Re: Some LaTeX commands, such as \boldsymbol and \hspace, can't be completed in AUCTeX.

[Denis Bitouzé]

Le 16/12/21 à 22h58, Arash Esbati a écrit :

>>> Agreed, but \includegraphics is indeed more an exception.
>>
>> In that case, could it be treated differently?
>
> I don't see how this can be achieved the way currently
> `LaTeX-arg-graphicx-includegraphics-key-val'[1] and the workhorse
> `TeX-read-key-val' are implemented.

Too bad!

>> IMO, `width=` should be the default completion and, in the rare cases
>> when the `=` is not wanted, it is easy to erase it.
>
> See above.

See above :)

>>> Ok, understood, so more like company.  For me, the way AUCTeX handles
>>> the completion was always great because I didn't need to hit the
>>> backslash which is hard to reach on a German keyboard.  I thought it is
>>> the same issue on French layout?
>>
>> Indeed, same issue. But what could be nice for `C-c C-m' would be to
>> type for us in the buffer (not the minibuffer) this damned backslash and
>> then, as soon as characters are typed, to provide a completion (à la
>> company).
>
> You can define a keystroke of your choice for backslash.  Not tested,
> but maybe something like this:
>
>     (defun my-backslash ()
>       (interactive)
>       (insert "\\"))
>
>     (with-eval-after-load "latex"
>       (define-key LaTeX-mode-map (kbd "C-c b") #'my-backslash))

Nice: already completion in point but limited. For instance: `\includeg`
completed in only `\includegraphics`

> Reg. completion à la company: There is company-auctex, have you tried
> it?[2]

Better but not as handy as TXS.

For instance: `\includeg` completed in `\includegraphics|[]{}` (where
`|` stands for the cursor position) whereas one could expect
`\includegraphics[•]{•}` where • are dummy placeholders, or better:
`\includegraphics[keyvals]{image file} ` where `keyvals` and `image
file` are non dummy placeholders.

Moreover, from `\includegraphics|[]{}`, moving the cursor in the
optional argument doesn't offer any completion.

> One other solution which looks promising to me is digestif[3].

Interesting. But not as handy as TXS.

For instance: `\includeg` completed in `\includegraphics|*[key-value
list]{filename}` (where `|` stands for the cursor position) but
`key-value list` and `filename` aren't placeholders.

Moreover, from `\includegraphics|*[key-value list]{filename}`, moving
the cursor in the optional argument doesn't offer any completion.

>>   ┌────
>>   │
>> https://htmlpreview.github.io/?https://github.com/texstudio-org/texstudio/master/utilities/manual/usermanual_en.html#SECTION110a
>>   └────
>> Sorry, the link above isn't accurate enough. What I wanted to point out
>> is not tables, but the screenshot about floating figures in section
>> ``2.10.1 Inserting a picture using a "wizard"'',. By contrast, what is
>> provided in the minibuffer by `C-c C-e fig TAB` is much less rich.
>
> Ah, I see.  Well, it seems to me that the wizard does all Emacs does
> through minibuffer query.

More than during my last try (I don't understand what happened). But:

- the user doesn't have an overview of what he will be asked,
- no label,
- he has to start a new process for `\includegraphics`.

> I can't tell if something like this is doable with Emacs.

At least something à la yasnippet would be helpfull. E.g., asking for
`figure` would insert:

\begin{figure}[⟨float position⟩]
  \centering
  \includegraphics[⟨key-value list⟩]{⟨filename⟩}
  \caption{⟨caption⟩}
  \label{⟨label⟩}
\end{figure}

where all the ⟨...⟩ are placeholders:

- easy to jump from one the the previous or next one,
- entirely selected once reached,
- with relevant completion when replaced.

>> Indeed, I always forget that the file has to be reload
>
> What do you mean by reload?  It should be sufficient to hit C-c C-n to
> reparse the file.

Indeed: much better but I always forget this keystroke. What would be
nice is that would be automatically done behind the scene as done by
TXS.

>> in order such macros are taken into account (no such a pain in TXS).
>>
>> What is missing compared to TXS is the placeholders: such placeholders
>> are:
>>
>> - highlighted: boxed, the selected one having a colored background,
>> - quite explicit in case of existing completion file and, in case of
>>   user macros such as the previous one, not so few explicit, here
>>   \congru[opt. arg1]{arg2}
>> - easy to reach: there are shortcuts to jump between them.
>
> Yupp, the company-style completion.  Have a look at the gif on digestif
> site, I think it does what you're looking for.

See above.

>>> Out of curiosity: Can TXS also parse new environments defined with
>>> fancyvrb package?  I mean something like this:
>>>
>>>     \documentclass{article}
>>>     \usepackage{fancyvrb}
>>>     \DefineVerbatimEnvironment{MyVerbatim}{Verbatim}{%
>>>       gobble=2,numbers=left,numbersep=2mm,
>>>       frame=lines,framerule=0.8mm}
>>>     \begin{document}
>>>     Try `C-c C-e MyVerbatim RET Optional fancyvrb keyvals RET'?
>>>     \end{document}
>>
>> Well:
>>
>> - with AUCTeX, even after the file being reload, no completion for
>>   `MyVerbatim` and the 1st RET inserts only the environment with the
>>   cursor inside: no "Optional", etc.
>
> When I put the code above in a .tex file, load it and hit C-c C-n, I see
> the following entries when looking at the variable
> `LaTeX-environment-list':
>
>  ("MyVerbatim" LaTeX-env-args
>   [TeX-arg-key-val
>    (LaTeX-fancyvrb-key-val-options)]
>   LaTeX-fancyvrb-env-reflabel-key-val)
>  ("MyVerbatim*" LaTeX-env-args
>   [TeX-arg-key-val
>    (LaTeX-fancyvrb-key-val-options)]
>   LaTeX-fancyvrb-env-reflabel-key-val)
>
> I.e., I get completion for optional arguments etc.

Still nothing with a new test :$

>> - with TXS, `\be + TAB` is completed into `\begin{` and, then, we have
>>   completion for all the known environments, including `MyVerbatim`
>>   (but nothing more).
>
> Thanks.

You're welcome!

> After all, I think we should start thinking about the direction AUCTeX
> should go in future.  It is a mature product, and adding styles for
> LaTeX packages is always a good thing, but that's only part of the
> answer.  Maybe company support is a first step.

Would be very, very nice! :)

Best.
-- 
Denis