TeX annotation in MathML output mode should be XML escaped

[physikerwelt]

It would be nice if the TeX annotation generted in the MathML output would be XML escaped.
Otherwise, XML validation tools might cosinder the characters & < >... to be invalid.
In https://gerrit.wikimedia.org/r/#/c/166556/ provides a workaround for most cases, but if the TeX string contains "" we could have a problem with the woraround.
As far as I know this is a MathJax and not a MathJax node bug, but feel free to move the bug if I'm wrong.

[physikerwelt]

I think it sould be fixed here.
https://github.com/physikerwelt/MathJax/blob/master/unpacked/extensions/toMathML.js#L164
However, I don't know how if there is already an XML escaping methoid build into mathjax that could be used.
@dpvc: Is there a method for XML escaping?

[pkra]

The reason is probably that we didn't expect MathJax-node, i.e., the only application would be "Show Source" in the menu. Here's what it's doing https://github.com/physikerwelt/MathJax/blob/master/unpacked/extensions/MathMenu.js#L760-L761

IIUC the best practice within XML is to use CDATA (e.g., http://jats.nlm.nih.gov/archiving/tag-library/1.1d1/n-skz0.html) but that wouldn't work in an HTML settings -- and I think we'd need both. So yes, I'd say that's reasonable but I'm not sure whether we need an XML / CDATA option.

[davidcarlisle]

On 14 October 2014 17:06, Peter Krautzberger notifications@github.com
wrote:

The reason is probably that we didn't expect MathJax-node, i.e., the only
application would be "Show Source" in the menu. Here's what it's doing
https://github.com/physikerwelt/MathJax/blob/master/unpacked/extensions/MathMenu.js#L760-L761

IIUC the best practice within XML is to use CDATA (e.g.,
http://jats.nlm.nih.gov/archiving/tag-library/1.1d1/n-skz0.html) but that
wouldn't work in an HTML settings -- and I think we'd need both. So yes,
I'd say that's reasonable but I'm not sure whether we need an XML / CDATA
option.

I think even in a pure xml context it's better to quote <>& rather than use
CDATA, apart from anything else it ensures ]]> is made safe

David

[physikerwelt]

I have a fix that works for me. I'll create a pull request soon.
https://github.com/physikerwelt/MathJax/tree/mathoid

[pkra]

Thanks, @davidcarlisle and thanks @physikerwelt for working on a pull request.

[pkra]

Is a treeReftest sufficient? E.g., \(&<>\) with

      <math xmlns="http://www.w3.org/1998/Math/MathML">
        <semantics>
          <merror>
            <mtext>&amp;&lt;&gt;</mtext>
          </merror>
          <annotation encoding="application/x-tex">&amp;&lt;&gt;</annotation>
        </semantics>
      </math>

[pkra]

Oh nevermind. I should remember my TeX...

[dpvc]

Is a treeReftest sufficient?

No, you will need to use a script test, because the change is in toMathML(), not the NativeMML output. So you need to check the results of toMathML() to see that the annotation contains the escaped characters.

[dpvc]

==> In Testsuite

UI/issue935.html