[please don't "Reply" to an existing thread.]
On Fri, Jan 06, 2006 at 02:07:22AM -0800, François Tissandier wrote:
> My problem is with the speed of the search. If the user looks for a very
> general word, it may give a huge amount of results. If you choose to
> display results 100 by 100, as you sort first with the property, you may
> have a very partial result on the first page. For example, if we choose
> 100 results per page, and if the search give 120 results for Universal,
> the first page will show only Universal movies, not the other studios.
I don't see that a partial result on a page is a problem.
> What I would like to do is to return first the number of results per
> Sony: 443 results
> Universal: 866 results
> Warner: 1254 results
> Then the user can choose to display the complete results for a studio,
> by relaunching a new search only on this studio. Or maybe can refine his
That's a database function, and swish isn't a database.
> Maybe I can do this by outputing only the property during my search, and
> then calculating the subtotal for each studio? It would be faster than
> the typical output, wouldn't it?
Internally, swish just has a list of document numbers in its result
list. There's not way to group those by a property without
sequentially going through all the results and looking up the
properties, sorting and then counting.
Even the presorted indexes can't help as even properties with the same
values have different values in those indexes.
Your only choice is to count up the results and then generate page
views in your own code. Not sure if it would be fast enough for your
Unsubscribe from or help with the swish-e list:
Help with Swish-e:
Received on Fri Jan 6 12:31:04 2006