Bug in formatting mathjax/LaTeX inconsistently depending on chosen delimiters, markdown interaction

[mpacer]

I just discovered that the notebook formats the same LaTeX input in different ways depending upon your choice of equation delimiter for LaTeX/mathjax which traditionally should not matter.

An example of this bug (plus a png of the correctly rendered incorrect rendering, since github apparently has its own bugs regarding displaying LaTeX/mathjax embedded in notebooks) can be found at https://gist.github.com/michaelpacer/2b627ffe8f10cde53926.

Put briefly, it's not supposed to matter whether you choose to delimit LaTeX formatted math with \(…\) and $…$ for inline equations or \[…\] or $$…$$ for displayed equations. Of course, because \ is used as an escape character we have to instead use\\(…\\) \\[…\\], which in theory is fine. However, because LaTeX formatted subscripts use the _ symbol, and that conflicts with markdown's interpretation of _…_ as an indicator of italics, it seems that notebook parses these two different ways of delimiting equations differently. If you use \\(…\\) \\[…\\] then you need to escape underscores \_ but that means that your LaTeX is now invalid outside of the notebook (as can be seen in the rendering of \_ when using $…$ and $$…$$ as the delimiters).

For now it seems safer to guide people to just use $…$ and $$…$$, since then at least it will interpret valid LaTeX properly (this has the unfortunate side-effect of conflicting with one of the LaTeX "deadly sins", but I think that's a less big issue than writing inaccurate LaTeX to get accurate formatting, imho).

I'm not sure if there is going to be an easy or even possible fix for this (since I know this kind of parsing is incredibly difficult to manage) but I was befuddled with why my properly formatted LaTeX was not being rendered at all, so I figured I'd bring it to people's attention.

[minrk]

It might be in how we escape math to avoid markdown mucking it up. Markdown shouldn't be touching the latex inside these blocks.

[mpacer]

Closed by #2220