What is a good name for `foo kv m = m >>= kv . return`?

Learn
#1

I’m currently writing code where

foo :: Monad m => (m a -> m a) -> m a -> m a
foo kv m = m >>= kv . return

occurs a lot as a partial application foo kv (currently inline as (>>= kv . return) $ ...).

foo kv m is not the same as kv m; the latter performs effects in kv before performing effects in m and it is crucial for my use case to “squeeze dry” m of any effects before calling kv.

I can’t restrict kv to the more specific type a -> m a; so simply bind = flip (>>=) and writing bind kv $ ... is not an option.

So what would be an appropriate or established name for foo? It’s kind of like call-by-value, but for effects. (Of course it is nothing like actual call-by-value, because the a is not necessarily evaluated.)

1 Like
What is a higher-order effect?
#2

I think I’d just write it like this:

kv . pure =<< ...

I feel like it would not be worth its own name. Although I do wonder what kinds of kvs you have that cannot be rewritten to have the type a -> m a.

2 Likes
#3

Interesting. What sort of kvs are you using this with? I don’t think I’ve come across the need for this pattern.

If I was using this I’d probably try to write it as an operator on m actions rather than as a higher order function, for example, as

squeezeDry :: Applicative m => m a -> m (m a)
squeezeDry = fmap pure

Then I’d use it as squeezeDry m >>= kv.

4 Likes
Lazy idempotent monadic actions
#4

I actually wrote a similar function recently, I called it flush.

#5

kv . pure =<< ...

(=<<) indeed seems like a good builtin way to make (>>= kv . return) $ bleh $ do ... blocks nicer.

Although I do wonder what kinds of kvs you have that cannot be rewritten to have the type a -> m a.

My kv is actually a parameter body to a type class method

class HasBind m v where
  bind :: (m v -> m v) -> (m v -> m v) -> m v

instance HasBind M V where
  bind rhs body = body . return =<< fix rhs -- can be much more complicated than a mere `fix rhs`

There are instances that just do bind rhs body = body (fix rhs) (“call-by-name”), hence I need this combinator.

#6

An interesting combinator and perhaps more principled, but I think I prefer @jaror’s solution because I don’t want to parenthesize m (which can be quite big)

Edit: Of course, f =<< m needs parens as well when m = bleh $ do .... Sigh, won’t get much better, I guess.

#7

With blockarguments you can write

kv . pure =<< bleh do ...
1 Like
#8

Hmm:

# ghci
GHCi, version 9.4.4: https://www.haskell.org/ghc/  :? for help
ghci> :t \ kv m -> m >>= kv . return
\ kv m -> m >>= kv . return
  :: (Monad m1, Monad m2) => (m2 a -> m1 b) -> m1 a -> m1 b
ghci>

Now if the type is limited to:

Monad m => (m a -> m b) -> m a -> m b

…it’s a monadic version of ($!):

ghci> :t ($!)
($!) :: (a -> b) -> a -> b
ghci>

Gofer had a primitive with that same type called strict, which suggests:

strictM :: Monad m => (m a -> m b) -> m a -> m b
strictM kv m = m >>= kv . return

…MPJ does it again!

3 Likes
#9

For any law-respecting monad these two definitions of bind should be equivalent, due to the right identity law. Also note that bind is a common spoken name for the >>= operator, so your new method name could be confusing.

#10

The right identity law says: m >>= return = m, but that does not imply m >>= f . return = f m. Counterexample:

Nothing >>= const (Just ()) . return 
=/= 
const (Just ()) Nothing
#11

Mea culpa. What I said is still worth pointing out, but let me be more pedantic. The two instances will be equivalent if the following chain of two equations holds:

m >>= f . return == f (m >>= return) == f m

The second equation is implied by the right identity law, the first is true for most common f and m, but it doesn’t hold for f == const or for another example (>>=) = undefined. I take the OP’s question to be inspired by an industrial problem, so they may not be bothered by this.

#12

I disagree, it is not just const, here’s another example:

tell [()] >>= (\m -> m *> m) . return
=/=
tell [()] *> tell [()]

Here’s another one:

get >>= (\m -> put True *> m) . return
=/=
put True *> get

I’d say the equality does not hold for most non-trivial functions.

1 Like
#13

Ok, you’re right. I’m not sure how to characterize non-trivial functions for this purpose. I how OP finds this tangent useful nevertheless.

#14

Addendum: I think gofer’s strict function is actually just $! and that is not the same as \f x -> x >>= f . return. Forcing x does not have anything to do with forcing its effect. Think of m = putStrLn "foo"; forcing m is not the same as running the IO action. In general, forcing evaluation in the host language is not the same as forcing an effect encoded in the host language.

Still, I like strictM, or perhaps byValueM.

#15

o_0 …you mean that wasn’t made clear in my previous post? Hmm:

#16

Well, great name either way :slight_smile: