Somehow I am not surprised. I may also try that on Oracle when the
code becomes available on trunk.
(BTW not sure that Stephane is subscribed to the dev list, so cc'ying
this message.)
Andrus
On May 24, 2009, at 12:52 PM, Andrey Razumovsky wrote:
> I should say my tests on Postgres and mysql didn't show any results.
> If I
> use small heap size, i get OutOfMemory no matter which fetch size
> was set.
> Fetch speed and memory usage are the same. Seems drivers just ignore
> this
> parameter. Stephane, did your workaround help you?
>
> Drivers of hsql, postres, mysql didn't throw any exceptions.. Maybe
> other
> DBMS drivers work well with this param. I'm going to commit today or
> tomorrow. Nothing will break anything if we add this ability.
>
> Andrey
>
> 2009/5/21 Andrus Adamchik <andru..bjectstyle.org>
>
>> Cool :-)
>>
>> On May 21, 2009, at 3:59 PM, Andrey Razumovsky wrote:
>>
>> Now when I know of this JDBC feature, I'd prefer to have it in
>> Cayenne
>>> sooner that later. We would also want to have it for other queries
>>> than
>>> just
>>> only SelectQuery.
>>>
>>
>> Absolutely. This has to be a part of the QueryMetadata on the
>> backend. On
>> the frontend any query that can potentially select data should have a
>> corresponding setter.
>>
>> Could someone help me testing it against different types
>>> of database if I commit?
>>>
>>
>> I can test it on almost all DB's that we support. Of course we
>> should have
>> Cayenne unit tests that will provide regression (i.e. driver XYZ
>> doesn't
>> throw UnsupportedOperationException when we call a corresponding JDBC
>> method).
>>
>> In addition to that I'd like to see if there's really memory/speed
>> savings
>> when using that (i.e. is it really worth it). For that I suggest
>> writing a
>> JDBC test outside of Cayenne, that can be run in profiler against
>> different
>> DB's.
>>
>> Andrus
>>
This archive was generated by hypermail 2.0.0 : Mon May 25 2009 - 02:39:06 EDT