Re: Indenting the code written in square brackets as optional parameters

[Sašo Živanović]

Hi!

Ikumi Keita je 8. 03. 22 ob 17:59 napisal:
> [ Including Sašo in To:, not to diverge discussion in two mail lists. ]
Thanks! I have subscribed to auctex-devel now, to make everybody's life 
a bit easier. ;-)

> > > > > > > Sašo Živanović <saso.zivanovic@guest.arnes.si> writes:
> > > > Especially, how does your patch work in the presence of valid interval
> > > > notation like [1,10[ or [1,10)?  Will that break indentation for the
> > > > remainder of the whole document?
> > > For [1,10[  yes it will break it
> > > but closing braces in the comment can prevent that:
> > > [1,10[%]]
> >
> > Unlike my patch, yours counts commented braces, which changes the
> > current AUCTeX behavior. I'm a bit afraid that breaks indentation in
> > other contexts. (Hmm, but such cases must be rather rare, so it wouldn't
> > matter much?)
> Arash Esbati je 8. 03. 22 ob 19:34 napisal:
> I agree, we shouldn't change the current behavior big time.
And I also agree, the default behaviour should stay the same.

I have to admit I have been using my own patch for such a long time I 
actually forgot that counting commented braces in not the default AuCTeX 
behaviour! And also, that I'm not sure whether I changed this behaviour 
on purpuse or by accident ;-)

However, as a long-time user of the []-based indentation, I can 
confidently say that counting *brackets []* (not braces {}) in comments 
is quite crucial for user's happiness. And not only in math mode, as 
noted by others in this thread, also in text (well, programming) mode, 
as in:
\def\memoize{%
  \@ifnextchar[\memoize@opt\memoize@noopt%]
}

There might even be another usage case for taking commented delimites 
into account, indenting \ifs:
\ifx ... %[
  ...
\fi %]
(I'm totally alergic to non-indented \ifs, and have even tried to hack 
into that, but due to the coexistence of TeX- and LaTeX-style \ifs, this 
seems impossible to implement in general.)

Summing up: while I find []-indenting indispensable for Forest trees, 
and TikZ in general, I would definitely want to have an way to override 
it for specific cases, and counting stuff in comments seems the simplest 
way to implement the override.

On the other hand, I have a hard time imagining a situation where I 
would want to override a {}-based indent via a comment, so I would say 
that the AuCTeX's current behaviour to ignore these in comments makes 
perfect sense and should be kept (even without the historical reasons). 
So I would propose that the extended indenting mechanism would ignore {} 
in comments but count all other delimiters.

Finally, I believe that if we're about to extend the indenting 
mechanism, it makes sense to implement in generally enough to allow for 
arbitrary, and configurable, delimiters. For one, we can then set the 
default to {} to guarantee unchanged default behaviour. Secondly, I got 
quite attached to seeing stuff like this:
------
develop a syntactic counterpart to conservativity (recall the
  discussion of generalized quantifiers in
  section~\ref{cha:generalized-quantifier-theory}), and ultimately our
understanding of determiners and the nature of quantification in natural
language itself.
------
And perhaps someone will develop a package using <> as delimiters for 
whatever, etc.

> Maybe I should modify `TeX-brace-count-line' instead of introducing a
> new function `LaTeX-brace-count-line'. Currently I'm not sure which is
> better.
Grep says that `TeX-brace-count-line' is only used in `latex.el'. So I 
guess that modifying this function should be better.
(By the way, in my patch, I didn't modify it as much as rewrite it 
completely. Seeing Ikumi's patch, I realize that was certainly an 
overkill stemming from my inadequate Lisp knowledge. So thanks Ikumi for 
the opportunity to learn some Lisp!)

Best,
Sašo