At least in this case where dates have a precision of day, the total number
of unique values should be relatively low (3,650 for a 10-year period or
even 18,250 for a 50-year period), so precision step probably won't matter
in this case much at all. The big benefit with tdate over old date here is
the fact that the string is stored as a long integer, which speeds
comparisons.
A precisionStep of zero ("0") means use the maximum value
(Integer.MAX_VALUE). The default (in trunk; not sure about 3.6 or earlier)
is "8", not 4 as the Javadoc indicates. Zero or high step values means more
precision, less index space consumed, but slower searching. Lower step
values, such as 8 or 4 or even 1, mean less precision, more index space
consumed, but faster searching.
I'm still struggling to figure out how to map values of precisionStep to
"precision" of the input data. For example, seconds, minutes, hours, days,
years for date values.
I'm also not sure what the implications, if any, are for faceting and the
FieldCache for Trie precision step.
Here's more detail:
http://lucene.apache.org/core/3_6_0/api/all/org/apache/lucene/search/NumericRangeQuery.html
-- Jack Krupansky
-----Original Message-----
From: Ahmet Arslan
Sent: Monday, May 14, 2012 10:16 AM
To: solr-user@lucene.apache.org
Subject: Re: Date format in the schema.xml
is it mandatory to use the date format yyyy-mm-ddThh:mm:ssZ
?
Yes.
I have a date with this format:
yyyymmdd
in my xml source file.
Where can I find more information, I found only these
definitions in the schema.xml
In schema.xml there is a xml comment about dates, starting with
<!-- The format for this date field is of the form 1995-12-31T23:59:59Z,
Could you explain me the PrecisionStep param also?
This presentation explains trie based fields.
http://www.thetaphi.de/share/Schindler-TrieRange.ppt
