On Mon, 2008-01-14 at 10:58 -0500, Alex Wang wrote: > Toke, you mentioned "Using a Collator works but does take a fair amount > of memory", can you please elaborate a little more on that. Thanks.
We have an index with 10 million records that takes up 37GB. Practically all records have a title, which can be used for sorting. The size of the titles is about 450MB in total. Using a Collator the usual way means that a CollatorKey is stored for all of those titles, among with the titles themselves. As for using a Collator, there's a neat tutorial at http://wiki.epc.ub.uu.se/display/~ronnie/International+Sorting Unfortunately for us, the memory requirements for this was too much for our machine. Instead of using the build-in caching mechanism, we used the fact that our index is currently static and calculated a single integer for every document, stating its position relative to all other documents in local sort order. That gives us fast sort with low memory overhead. Of course we'll have to discard that idea when we move to a non-static index. --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]