On 08/05/2016 02:15 PM, Christian Ohler wrote:
> On Fri, Aug 5, 2016 at 12:55 PM, Rob Sargent <robjsargent(at)gmail(dot)com> wrote:
>> What sort of interface are you looking for. Where/When would you grab the information? Do what with it? Log triggers are the typical pattern here (with packages just for that sort of thing).
> I'm looking for a statement (or sequence of statements) that, when run
> within a transaction, tells the client if any writes are happening in
> that transaction – basically an interface similar to my proposed
> solution. I have some database wrapper code on the client that passes
> along arbitrary statements to Postgres, and I'd like for that wrapper
> code to be able to determine whether the current transaction is a
> write (for various purposes including logging). It would be nice not
> to make the client-side wrapper code dependent on instrumentation of
> the database schema itself.
>
> What can you tell me about my proposed solution? Does it do what I
> describe I want from it? Are there limitations I should be aware of?
At what point do you intend to inform the client that the db will be
(significantly) altered? You're planned call is within the transaction
and presumably late in the sequence (so the locks have been created).
Not sure if your client can see any results until after the transaction
has been largely executed. Does that matter?