Bruce Momjian wrote:
>
> > "Thomas G. Lockhart" <lockhart@alumni.caltech.edu> writes:
> > >>>>>> create table authors (
> > >>>>>> zip char(5) null
> > >>>>>> );
> >
> > > Sheesh. After that long song and dance about why we can't implement
> > > this, it turns out that it works fine. We had been trying to implement a
> > > slightly different syntax, "WITH NULL", which conflicted with the
> > > SQL92-defined data type declaration "TIMESTAMP WITH TIME ZONE".
> >
> > > The "Practical SQL Handbook"-compatible form will be available in the
> > > next full release of Postgres. Thanks.
> >
> > Now that we have the syntax problem straightened out: I'm still confused
> > about the semantics. Does a "NULL" constraint say that the field
> > *must* be null, or only that it *can* be null (in which case NULL is
> > just a noise word, since that's the default condition)? I had assumed
> > the former, but Bruce seemed to think the latter...
>
> Can be null. Noise word. At least that is what I rememeber Thomas
> saying, and because it was noise, we removed it. In fact, it doesn't
> look like the standard accepts it, but there is no reason we can't.
This NULL clause is not part of constraints it is a default option and
we already support it,
there's nothing like:
CREATE TABLE table1 (field1 type NULL) in SQL92.
but the following is SQL92 and it works on PostgreSQL:
prova=> CREATE TABLE table1 (field1 INTEGER DEFAULT NULL);
CREATE
SQL92 syntax:
11.5 <default clause>
Function
Specify the default for a column or domain.
Format
<default clause> ::= DEFAULT <default option>
<default option> ::= <literal> | <datetime value function> | USER
| CURRENT_USER | SESSION_USER | SYSTEM_USER | NULL
-Jose'-