You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@lucene.apache.org by "Doron Cohen (JIRA)" <ji...@apache.org> on 2007/02/01 09:05:10 UTC
[jira] Updated: (LUCENE-790) contrib/benchmark - few improvements
and a bug fix
[ https://issues.apache.org/jira/browse/LUCENE-790?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Doron Cohen updated LUCENE-790:
-------------------------------
Attachment: TrecDocMaker.patch
Attached TrecDocMaker.patch also contains the changes in current patch in 788 - because both patches modify ReutersDocMaker - so it is sufficient to apply this patch only. I will add a comment on that in 788. Once this is committed, will mark 788 as duplicate of this.
Some TODO items are in byTask/Benchmark.java's javadocs - comments are welcome.
> contrib/benchmark - few improvements and a bug fix
> --------------------------------------------------
>
> Key: LUCENE-790
> URL: https://issues.apache.org/jira/browse/LUCENE-790
> Project: Lucene - Java
> Issue Type: Improvement
> Components: Other
> Affects Versions: 2.1
> Reporter: Doron Cohen
> Assigned To: Doron Cohen
> Priority: Minor
> Fix For: 2.1
>
> Attachments: TrecDocMaker.patch
>
>
> Benchmark byTask was slightly improved:
> 1. fixed a bug in the "child-should-not-report" mechanism. If a task sequence contained only simple tasks it worked as expected (i.e. child tasks did not report times/memory) but if a child was a task sequence, then its children would report - they should not - this was fixed, so this property is now "penetrating/inherited" all the way down.
> 2. doc size control now possible also for the Reuters doc maker. (allowing to index N docs of size C characters each.)
> 3. TrecDocMaker was added - it reads as input the .gz files used in Trec - e.g. .gov data - this can be handy to benchmark Lucene on these large collections. Similar to the Reuters collection, the doc-maker scans the input directory for all the files and extracts documents from the files. Here there are multiple documents in each input file. Unlike the Reuters collection, we cannot provide a 'loader' for these collections - they are available from http://trec.nist.gov - for research purposes.
> 4. a new BasicDocMaker abstract class handles most of doc-maker tasks, including creating docs with specific size, so adding new doc-makers for other data is now much simpler.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
---------------------------------------------------------------------
To unsubscribe, e-mail: java-dev-unsubscribe@lucene.apache.org
For additional commands, e-mail: java-dev-help@lucene.apache.org