Wither Self

[Raimo Niskanen]

If the compiler would check for unbound variables in the right side of 
match expression *after* the pattern match, how would the following be 
treated?

     {Even, Odd, Strange} =
         {fun(0) -> true;  (N) -> Odd( N-1) end,
          fun(0) -> false; (N) -> Even(N-1) end,
	 Strange},

when Strange is bound vs unbound, or

     {Even, Odd, Strange} =
         {fun(0) -> true;  (N) -> Odd( N-1) end,
          fun(0) -> false; (N) -> Even(N-1) end,
	 {very,Strange}},

or even

     {Even, Odd, Strange} =
         {fun(0) -> true;  (N) -> Odd( N-1) end,
          fun(0) -> false; (N) -> Even(N-1) end,
	 case Strange() of
		true -> fun()->Odd(0)end;
		false -> fun()->Even(0)end
	 end},

or do you mean that it is just fun()s unbound variables that should be 
checked after the pattern match binding?

/ Raimo Niskanen, Erlang/OTP, Ericsson AB



Richard A. O'Keefe wrote:
> Joe Armstrong <joe@REDACTED> wrote:
> 	    Shouldn't there be a Self in funs?
> 	
> 		Fac = fun(0) -> 1;	
> 			 (N) -> N*Self(N-1) end
> 	 
> Well, if the compiler were just a touch smarter, you wouldn't need it.
> 
> If you have
>     <pattern> = ... <fun> ...
> then the check for unbound variables in <fun> should ask whether
> any variable will still be unbound *after* the pattern is matched
> instead of whether any variable is unbound *before* the pattern match.
> 
> Then you could just write
> 
> 	Fac = fun(0) -> 1; N -> N * Fac(N-1) end,
> 
> without needing Self.  This would also handle mutual recursion:
> 
>     {Even, Odd} =
>         {fun(0) -> true;  (N) -> Odd( N-1) end,
>          fun(0) -> false; (N) -> Even(N-1) end},
> 
> which the Self proposal would not deal with at all.
>