[
https://issues.apache.org/jira/browse/LUCENENET-103?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_12535663
]
Digy commented on LUCENENET-103:
--------------------------------
Hi Jon,
I tested the code you sent with some "additions and deletions into the original
index and checking the clones" and it seems OK.
(Note: copying "sizeInBytes" field to clone is also needed).
(I prepared a patch for this but I am not sure about sending this code,
because I don't know if this action results in a Copyright violation or not.)
My Test results are:
using orgDir.Clone(false) -> 3t
using orgDir.Clone(true) -> 7t
using new RAMDirectory(orgDir) -> 9.5t
with ~150,000 docs containing ~1,200,000 terms (t:100ms).
But I have some hesitations on applying the patch to Lucene.Net
First, although I couldn't generate an error in "Clone(false)" case, I think it
can be dangerous since it does not clone the byte array, instead just put the
address into the buffer(and this is the reason of speed-up).
Second, the original java-code from Nick Smith(Adding clone method to
RAMDirectory+RAMFile) is not included in the current release of the Java
Lucene. Adding this functionality to Lucene.Net means divergence from the Java
code.
So I leave that hot potato for George Aroush(the only commiter of this project).
DIGY
> Need to implement ICloneable on RAMDirectory
> --------------------------------------------
>
> Key: LUCENENET-103
> URL: https://issues.apache.org/jira/browse/LUCENENET-103
> Project: Lucene.Net
> Issue Type: Improvement
> Environment: C# 2.0
> Reporter: Jon Davis
> Priority: Minor
>
> IClonable needs to be added to Lucene.net's RAMDirectory.
> See Lucene (Java) item resolution at:
> http://www.mail-archive.com/[EMAIL PROTECTED]/msg03725.html
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
