[pacman-dev] [PATCH] Feedback from installed package while searching

[Xavier]

On Fri, Sep 4, 2009 at 1:09 AM, Allan McRae<allan at archlinux.org> wrote:
> Xavier wrote:
>>
>> I submitted a similar patch 1 year and a half ago, but it was never
>> accepted :
>> http://www.mail-archive.com/pacman-dev@archlinux.org/msg00109.html
>>
>> So I am afraid you are stuck with yaourt / pacsearch / any other
>> pacman wrapper to get that feature.
>>
>
> Well...  this is nothing if not persistent given Laslo had a bug report
> closed as "Won't Implement" for this feature request...
>
> So lets take a step back and see why this feature was dismissed.  As far as
> I can tell, it was rejected because it is slower as we need to parse the
> local db as well as the sync db - of course only applies before the local db
> is cached.
>
> I actually had Xavier's patch bookmarked because I was going to revisit this
> idea with this awesome suggestion....   How about adding a configuration
> option called "SearchLocal" or something that directs whether this is
> enabled.
>

I did some more testing and the performance differences are pretty
insignificant.
For the -Sl operating, without fs cache, it is 2.2s vs 2.6s
with fs cache, I get the same result than before : 0.1s vs 0.2s
but who is going to notice that ? :)

besides, this should only be done for the non-quiet output, so the
quiet output should remain the same.

In the case of -Sl, we are already looking at 4300 dir for the sync
databases. Considering local db just add a few hundreds more.

In the case of -Ss (Laszlo patch), the difference is even smaller,
because we already have to look at 4300 dir + 4300 desc files!
Considering local database only add a few hundred directories to that.

If we have a performance problem, I think it's already here, and it's
not going to change much with that patch
("pacman -Ss" after dropping fs caches takes ~10 seconds in both cases here :) )