lucene-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From David Smiley <>
Subject Custom DocValues Source -- what's involved?
Date Tue, 20 Nov 2012 05:03:22 GMT
Hi folks,

I'm looking at DocValues to store information per-document.  Spatial data.  The existing DocValues
Source types are mostly just one primitive value per document, although there are byte array
options and I'm starting to use that.  For DocValues usage in which the information is needed
in memory, not on-disk ("not direct" in DocValues terminology, I believe), it's unfortunate
I still have to constantly decode the byte array data into the integers, floats, and doubles
that are stored in those bytes for each document as they are accessed.  Ideally the decoding
would happen once and it would be those decoded values cached in memory, not byte arrays.
 Lets say I simply wanted to have an array of floats instead of a byte array.  Can someone
(Simon?) give me a quick rough plan of what it takes to do this?  Do I actually need a custom
codec?  I fear the answer is yes but maybe that's no so bad as it sounds.  I've poked around
a bit and its encouraging to see DocValues used internally for misc purposes like Lucene 3x
norms, but I haven't found a real exemplar to imitate.

	~ David
To unsubscribe, e-mail:
For additional commands, e-mail:

View raw message