[Koha] Ghost records

Tomas Cohen Arazi tomascohen at gmail.com
Thu Apr 28 13:12:39 NZST 2011


On Wed, Apr 27, 2011 at 9:55 PM, Jared Camins-Esakov
<jcamins at cpbibliography.com> wrote:
> Tomas,
>>
>> > Our problem was :
>> > melm 001        local-number,local-number:n
>> > ......
>> > melm 035$a      local-number,local-number:n
>> > We had two fields indexed in "local-number" and we had "ghost entries".
>> > We have corrected in :
>> > melm 001        local-number,local-number:n
>> > ......
>> > melm 035$a      identifier-other,identifier-other:n
>> > Reindex and it was the end of the fabulous story of ghost entries.
>>
>> Perhaps its just out of my knowledge's scope, by I don't see the
>> problem of having two different fields indexed under the same name, I
>> mean, I'd think both would match the same record and thus not be shown
>> twice. I'd rather think of a weird coincidence. I'm sorry if my
>> confusion just makes the situation more obscure.
>
> The issue here is that the local-number index has "magic" properties. I
> don't really understand how Zebra works, but I do know that every entry in
> the local-number index is considered a "separate" record. So if record has
> two fields that appear in the local-number index, the Zebra index will store
> two copies of the record (this is where the search results in Koha come
> from, not from the database). At some point I will submit a patch changing
> the suggestion for the 035 index, since uncommenting the "melm 035" line as
> it stands now in the MARC21 record.abs is pretty much guaranteed to corrupt
> your Zebra indexes.

I guess the magic property might have to do with this in zebra-*.cfg:

recordId: (bib1,Local-number)

i suppose it is the value used as interal ID for the record. I'd think
it is a bug in zebra or th eway we use it.

Regards
To+


More information about the Koha mailing list