Enrico Olivelli - Diennea wrote on 19.09.2014 12:44:
> Hi,
>
> we are moving from Oracle to PostgreSQL and we hit a serious
> problem.
>
> When executing a query with a large result set your driver loads each
> record in memory and this is a blocking problem for us.
>
> The work-around is to use server-side cursors
> (http://jdbc.postgresql.org/documentation/head/query.html) but in
> order to do this we have to change autocommit mode of the transaction
> and this will change transaction semantics of out code.
>
> Many JDBC drivers provide some swap-to-disk function in oredr to deal
> with huge result sets
>
> Something like “if we are loading in RAM more than X MBs than swap on
> disk”
>
> or “if we are loading in RAM more than X rows than swap on disk”
> (more simple to implement)
>
Wouldn't it make more sense to allow for a non-buffered result when using auto-commit
(something like "bufferResults=false") instead?
Or is that technically not possible due to the Postgres wire-protocol?
Thomas
