Re: identity columns

On 4/6/17, Peter Eisentraut <peter(dot)eisentraut(at)2ndquadrant(dot)com> wrote:
> On 4/4/17 22:53, Vitaly Burovoy wrote:
>> The next nitpickings to the last patch. I try to get places with
>> lacking of variables' initialization.
>> All other things seem good for me now. I'll continue to review the
>> patch while you're fixing the current notes.
>
> Committed with your changes (see below) as well as executor fix.

Thank you very much!

> As I tried to mention earlier, it is very difficult to implement the IF
> NOT EXISTS behavior here, because we need to run the commands the create
> the sequence before we know whether we will need it.

In fact with the function "get_attidentity" it is not so hard. Please,
find the patch attached.
I've implement SET GENERATED ... IF NOT EXISTS. It must be placed
before other SET options but fortunately it conforms with the
standard.
Since that form always changes the sequence behind the column, I
decided to explicitly write "[NO] CACHE" in pg_dump.

As a plus now it is possible to rename the sequence behind the column
by specifying SEQUENCE NAME in SET GENERATED.

I hope it is still possible to get rid of the "ADD GENERATED" syntax.

>
> --
> Peter Eisentraut http://www.2ndQuadrant.com/
> PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services

--
Best regards,
Vitaly Burovoy