Re: [Haskell-cafe] How is laziness defined?

I said:

Although it covers irrefutable (lazy) pattern matching in the second section, it does appear to miss the point that let bindings are always irrefutable.

Thus, there is no difference between these two:

let (x,y) = foo in ... let ~(x,y) = foo in ...

Andrew Bromage said:

let (x,()) = (1,undefined) in x let (x,~()) = (1,undefined) in x

In other words, the irrefutability of a pattern match does not distribute inside the top-level data constructor of the pattern. See also the second rule in section 3.17.2 of the Revised Haskell 98 Report (Informal Semantics of Pattern Matching).