Non-portable use of String method .includes

[ccorn]

KaTeX v0.9.0-beta chokes on \mathbb{C} (and \mathbb{R} so it's probably the macro), telling me "undefined not callable".
KaTeX v0.8.3 and v0.9.0-alpha2 render that without problems.
Found using katex.renderToString with Duktape.

[kevinbarabash]

Hmm... \mathbb{C} works in the test page on master, but that uses katex.render. I wonder if there's a difference between that and katex.renderToString.

[kevinbarabash]

I tried katex.renderToString("\\mathbb{C}") on master and that works as well.

[kevinbarabash]

I checked v0.9.0-beta as well. What's Duktape? Could you post a more complete code example?

[ccorn]

Hmm, when I try it with less intervening Ruby layers, it works. Sorry, that's not your bug, even though it manifests at the Javascript execution level. Have to track it down myself.

[ccorn]

Use of the string method includes seems to be the culprit. Using the polyfill from Mozilla's documentation makes the error go away.
Found by using Spidermonkey 1.8.5 instead of Duktape, which gives more detailed error messages: TypeError: fontFamily.includes is not a function
I still have not figured out why a bare call to katex.renderToString succeeds however.

[ccorn]

I'd propose to use .indexOf(...) !== -1 instead, for compatibility. Or including the polyfill.

[kevinbarabash]

Nice find!

[kevinbarabash]

We'll definitely want to fix this before the next release.

[kevinbarabash]

@ccorn if you have time, would definitely accept a PR to fix this.

[ccorn]

I'd like to do that, but I am not a Javascript expert, and I am not familiar with the build system around KaTeX. I can fork the repo, do a minimal .indexOf replacement, but then I would not know how to test that change. I could launch a PR and hope that some CI kicks in and that everything turns out well, but if not, then anyone familiar with KaTeX development should better take over. Also, compatibility issues such as this one are bound to happen again, and the preferred strategy might be different from fixing the instances as they are discovered.

[ccorn]

Hmm, there is a Makefile. Known territory. OK, I'll ssh into a server that has Node available, and then try and clone and mess around a bit.
Update: Oh wow, that was easy. A git clone, a git submodule update --init --recursive, and a make seem to do what I need. Great. I'm on my way.