You could also think about pre-building a set of Filters once your index has been built, perhaps at warm-up time. One filter for each class and use those in your queries....
Erick On 5/30/07, karl wettin <[EMAIL PROTECTED]> wrote:
30 maj 2007 kl. 12.21 skrev Laxmilal Menaria: > On 5/30/07, karl wettin <[EMAIL PROTECTED]> wrote: >> 30 maj 2007 kl. 10.51 skrev Laxmilal Menaria: >> >> > What's the problem with a hit list iteration ( it should be >> >> > very fast ) ? >> > >> > Thats okay for short index, But if index have millions of records >> > or GB's data then it will get slow . >> >> Iterate only the top n results when you gather the unique values. If >> you get a million hits, ask the user to narrow down the search a bit. >> >> Searching the forum archives for facets or faceted classification >> might also be helpful. > > But if I implement faceted classification, then I know whats our > classes > name, but if I don't know classes name, then what should I do ? Then my recommend my first suggestion, that you gather the information by iterating the top n results. But you know the classes. There is a finite number of them and at some point in time you have inserted them to your index. Depending on how you store and index documents you might even be able to extract them from the index. For instance, you could index the class name untokenized in field used only for that purpose. The terms in this field now represents the available classes. You access them by seeking and iterating terms using the IndexReader. -- karl --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
