Generate <mo>+<mrow> instead of <mfenced>

[fred-wang]

See https://groups.google.com/forum/#!topic/mathjax-dev/GIvpWSNJrgA/discussion about various options regarding what the input/output modes could do and whether there are benefits to MathJax code.

For Gecko, this avoids a bug regarding size of open/close operators in mfenced. If this change is made, an explicit <mrow> should be added to element to workaround this bug: https://bugzilla.mozilla.org/show_bug.cgi?id=236963

That could also solve issue in Opera when multiple children are used. I don't know whether that make the situation of operator stretching better or worse, though

[dpvc]

I did some testing, and Opera only stretches delimiters if they are in <mfenced> not normal <mo> operators in an <mrow>.

[fred-wang]

I've created an issue359 branch to make the TeX input Jax uses mrow+mo instead of mfenced. The output of LaTeXToMathML seems to be as expected. Here are the tests that use mfenced:

LaTeXToMathML/delimiters/left-right-1.html
LaTeXToMathML/frac/abovewithdelims-1.html
LaTeXToMathML/frac/atopwithdelims-1.html
LaTeXToMathML/frac/binom-1.html
LaTeXToMathML/frac/brace-1.html
LaTeXToMathML/frac/brack-1.html
LaTeXToMathML/frac/choose-1.html
LaTeXToMathML/frac/dbinom-1.html
LaTeXToMathML/frac/genfrac-1.html
LaTeXToMathML/frac/tbinom-1.html
LaTeXToMathML/frac/overwithdelims-1.html
LaTeXToMathML/layout/newline-1.html
LaTeXToMathML/macro/let-1.html
LaTeXToMathML/macro/macro-3.html
LaTeXToMathML/matrices/bmatrix-1.html
LaTeXToMathML/matrices/bmatrix-2.html
LaTeXToMathML/matrices/pmatrix-1.html
LaTeXToMathML/matrices/pmatrix-2.html
LaTeXToMathML/matrices/vmatrix-1.html
LaTeXToMathML/matrices/vmatrix-2.html

so I mark this as "unit test wanted" to remember that I need to update them later.

The ASCIIMath input does not seem to generate mfenced and I have not modified the MathML input (or the NativeMML ouput) yet. For Firefox, the NativeMML ouput should be modifed to add an explicit mrow to each mtd cell.

[fred-wang]

I opened #385 but actually my branch replaces the mfenced by an mrow, so the open/close operators are not direct children of the mtd and consequently that does not make the situation worse for Firefox.

[fred-wang]

OK, I've updated the tests.

=> In testsuite.

[dpvc]

I was trying to review this branch, but it is a bit difficult. This has issue356 merged into it, and then some additional work was done on issue356 which isn't merged (namely the renaming of properties to include an HTML prefix). I think those changes to issue356 need to be incorporated into issue359 before it can be properly evaluated.

[fred-wang]

Yes, sorry it's a bit messy. I've updated my issue359 branch:

https://github.com/fred-wang/MathJax/compare/issue356...issue359

[dpvc]

I've made a pull request to include some mostly cosmetic changes. Once that is done, I think it is ready to go, and I'll merge it.

=> Ready for Release

[fred-wang]

You can merge this.

[dpvc]

=> Merged as 0891402