* Claudio Freire (klaussfreire@gmail.com) wrote:
> On Thu, Oct 2, 2014 at 4:40 PM, Stephen Frost <sfrost@snowman.net> wrote:
> > The downside of the 'explain' approach is that the script then has to be
> > modified to put 'explain' in front of everything and then you have to go
> > through each statement and consider it. Having a 'dry-run' transaction
> > type which then produces a report at the end feels like it'd be both
> > easier to assess the overall implications, and less error-prone as you
> > don't have to prefex every statement with 'explain'. It might even be
> > possible to have the local "view" of post-alter statements be available
> > inside of this 'dry-run' option- that is, if you add a column in the
> > transaction then the column exists to the following commands, so it
> > doesn't just error out. Having 'explain <whatever>' wouldn't give you
> > that and so you really wouldn't be able to have whole scripts run by
> > just pre-pending each command with 'explain'.
>
> That sounds extremely complex. You'd have to implement the fake
> columns, foreign keys, indexes, etc on most execution nodes, the
> planner, and even system views.
Eh? We have MVCC catalog access.
> IMO, dry-run per se, is a BEGIN; stuff; ROLLBACK. But that still needs
> locks. I don't think you can simulate the side effects without locks,
Why? If you know the transaction is going to roll back and you only add
entries to the catalog which aren't visible to any other transactions
than your own, and you make sure that nothing you do actually writes
data out which is visible to other transactions..
> so getting the local view of changes will be extremely difficult
> unless you limit the scope considerably.
I agree that there may be complexities, but I'm not sure this is really
the issue..
Thanks,
Stephen