Eric Wilson (2/18/2008)
With due respect, I'm afraid you missed my point. One (major) goal of the Relational Model was to be as uncluttered as possible. To that end, relations (a.k.a. tables) provide all the necessary complexity we need. Arrays, or any other data structure, only add complexity without adding additional power! This leads to arbitrariness of solutions and additional confusion between the logical and physical layers. No good can come of it.
I'm not arguing that relations are more or less powerful than arrays, in fact I fully agree that arrays are a poor substitute for relations -- that's a no-brainer. I wasn't suggesting that they *are* a substitute for relations. I was suggesting that the array is an incredibly useful feature in modern programming languages that is wholly missing from T-SQL, and one that, as T-SQL *is now implemented*, would be quite useful to solve a number of problems that we've had to hack around for many years.
As a programmer, I most certainly *do not* use a different solution simply "because I can". I look for what I believe to be the best solution to a given problem, and preferably look for a solution that someone else has already come up with. More "options" does not necessarily mean "more complexity", but more options just might give me the *option* to deliver a better solution. And delivering the "best" solution is my job (whatever the definition of "best" may be for the task at hand).
Anyway, I really don't disagree with you. I think you nailed it when you said tables "...are more powerful (if fully implemented...)". If we were able to toss tables around in TSQL just like any other datatype, that would certainly go a long way.