Re: patch: Add columns via CREATE OR REPLACE VIEW

"Robert Haas" <robertmhaas(at)gmail(dot)com> writes:
> On Thu, Aug 7, 2008 at 5:26 PM, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
>> Because it sidesteps the problem of tracking which column is supposed to
>> be which. If you try to do it through CREATE OR REPLACE VIEW, you have
>> to either be extremely restrictive (like probably not allow renaming
>> of columns at all), or write some AI-complete algorithm to guess at what
>> the user intended.

> The current code takes the approach of being extremely restrictive -
> it doesn't let you change anything at all. The code I'm proposing
> manages to relax that restriction without creating any ambiguity that
> anyone has been able to point out. All of the ambiguities that have
> been mentioned are problems that might be created by some other,
> entirely hypothetical patch.

Well, my feeling is that if we are inventing a new feature we ought not
paint ourselves into a corner by failing to consider what will happen
when obvious extensions to the feature are attempted. Whether the
present patch is self-consistent is not the question --- the question
is do we have a self-consistent vision of how we will later do the
other stuff like renaming, changing column type, etc.

regards, tom lane