Great!
Thanks !
--
Antoine Baudoux
Development Manager
[EMAIL PROTECTED]
Tél.: +32 2 333 58 44
GSM: +32 499 534 538
Fax.: +32 2 648 16 53
Le 31 Aug 2007 à 09:45, Michael Busch a écrit :
Antoine Baudoux wrote:
From what I have seen in the patch, It re-opens the segments tha
have changed.
So Imagine I always change the biggest sement (because that's
where
most docs are and i need to update them frequently) . Will there
still
be a benefit of IndexReader.reopen()?
Yes, because the patch also improves how segments that have changed
are
loaded. If documents were deleted from a segment since it was
(re)opened, then IndexReader.reopen() will only load the deleted docs
file from the disk, and won't load the other data structures again,
like
the dictionary, that haven't changed.
There are actually some performance numbers in LUCENE-743. The index
size in these experiments was 4.5M docs in a single segment. In a loop
random docs were deleted and the IndexReader was reopened afterwards.
Compared to closing/opening the reader, reopen() was ~47 times faster.
It's likely that the improvement will be a little smaller in the final
version of the patch due to some changes that I have to make, but it
should still be significant.
- Michael
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
