Re: [Koha-zebra] Investigating zebra config files, a question

[Sebastian Hammer]

Paul POULAIN wrote:

> Hello guys,
>
> I'm trying to understand some things about zebra parameters, CQL,...
>
> in CQL, you can write "title=*tower* and author=*tolkien*"
> to find "the two towers, from JRR Tolkien".

You *could*, but I hope you won't. That is a terrible SQL-esque thing to 
say.. Zebra doesn't optimize for left-truncation, so the search above 
would be horribly inefficient. Since searches buy default are for 
keywords, not whole fields, you could have said "title=tower* and 
author=tolkien" and performance would be much better. The other search 
would work, but it would scale horribly.

> The question now is : how did zebra know what is a "title" and what is 
> an "author". Our zebra config files says (I mean in France, UNIMARC) :
>
> > attset: bib1.att               <<<< bib-1 attributes are loaded
> > attset: explain.att            <<<< not sure it is useful here
> > recordType: grs.marc.unimarc   <<<<  loads unimarc.abs if I don't mind
>
> (some useless lines removed)
>
> unimarc.abs says :
> > name unimarc
> > reference Usmarc
> > attset bib1.att
> > tagset unimarc.tag
>
> Then,
>
> > elm 700            Author                  -
> > elm 700/?          Author                  -
> > elm 700/?/a        Author                  !:w
>
>
> Does that means the UNIMARC 700$a field is related to "Author" ? I bet 
> yes.

You won your bet. ! is a shorthand for 'Author' in this case.

> And if I add
> > elm 701            Author                  -
> > elm 701/?          Author                  -
> > elm 701/?/a        Author                  !:w
>
> I say 701 is ALSO related to author.

You won again.

> Thus, "author" is translated to "Author" translated to "700$a or 701$a"
>
> Am I right ?

Yep.

> The next question will be : If I want to provide records in another 
> format, I will have to write a .map file
> I investigate a little bit more, and come back with next question ;-)
>
> (joshua/chris : if my idea is correct, we should be able to provide 
> Dublin Core output in Koha 3.0 I bet ;-) )
The schema mapping system is capable of providing a reasonable mapping 
from MARC to DC, but it is clunky and not very flexible.. I would rather 
hold out for the upcoming facility in Zebra to run XSLT transformations 
on records before retrieval, and defer any fancy mapings that would 
require boring and frustrating work with the old .map feature.

--Sebastian

--
Sebastian Hammer, Index Data
address@hidden   www.indexdata.com
Ph: (603) 209-6853