Re: aliasing table name in update

Holger Klawitter <holger@klawitter.de> writes:
> Upgrading from 7.0.3 to 7.1(rc4) went very smooth - but a small
> issue came up:

> The following statement does not work any more:
>     update l set i=2 from longname l where l.i=1;

I'd say it's a fluke that that ever was accepted.  Given the obvious
ambiguity of interpretation (suppose there is a table named l?), it
should be rejected IMHO.

There has been some talk of allowing

    update longname [as] l set ...

but we do not currently accept that.

            regards, tom lane