lucene-java-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Erick Erickson" <>
Subject Re: merge factor and real time indexing
Date Tue, 30 May 2006 23:52:23 GMT

I realize this is test code, but is it really measuring something
meaningful? If you really need to open/modify/close for each document, then
I've got to assume that the rate they're coming in is slow enough that the
time it takes for this operation is not very important.

And if they *are* coming in quickly, then you could hold open the modifier
longer and amortize the time it takes to open etc. over indexing multiple
documents. I realize that's "near real time" and not real-time, but....

Otis: What are the rules for reading an index that's being modified? Here's
the sequence that I'm unclear about.

open reader0
open writer
add document 1
***at this point, am I right that document1 is not visible to reader0?
close reader0
open reader1
***Now, is document 1 visible to reader1? Even though the writer is still

Where I'm going with this is wondering whether John needs to close/open the
writer except very rarely but still be able to pick up modifications from a
newly opened reader.

Which, of course, leads to the question where his performance needs to be.
It may be real-time, but how close is "close enough" to real-time?



  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message