db-derby-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Dibyendu Majumdar <dibye...@mazumdar.demon.co.uk>
Subject Re: BTree package documentation
Date Tue, 06 Sep 2005 23:50:58 GMT
Mike Matrigali wrote:

>One could create compressed branch entries if the RowLocation were
>not needed to differentiate between 2 adjacent entries (ie.
>for 2 adjacent branch entries: mike, row loc 1, leaf page 113; stan,
>row loc 2, leaf page 112 -- for these 2 branch rows the minimum
>suffix of mike/stan would be enough without the row location).
>Because the page/row format in derby does not require constant number
>of column rows such a change would not be too hard.  One could also
>do suffix compression within a column - for instance in the previous
>example only m/s is really necessary, but such a decision is very
>datatype dependent and needs to be supported by the datatype itself -
>not a decision by store looking at the bytes.
>  
>
I guess it would be more fruitful to do compression of duplicate keys in 
non-unique indexes at leaf level pages, unless this is already done.

Attached is a revised package.html with Javadoc style links. I haven't 
been able to test the links - but I think they should work.

Regards

Mime
View raw message