Katex: \bmod in inline math mode adds extra following space

[akalin]

Test file: https://cdn.rawgit.com/akalin/98bc47f5e54f7529462f12a1d7336858/raw/28a50452296c9d7bd4dc9e5023056cf64412a913/test-0.8.3.html

Using \bmod in inline math mode seems to add some extra following space, which seems like a bug. Not sure if it's a problem with other binary operators.

[edemaine]

This is a consequence of #836, I'm afraid. (Not quite sure why \bmod is implemented with negative space, but it is.) We still don't know how to fix it (or the exact cause).

[akalin]

Hmm...but does it need negative space? If I use \mathbin{\mathrm{mod}} that works perfectly fine: see
https://cdn.rawgit.com/akalin/98bc47f5e54f7529462f12a1d7336858/raw/b71b173e7fdc3459938d7f362574da51eacdad6a/test-0.8.3.html

So I guess that's a workaround for now.

[edemaine]

For some reason, LaTeX (latex.ltx) defines \bmod as follows:

\def\bmod{%
  \nonscript\mskip-\medmuskip\mkern5mu%
  \mathbin{\operator@font mod}\penalty900\mkern5mu%
  \nonscript\mskip-\medmuskip}

So KaTeX is mimicking that. Basically, in display/textstyle, we want 5mu minus \medmuskip (which is 4mu plus 2mu minus 1mu skip), while in script/scriptscriptstyle, we want 5mu skip. Given the #836 bug, I think it'd be nice to remove the negative space in \bmod. I've created PR #984 to this effect.

Meanwhile, you can define a macro of "\\bmod": "\\kern1mu\\mathbin{\\mathrm{mod}}}\\kern1mu", assuming you're not using \bmod in subscripts.

[akalin]

Hmm, that \bmod macro seems to have too little space after mod. With

\begin{aligned}
& (a \cdot b) \bmod 257 \\
& (a \cdot b) \mathbin{\mathrm{mod}} 257 \\
& (a \cdot b) \kern1mu\mathbin{\mathrm{mod}}\kern1mu 257
\end{aligned}

I get

\mathbin{\mathrm{mod}} seems closer to \bmod.

[akalin]

Although the 2nd \kern1mu should _add_ space, right? It's weird that putting it in _removes_ space...

[edemaine]

I saw that too. Must be a bug of some sort... Should track it down.

The PR has the right space though.