Re: [Haskell-cafe] is proof by testing possible?

Neil Brown writes:

...

swap :: (a,b) -> (b,a)

...

We don't need to test it at all, it can only do one thing, swap its arguments. (Assuming it terminates.)

swap = undefined

Terminates and does not swap its arguments :-)

I think this counts as non-termination, and that for semantic purposes, any bottom value -- infinite loops, exceptions, undefined -- is treated equally. -k -- If I haven't seen further, it is by standing in the footprints of giants