Re: Lookup Tables, the right way?

> Why do you need "nulls" to represent "missing values"?

You don't ever "need" nulls. There always exists an equivalent RM schema that does not contain any nullable attributes. Nulls and outer joins are conveniences that make it easier to answer existential queries.

<michael_at_preece.net> wrote in message news:1142463134.599507.84540_at_z34g2000cwc.googlegroups.com...
>
> Frank Hamersley wrote:
>
>> michael_at_preece.net wrote:
>> > In the Pick environment you can use the Pick DBMS to persist data or
>> > you can use the host OS's file system - or any other physical media,
>> > including that accessible through a relational database. Sometimes it
>> > makes sense to organise data into a two-dimensional matrix with, say,
>> > columns represting one dimension and columns the other. Sometimes it
>> > doesn't. The designer is free to choose whichever model is most
>> > logical. A willingness to be constrained to a two-dimensional model is
>> > what we call adherence to the illogical model.
>>
>> What makes you so sure it (the RM to be explicit) is two-dimensional?
>>
>> Cheers, Frank.
>
> This is your ball-park, your level playing field, not mine, but...
>
> Why do you need "nulls" to represent "missing values"? Isn't it because
> there can sometimes be no known value for a "cell", which we can think
> of as the intersection of a "column" (one dimension) and a "row"
> (t'other one)? In other words, every "n-dimensional tuple" (row) must
> have the same number of atomic values as every other "n-dimensional
> tuple" in the same table. Don't we then have a value to ascribe to "n"?
> Isn't "n" equal to the number of "columns"?
>
> Mike.
>
>
> Mike.
>
Received on Thu Mar 16 2006 - 13:35:04 CET