[Freeipa-devel] [PATCH] enhanced user search

[Kevin McCarthy]

Hi Rob,

These are good comments.  I'd ideally like to get this patch committed
and we can iterate on the best solution (I'm currently wrestling with
some other big changes in my client to get a large data demo out).

Comments inline below.

Rob Crittenden wrote:
> Rob Crittenden wrote:
>> We will need to do searches of groups as well. Can you break out the 
>> search_pattern-creation code block into a function?

Yes, this is a good idea.

>> Do we want to do any approximate searching (=~)?

Pete mentioned this for later testing.  We need to set up special
indices and only do it for certain fields so we'll have to think
carefully about it.

>> What implication does this have if I know the entry I want.
>> What I mean is that I want to bring up the record for 'rcrit' so I do a 
>> search on it and I'd like it to just pop up that record since I have an 
>> exact match on it. What happens if there is a jrcritenham too? That record 
>> is going to pop up as well because of the partial searches.

That's true, but your 'rcrit' record will come up first in the list.  I
think that's a pretty good compromise.

>> I see why you drop "*" from user input but will that make users mad?

Maybe, but the idea is to have the server figure out the best search.
'*' is a pretty antiquated search ui paradigm now (IMHO), and I think
it's better to not let the user control it.

>> This is, of course assuming that if only one record is found rather than 
>> displaying a list of users, that user's record is brought up. Not sure 
>> that is the right thing to do either.
>> The code generally looks good. We can always just stick it in and see if 
>> it annoys us too much :-)

Yes, that's what I'd prefer.  I'm putting together a demo with 10,000
users so we can see how well it works.  I'd like to get this patch in
and then iterate on ideas, if the team is okay with that.

> As I thought about this more perhaps we should pass in some flags outlining 
> how the search should work. Exact match, partial match, etc...
>
> This gives the client implementor control.

Perhaps, but I'm not sure we necessarily want to give the client control
here.  We'll have to make the search mechanism much more flexible, and I
think it would be better for us to just experiment with what we think
works best (at least for V1) and then see if our users scream.

-Kevin

-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/x-pkcs7-signature
Size: 2228 bytes
Desc: not available
URL: <http://listman.redhat.com/archives/freeipa-devel/attachments/20070824/65202533/attachment.bin>