On Oct 5, 2009, at 12:23 AM, Yonik Seeley wrote:
On Sun, Oct 4, 2009 at 10:53 PM, Israel Ekpo <israele...@gmail.com>
wrote:
Smaller precisionStep values (specified in bits) will lead to more
tokens
indexed per value, slightly larger index size, and faster range
queries
It also states that for faster range queries, consider the
tint/tfloat/tlong/tdouble types.
Now, the tint/tfloat/tlong/tdouble have a precisionStetp of 8 while
the
int/float/long/double types have a precisionStep of 0
precisionStep of 0 means don't do precision steps at all (i.e. no
acceleration of range queries).
I'll take a shot at clearing this up in the example schema.xml.
It seems like we should have tint4, tint8 instead of just 8, no? My
impression of the Lucene javadocs is that a step of 4 for ints/floats
was more appropriate. Or am I missing something (I haven't followed
the numerics stuff that closely)
-Grant
