Skip to main content.
home | support | download

Back to List Archive

Re: [swish-e] Max records in sort buffer

From: Peter Karman <peter(at)>
Date: Sat Oct 17 2009 - 03:31:14 GMT
Thomas den Braber wrote on 8/18/09 2:58 AM:
> Peter,
> I use the Perl API with swish-e 2.4.7.
> Yes I do sort on properties but even swishrank will slow down if there are
> many results.
> I only show about 50 results and go with 'seek' to the next 50 results.
> Browsing trough a search that returns 100000 hits and a search that
> returns only 1000 hits is about 900% slower  (300ms vs 40ms).
> With 500000 hits is takes seconds to complete the search.
> Any idea how to speed this up for ?

sorry it took me a couple of months to carve out some time to experiment.

I created a corpus of 1M tiny docs. Then I ran the query:

 % swish-e -w 'a or (not a)'

which should return all the docs in the index:

# SWISH format: 2.5.8
# Search words: a or ( not a )
# Removed stopwords:
# Number of hits: 1000000
# Search time: 0.410 seconds
# Run time: 0.416 seconds

so yes, I expect that the more hits you get, the longer it will take to iterate
over them all and calculate a rank for each.

Peter Karman  .  .  peter(at)
Users mailing list
Received on Fri Oct 16 23:31:16 2009