[Koha-devel] Keyword searches

[Chris Cormack]

Hi all

Hope everyones holidays are going well.

I ended up taking a look at the keyword searching in koha the other day,
it was running slow and was causing much consternation in the holiday
rush.
What was happening is because of the likes in the search and the
wildcarding from the front of the search terms. Mysql couldnt use any
indices to make the search faster.
Which is fine until you have to join the biblio, biblioitems and
bibliosubtitle tables together to search over. Which is what the keyword
search needs.
It meant that for the Horowhenua library's for every keyword entered
mysql had to scan 160955 rows. Which isnt pretty when a few ppl are
doing it at the same time with a few keywords.
What I did was pretty much a quick bodge
I ended up using Set::Scalar, and breaking the queries down into one per
table, then using set unions to join the results together.
As bodges go it worked nicely, searches that previously took 10+ seconds
now take less than 1.

But im hesistant to commit my change to cvs as I think that with steves
work the database structure is going to change and the searches will
need to be rewritten anyway. Plus it also means another module that ppl
would have to install to get koha going.
So if we can fix the broken database design, (putting subtitles as just
a column in biblio would be one way to drop an unnessecary table for a
start) and do the searches without the uses of Set::Scalar i think that
would be far preferable.

What do others think? Is it worth committing the change to the
sourceforge cvs? Would anyone find it useful as a quick fix?

If not ill just keep it in my local copy.

Hope everyone is well
and Happy new year

Chris
-- 
Chris Cormack                                                    
Programmer
025 500 789                                        Katipo Communications
Ltd
chris at katipo.co.nz                                         
www.katipo.co.nz