lucene-java-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Grant Ingersoll <gsing...@apache.org>
Subject Re: index conversion
Date Mon, 24 Sep 2007 14:25:42 GMT

On Sep 24, 2007, at 9:54 AM, Bogdan Ghidireac wrote:

> On 9/24/07, Doron Cohen <DORONC@il.ibm.com> wrote:
>>
>> For an already optimized index calling optimize() is a no-op.
>>
>> You may try this: after opening the writer and setting  
>> compound=false, add
>> a dummy (even empty) document to the index, then optimize(), and  
>> finally
>> optionally remove the dummy document.
>>
>> Note that calling optimize() might be lengthy as well for a large  
>> index.
>> In
>> any case, try this first on a trial index, and also make a backup  
>> of the
>> existing index, just in case.
>
>
> thanks, that worked smoothly ..
>
>
> Why do you want to transform the index from compound to non  
> compound form?
>
>
> My index is quite large (14GB) and this is because I use stored  
> fields. The
> requirements of the project force me to have a response time below  
> 500ms and
> I cannot do this at this moment (the tp99.9 is 700ms).
>
> What I am planning to do is to use the standard format so I can  
> move all
> relevant files except .fdt and .fdx to a memory disk. After  
> conversion the
> .fdt + .fdx files have around 11.5GB.
>
> I don't think there is an index reader that supports two locations  
> (one for
> index data and another one for stored data) but I will modify an  
> existing
> one and check the results.

I believe this is the use case for ParallelReader (although I have  
never used it) but it comes with caveats.  Search the archives and  
checkout the javadocs for info.


---------------------------------------------------------------------
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