Re: /srv/bzr/emacs/trunk r101829: (regexp-opt): Add `symbols' mode.

Re: /srv/bzr/emacs/trunk r101829: (regexp-opt): Add `symbols' mode.

[Glenn Morris]

Miles Bader wrote:

> revno: 101829
> committer: Miles Bader <miles@gnu.org>
> branch nick: trunk
> timestamp: Thu 2010-10-07 16:24:21 +0900
> message:
>   (regexp-opt): Add `symbols' mode.

Please see

http://debbugs.gnu.org/cgi/bugreport.cgi?bug=4392#12

Re: /srv/bzr/emacs/trunk r101829: (regexp-opt): Add `symbols' mode.

[Miles Bader]

Glenn Morris <rgm@gnu.org> writes:
>>   (regexp-opt): Add `symbols' mode.
>
> http://debbugs.gnu.org/cgi/bugreport.cgi?bug=4392#12

Oh brother.  I thought "oh there's no point in discussing this first
because it's _so obvious and trivial_..."

I made that chance in response to seeing some (not in emacs) elisp
code that -- incorrectly, since it's a programming language mode --
switched from concatenating \\< and \\> to using `words'.  The
result of course was exactly identical, but apparently the mode
author felt that using `words' was more blessed or something.

Thus I think regexp-opt should have support/advertise both
`words' and `symbols' or it should do neither, because having
only `words' lends a subtle advocacy.

Alternatively, the support could be left in, but the docstring and
documentation changed to say something like:

   "... If you want to match only whole words or symbols, then
   surround the return value with \\< and \\> or \\_< and \\_>
   (the latter being more appropriate for matching programming
   language keywords and the like), using `concat'.  Previously
   this function advertised a magic value of `words' for the PAREN
   argument, but it is now deprecated, and should not be used."

-Miles

-- 
Genealogy, n. An account of one's descent from an ancestor who did not
particularly care to trace his own.

Re: /srv/bzr/emacs/trunk r101829: (regexp-opt): Add `symbols' mode.

[Stefan Monnier]

>>> (regexp-opt): Add `symbols' mode.
>> http://debbugs.gnu.org/cgi/bugreport.cgi?bug=4392#12
> Oh brother.  I thought "oh there's no point in discussing this first
> because it's _so obvious and trivial_..."

I didn't bother reacting, indeed.

> Thus I think regexp-opt should have support/advertise both
> `words' and `symbols' or it should do neither, because having
> only `words' lends a subtle advocacy.

How'bout we set-advertised-calling-convention so as to obsolete the
second argument?


        Stefan

Re: /srv/bzr/emacs/trunk r101829: (regexp-opt): Add `symbols' mode.

[Miles Bader]

Stefan Monnier <monnier@iro.umontreal.ca> writes:
> How'bout we set-advertised-calling-convention so as to obsolete the
> second argument?

But that's only reasonable if regexp-opt adds "significant" parens by
default, and doing so actually does run a risk of screwing up existing
users (because their match-data indices can be effected).

-Miles

-- 
Cynic, n. A blackguard whose faulty vision sees things as they are, not as
they ought to be. Hence the custom among the Scythians of plucking out a
cynic's eyes to improve his vision.

Re: /srv/bzr/emacs/trunk r101829: (regexp-opt): Add `symbols' mode.

[Miles Bader]

Miles Bader <miles@gnu.org> writes:
>> How'bout we set-advertised-calling-convention so as to obsolete the
>> second argument?
>
> But that's only reasonable if regexp-opt adds "significant" parens by
> default, and doing so actually does run a risk of screwing up existing
> users (because their match-data indices can be effected).

Hmm, or are you thinking that callers should just always add \\( and \\)
themselves?

-miles

-- 
Congratulation, n. The civility of envy.