On Sun, Mar 14, 2010 at 3:39 PM, Peter Sturge <peter.stu...@googlemail.com> wrote: > I'm a bit confused now. In your recent Mastering Solr webinar (great stuff, > btw, thank you!), the slides imply using tdate fields with a precisionStep > of 8 for faster range queries: > > - Use tint, tfloat, tlong, tdouble, tdate for faster range queries > - <fieldTypename="tint" class="solr.TrieIntField" precisionStep="8" > omitNorms="true" positionIncrementGap="0"/> > - Date faceting also uses range queries
Correct. precisionStep of 8 compared to no precision step at all. Precision step of 0 is like normal - a single token is indexed for a single value. Precision step of 8 would index a 32 bit int with 4 tokens (bigger index, faster range queries) Precision step of 4 would index a 32 bit int with 8 tokens (even bigger index, even faster ranges) > Can you clarify, for fast 'wide' date range queries (for date faceting and > otherwise), what is the best precisionStep value to use for tdate? It's a tradeoff - I don't know what the best is. By "wide"... it means a range that covers a lot of unique values. If the ranges only cover a few unique values, the trie strategy doesn't help much. -Yonik http://www.lucidimagination.com
