Re: adding support for zero-attribute unique/etc keys

On 27/03/13 06:14, Darren Duncan wrote:
> On 2013.03.26 1:40 AM, Albe Laurenz wrote:
>> Darren Duncan wrote:
>>> So, determining if 2 rows are the same involves an iteration of
>>> dyadic logical
>>> AND over the predicates for each column comparison. Now logical AND
>>> has an
>>> identity value, which is TRUE, because "TRUE AND p" (and "p AND
>>> TRUE") results
>>> in "p" for all "p". Therefore, any 2 rows with zero columns each
>>> are the same.
>>>
>>> Since any 2 rows with zero columns are the same, the "UNIQUE
>>> predicate" is FALSE
>>> any time there is more than 1 row in a table.
>>>
>>> Does anyone agree or disagree with this logic?
>>
>> Yes :^)
>>
>> You could use the same kind of argument like this:
>>
>> UNIQUE is true iff any two rows in T satisfy for each column:
>> the column in row 1 is null OR the column in row 2 is null OR
>> the column in row 1 is distinct from the column in row 2
>>
>> Now you you iterate your logical AND over this predicate
>> for all columns and come up with TRUE since there are none.
>> Consequently UNIQUE is satisfied, no matter how many rows there are.
>>
>> In a nutshell:
>> All members of the empty set satisfy p, but also:
>> all members of the empty set satisfy the negation of p.
>>
>> You can use this technique to make anything plausible.
>
> Consider the context however. We're talking about a UNIQUE constraint
> and so what we want to do is prevent the existence of multiple tuples
> in a relation that are the same for some defined subset of their
> attributes. I would argue that logically, and commonsensically, two
> tuples with no attributes are the same, and hence a set of distinct
> tuples having zero attributes could have no more than one member, and
> so a UNIQUE constraint over zero attributes would say the relation
> can't have more than one tuple. So unless someone wants to argue that
> two tuples with no attributes are not the same, my interpretation
> makes more sense and is clearly the one to follow. -- Darren Duncan
>
>
>
Hmm as a user, I would like at most one row with empty fields covered by
a unique index.

Logical arguments to the contrary, remind me of the joke of the school
boy who told his unlearned father that he had learnt logic and could
prove that his father actually had 3 fish in his basket despite both
seeing only 2 fish. His unlearned father did not try to argue, and
simply said: well your mother can have the first fish, I'll have the
second, and that his learned son could have the third...