Re: `save-excursion' defeated by `set-buffer'

[Uday Reddy]

On 3/12/2011 3:25 PM, David Kastrup wrote:

> > Frankly, I think the warning message is quite fine.  save-excursion is
> > trying to preserve the point (and mark), but only for the
> > current-buffer.  set-buffer is changing the current-buffer and, so,
> > the preservation of the point in the current-buffer is useless.
>
> There is little reason to warn about "useless" code.

Why is that?  I am quite happy to receive warnings about bad code.

> > If one wants a completely plain warning message, it could be:
> >
> >    "save-excursion has the effect of save-current-buffer"
> >
> > It doesn't say very much, but nobody will presumably argue about it.
>
> The most plausible theory I have for this thread is that you are
> collectively trying to pull my leg.
>
> If you give the user that warning, he'll say "great, just like I wanted
> it to do".

Ok, how about

   "save-current-buffer is a better choice than save-excursion"

> Why warn the user that "save-excursion" is doing what he wants it to do,
> when the actual problem is that it may do more, namely reverting an
> excursion happening unintendedly in the original buffer?

If an excursion is happening unintentionally that is definitely a bug! 
You probably mean that reverting it unintentionally is not so bad.  But 
I happen to think that it is.  save-excursion's should be as close as 
possible to the point movements they are protecting.  One shouldn't 
randomly wrap large pieces of bad code in save-excursion and hope that 
nobody will notice.  It is really hard to maintain such code.  The 
compiler warning is really welcome from my point of view.  I only wish 
that the compiler warning was there 20 years ago so that I wouldn't have 
to deal with that mess now.  But better late than never!

Cheers,
Uday