Sunday, December 29, 2013

Someone has done this before...

Happened across it while looking for Haskell-related blogs to link to: from Derek Wyatt's blog, a post titled "Haskell sequence over functions -- explained" gives this example:

sequence [(> 4), (< 10), odd] 7

gives the result we would get from our wonkyMap with the same arguments:

[True, True, True]

What is this sequence?

:t sequence
sequence :: Monad m => [m a] -> m [a]

If m is a monad, then sequence will map a [m a] to a m [a].

So, our functions must have the type m a for some m and a. Let's inquire with ghci once again.

:t sequence [(> 4), (< 10), odd]
sequence [(> 4), (< 10), odd] :: Integral a => a -> [Bool]

Aha. That reminds me of the exercise in Typeclassopedia: show that (->) r is a functor.  (WTF? That's the type of functions that take an argument of type r.) Turns out that not only is it a functor, it's a monad; Typeclassopedia refers to it as the "reader monad".

Unfortunately, we have dueling definitions here. Mr. Wyatt's blog post uses the Learn You a Haskell for Great Good! definition, as opposed to what we see from ghci, though clearly the end result is the same. Let's carry on with the version that ghci uses, and you neither want nor need to see my long, rambling, not-yet-successful attempt to convince myself that (->) r really is a monad. We'll just suppose it is for now, and see what the consequences are.

Haskell's type inference takes the most general possible type. The functions we have here are all Integral a => a -> Bool.  In fact, I suspect Integral b => b is r for this example, using b because the sequence declaration has already used a, and I don't need help getting more confused. For us, the a mentioned in sequence's declaration is Bool, and in our case, we have

sequence :: Integral b => [b -> Bool] -> b -> [Bool]

which is exactly what we wanted for our solution to the problem in that online Haskell course.

So, bye-bye wonkyMap; we can rewrite strong as

strong password = all $ sequence constraints password
    where constraints = [(>= 15) . length,
                         any isUpper,
                         any isLower,
                         any isNumber]

...and I definitely need to get my head around monads, if for no other reason than to come up with alternative Hoogle search strings. :)

No comments:

Riddler Classic, May 23, 2020—Holy Mackerel!

Another one using Peter Norvig's word list . It turns out that the word "mackerel" has a curious property: there is exactly ...