Autoloading comment-forward

Autoloading comment-forward

[Ralf Angeli]

Hi,

RefTeX and AUCTeX use `comment-forward' when skipping over multiple TeX
groups.  The function -- while being quite useful -- is not autoloaded
at the moment.  Perhaps it should be?

The current situation is not a big problem for AUCTeX since it contains
a compatibility function for `comment-forward' anyway.  The reason for
this being XEmacs not containing `comment-forward' at all.  RefTeX
likely will get a compatibilty function as well due to the same reason.
Currently some RefTeX functionality is not working as long as
newcomment.el is not loaded.

-- 
Ralf

Re: Autoloading comment-forward

[Stefan Monnier]

> RefTeX and AUCTeX use `comment-forward' when skipping over multiple TeX
> groups.  The function -- while being quite useful -- is not autoloaded
> at the moment.  Perhaps it should be?

Before calling comment-forward, you need to call comment-normalize-vars
which *is* autoloaded.


        Stefan

Re: Autoloading comment-forward

[Ralf Angeli]

* Stefan Monnier (2007-07-07) writes:

>> RefTeX and AUCTeX use `comment-forward' when skipping over multiple TeX
>> groups.  The function -- while being quite useful -- is not autoloaded
>> at the moment.  Perhaps it should be?
>
> Before calling comment-forward, you need to call comment-normalize-vars
> which *is* autoloaded.

Oh, okay.  Thanks for the hint.  Is this documented somewhere?  Should
it be done everytime `comment-forward' is called?

I saw that other functions in newcomment.el call
`comment-normalize-vars' themselves.  Would it make sense if
`comment-forward' did the same (and was autoloaded)?

-- 
Ralf

Re: Autoloading comment-forward

[Richard Stallman]

It seems confusing to have both `comment-forward' and
`forward-comment'.  Is it possible to make `comment-forward' support
backward motion?  Then we could rename `forward-comment' to
`comment-forward-internal' and alias it to `comment-forward'.

Re: Autoloading comment-forward

[Stefan Monnier]

>>> RefTeX and AUCTeX use `comment-forward' when skipping over multiple TeX
>>> groups.  The function -- while being quite useful -- is not autoloaded
>>> at the moment.  Perhaps it should be?
>> 
>> Before calling comment-forward, you need to call comment-normalize-vars
>> which *is* autoloaded.

> Oh, okay.  Thanks for the hint.  Is this documented somewhere?  Should
> it be done everytime `comment-forward' is called?

Once is enough: It simply should be called at some point after the major
mode is set but before internal functions such as comment-forward are used.

> I saw that other functions in newcomment.el call
> `comment-normalize-vars' themselves.  Would it make sense if
> `comment-forward' did the same (and was autoloaded)?

Could be.  What kind of situation do you have in mind where this is useful?


        Stefan

Re: Autoloading comment-forward

[Ralf Angeli]

* Stefan Monnier (2007-07-08) writes:

>> Oh, okay.  Thanks for the hint.  Is this documented somewhere?  Should
>> it be done everytime `comment-forward' is called?
>
> Once is enough: It simply should be called at some point after the major
> mode is set but before internal functions such as comment-forward are used.

Okay.

>> I saw that other functions in newcomment.el call
>> `comment-normalize-vars' themselves.  Would it make sense if
>> `comment-forward' did the same (and was autoloaded)?
>
> Could be.  What kind of situation do you have in mind where this is useful?

I'm not sure I understand the question.  AUCTeX uses `comment-forward'
to skip empty or commented lines when traversing LaTeX macro arguments.
So for me `comment-forward' is not so much an internal function of
newcomment.el but rather one which is useful for other parts of Emacs or
packages like AUCTeX.  That's also why I think it makes sense for it to
be autoloaded.

Anyway, I will put a call to `comment-normalize-vars' into the
compatibility function for `comment-forward' in AUCTeX before
`comment-forward' is actually called (if available).  This will get rid
of the issue.  And the call to `comment-forward' in RefTeX has actually
only been present in my local copy.  RefTeX actually uses a different
mechanism for skipping over comments.

-- 
Ralf

Re: Autoloading comment-forward

[Stefan Monnier]

> I'm not sure I understand the question.  AUCTeX uses `comment-forward'
> to skip empty or commented lines when traversing LaTeX macro arguments.

Why doesn't it use forward-comment instead?


        Stefan

Re: Autoloading comment-forward

[Ralf Angeli]

* Stefan Monnier (2007-07-09) writes:

>> I'm not sure I understand the question.  AUCTeX uses `comment-forward'
>> to skip empty or commented lines when traversing LaTeX macro arguments.
>
> Why doesn't it use forward-comment instead?

Because it has to deal with Texinfo files as well.

-- 
Ralf

Re: Autoloading comment-forward

[Stefan Monnier]

>>> I'm not sure I understand the question.  AUCTeX uses `comment-forward'
>>> to skip empty or commented lines when traversing LaTeX macro arguments.
>> Why doesn't it use forward-comment instead?
> Because it has to deal with Texinfo files as well.

I see.  Then I think Richard's suggestion makes a lot of sense.  Except that
it means comment-forward (and hence newcomment.el) should not be autoloaded
but preloaded.  I'm all in favor of such a change (after all, the
functionality provided by newcomment.el used to be preloaded (it was in
simple.el) before I rewrote it).

Richard, do you have an objection to preloading newcomment.el (after
changing forward-comment to comment-forward-internal, adding backward
motion to comment-forward and making forward-comment an alias of
comment-forward).


        Stefan

Re: Autoloading comment-forward

[Richard Stallman]

    Richard, do you have an objection to preloading newcomment.el (after
    changing forward-comment to comment-forward-internal, adding backward
    motion to comment-forward and making forward-comment an alias of
    comment-forward).

It is ok if it can't cleanly be avoided.  But I'd rather look
for a way to avoid it, and I expect that should be possible.

For instance, the code of comment-forward could call a new
preloaded function equivalent to `comment-normalize-vars',
and `comment-normalize-vars' itself could call that new function
to do the work.  This way, calls to `comment-normalize-vars'
will still be a good way to cause newcomment.el to get loaded,
and everything that works now will continue to work.

Re: Autoloading comment-forward

[Stefan Monnier]

>     Richard, do you have an objection to preloading newcomment.el (after
>     changing forward-comment to comment-forward-internal, adding backward
>     motion to comment-forward and making forward-comment an alias of
>     comment-forward).

> It is ok if it can't cleanly be avoided.  But I'd rather look
> for a way to avoid it, and I expect that should be possible.

> For instance, the code of comment-forward could call a new
> preloaded function equivalent to `comment-normalize-vars',
> and `comment-normalize-vars' itself could call that new function
> to do the work.  This way, calls to `comment-normalize-vars'
> will still be a good way to cause newcomment.el to get loaded,
> and everything that works now will continue to work.

If someone else wants to do that, I can't prevent it, but spreading the code
like this just in the hope to avoid a few KB in the default build (most
people will load newcomment.el sooner or later since it's loaded as soon as
you hit M-; or use auto-fill-mode or ...), is just a waste of time IMNSHO.


        Stefan

Re: Autoloading comment-forward

[Richard Stallman]

    > For instance, the code of comment-forward could call a new
    > preloaded function equivalent to `comment-normalize-vars',
    > and `comment-normalize-vars' itself could call that new function
    > to do the work.  This way, calls to `comment-normalize-vars'
    > will still be a good way to cause newcomment.el to get loaded,
    > and everything that works now will continue to work.

    If someone else wants to do that, I can't prevent it, but spreading the code
    like this just in the hope to avoid a few KB in the default build (most
    people will load newcomment.el sooner or later since it's loaded as soon as
    you hit M-; or use auto-fill-mode or ...), is just a waste of time IMNSHO.

Why does auto-fill-mode load newcomment when used in Text mode?
That seems like a bug, since Text mode has no comment syntax.

Re: Autoloading comment-forward

[Stefan Monnier]

>> For instance, the code of comment-forward could call a new
>> preloaded function equivalent to `comment-normalize-vars',
>> and `comment-normalize-vars' itself could call that new function
>> to do the work.  This way, calls to `comment-normalize-vars'
>> will still be a good way to cause newcomment.el to get loaded,
>> and everything that works now will continue to work.

>     If someone else wants to do that, I can't prevent it, but spreading
>     the code like this just in the hope to avoid a few KB in the default
>     build (most people will load newcomment.el sooner or later since it's
>     loaded as soon as you hit M-; or use auto-fill-mode or ...), is just
>     a waste of time IMNSHO.

> Why does auto-fill-mode load newcomment when used in Text mode?

Not sure.  Maybe it doesn't in this case.


        Stefan

Re: Autoloading comment-forward

[Richard Stallman]

    > Why does auto-fill-mode load newcomment when used in Text mode?

    Not sure.  Maybe it doesn't in this case.

There is a bug.  The doc string of comment-line-break-function
says it is called only in comments, but in fact it is called
unconditionally.

I will make a change that takes a step in avoiding calling
it when you're not in a comment.