Re: Example of expression bias?

Neo wrote:
> What would be the disadvantage of using untyped lamda calculus? By
> syntactically sugared implementations of the typed lambda calculus, do
> you mean the likes of Common LISP?
>

Using the untyped lambda calculus would simply mean you have a lot more work to do; for example, you'll have to define integers for yourself to begin with (remember that thread where I mentioned "frightening degrees of circumlocution" ?) So, it's doable, but it's the sort of mechanical grunt you'd be far better off letting a compiler take care of for you. The important thing to note is that languages like Haskell can be reduced to the lambda calculus straightforwardly; they really are *only* syntactic sugar.

I don't have much experience of Common LISP; but it does involve objects, which is a bad sign, and many LISP dialects include updateable variables, so no, I wouldn't go there. Standard ML would be better, but as mentioned in a thread with Marshall it too has side effects and updateable variables, and more unfortunately eager evaluation. Haskell would be a fine choice for me, Miranda, HOPE or maybe even Lazy ML perhaps. But, as I say, it's basically a choice of which syntactic sugar you prefer.

> If one came along, what significant/distinguishing characteristic would
> it exhibit?
>

Use of the appropriate FP language for data type and operator definition. Direct use of the relational algebra. Relations, tuples and attributes exactly as defined in the relational model, and obviously, no SQL or SQL mis-features. (But you probably guessed that :)

> What should the inteface for such a db resemble in your opinion? Which
> existing product's interface comes closest?

Speaking very personally, I'd just like to be able to fire up something like Hugs and be able to use relations & the relational algebra straight away. Interfaces are very much YMMV territory though. Received on Wed Jun 21 2006 - 02:31:28 CEST