Re: Using transactions with plpythonu

Tom Lane wrote:
> "imageguy" <imageguy1206(at)gmail(dot)com> writes:
> > So... unless I am missing something, I would suggest you CANNOT us
> > plpython (or perhaps any other pl language ??) to process transactions
>
> I think the point you are missing is that every function already runs
> within a transaction. You can't issue BEGIN/COMMIT from within a
> function because that would represent destroying the transaction that
> supports your execution of the function.

Tom, I think this is what I was missing, and of course this makes
sense.

> AFAICT you are worried about whether several different updates issued by
> your function will all be committed atomically. They will be; you don't
> need to, and indeed can't, do anything to adjust that. If there was
> some other issue you had, you need to be more specific...
>
You are exactly right on this. I was concered about the atomicity of
the updates.
In the test environment I setup, it appeared to me that each "INSERT"
was being committed immediately becuase and if the next transaction
failed it seemed I could still see the the previous transaction after
the function ended (ie. using pgAdminIII)

After reading your post, I will reset my "test" environment. Clearly I
had a flaw somewhere or didn't understand the results I was seeing.

Thanks so much for your guidance on this. I will post back here once I
have completed the retesting.

Thanks.
Geoff.