Question/Feature request: dealing with tags that are not self/closing (sgml)

[guraltsev]

This is more of a question than a bug report. Please tell me if I am using the wrong tool for the job. In SGML some tags are not allowed to be self-closing. e.g. all modern browsers error out if they encounter

<script src="somescript.js"/>

even is the document is of type html5. The script must be

<script src="somescript.js"></script>

For context:

I use esxml in several filter functions that I execute when exporting org files to html. I do not want to redefine a new exporter, just to filter some things. I use libxml to parse the output of the org exporter at different steps, I modify it using dom, and then I re-output it using esxml. I actually found it very surprising that esxml is not built into emacs proper.

One example is that I define an option (variable) that decides whether local css and JS scripts should be inlined in the html.

I added a hook function into org-export-filter-final-output-functions (see (https://orgmode.org/manual/Advanced-Export-Configuration.html)[Advanced Export Configuration]) that parses the output html, changes the contents of the script tag if inlining is required, then finally pushes everything back to html. However, if I am not inlining the scripts, the roundtrip html->emasc xml->html
changes <script>[...]</script> pairs into <script/> making the html output incorrect. I temporary solved the issue by adding a whitespace as a string inside empty <script></script> tags.

Do you have any comments about this? Now that there is a pcase would it be appropriate to implement such behavior? Technically this is not an XML behavior but de-facto I see no other library to programmatically edit HTML documents in emacs.

[tali713]

Okay, so this is a problem also for textarea. When I was originally designing, I decided that the simplest way to address this problem was simply to allow (tag attr-list) transform to <tag attributes/> but (tag attr-list "") will transform to <tag attributes></tag>. the empty string being equivalent to the empty-body. So your solution is indeed the intended solution.

However, let's say this was not good enough for you, and you really wanted to have a personal sublanguage with any special cases you like,

esxml/esxml.el

Lines 194 to 213 in 2656460

	(defun sxml-to-esxml (sxml)
	"Translates sxml to esxml so the common standard can be used.
	See: http://okmij.org/ftp/Scheme/SXML.html."
	(pcase sxml
	(`(,tag (@ . ,attrs) . ,body)
	`(,tag ,(mapcar (lambda (attr)
	(cons (first attr)
	(or (second attr)
	(prin1-to-string (first attr)))))
	attrs)
	,@(mapcar 'sxml-to-esxml body)))
	(`(,tag . ,body)
	`(,tag nil
	,@(mapcar 'sxml-to-esxml body)))
	((and sxml (pred stringp)) sxml)))
	(defun sxml-to-xml (sxml)
	"Translates sxml to xml, via esxml, hey it's only a constant
	factor. :)"
	(esxml-to-xml (sxml-to-esxml sxml)))

shows how.