lucene-java-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Jason Rutherglen <jason.rutherg...@gmail.com>
Subject Re: Tag Index patch (LUCENE-1292) status?
Date Fri, 22 Jan 2010 04:36:03 GMT
Chris,

Great questions....

TI is for untokenized fields only and so is probably for a
different use case than LUCNE-1879 (which is very cool, does it
work yet?). TI should just handle everything underneath, so one
wouldn't manually worry about it.

The only hard part, or untested performance wise part, is the if
a posting "block" becomes too small, how will that affect
performance of posting list iteration (i.e. skipping). Instead
of vint decoding, one'd probably implement TI postings using
PFOR or some other naturally "block"ed posting algorithm, which
PFOR happens to be. So if I started again, which would take a
lot of time to work on, I'd start benchmarking there to see if
the idea is even tenable. Though I'm assuming somehow it works,
as GMail labels appear to implement this type of thing. As more
current responsibilities are far removed from these lower Lucene
levels, I don't see working on it in the near future.

Jason

On Thu, Jan 21, 2010 at 8:18 PM, Chris Harris <ryguasu@gmail.com> wrote:
> I'm probably not going to work on it right now.
>
> It might be nice, though, to make sure I have the right big-picture
> idea of the tag index patch. I can think of two ways to ask it:
>
> 1. What's the relationship between the tag index patch and LUCENE-1879
> ("Parallel incremental indexing", which also has the associated page
> http://wiki.apache.org/lucene-java/ParallelIncrementalIndexing)? Is
> the tag index an alternative idea, or a complementary one?
>
> 2. Does the tag index patch means to take care of all of my fields, or
> does it only apply to my updateable (and non-tokenized) fields? For
> example, if I use the tag index patch (and suppose I don't use any
> LUCENE-1879 stuff), then do I need to manually create two indexes, one
> for my static fields and one for my tags? (I would need to be careful
> about how I coordinated these indexes, so I could use a ParallelReader
> with them.) Or is there only one index, and the tag fields are
> updateable simply by virtue of the patch's new segment file types?
>
> Thanks,
> Chris
>
> On Tue, Jan 19, 2010 at 8:10 PM, Jason Rutherglen
> <jason.rutherglen@gmail.com> wrote:
>> Hi Chris,
>>
>> It's not actively being worked on.  Are you interested in working on it?
>>
>> Jason
>>
>> On Tue, Jan 19, 2010 at 4:42 PM, Chris Harris <ryguasu@gmail.com> wrote:
>>> I'm interested in the Tag Index patch (LUCENE-1292), in particular
>>> because of how it enables you to modify certain fields without
>>> reindexing a whole document. However, that issue is marked Lucene
>>> 2.3.1 and hasn't been updated since July 2008. Can anyone provide any
>>> status updates on this patch? Questions include:
>>>
>>> - Is anyone using this in production?
>>>
>>> - Does this work with Lucene 2.9/3.0?
>>>
>>> - Is anyone actively working on the patch now?
>>>
>>> - Are there any competing approaches now for doing updates to
>>> particular fields? (I haven't found any yet.)
>>>
>>> Thanks,
>>> Chris
>>>
>>> ---------------------------------------------------------------------
>>> To unsubscribe, e-mail: java-user-unsubscribe@lucene.apache.org
>>> For additional commands, e-mail: java-user-help@lucene.apache.org
>>>
>>>
>>
>> ---------------------------------------------------------------------
>> To unsubscribe, e-mail: java-user-unsubscribe@lucene.apache.org
>> For additional commands, e-mail: java-user-help@lucene.apache.org
>>
>>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: java-user-unsubscribe@lucene.apache.org
> For additional commands, e-mail: java-user-help@lucene.apache.org
>
>

---------------------------------------------------------------------
To unsubscribe, e-mail: java-user-unsubscribe@lucene.apache.org
For additional commands, e-mail: java-user-help@lucene.apache.org


Mime
View raw message