Adds \notni character ∌

[SuzanneSoy]

No description provided.

[khanbot]

Hey @jsmaniac,

Thanks for the PR! Mind signing our Contributor License Agreement?

When you've done so, go ahead and comment [clabot:check] and I'll check again.

Yours truly,
khanbot

[SuzanneSoy]

[clabot:check]

[khanbot]

CLA signature looks good 👍

[ronkok]

I don’t think this approach will work, because the KaTeX-Main font does not contain a glyph for U+220C. What could work is:

defineMacro("\\notni", "\\mathrel{\\rlap{\\kern{0.04em}/}{\\ni}}");

That results in a character that fits in pretty well:

[SuzanneSoy]

@ronkok Thanks a lot for the solution! I hadn't thought about the font issue.

[SuzanneSoy]

Note that the best might be to support \not\ni, which works with AMSMath IIRC, but I'm not sure if KaTeX supports defining macros with arguments. So \notni seems a good compromise (its meaning is clear, it exists in (non-AMS) LaTeX packages).

[ronkok]

@jsmaniac's original code would give a MathML result that is much superior to the macro that I proposed. Perhaps the thing to do is accept the macro for now, but put U+220C on to the font To Do list, along with \euro and \colonequals. Also, the macro code should include a TODO comment that suggests that the macro is temporary and that the font glyph should be used when it is available.

[SuzanneSoy]

@ronkok I added the TODO.

[edemaine]

@ronkok Thanks for the input! @jsmaniac Thanks for being responsive to the alternative approach!

If it's not too much trouble, could we see here output from texcmp (available in dockers/texcmp) comparing this symbol to one of the LaTeX versions of the symbol? (This involves adding a tiny test to test/screenshotter/ss_data.yaml.) Let me know if you have trouble running texcmp.

Incidentally, general \not support is in process/almost complete in #140. (Hi @kevinbarabash!) But I think there's still value to this PR, given that it is a supported command in some packages. I guess we could later change the macro to expand to just \not\ni... though that might have slightly inferior spacing?

P.S. You're currently failing tests because of long lines. Run npm test locally to see this output.

[kevinbarabash]

I'll have a look at what's left to do in #140 this evening.

[SuzanneSoy]

@edemaine I know about docker but don't use it in my regular development so I tend to forget the syntax, and google is proving itself unhelpful at telling me how to start a Dockerfile. I tried docker start dockers/texcmp and docker exec -ti dockers/texcmp /bin/sh, but both fail. Care to remind me what's the one-liner?

[edemaine]

@jsmaniac You can try sudo dockers/texcmp/texcmp.sh YourTestName which does the launching for you. (But see #708) Personally I've had more success with just running node dockers/texcmp/texcmp.js YourTestName.

[SuzanneSoy]

@edemaine Sorry, I tried running sudo dockers/texcmp/texcmp.sh Sqrt on my local machine, and inside a fresh Ubuntu 14.04 machine (with the default NodeJS and with NodeJS 6 too), and it fails in both cases with SyntaxError: Use of const in strict mode..

I also tried with the node dockers/texcmp/texcmp.js Sqrt command directly, and it throws module.js:471 throw err;.

You'll have to run texcmp, as I cannot.

[edemaine]

Regarding your latter error, you probably need to run npm install (which installs all dependencies) first. (Though hard to tell without seeing the actual error message, which is in the line after throw err;.)

I forgot you need to run screenshotter first, though, which I've only gotten to run using Docker. So you might have had trouble anyway.

Anyway, here is what I got with \usepackage{newtxmath} added to test/screenshotter/test.tex:

There's a difference here, but it seems like a consistent difference. Here's a comparison of \in\ni\notin\notni:

[edemaine]

@kevinbarabash Great! Perhaps we should wait to see if #140 is easy to finalize, and then modify this PR accordingly...

[kevinbarabash]

@jsmaniac I'm getting close with #140. With any luck it should be merged sometime this week and that will unblock this PR. Thanks for your patience.

[kevinbarabash]

@jsmaniac I finally got \not merged. Can you update your diff to us \not instead of /?

[SuzanneSoy]

@kevinbarabash Sure! I'll look into this next week if you don't mind. Thanks for the update!

[edemaine]

@jsmaniac A reminder on this -- should be straightforward. Also, you can use { and } instead of \bgroup and \egroup.

[kevinbarabash]

LGTM. Thanks for updating the diff. Sorry for making you wait so long for \not to be added.

[kevinbarabash]

I think this line should be removed b/c we don't have that symbol in the font yet.

[kevinbarabash]

Should the definition be wrapped in a \mathrel?

[kevinbarabash]

After doing some checking this doesn't look to be necessary as \ni is already a mathrel and the \not has no effect on the atom type.

[kevinbarabash]

@jsmaniac thanks for the PR. Sorry this took so long to get merged.

[SuzanneSoy]

Hi @kevinbarabash, sorry I didn't remove the defineSymbol quick enough. I thought it would allow typing the unicode character in the TeX source, and wanted to ask why it wasn't working when testing with the npm test server. I hadn't realised it was meant to translate the other way round (from the TeX command to the unicode char for MathML).

Thanks for fixing this and merging!

[kevinbarabash]

@jsmaniac sorry for being over eager to merge your PR. I think in order to support the unicode char for this you'll have to define a macro, see https://github.com/Khan/KaTeX/pull/950/files for some examples.