The semantics of irrefutable pattern matching is given as,
case e0 of { ~p -> e ; _ -> e' }
= let { y = e0 }
in
let { x1' = case y of { p -> x1 } }
in ...
let { xn' = case y of { p -> xn } }
in
e[x1'/x1,x2'/x2 .. ,xn'/xn]
x1,..,xn are all variables in p ;
y,x1',...,xn' are completely new variables
(Pg. No. 22 , Haskell 1.2 )
Does this mean that whenever a variable from the pattern p is required
(i.e x1,..xn) y is pattern matched against p, and the required
variable is returned.
Consider the following eg.
Using the above tanslation ,
case (1,2) of { ~(a,b) = a + b ; _ -> 0 }
== let { y = (1,2) }
in
let { x1' = case y of { (a,b) -> a }}
in
let { x2' = case y of { (a,b) -> b }}
in
x1' + x2'
So, when x1' + x2' is evaluated,
Is (1,2) pattern matched against (a,b) twice -- once for x1' and once
for x2' ??
I feel this is what the above let translation indicates.
But,
When (1,2) is pattern matched with (a,b) for the first time ,
bindings for both a and b can be created. So when x2'is needed,
pattern matching need not be done - the binding created
during the first pattern match can be used.
-- Namrata
[EMAIL PROTECTED]
