Re: [Dev-luatex] \mathchar feature request

Produces three Γ with lualatex TL2018. First one is missing char. And I can't reproduce bad example with the wrong symbol because it was a huge file and I can't find it now. \documentclass{article} \usepackage{amsfonts} \begin{document} \Gamma $\Gamma$ \Gamma $\Gamma$ \end{document} On 3/21/2019 4:14 PM, Marcel Krüger wrote:

---- On Thu, 21 Mar 2019 14:51:25 +0100 Linas Stonys wrote ----

...

the problem is that this command appears not in math mode. Then it takes \textfont for character to output. For example one forgot \Gamma to write as $\Gamma$ and he/she used some packages where \Gamma appeared in pdf. But such a file comes to publishing system where different packages and different fonts are used this \Gamma will silently become some other character.

Sorry, I think I am missing something here but I still do not see your point. If someone forgot to write $\Gamma$ instead of \Gamma and \Gamma is a \mathchar, then \Gamma is printed outside math mode with the appropriate \textfont. But if he would have written $\Gamma$, basically the same thing would have happened: The Gamma would still be written with the same \textfont, so the same slot + the same font => The same character. How can this silently become another character? The only difference is that there are no mathon/mathoff nodes around it and \everymath is not called.

Is the problem that `\everymath` would have changed your math fonts but this doesn't happen for non math mode mathchars?

...

lets say \mathaccent doesn't compile in text mode but \mathchar does.

What does \mathaccent has to do with this? Are you concerned that a \mathchar might be implemented using \mathaccent in other packages?

...

luatex sources just outputs: void math_char_in_text(mathcodeval mval) { .... p = new_char(fam_fnt(mval.family_value, text_size), mval.character_value); }

that's why I'm asking about some other subtype that I could control such situations.

I do not think that this would be a bad idea, I just do not see why you need it in this situation.

But if this LuaTeX team does not want to implement this, you might be able to detect such characters by scanning for glyph nodes using a math font and not being inside a mathon/mathoff pair

Best regards Marcel Krüger

...

...

subtypes already have a defined meaning

but you can consider setting an attribute in mathmode (with everymath) so that you can then (wherever) check if a character originates in math