Math symbols

Math symbols

[Augusto Stoffel]

Since Emacs now knows all the Emojis, I was wondering about the next
coolest thing, math symbols.  Would it make sense to make their TeX
names baked into Emacs, next to all other information show in
`describe-char?

There are around 2500 math symbols in Unicode.  Here is a comprehensive
listing of their TeX names:

    https://ctan.mirror.norbert-ruehl.de/macros/unicodetex/latex/unicode-math/unimath-symbols.pdf

This data could be shared by half a dozen different features.

Now, the answer to the question above might well be No, each of these
potential features should have their own copy of the data in their
preferred format.  If that's the case, then I would like to add the
following package to ELPA:

    https://github.com/astoff/unicode-math-input.el

PS: I don't know if anyone can make any serious intellectual property
claim on those character names, but since I obviously copied them from
some place, let me mention that, to my best knowledge, they originate
from the STIX project, see http://www.ams.org/STIX/.  There are no terms
of use on that page.

Re: Math symbols

[Andreas Schwab]

On Sep 23 2022, Augusto Stoffel wrote:

> Since Emacs now knows all the Emojis, I was wondering about the next
> coolest thing, math symbols.  Would it make sense to make their TeX
> names baked into Emacs, next to all other information show in
> `describe-char?

There is already the TeX input method.

-- 
Andreas Schwab, schwab@linux-m68k.org
GPG Key fingerprint = 7578 EB47 D4E5 4D69 2510  2552 DF73 E780 A9DA AEC1
"And now for something completely different."

Re: Math symbols

[Augusto Stoffel]

On Fri, 23 Sep 2022 at 21:43, Andreas Schwab wrote:

> On Sep 23 2022, Augusto Stoffel wrote:
>
>> Since Emacs now knows all the Emojis, I was wondering about the next
>> coolest thing, math symbols.  Would it make sense to make their TeX
>> names baked into Emacs, next to all other information show in
>> `describe-char?
>
> There is already the TeX input method.

Yes, but it's pretty incomplete.  It also has some translations that
interfere with normal typing, such as ^1 → ¹.

Re: Math symbols

[Eli Zaretskii]

> From: Augusto Stoffel <arstoffel@gmail.com>
> Cc: emacs-devel@gnu.org
> Date: Fri, 23 Sep 2022 21:54:11 +0200
> 
> On Fri, 23 Sep 2022 at 21:43, Andreas Schwab wrote:
> 
> > On Sep 23 2022, Augusto Stoffel wrote:
> >
> >> Since Emacs now knows all the Emojis, I was wondering about the next
> >> coolest thing, math symbols.  Would it make sense to make their TeX
> >> names baked into Emacs, next to all other information show in
> >> `describe-char?
> >
> > There is already the TeX input method.
> 
> Yes, but it's pretty incomplete.  It also has some translations that
> interfere with normal typing, such as ^1 → ¹.

Then let's improve/expand/fix it.

Or are you saying that input method is not up to the job?  If so,
please explain why you think so.

Re: Math symbols

[Augusto Stoffel]

On Fri, 23 Sep 2022 at 23:14, Eli Zaretskii wrote:

>> From: Augusto Stoffel <arstoffel@gmail.com>
>> Cc: emacs-devel@gnu.org
>> Date: Fri, 23 Sep 2022 21:54:11 +0200
>> 
>> On Fri, 23 Sep 2022 at 21:43, Andreas Schwab wrote:
>> 
>> > On Sep 23 2022, Augusto Stoffel wrote:
>> >
>> >> Since Emacs now knows all the Emojis, I was wondering about the next
>> >> coolest thing, math symbols.  Would it make sense to make their TeX
>> >> names baked into Emacs, next to all other information show in
>> >> `describe-char?
>> >
>> > There is already the TeX input method.
>> 
>> Yes, but it's pretty incomplete.  It also has some translations that
>> interfere with normal typing, such as ^1 → ¹.
>
> Then let's improve/expand/fix it.
>
> Or are you saying that input method is not up to the job?  If so,
> please explain why you think so.

Yes, it can be improved, and the first question to answer here is
whether Emacs will have the TeX symbol names built-in or each package
should contain its own data.

Now, I personally dislike certain details about the built-in TeX IM
(mainly the extra symbols starting with "^" and "_").  So I am not too
inclined to work on it.

Re: Math symbols

[Eli Zaretskii]

> From: Augusto Stoffel <arstoffel@gmail.com>
> Cc: schwab@linux-m68k.org,  emacs-devel@gnu.org
> Date: Fri, 23 Sep 2022 22:54:23 +0200
> 
> >> > There is already the TeX input method.
> >> 
> >> Yes, but it's pretty incomplete.  It also has some translations that
> >> interfere with normal typing, such as ^1 → ¹.
> >
> > Then let's improve/expand/fix it.
> >
> > Or are you saying that input method is not up to the job?  If so,
> > please explain why you think so.
> 
> Yes, it can be improved, and the first question to answer here is
> whether Emacs will have the TeX symbol names built-in or each package
> should contain its own data.

Maybe I don't understand in enough detail what you want to propose.
What other features, except input methods, may need the TeX names of
these characters and for what purposes?  (And the input method doesn't
supply names of the symbols, it provides strings that are converted to
characters.)

> Now, I personally dislike certain details about the built-in TeX IM
> (mainly the extra symbols starting with "^" and "_").  So I am not too
> inclined to work on it.

We could, I think, add an alternative input method to latin-ltx.el,
which will be identical to the existing one, except for changes in the
above-mentioned areas.

Re: Math symbols

[Augusto Stoffel]

On Sat, 24 Sep 2022 at 09:05, Eli Zaretskii wrote:

> Maybe I don't understand in enough detail what you want to propose.
> What other features, except input methods, may need the TeX names of
> these characters and for what purposes?

There can be a variant of `insert-char' that reads the TeX name instead
of the Unicode name.  Also a hierarchical menu like `emoji-insert'.
Then tex-mode probably wants to suggest those symbols in its
completion-at-point.  And so on.

> We could, I think, add an alternative input method to latin-ltx.el,
> which will be identical to the existing one, except for changes in the
> above-mentioned areas.

This is done: it's the package I linked in my original message.

An alternative is to make the TeX IM customizable.  But this, again,
raises the question: should there be an alist of 2500 symbols defined in
latin-ltx.el, or should this info be stored in some shared place so
various features can benefit from it?  (I'm not sure, and I'm not
advocating for either.)

Re: Math symbols

[Eli Zaretskii]

> From: Augusto Stoffel <arstoffel@gmail.com>
> Cc: schwab@linux-m68k.org,  emacs-devel@gnu.org
> Date: Sat, 24 Sep 2022 08:36:30 +0200
> 
> On Sat, 24 Sep 2022 at 09:05, Eli Zaretskii wrote:
> 
> > Maybe I don't understand in enough detail what you want to propose.
> > What other features, except input methods, may need the TeX names of
> > these characters and for what purposes?
> 
> There can be a variant of `insert-char' that reads the TeX name instead
> of the Unicode name.  Also a hierarchical menu like `emoji-insert'.
> Then tex-mode probably wants to suggest those symbols in its
> completion-at-point.  And so on.

I'd prefer not to add ways of input beyond what we already have,
unless the IM framework cannot support it.  Since the TeX input method
demonstrates (I think) that IM framework can support these symbols,
introducing yet another way of inputing those characters sounds like
unnecessary complication to me.

Emoji input may justify a special solution because there are no
acceptable short names for its large set of sequences (which grows
with each new version of Unicode), something that is not true for the
math symbols.  And if you'd like to offer an input facility similar to
Emoji, i.e. one that uses transient.el and the Unicode names of the
characters, I don't think I'd object (although the question of
necessity and whether it's justified given the existing IM would still
be pertinent).  But that's not what your package does, AFAICT: it's a
proper input method, just a different one.

As for completion-at-point: couldn't that use the input-method as
well?  Isn't it reasonable to assume that people who use tex-mode also
uset the TeX input method?

> > We could, I think, add an alternative input method to latin-ltx.el,
> > which will be identical to the existing one, except for changes in the
> > above-mentioned areas.
> 
> This is done: it's the package I linked in my original message.

I meant to have such an alternative input method as part of
latin-ltx.el.

If you don't want to add this to Emacs, that's fine as well.  I guess
I misinterpreted your original proposal, then

> An alternative is to make the TeX IM customizable.

I propose to have a separate input method in latin-ltx.el.  Call it
TeX-alternative or something.  I think this is better than somehow
customizing a single IM.

> But this, again, raises the question: should there be an alist of
> 2500 symbols defined in latin-ltx.el, or should this info be stored
> in some shared place so various features can benefit from it?  (I'm
> not sure, and I'm not advocating for either.)

I don't think it matters where the names are stored.  Strings used by
input methods are not usually shared, but if you want to suggest to
have a list of names somewhere, and then code which generates the
relevant input methods from that list of names (and other features
that use the same list), I don't think I'd object, provided that we
find a good solution for avoiding to load that until it's actually
used.

Re: Math symbols

[Augusto Stoffel]

On Sat, 24 Sep 2022 at 10:21, Eli Zaretskii wrote:

> I'd prefer not to add ways of input beyond what we already have,
> unless the IM framework cannot support it.  Since the TeX input method
> demonstrates (I think) that IM framework can support these symbols,
> introducing yet another way of inputing those characters sounds like
> unnecessary complication to me.

The input method is good when, say, you want to add some simple formulas
in plain text in an email.

There are other use cases, though.  Suppose you are writing a paper in
TeX and you want to insert (the command that generates) a × with a box
around it.  How do you figure out the command is called “\boxtimes”?
The analogue of C-x 8 RET for TeX names is helpful for this.

Re: Math symbols

[Eli Zaretskii]

> From: Augusto Stoffel <arstoffel@gmail.com>
> Cc: schwab@linux-m68k.org,  emacs-devel@gnu.org
> Date: Sat, 24 Sep 2022 09:41:09 +0200
> 
> There are other use cases, though.  Suppose you are writing a paper in
> TeX and you want to insert (the command that generates) a × with a box
> around it.  How do you figure out the command is called “\boxtimes”?
> The analogue of C-x 8 RET for TeX names is helpful for this.

The IM framework's response to this challenge is twofold:

  . typing characters that could be part of a recognized input
    sequence shows in the echo area the candidates for the next
    character;
  . C-h C-\ RET shows the full chart of the supported sequences and
    the corresponding characters.

The latter possibility is a close equivalent of "C-x 8 RET" for the
case that the use doesn't know the name of the symbol even
approximately.  (In the "C-x 8 RET" case, the user would just type
"C-x 8 RET TAB" to get the list of possible completions.)

Re: Math symbols

[Augusto Stoffel]

On Sat, 24 Sep 2022 at 10:57, Eli Zaretskii wrote:

>   . C-h C-\ RET shows the full chart of the supported sequences and
>     the corresponding characters.

IMO this is not more helpful than opening that big PDF listing in a PDF
viewer and searching through it.  (It's also not less helpful; it's okay
if you only need it once or twice.)

Re: Math symbols

[Eli Zaretskii]

> From: Augusto Stoffel <arstoffel@gmail.com>
> Cc: schwab@linux-m68k.org,  emacs-devel@gnu.org
> Date: Sat, 24 Sep 2022 16:02:11 +0200
> 
> On Sat, 24 Sep 2022 at 10:57, Eli Zaretskii wrote:
> 
> >   . C-h C-\ RET shows the full chart of the supported sequences and
> >     the corresponding characters.
> 
> IMO this is not more helpful than opening that big PDF listing in a PDF
> viewer and searching through it.  (It's also not less helpful; it's okay
> if you only need it once or twice.)

That's okay, IMO, for the case where the user doesn't have any idea
how the symbol is called in TeX.

Re: Math symbols

[Juri Linkov]

>>   . C-h C-\ RET shows the full chart of the supported sequences and
>>     the corresponding characters.
>
> IMO this is not more helpful than opening that big PDF listing in a PDF
> viewer and searching through it.  (It's also not less helpful; it's okay
> if you only need it once or twice.)

You can also use completion during IM input, e.g. after 'C-\ tex RET',
'\ TAB TAB TAB p TAB i TAB …'

Re: Math symbols

[Stefan Monnier]

> Yes, but it's pretty incomplete.  It also has some translations that
> interfere with normal typing, such as ^1 → ¹.

Indeed, e.g. I use a tweaked version:

    (defun my-quail-activate-hook ()
      (unless (member (quail-name) sm-quail-activate-hook-done)
        (push (quail-name) sm-quail-activate-hook-done)
        (when (member (quail-name) '("TeX"))
          ;; Copy the "_..." bindings to "\_...".
          (setf (alist-get ?_ (cdr (alist-get ?\\ (quail-map))))
                (alist-get ?_ (quail-map)))
          ;; Remove the "_..." bindings.
          (setf (alist-get ?_ (quail-map)) nil)
          )))
    (add-hook 'quail-activate-hook #'my-quail-activate-hook)


-- Stefan

Re: Math symbols

[Eli Zaretskii]

> From: Stefan Monnier <monnier@iro.umontreal.ca>
> Cc: Andreas Schwab <schwab@linux-m68k.org>,  emacs-devel@gnu.org
> Date: Fri, 23 Sep 2022 16:56:49 -0400
> 
> > Yes, but it's pretty incomplete.  It also has some translations that
> > interfere with normal typing, such as ^1 → ¹.
> 
> Indeed, e.g. I use a tweaked version:

Why not add a variant input method to latin-ltx.el, so everyone could
enjoy that?

Re: Math symbols

[Augusto Stoffel]

On Sat, 24 Sep 2022 at 09:06, Eli Zaretskii wrote:

>> From: Stefan Monnier <monnier@iro.umontreal.ca>
>> Cc: Andreas Schwab <schwab@linux-m68k.org>,  emacs-devel@gnu.org
>> Date: Fri, 23 Sep 2022 16:56:49 -0400
>> 
>> > Yes, but it's pretty incomplete.  It also has some translations that
>> > interfere with normal typing, such as ^1 → ¹.
>> 
>> Indeed, e.g. I use a tweaked version:
>
> Why not add a variant input method to latin-ltx.el, so everyone could
> enjoy that?

Everyone can enjoy ELPA packages, right?

Re: Math symbols

[Eli Zaretskii]

> From: Augusto Stoffel <arstoffel@gmail.com>
> Cc: Stefan Monnier <monnier@iro.umontreal.ca>,  schwab@linux-m68k.org,
>   emacs-devel@gnu.org
> Date: Sat, 24 Sep 2022 08:37:14 +0200
> 
> On Sat, 24 Sep 2022 at 09:06, Eli Zaretskii wrote:
> 
> >> From: Stefan Monnier <monnier@iro.umontreal.ca>
> >> Cc: Andreas Schwab <schwab@linux-m68k.org>,  emacs-devel@gnu.org
> >> Date: Fri, 23 Sep 2022 16:56:49 -0400
> >> 
> >> > Yes, but it's pretty incomplete.  It also has some translations that
> >> > interfere with normal typing, such as ^1 → ¹.
> >> 
> >> Indeed, e.g. I use a tweaked version:
> >
> > Why not add a variant input method to latin-ltx.el, so everyone could
> > enjoy that?
> 
> Everyone can enjoy ELPA packages, right?

Yes.  But then what was your original proposal and questions about?
You said there:

> I was wondering about the next coolest thing, math symbols.  Would
> it make sense to make their TeX names baked into Emacs, next to all
> other information show in `describe-char?

Was this also about adding something to ELPA?  I just assumed that
"baked into Emacs" means to have something in core.

Re: Math symbols

[Augusto Stoffel]

On Sat, 24 Sep 2022 at 10:24, Eli Zaretskii wrote:

>> From: Augusto Stoffel <arstoffel@gmail.com>
>> Cc: Stefan Monnier <monnier@iro.umontreal.ca>,  schwab@linux-m68k.org,
>>   emacs-devel@gnu.org
>> Date: Sat, 24 Sep 2022 08:37:14 +0200
>> 
>> On Sat, 24 Sep 2022 at 09:06, Eli Zaretskii wrote:
>> 
>> >> From: Stefan Monnier <monnier@iro.umontreal.ca>
>> >> Cc: Andreas Schwab <schwab@linux-m68k.org>,  emacs-devel@gnu.org
>> >> Date: Fri, 23 Sep 2022 16:56:49 -0400
>> >> 
>> >> > Yes, but it's pretty incomplete.  It also has some translations that
>> >> > interfere with normal typing, such as ^1 → ¹.
>> >> 
>> >> Indeed, e.g. I use a tweaked version:
>> >
>> > Why not add a variant input method to latin-ltx.el, so everyone could
>> > enjoy that?
>> 
>> Everyone can enjoy ELPA packages, right?
>
> Yes.  But then what was your original proposal and questions about?
> You said there:
>
>> I was wondering about the next coolest thing, math symbols.  Would
>> it make sense to make their TeX names baked into Emacs, next to all
>> other information show in `describe-char?
>
> Was this also about adding something to ELPA?  I just assumed that
> "baked into Emacs" means to have something in core.

I meant that I see two mutually exclusive possibilities:

A. Add my package to ELPA.
B. Make the symbols names built into Emacs somehow.

In case B my package becomes irrelevant because there will be better
ways to implement its functionality in Emacs directly.

Re: Math symbols

[Eli Zaretskii]

> From: Augusto Stoffel <arstoffel@gmail.com>
> Cc: monnier@iro.umontreal.ca,  schwab@linux-m68k.org,  emacs-devel@gnu.org
> Date: Sat, 24 Sep 2022 09:45:35 +0200
> 
> >> I was wondering about the next coolest thing, math symbols.  Would
> >> it make sense to make their TeX names baked into Emacs, next to all
> >> other information show in `describe-char?
> >
> > Was this also about adding something to ELPA?  I just assumed that
> > "baked into Emacs" means to have something in core.
> 
> I meant that I see two mutually exclusive possibilities:
> 
> A. Add my package to ELPA.
> B. Make the symbols names built into Emacs somehow.
> 
> In case B my package becomes irrelevant because there will be better
> ways to implement its functionality in Emacs directly.

I think my responses covered what I have to say about B.

Re: Math symbols

[Visuwesh]

[சனி செப்டம்பர் 24, 2022] Augusto Stoffel wrote:

> I meant that I see two mutually exclusive possibilities:
>
> A. Add my package to ELPA.
> B. Make the symbols names built into Emacs somehow.
>
> In case B my package becomes irrelevant because there will be better
> ways to implement its functionality in Emacs directly.

For (B), we can create a "core" TeX input method that only has
translations for \thingies then use the decode map generated by Quail
(see quail-decode-map) to create a "fancy" TeX input method which
inherits the "core" IM and adds other stuff like ^1, ---, etc.
(Currently, CREATE-DECODE-MAP is set to nil for the TeX IM.)

Re: Math symbols

[Eric S Fraga]

On Friday, 23 Sep 2022 at 21:54, Augusto Stoffel wrote:
> It also has some translations that interfere with normal typing, such
> as ^1 → ¹.

These are annoying; even more so is that you cannot back up in the input
of a symbol when you've made a mistake, e.g. typing "\alpja BS BS ha"
doesn't give you α.  The combination of this and issues with _ and ^
make the TeX IM not frictionless.

-- 
Eric S Fraga via gnus (Emacs 29.0.50 2022-09-22) on Debian 11.4

Re: Math symbols

[Eli Zaretskii]

> From: Eric S Fraga <e.fraga@ucl.ac.uk>
> Date: Sun, 25 Sep 2022 12:14:46 +0100
> 
> On Friday, 23 Sep 2022 at 21:54, Augusto Stoffel wrote:
> > It also has some translations that interfere with normal typing, such
> > as ^1 → ¹.
> 
> These are annoying; even more so is that you cannot back up in the input
> of a symbol when you've made a mistake, e.g. typing "\alpja BS BS ha"
> doesn't give you α.  The combination of this and issues with _ and ^
> make the TeX IM not frictionless.

Patches are welcome to make this better, thanks.

Re: Math symbols

[Eric S Fraga]

On Sunday, 25 Sep 2022 at 16:55, Eli Zaretskii wrote:
> Patches are welcome to make this better, thanks.

Sure but likely beyond my meagre elisp-fu...

-- 
Eric S Fraga via gnus (Emacs 29.0.50 2022-09-22) on Debian 11.4

Re: Math symbols

[Stefan Monnier]

Augusto Stoffel [2022-09-23 21:25:36] wrote:
>     https://ctan.mirror.norbert-ruehl.de/macros/unicodetex/latex/unicode-math/unimath-symbols.pdf
> This data could be shared by half a dozen different features.

Is this data available in more convenient form (for easy automatic
processing)?

If so, we could add it to Emacs's source code, then auto-build a `.el`
file out of it, and change latin-ltx.el to make use of it.


        Stefan

Re: Math symbols

[Augusto Stoffel]

On Sat, 24 Sep 2022 at 09:34, Stefan Monnier wrote:

> Augusto Stoffel [2022-09-23 21:25:36] wrote:
>>     https://ctan.mirror.norbert-ruehl.de/macros/unicodetex/latex/unicode-math/unimath-symbols.pdf
>> This data could be shared by half a dozen different features.
>
> Is this data available in more convenient form (for easy automatic
> processing)?
>
> If so, we could add it to Emacs's source code, then auto-build a `.el`
> file out of it, and change latin-ltx.el to make use of it.
>
>
>         Stefan

I believe this is the primary source:

  http://www.ams.org/STIX/bnb/stix-tbl.ascii-2006-10-20

More info, including the layout description, here:

  http://www.ams.org/STIX/