Обсуждение: Use Declare/Fetch and PG 7.4

Hi,


When the ODBC´s "Use Declare/Fetch" option is enabled the automatic
declaration of cursor great improve the performence for me.
But testing this feature in PG 7.4, causes the following error:

When the application makes a COMMIT, the cursors are automatically closed
but the ODBC try to close this cursors defined inside transaction...
In PG 7.3, this causes a WARNING only, but the behavior of this event is
changed to ERROR in PG 7.4, aborting the application.

Thank´s all

Alexandre

Le Mon, Nov 10, 2003, à 04:30:42PM -0200, alexandre :: aldeia digital a écrit:

> When the ODBC´s "Use Declare/Fetch" option is enabled the automatic
> declaration of cursor great improve the performence for me.

I second this observation with the workloads I'm confronted with.

> In PG 7.3, this causes a WARNING only, but the behavior of this event is
> changed to ERROR in PG 7.4, aborting the application.

I'm not going to contribute much here, except express interest in seeing
a solution to this (I didn't plan to move to 7.4 very quickly, but this
would be a showstopper to me -- it's not a totally rare event for me to
need datasets beyond what 32-bit Win32 will allow a single process to
allocate, thus killing postgres.exe (ouch!)).

    -- Cyrille

--
FRA: 61-18(FIJ) 51-29(JPN) 51-9(SCO) 41-14(USA) >> 43-21(IRL)

Hi,

I will move to PG 7.4 quickly. I change the PG code from ERROR to WARNING
to aviod this error, but I think that ODBC driver must correct this.
Searching the mailing lists, I see a request from PG developer about the
change CURSORs behavior, from WANRING to ERROR. Anyone report problemns
and the code was changed... :(

Alexandre


> Le Mon, Nov 10, 2003, à 04:30:42PM -0200, alexandre :: aldeia digital a
> écrit:
>
>> When the ODBC´s "Use Declare/Fetch" option is enabled the automatic
>> declaration of cursor great improve the performence for me.
>
> I second this observation with the workloads I'm confronted with.
>
>> In PG 7.3, this causes a WARNING only, but the behavior of this event is
>> changed to ERROR in PG 7.4, aborting the application.
>
> I'm not going to contribute much here, except express interest in seeing
> a solution to this (I didn't plan to move to 7.4 very quickly, but this
> would be a showstopper to me -- it's not a totally rare event for me to
> need datasets beyond what 32-bit Win32 will allow a single process to
> allocate, thus killing postgres.exe (ouch!)).
>
>     -- Cyrille
>
> --
> FRA: 61-18(FIJ) 51-29(JPN) 51-9(SCO) 41-14(USA) >> 43-21(IRL)
>