Interesting reduction not caught by hlint.

[sheganinans]

So I have this function:

-- "Dual" of map, takes an object and a list of functions
-- and applies the object to each function in the list.
pam :: a -> [a -> b] -> [b]
pam a (f:fs) = f a : pam a fs
pam a [] = []

I was able to reduce it to:

pam :: a -> [a -> b] -> [b]
pam a = map (\f -> f a)

Sadly hlint didn't catch this reduction(Although it's able to catch a whole bunch of other reductions in my code any other day!), maybe this might lighten the way for a whole set of new reductions? Thanks for the awesome tool!

[ndmitchell]

Thanks for the bug report. Running your example, I get:

Sample.hs:4:1: Error: Use map
Found:
   pam a (f : fs) = f a : pam a fs
   pam a [] = []
Why not:
  pam a fs = map (\ f -> f a) fs

That's using version 1.9.16. Do you see any suggestions for your original input?

[sheganinans]

Oops, so I guess it seems when the function has a wildcard instead of a named variable is when it fails.

This works:

pam :: a -> [a -> b] -> [b]
pam a (f:fs) = f a : pam a fs
pam a [] = []

This doesn't:

pam :: a -> [a -> b] -> [b]
pam a (f:fs) = f a : pam a fs
pam _ [] = []

[ndmitchell]

I have a custom recogniser for these types of expression, it just needs generalizing a little. Three examples which don't match but which should are:

pam :: a -> [a -> b] -> [b]
pam a (f:fs) = f a : pam a fs
pam _ [] = []

pam2 :: a -> [a -> b] -> [b]
pam2 a (f:fs) = f a : pam2 a fs
pam2 _ _ = []

pam3 :: a -> [a -> b] -> [b]
pam3 a (f:fs) = f a : pam3 a fs
pam3 b _ = []

[sheganinans]

Interesting! Seems like you know about it and this issue is probably redundant. On my end it's really not a game-ending issue, so no reason to keep this thing around. Thanks for the speedy replies!

Well either way thanks a bunch for this super great project! I can't imagine life without it!

[ndmitchell]

Nope, I only investigated after your report, and I'll use this ticket to record the cases I should still do better.

[sheganinans]

Oh! Sounds great!

[mishun]

Should hlint further suggest

pam a = map ($ a)

or even

pam = map . flip ($)

?

[sheganinans]

Yes, I would agree if generalize is turned on, it would be good to have those suggestions recommended.

[ndmitchell]

The final map . flip ($) seems pretty confusing to my eye - HLint used to go that far with point-free operations, but generally it doesn't make code more readable in general.

For reducing \ f -> f a, it's certainly a hint I could add, and then HLint when run twice would do the replacement in two steps. At the moment I don't have it though, since it feels a little complex - it's in the ballpark of what I think would be reasonable though.