[Koha-devel] searching for phrases

[LAURENT Henri-Damien]

Galen Charlton a écrit :
> Hi,
>
> On Mon, Mar 1, 2010 at 3:57 PM, Nicole Engard <nengard at gmail.com> wrote:
>   
>> Correct me if I'm wrong- but I've been told that Zebra is the reason
>> the facets work the way they do - and there are issues there as well.
>>     
>
> It's more correct to say that since Koha's facets *don't* use Zebra's
> native facet functionality that they don't work as well as they
> should.
>
>   
Well actually, I have made some tests on zebra facets.
And found many reasons why we could/should not use them.
Especially because diacritics are messed around.
Even if simple charmap files. I donot even talk with icu configuration 
files.
So It is not only the fault of C4/Search.pm but also poorly supported 
feature in zebra.
I was told that pazpar2 was better handling facets and diacritics but I 
took no time to test that.
So whatever search engine we choose we would have to enhance the support 
so that every feature we announce is really well documented and achieved.

Moreover, we should have some way to setup the facets and indexes so 
that ppl can add or remove those quite easily. This is not a problem of 
Search Engine, this is a matter of Koha configuration and Koha 
collaboration with search Engine configuration. But at the moment Koha 
Configuration is so tighted with zebra, that it would be dangerous to 
change.

But I just want to remind that whatever the search engine, it is not its 
name or its "ggl trend" which makes the configuration or the feature works.
We invested much time in taming the zebra.
We would still have to invest : (dom configuration is one thing we donot 
use thoroughly. We could also try and index items, subscriptions and 
acquisition data.)

But there are also some features that are clearly missing (left 
truncation for icu, facets). And I think that changing to pazpar2 or to 
solr could require quite the same amount of work.

One big point for solr though is that documentation is really accessible 
online, and that the project lists are pretty much active.

My 2 cents.
-- 
Henri-Damien LAURENT


>> I don't find the only issue with Zebra to be phrase searching, but
>> also boolean searches, the ordering of the facets, the inability to
>> show the number of hits next to facets or order them in alpha order
>> ... and maybe even more I'm forgetting now.
>>     
>
> All of those issues have more to do with C4::Search than Zebra itself.
>
>   
>> But I do agree that there will be problems no matter what - or maybe
>> not problems but challenges - this is the nature of technology.
>>     
>
> Don't get me wrong - I'm not speaking as an apologist for Zebra,
> except possibly to counter-balance what I think are some
> mischaracterizations of where the problems lie.  I would be quite
> happy to have Koha work with Solr as well.
>
> Regards,
>
> Galen
>