Change in fill-nobreak-predicate

Change in fill-nobreak-predicate

[Chong Yidong]

In 21.3, the variable fill-nobreak-predicate stored a function symbol.
This was changed in CVS, long ago, into a hook:

2001-10-30  Stefan Monnier  <monnier@cs.yale.edu>

	* textmodes/fill.el (sentence-end-double-space)
	(sentence-end-without-period): Move to paragraphs.el.
	(fill-indent-according-to-mode): Change default to t.
	(fill-context-prefix): Simplify control-flow and use a more
	sophisticated merge that unifies both previous checks.
	(fill-single-word-nobreak-p, fill-french-nobreak-p): New funs.
	(fill-nobreak-predicate): Make it into a defcustom'd hook.

This breaks third-party code that relies on fill-nobreak-predicate (such
as longlines.el), but never mind; the problem is that the NEWS entry does
not give any indication of this:

** You can now customize fill-nobreak-predicate to control where
filling can break lines.  We provide two sample predicates,
fill-single-word-nobreak-p and fill-french-nobreak-p.

+++

This makes it sound like the only change was to make it customizable, when
in fact its data type has changed.

The entry in lispref/text.texi should also be corrected:

@defvar fill-nobreak-predicate
This variable gives major modes a way to specify not to break a line at
certain places.  Its value should be a function.  This function is
called during filling, with no arguments and with point located at the
place where a break is being considered.  If the function returns
non-@code{nil}, then the line won't be broken there.
@end defvar

Re: Change in fill-nobreak-predicate

[Stefan Monnier]

> This breaks third-party code that relies on fill-nobreak-predicate (such
> as longlines.el), but never mind;

Actually, hooks (for some old historical backward compatibility reason) can
contain either a list of functions or a single function.
So fill-nobreak-predicate can still be set to a single function, as in
the past.

I do not doubt that my change introduced some incompatibility, but most
normal uses should work just as well as before.
If not, please give us more information about the problem.


        Stefan

Re: Change in fill-nobreak-predicate

[Chong Yidong]

>> This breaks third-party code that relies on fill-nobreak-predicate (such
>> as longlines.el), but never mind;
>
> Actually, hooks (for some old historical backward compatibility reason)
> can contain either a list of functions or a single function.
> So fill-nobreak-predicate can still be set to a single function, as in
> the past.
>
> I do not doubt that my change introduced some incompatibility, but most
> normal uses should work just as well as before.
> If not, please give us more information about the problem.

I had code calling (funcall fill-nobreak-predicate), which failed because
some modes, such as TeX mode, use the new hook mechanism and set
fill-nobreak-predicate to a list. Changing my code to
(run-hook-with-args-until-success fill-nobreak-predicate) fixed it.

Another thing: maybe the doc-string of run-hook-with-args-* should reflect
the fact that hook can be a function:

 run-hook-with-args-until-success is a built-in function in `C source code'.
 (run-hook-with-args-until-success hook &rest args)

 Run hook with the specified arguments args.
 hook should be a symbol, a hook variable.  Its value should
 be a list of functions.  We call those functions, one by one,
 passing arguments args to each of them, until one of them
 returns a non-nil value.  Then we return that value.
 If all the functions return nil, we return nil.

The docstring for run-hook-with-args is already correct:

 run-hook-with-args is a built-in function in `C source code'.
 (run-hook-with-args hook &rest args)

 Run hook with the specified arguments args.
 hook should be a symbol, a hook variable.  If hook has a non-nil
 value, that value may be a function or a list of functions to be
 called to run the hook.  If the value is a function, it is called with
 the given arguments and its return value is returned.  If it is a list
 of functions, those functions are called, in order,
 with the given arguments args.

Re: Change in fill-nobreak-predicate

[Richard Stallman]

    Another thing: maybe the doc-string of run-hook-with-args-* should reflect
    the fact that hook can be a function:

Thanks.

Re: Change in fill-nobreak-predicate

[Richard Stallman]

Thanks.