On 4/9/2010 7:35 PM, Lance Norskog wrote:
The example function seems to round time to years, so you're boosting by year?
Your dates are stored as UTC 64-bit longs counting the number of
milliseconds since Jan 1, 1970. That's it. They're in milliseconds
whether you supplied them that way or not. So I think the example is
what you want.
Function queries are notoriously slow. Another way to boost by year is
with range queries:
[NOW-6MONTHS TO NOW]^5.0 ,
[NOW-1YEARS TO NOW-6MONTHS]^3.0
[NOW-2YEARS TO NOW-1YEARS]^2.0
[* TO NOW-2YEARS]^1.0
Notice that you get to have a non-linear curve when you select the
ranges by hand.
I hadn't figured out yet exactly what I wanted to do as far as the
actual boost value and what ranges to use. With our current indexing
solution (expensive), we don't use relevancy at all, we sort by
post_date, which is just an integer with seconds since epoch. In the
Solr replacement we are hurriedly trying to get into production, I am
indexing/storing it as a tlong, and I'm pretty sure it is required in
integer format by our application. If you can help me set up schema.xml
and DIH to index an additional field in a Solr Date format, then I think
I could use the range query you are suggesting, once I know how to
actually incorporate it.
What I would like to do is find a way to use dismax to handle all this
automatically without requiring the application to do much beyond supply
the query text that the user typed in. The first iteration of code
they've come up with doesn't "AND" the terms together, so the results
were not what we expected. I found from the 1.4 solr book that dismax
can do a number of things to improve this situation. What I'd like to
do is have boost values for all the default fields, turn on query slop
and phrase slop, and have the post_date give newer documents a little
boost, exactly how much is yet to be determined because I don't have any
of this working yet.
Should I open a new thread for help with the dismax requestHandler
config? I'm having trouble finding good examples, and what I've been
able to find so far just has me confused.
Thanks,
Shawn
