,/<xsl:param/s#<xsl:(param)#\n  <xsl:\1#'
}

outputspace() {
    sed -E '1,/<xsl:output/s#<xsl:(output)#\n  <xsl:\1#'
}

variablespace() {
    sed -E '1,/<xsl:variable/s#<xsl:(variable)#\n  <xsl:\1#'
}

nowhitespaceblanklines() {
    sed -E 's#^[ \t]*$##'
}

basic | nltemplate | stylesheet | stylespace | paramspace | outputspace | variablespace | nowhitespaceblanklines

[prettydiff]

Thanks for opening the issue. Would you also be kind of to create a gist with a complete XSLT code sample that includes a bunch of possible edge cases? All that you asked for is easily achievable, but I want to make sure I get it right and do not introduce any harmful regression along the way.

[spodzone]

No worries. See https://gist.github.com/spodzone/675e40505b70c34fe5ee
I think that's survived intact (md5sum==5b68157d09dba0fa1ed1e53bb5bf0bb1). It's not really in use anywhere, but does contain several aspects I like in more realistic code.

No doubt some folks consider me a heathen for some of these, but a few things I would particularly draw your attention to:

• The indents on lines 3-9, 12-16, 34-56
• The blank line before comments attached to a subsequent xsl:template (lines 17, 25, 30 etc)
• the space at the end of self-closing tags />
• the uses of xsl:text between lines 68-73

Thanks!

[prettydiff]

Here is what I can do:

• a new option to always force an empty line above a comment (or run of comments)
• a new option to always indent attributes, each on its own line
• a new option to impose a space before />
• treat xsl:text exactly like HTML's pre
• force a blank line above tags with the xsl namespace
• I won't mess with indentation definitions (2 spaces) as this can already be customized
• if the root element has the xsl namespace then arbitrarily set these new options to true

[spodzone]

Yep, that's a pretty good start. The n-space indent should be taken from the global config, agreed.

We might need to be careful about two of the options: I don't always newline+indent attributes, as the ul/li/a subtree shows. It's only when the line is going to wrap otherwise, then I will. Second: "force a blank line above xsl tags", well, careful how that interacts with forcing an empty line around comments...

I was thinking this morning:
0) blank lines are truly blank - not a space or tab on them - so I can bounce on { and } in vi if I have to

1. XSLT (at least in my world) tends to be a section of preparatory / header blocks and then a load of templates, where the headers are groups of xsl:include URIs and some xsl:param and some xsl:variable definitions, each of which forms a tight paragraph.

It would make an interesting test to join all the above XSLT into one long line, beautify it and see what falls out.

[prettydiff]

This is available for you to test out via my web tool. http://prettydiff.com/ignore/testlocation/prettydiff.com.xhtml?m=beautify

I created two new options:

• commline - Forces a new line above comments
• spaceclose - Forces self-closing tags to close with a space: " />"

You can find these options if you scroll down the page and look for the markup specific options. I am not actually setting these options to true yet, but I am automatically putting newlines above xsl tags.