[erlang-questions] nand problem

[Richard A. O'Keefe]

On 27/01/2015, at 6:36 am, Roelof Wobben <r.wobben@REDACTED> wrote:
> I thought I have read somewhere that using _ for defensive programming was nog good practice
> but that was on using other on case on the next chapter.

It all depends.

In languages like SML, Haskell, and F#, the type system ensures that the
arguments *cannot* be anything other than the declared type.  So

b_nand :: Bool -> Bool -> Bool

b_nand True True = False
b_nand _    _    = True

is as safe as it could possibly be.

In languages like Prolog and Erlang, the absence of a static type system
means that the arguments could be anything, so writing

nand(true, true, false).
nand(_,    _,    true).

in Prolog would allow the query nand(1, 1, X) to succeed with X = true.

However, suppose you have a data type with N possible values.
Think of the abstract syntax trees that the Erlang compiler uses,
where N is in the low dozens.

Now suppose you want to write a function that takes two of these
values, but only has O(N) cases that make sense.  You *could* write
out all N**2 combinations, but you would have to be mad.

It's only when N is small (like in your example where N = 2) that
avoiding _ begins to make sense.