Re: [Evolution] Address completion, is it magic, or just slow

[Scott Dixon <scott metaphorics com>]

Hi Jon,
I don't think the problem with completion is just speed.  For example,
if I type "da" in the From box I get just one name (David ...).  If I
then continue to type by adding "v" I get *more* possible completions
including various other David or Dave names.  There seems to be no
simple pattern to it.  If I just type "d" I get all the names (I think)
starting with D.  Then if I continue to type "a" and then "v" the list
narrows down as you would expect.  But if I type "da" before the first
completion comes up, I get the strange behavior.  I'll send you a copy
of my Contacts list (which is only 522 names).  It is possible there is
something funny with it.  It has been through the various permutations
of db since the early snapshots so something could have gotten munged in
it along the way.
Scott
On Wed, 2001-12-12 at 15:23, Jon Trowbridge wrote:
> On Wed, 2001-12-12 at 15:19, Zot O'Connor wrote:
> > What is really going on?  I know the DB lookup is slow, but can it be
> > cached per session (with a check for changed contactdb?).
> >
> > Is it just really slow (like in 30-40 seconds) and I and finding
> > behavior? 
> >
> > Is any one else seeing this or just me?
>
> Well, one of two things could be happening here.
> (1) Something is terribly wrong on your system.
> (2) Your addressbook is really big.
>
> Let's discuss case (2) first.
>
> Right now, the addressbook search scales very poorly -- it is fine for
> "typical" sets of contacts (like many a hundred or so) but gets ugly
> when the addressbook is big.  I used an addressbook with 1500 contacts
> in it for stress-testing, and added some optimizations based on that
> experience, but using an addressbook that big is still pretty painful.
>
> FYI, I'm currently working on a redesign of the completion searching for
> Evolution 1.2 that will cache and index everything, so that our searches
> will be O(log N) instead of O(N) --- so there is hope for those of you
> with big addressbooks.  And you'll be able to do completions off of any
> addressbook, not just the main one.
>
> We do cache things now, but only by the first letter of the search.  So
> when you type 'g' we cache all of the matches, and then search that
> subset when you type 'e', 'o', etc.  This helps a lot, but that initial
> query can still be painful.
>
> Question: if you type 'g' into an address entry and then walk away, does
> anything ever pop up?  If not, something is wrong.
>
>
> Now if your addressbook is "small" (as defined above), I suspect that
> (1) holds.  Are all addressbook searches slow, or just the ones related
> to completion?  You can test this by going into the Contacts folder and
> doing a few queries by typing into the search bar.  If they come back
> very slowly, then something funny is going on.
>
> -JT
>
>
>
>
>
>
>
>
> _______________________________________________
> evolution maillist  -  evolution ximian com
> http://lists.ximian.com/mailman/listinfo/evolution