Tom Lane wrote:
Mike Benoit <[EMAIL PROTECTED]> writes:
On Thu, 2004-07-01 at 18:38 -0400, Alvaro Herrera wrote:
If we change the syntax, say by using SUBCOMMIT/SUBABORT for
subtransactions, then using a simple ABORT would abort the whole
transaction tree.
But then we're back to the application having to know if its in a
regular transaction or a sub-transaction aren't we? To me that sounds
just as bad.
Someone (I forget who at this late hour) gave several cogent arguments
that that's *exactly* what we want. Please see the prior discussion...
Right at the moment I think we have a consensus that we should use
SUBBEGIN/SUBEND or some such keywords for subtransactions. (I do not
say we've agreed to exactly those keywords, only that it's a good idea
to make them different from the outer-level BEGIN/END keywords.)
Either approach still needs some mechanism to clear the current stack of
transactions and subtransactions. That's why I was thinking ABORT ALL
and ROLLBACK ALL would be sufficient to cover that and be clear enough
to the user/programmer.
There was also some talk of offering commands based around the notion of
savepoints, but I'm not sure that we have a consensus on that yet.
regards, tom lane
---------------------------(end of broadcast)---------------------------
TIP 9: the planner will ignore your desire to choose an index scan if your
joining column's datatypes do not match