Old display_matrix macro from PGmatrixmacros causes issues with BEGIN_TEXT/END_TEXT

[somiaj]

There are probably some problems like this in the OPL, but something broke this macro when using old problem methods.

This fails to render correctly.

DOCUMENT();
loadMacros(qw(
    PGstandard.pl
    PGML.pl
    PGmatrixmacros.pl
));

BEGIN_TEXT
\{ display_matrix([
   [ans_rule(5),ans_rule(5),ans_rule(5),ans_rule(5)],
   [ans_rule(5),ans_rule(5),ans_rule(5),ans_rule(5)],
   [ans_rule(5),ans_rule(5),ans_rule(5),ans_rule(5)]
],
'align'=>"ccc|c") \}
END_TEXT
ENDDOCUMENT();

For some reason mathjax is creating a very large vertical bar and the matrix doesn't really render.

This renders fine:

DOCUMENT();
loadMacros(qw(
    PGstandard.pl
    PGML.pl
    PGmatrixmacros.pl
));

BEGIN_PGML
[@ display_matrix([
   [ans_rule(5),ans_rule(5),ans_rule(5),ans_rule(5)],
   [ans_rule(5),ans_rule(5),ans_rule(5),ans_rule(5)],
   [ans_rule(5),ans_rule(5),ans_rule(5),ans_rule(5)]
],
'align'=>"ccc|c") @]*
END_PGML
ENDDOCUMENT();

We should probably fix this method as it is used in the OPL.

[somiaj]

I'm unsure why BEGIN_TEXT / END_TEXT will break this macro when it works fine in PGML. I am also unsure where to look, but I suspect this has broken things in the OPL as a result.

[drgrice1]

This seems to be something that is browser specific. I tested with Firefox first, and though you were full of **** as usual! But then I tested in Google Chrome, and see something like your screen shot. I also see console errors with several messages like Error: <svg> attribute viewBox: Expected number, "0 -3295 NaN 6090". So it looks like this is a MathJax issue.

[somiaj]

I'm using firefox-esr and the problem is there too. So probably something that newer firefox fixed, but is still an issue firefox-esr and chromium. The odd part to is the same macro only has issues when inside BEGIN/END_TEXT, so there is something about how that is wrapping things causing the problem.

@dpvc do you have any insights as to what is going on? Seems to be some strange mathjax interaction causing a NaN to appear.

[somiaj]

Further testing, if I change the render to CHTML it works, but fails with SVG. @drgrice1 want to test firefox again, maybe it is the mathjax render and not the browser?

[drgrice1]

Yes, I did have the CHTML renderer set in Firefox for some reason. After switching to the SVG rendering, I see the problem in Firefox as well.

Note, that the main difference between how these are inserted into the page is that without PGML the math is wrapped in <script type="math/tex">...</script>. PGML does not do that and just uses \(...\).

If I change line 1924 of PGbasicmacros.pl to $out = ($mode eq 'display' ? '\[' : '\(') . $in . ($mode eq 'display' ? '\]' : '\)');, then the issue goes away.

[drgrice1]

Maybe it is time to eliminate that script tag wrapping approach. Either that or an adjustment is needed in the mathjax-config.js file where this script tag processing is done.

[drgrice1]

In any case, I am tagging @dpvc in case he has some ideas on this.

[somiaj]

Seems that line in PGbasicmacros.pl and a single test in t/pg_problems/problem_file.t are the only two places I see in PG that the script tag is still used this way.

[drgrice1]

It is also done on line 1209 of PG.pl for feedback.

[somiaj]

Ahh missed that as I was looking for <script type="math, forgot there are other ways to create tags.

[dpvc]

Because you are seeing a NaN in the viewBox for the SVG, that may mean that this is the bug from mathjax/MathJax#3568 that is fixed in v4.1.3. What version of MathJax is being used?

[drgrice1]

Version 4.1.3 is the version being used.

[drgrice1]

Is it possible that the script tag handling defined in mathjax-config.js is not doing something right that causes that bug to still persist?

What do you think about removing the code in PG and making it just use \(...\) and \[...\] instead of wrapping math in <script type="math/tex">...</script>? It would be easy to do, and I don't think that is needed anymore.

[dpvc]

What do you think about removing the code in PG and making it just use \(...\) and \[...\] instead of wrapping math in <script type="math/tex">...</script>?

Yes, that would be a help, but you will probably want to HTML-escape < and &. One of the reasons for the original <script type="math/tex'> was because the contents of <script> tags are CDATA, so you don't have to quote HTML special characters. You do with \(...\) and \[...\] in plain HTML text. I haven't looked at the code to see if that is taken care of anywhere.