Happily, all of this has nothing to do with monads or Haskell's special syntax for monads. :) I would not call it static-types fanaticism, but rather, pattern-matching fanaticism. To me, "boolean blindness" is a cute way to describe how this: -- | Sum the first two ints of a list, if possible foo1 :: [Int] -> Maybe Int foo1 (x:y:_) = Just (x + y) foo1 [x] = Just x foo1 [] = Nothing is better than this: -- | Same thing, but with "blind bools" foo2 :: [Int] -> Maybe Int foo2 xs = if length x >= 2 then Just ((x!!1) + (x!!2)) else if length x == 1 then Just (x!!1) else Nothing NB: I already had to correct one mistake in that second code while writing it, and there may be others (I forget if it's (!!) or (!), and I also forget if it's 0-based or 1-based). The first is a description of what I want. It is correct by construction. The second requires converting the description of what I want (sum of the first two ints) into a description of how it is computed, with steps like "find the length and compare it to 2". By using pattern matching, I get both a guarantee of the structure I need, as well as information from within the structure at the same time. With an operational description, I have to do it in separate steps. To me, that's the crux of boolean blindness: Use pattern matching whenever possible! On 07/05/2018 11:04 AM, PY wrote:
So, the problem is that test of emptiness does not force you to something right. And possible errors are:
if empty: # do if NOT empty - BUG! else: # do if EMPTY - BUG TOO!
or
# do if NOT empty - BUG! if NOT empty: # now nothing or old "do if NOT EMPTY"
OK, I understand it. But how is it related to Booleans? :) Sure, if you use Maybe or Either you are forced with nature of ">>=": it cuts off incorrect branches. With if-then - it does not. But it's not related to Bool: Bool is result for predicates. Maybe/Either forces you with magic operation ">>=" (which is hidden by do-sugar). Bool does not force you - right. But it's problem of Haskell implementation. For example, Prolog "forces" you:
Haskell forces you in monad "do": do someInt <- someMaybe -- will not be executed if someMaybe is Nothing
Prolog forces you too but on success/fail (Boolean?): someGoal, anotherGoal % anotherGoal will not be executed if someGoal is False
Haskell adds only "bool" function which is close to ">>=" in terms of it hides semantic of right bool's processing, avoid those possible errors. If you will use "bool" anywhere when you use Bool as result - all will be fine, or? Sure, you can move "head" usage out of "bool" but you will get empty "bool"s argument. So, IMHO examples of problem with booleans is not related to Bool type at whole, but related to problem that Bool has kind * but not * -> * to be processed in monadic style (and to has needed ">>=" semantic to force you).
OK, but original article was not about Haskell's monads, but about Bool in general :) Also what I can't understand: if we will think in such manner does it mean that "if..test" construct is "boring"/"blindness" (excuse my English:) at whole? And all predicates must be removed from the language? What will happen to `filter` function without predicates? And no way to avoid "if..else" construct and predicates functions.
As for me, this question is related to static-types fanaticism and "How many angels could dance on the head of a pin". Example with "head" is totally incorrect - it can deconstruct list and no need to predicate function. But what I should do with isSpace, isLower, etc? How to use predicates at whole? :) To map a -> Bool to a -> Maybe a ? What about function which returns IO Bool? Action which can ends with non-critical failure (and need optionally logging, for example) ?
05.07.2018 17:27, Stefan Monnier wrote:
There is an opinion that Bool type has problems. It's "dangerous", because it's not good to be used as flag for success/fail result. I read this post: https://existentialtype.wordpress.com/2011/03/15/boolean-blindness/ and was shocked. How popular is such opinion? Is it true, that bool is "bad" type? To me the argument boils down to the `head` case mentioned by Alex.
Most programming languages force you to write code like
if List.empty l then ... else ... x = List.head l ...
where the problem is that the fact that the List.head call will find the list non-empty is not obvious (in the sense that it requires reasoning).
In contrast
case l | nil => ... | cons x xs => ...
makes it trivially obvious that `x` is extracted from a non-empty list without any reasoning needed at all.
Stefan
_______________________________________________ Haskell-Cafe mailing list To (un)subscribe, modify options or view archives go to: http://mail.haskell.org/cgi-bin/mailman/listinfo/haskell-cafe Only members subscribed via the mailman list are allowed to post.
_______________________________________________ Haskell-Cafe mailing list To (un)subscribe, modify options or view archives go to: http://mail.haskell.org/cgi-bin/mailman/listinfo/haskell-cafe Only members subscribed via the mailman list are allowed to post.