At 4:50 PM +0200 9/22/99, Martin Ramsch wrote:
>On Wed, 1999-09-22 06:42:14 -0700, Rick Moore wrote:
>> id tinyint unsigned not null auto_increment );
>[...]
>> show columns from test2;
>> +-------+---------------------+------+-----+---------+----------------+...
>> | Field | Type | Null | Key | Default | Extra |...
>> +-------+---------------------+------+-----+---------+----------------+...
>> | id | tinyint(3) unsigned | | PRI | NULL | auto_increment |...
>> +-------+---------------------+------+-----+---------+----------------+...
>> 1 row in set (0.00 sec)
>>
>> I thought I specified "not null?" I haven't researched the manual
>> thoroughly, but I assume this is an intentional silent column
>> change...
>
>No, it's no column changes, because the "NOT NULL" of the column
>definition is reflected in the 'Null' column above: being empty means,
>no NULL values.
>
>But NULL as default values is--as I've recently learned here on the
>list--to make ODBC happy and have it really insert NULL into
>AUTO_INCREMENT columns. Otherwise you wouldn't get the automatic
>increments ...
>(Hope, I got that explanation right ...)
Close enough!
It's like TIMESTAMP, for which the same information is reported.
What it really indicates is that you are allowed to store a NULL
into either type of column (which is true), not that the column
will actually have NULL values stored in it.
I usually have to think about this a bit each time I see it, but
having SHOW COLUMNS/DESCRIBE report NOT NULL would have a similar
problem. Just in reverse. :-)
--
Paul DuBois, paul@stripped
