Native MathML: add a workaround for <mlabeledtr>, rowspacing/columnspacing

[fred-wang]

is used in the code generated by MathJax from LaTeX (for equation labeling) but browsers like Webkit, Gecko or Opera do not support this element:

https://bugzilla.mozilla.org/show_bug.cgi?id=689641
https://bugs.webkit.org/show_bug.cgi?id=85732
http://www.w3.org/TR/mathml-for-css/#d2e2181

cf the discussion on mathjax-dev

[fred-wang]

One of the most important missing features in Gecko/Webkit is mtable@rowspacing/mtable@columnspacing. These attributes are used a lot by the TeX input Jax so I'm stretching a bit this issue to include them.

https://bugzilla.mozilla.org/show_bug.cgi?id=330964
https://bugs.webkit.org/show_bug.cgi?id=85731

I suspect this could be implemented by editing CSS properties on the table cells. The browser's default rules are:

http://mxr.mozilla.org/mozilla-central/source/layout/mathml/mathml.css#301
http://trac.webkit.org/browser/trunk/Source/WebCore/css/mathml.css#L187

and be dynamically updated. The MathML CSS profile used by Opera contains mtd { display:table-cell; padding:0 0.5ex; }.

[fred-wang]

@dpvc You said that you did some experiments with mabeledtr. Can you publish your work? I think I can take this issue and try something for rowspacing/columnspacing too.

[dpvc]

I had one small piece to add before committing it. I'll see if I can get to that today.

[dpvc]

OK, my issue356 branch has the fix to this point.

[fred-wang]

Thanks! I didn't look the code in details, but that looks great. I've verified that it works as expected for the basic tests from MathMLToDisplay/Presentation/TablesAndMatrices/mlabeledtr/ (essentially those of the MathML Test suite). I'll see if I can do something for rowspacing/columnspacing.

[dpvc]

Note that this commit also include adding the <mrow> to <mtd>, and an attempt to fix <mspace> by setting the style to the proper width. I had hoped to be able to handle negative widths by using a negative margin-right, but this doesn't work (FF's width computations are still wrong for this). I may try to do more to work around that, but I'm not sure if it can be done.

[fred-wang]

OK, thanks. The bug with mspace/mpadded in table will be fixed, so no need to attach a style attribute for Gecko >= 20. I didn't get any update yet regarding the patch I submitted last November, to implement negative mspace@width (or at least to implement it partially, but I think it is enough for MathJax use cases, there are experimental builds here: http://www.wg9s.com/mozilla/firefox/)

[fred-wang]

I've created a fork of your issue356 branch and commited some changes to implement rowspacing/columnspacing.

I realized that your commit says it implements malebeledtr for "Firefox and other browsers" but there is still a "HUB.Browser.isFirefox" around the section you added. I don't know if that's on purpose...

[dpvc]

Well, I hadn't completely finished the code, and committed it so you could start to work with it. I guess that needs to be something like nMML.mlabeledtrBug instead, and the mtd and mspace code need their own bug values as well.

[fred-wang]

OK, I can try to fix this and finish the work. But the code to set the padding on the cell didn't seem to work very well with Chrome, at least when I tried this morning.

[fred-wang]

It turned out that my code used the DOM element interface but for some reason I have to use the DOM node interface to make it work with Chrome.

I updated my branch with these changes as well as those mentioned in the review. I also added the nMML bug properties.

It does not seem possible to make this work on Opera, so let's not do it.

[dpvc]

OK, good to know about Chrome.

About Opera, I'm not surprised. The question is whether the original rendering of <mlabeledtr> is better or worse than our modified version. Even though the modified one doesn't do exactly what we want, it might still be better than the original. What do you think?

[fred-wang]

Actually, I think mlabeledtr is not implemented at all in Opera:

http://www.opera.com/docs/specs/presto2.12/mathml/

so I guess we can enable our workaround for mlabeledtr Opera too (the one for rowspacing/columnspacing does not change anything)