Re: [HACKERS] Re: TODO items

Just my 0.02 kronor . . .

On Sep 27, Bruce Momjian noted:

> > * Update table SET table.value = 3 fails
> >
> > AFAICS, the SQL92 syntax allows only a bare <column name> as the
> > target of a SET clause. Not sure it's worth expending any effort
> > on this one...
>
> Marked now as:
>
> * Update table SET table.value = 3 fails(SQL standard says this is OK)

In my opinion this should definitely _not_ be allowed. Let's be glad the
UPDATE command is so conceptually simple (cf. SELECT). The next thing they
want is ALTER TABLE foo RENAME foo.colum [ TO bar.something ??? -- moving
columns between tables, why not :) ] and then CREATE TABLE foo (foo.a int,
...); and it won't stop :)

> > MISC
> >
> > * Do autocommit so always in a transaction block(?)
> >
> > Huh? What is this supposed to mean?
>
> Some people want the SQL session to start inside a transaction, and you
> have to explicity use COMMIT, at which point you are in a new
> transaction that lasts until the next commit. Ingres SQL does this, and
> it is a pain, I think.

I have been wondering about this, too. Oracle does this as well. This is
also how they taught me SQL in university, so it is probably not out of
the blue. What do the standards say?

Then again, while I think that client programmers won't die if they type
an extra BEGIN here or there, this might be useful as a psql feature. Too
many times I've seen people type DELETE FROM <table>; by accident.

What do y'all think? (Besides the fact that this might be a pain to
implement.)

Peter

--
Peter Eisentraut - peter_e(at)gmx(dot)net
http://yi.org/peter-e