[AUCTeX] Re: Comment filling : I would like to make a contribution

[Vincent Belaïche]

Ralf Angeli a écrit :
> > > I don't think that this would be enough, because even though by some 
> > > tampering with variable `paragraph-separate' one line of preprocessor 
> > > code could be recognized as a single paragraph, this would not prevent 
> > > that this line may be folded into two lines if it is too long, which 
> > > would cause the absence of an Eupp end-mark for the first line, and of 
> > > an Eupp start mark for the second line.
> > >       
>
> The idea was not to make each line a of the preprocessor code its own
> paragraph separator, but to establish some separation between a normal
> comment and the preprocessor code.  Then when you fill a comment
> paragraph it will not mess up the preprocessor code.
>
> The solution for the preprocessor code is simply not to type `M-q' when
> point is in it.  In case there should be more control about how the code
> is to be indented or filled, using a separate major mode which gets
> switched on if the preprocessor code is entered and off when left would
> probably be a better idea than shoehorning filling restrictions into
> AUCTeX.
>
>   
Then there are two drawbacks for that
1) the user has to care not to make a M-q when the point is in the 
pre-preprocessor code
2) the pre-processor code will split paragraphs into pieces, which I 
think is awkward.
> I can still see some use for the definition of sections which should be
> spared from filling (especially if a larger region is to be filled), but
> perhaps there should better be a general mechanism in Emacs which could
> be used or adopted by AUCTeX once it is there.
>
>   
I do not object to this suggestion, however what you say is 
contradictory with the fact that currently AUCTeX does not rely on Emacs 
general filling procedure : AUCTeX uses function 
`LaTeX-fill-region-as-para-do' instead of function 
`fill-region-as-paragraph', just as if people wanted AUCTeX to be 
independant from Emacs fill.el.

If AUCTeX would use fill-region-as-paragraph, then I could modify this 
function to make into it some support for unfillable sections, but as 
AUCTeX does not use it (but use a copy `LaTeX-fill-region-as-para-do' of 
it), then I had rather make the checks and skips one level above (in 
function `LaTeX-fill-region-as-paragraph') as this seemed easier.

What you suggest would need that AUCTeX filling mechanism should be a 
layer on top of some general filling kernel provided by Emacs, whereas 
it seems to be currently a complete overload of Emacs filling procedure 
(w/o reuse).

*Please confirm your intention* :
do you object to my contribution ?
If yes, you seem however to see some need for unfillable sections, could 
you please tell me how I could make this function available in an 
acceptable manner, as I do need for it, and I am sure that other people 
may need for it too.

> <URL:http://www.gnu.org/software/auctex/mailing-lists.html>
>
>   
Thanks for the link.

By the way I have checked that there is no infinite loop in the code 
that I have provided, because when LaTeX-fill-region-as-paragraph is 
called, the calling code has already made sure that a paragraph does not 
end at buffer end, ie a newline has been inserted before buffer end when 
need be.

Still I could make a modification in the code that I have already 
provided to you in order to make it more robust (actually I have already 
made this modification, and I can make it available on request). But I 
am not sure whether this is needed, because many other part of the code 
may assume that the last char in a paragraph is not at eob, and that a 
(beginning-of-line 2) would always move one line forward.

BR,
 Vincent.