Running process with grep-find interface is very slow

Running process with grep-find interface is very slow

[Tom]

I often have to search large codebases and I use Ack for that
which is a better grep as it calls itself. I want to use the grep-find 
interface with ack, so I can jump to matches easily and it's working, 
but it's really slow.

When I run ack on the command line then it returns the
results in seconds, and if the files are cached by the OS then
it returns the results instantly.

But if I'm running the same command with the grep interface:

   (grep-find "c:/xampp/perl/bin/perl.exe c:/bin/ack -i somestuff")


then it takes minutes for the results appear in the grep buffer,
and when I'm running the same search again immediately when the OS
still have the files cached, even then it takes 30-40 seconds for
the results to appear.

Why is it so much smaller to run the same command via grep than
via the command line?


It's GNU Emacs 23.2.1 (i386-mingw-nt6.1.7601) on Windows.

Re: Running process with grep-find interface is very slow

[Tom]

Tom <adatgyujto <at> gmail.com> writes:
> 
> Why is it so much smaller to run the same command via grep than
> via the command line?
> 

I mean slower, of course.

Re: Running process with grep-find interface is very slow

[Tom]

Tom <adatgyujto <at> gmail.com> writes:
> 
> Why is it so much slower to run the same command via grep than
> via the command line?
> 

Turns out it was the way I ran the command.

I created a small convenience function which always started the
search from the root directory of the project


(let ((default-directory rootdir))
  (grep-find (concat "c:/xampp/perl/bin/perl.exe c:/bin/ack -i " pattern)))


Looks like default-directory caused the problem, because if I change it
to this:

  (with-current-buffer (find-file-noselect rootdir)
    (grep-find (concat "c:/xampp/perl/bin/perl.exe c:/bin/ack -i " pattern)))

then it's not slow anymore.


It fixed the problem, though I don't really get the difference. Isn't
setting default-directory temporarily a usual way to make a directory
current for the duration of a function?

Re: Running process with grep-find interface is very slow

[Kevin Rodgers]

On 8/31/11 2:18 AM, Tom wrote:
> Tom<adatgyujto<at>  gmail.com>  writes:
>>
>> Why is it so much slower to run the same command via grep than
>> via the command line?
>>
>
> Turns out it was the way I ran the command.
>
> I created a small convenience function which always started the
> search from the root directory of the project
>
>
> (let ((default-directory rootdir))
>    (grep-find (concat "c:/xampp/perl/bin/perl.exe c:/bin/ack -i " pattern)))
>
>
> Looks like default-directory caused the problem, because if I change it
> to this:
>
>    (with-current-buffer (find-file-noselect rootdir)
>      (grep-find (concat "c:/xampp/perl/bin/perl.exe c:/bin/ack -i " pattern)))
>
> then it's not slow anymore.
>
>
> It fixed the problem, though I don't really get the difference. Isn't
> setting default-directory temporarily a usual way to make a directory
> current for the duration of a function?

Yes.  Could you profile the grep-*, compile-*, and compilation-*
functions with elp.el under both scenarios to show where the performance
problem occurs?

-- 
Kevin Rodgers
Denver, Colorado, USA