[ 
https://issues.apache.org/jira/browse/LUCENE-6863?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Adrien Grand updated LUCENE-6863:
---------------------------------
    Attachment: LUCENE-6863.patch

Here is a patch. It detects sparsity by checking if less than 10% documents 
have a value. When this happens, it stores doc IDs that have a value using a 
DirectMonotonicWriter and non-missing values as a regular numeric field (so 
values will still be compressed with table/gcd compression if applicable).

I first wanted to update the format javadocs but noticed that the low-level 
javadocs were very outdated (still documenting that sorted doc values were 
stored in a FST) so I decided to remove the low-level description in favour of 
the high-level one, which is much more interesting in my opinion.

This is just a first iteration, tests pass but maybe the heuristic needs to be 
better thought. I will also do some benchmarking.

> Store sparse doc values more efficiently
> ----------------------------------------
>
>                 Key: LUCENE-6863
>                 URL: https://issues.apache.org/jira/browse/LUCENE-6863
>             Project: Lucene - Core
>          Issue Type: Improvement
>            Reporter: Adrien Grand
>            Assignee: Adrien Grand
>         Attachments: LUCENE-6863.patch
>
>
> For both NUMERIC fields and ordinals of SORTED fields, we store data in a 
> dense way. As a consequence, if you have only 1000 documents out of 1B that 
> have a value, and 8 bits are required to store those 1000 numbers, we will 
> not require 1KB of storage, but 1GB.
> I suspect this mostly happens in abuse cases, but still it's a pity that we 
> explode storage requirements. We could try to detect sparsity and compress 
> accordingly.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org
For additional commands, e-mail: dev-h...@lucene.apache.org

Reply via email to