Is 'flip' really necessary?

aditya siram

25 Jul 2010 25 Jul '10

5:13 p.m.

Eta-reducing is nice, and sometimes it makes code more readable. But 'flip' is one of those functions that always seems to hinder rather than help readability, conversely factoring out flip always makes code easier to comprehend. I don't see a need for its existence - maybe I'm missing something and more experienced Haskellers can comment. -deech

Attachments:

attachment.html (text/html — 382 bytes)

Show replies by date

Antoine Latter

25 Jul 25 Jul

5:23 p.m.

On Sun, Jul 25, 2010 at 4:13 PM, aditya siram wrote:

...

Eta-reducing is nice, and sometimes it makes code more readable. But 'flip' is one of those functions that always seems to hinder rather than help readability, conversely factoring out flip always makes code easier to comprehend. I don't see a need for its existence - maybe I'm missing something and more experienced Haskellers can comment.

I'm a fan of (flip map), (flip mapM) and the like. These are usually renamed 'for' and 'forM' (and then 'forM_', of course). It makes for nice in-line function literals:

...

flip map xs $ \elem -> do something long and complex with xs

Antoine

Daniel Fischer

5:27 p.m.

On Sunday 25 July 2010 23:13:16, aditya siram wrote:

...

Eta-reducing is nice, and sometimes it makes code more readable. But 'flip' is one of those functions that always seems to hinder rather than help readability, conversely factoring out flip always makes code easier to comprehend. I don't see a need for its existence - maybe I'm missing something and more experienced Haskellers can comment.

-deech

For example, sortBy (flip compare) is very convenient (easier to parse¹ and it provides fewer opportunities for error than (\a b -> compare b a)). Also foldl' (flip Set.insert) Set.empty values and such stuff. Sometimes, functions have the wrong argument order for what you want to do, then flip is very handy. ¹ This is of course not a universal truth but a personal preference.

Edward Z. Yang

5:53 p.m.

Excerpts from aditya siram's message of Sun Jul 25 17:13:16 -0400 2010:

...

Eta-reducing is nice, and sometimes it makes code more readable. But 'flip' is one of those functions that always seems to hinder rather than help readability, conversely factoring out flip always makes code easier to comprehend. I don't see a need for its existence - maybe I'm missing something and more experienced Haskellers can comment.

An interesting alternate spin on flip is infix notation combined with partial application, such as: (`foobar` 3) which is equivalent to \x -> foobar x 3 I frequently use this, although the jury's out on whether or not it's more readable. Cheers, Edward

David Virebayre

26 Jul 26 Jul

2:33 a.m.

On Sun, Jul 25, 2010 at 11:53 PM, Edward Z. Yang wrote:

...

An interesting alternate spin on flip is infix notation combined with partial application, such as:

(`foobar` 3)

which is equivalent to

\x -> foobar x 3

I frequently use this, although the jury's out on whether or not it's more readable.

I had HLint suggest me this : before : listeEtagTot = concatMap (flip listeEtagArm cfgTypesTringle) listeArmOrd after : listeEtagTot = concatMap (`listeEtagArm` cfgTypesTringle) listeArmOrd David.

Ivan Miljenovic

2:39 a.m.

On 26 July 2010 16:33, David Virebayre wrote:

...

On Sun, Jul 25, 2010 at 11:53 PM, Edward Z. Yang wrote:

...
An interesting alternate spin on flip is infix notation combined with partial application, such as:

(`foobar` 3)

which is equivalent to

\x -> foobar x 3

I frequently use this, although the jury's out on whether or not it's more readable.

I had HLint suggest me this :

before :

listeEtagTot = concatMap (flip listeEtagArm cfgTypesTringle) listeArmOrd

after :

listeEtagTot = concatMap (`listeEtagArm` cfgTypesTringle) listeArmOrd

However, if you had something like this, I think the flip version is nicer: With flip: foo = map (f . flip g x) Without flip: foo = map (f . (`g` x)) -- Ivan Lazar Miljenovic Ivan.Miljenovic@gmail.com IvanMiljenovic.wordpress.com

Nils

4:42 p.m.

On 26.07.2010 08:33, David Virebayre wrote:

...

listeEtagTot = concatMap (`listeEtagArm` cfgTypesTringle) listeArmOrd

...

listeEtagTot = concatMap (listeEtagArm `flip` cfgTypesTringle)

You can use flip as a "wildcard" aswell: listeArmOrd Makes it even more readable in my opinion, since this really "shows" you where the value belongs to.

aditya siram

4:53 p.m.

That's just cool. I now reverse my original statement - 'flip' does have it's place in the pantheon of standard Haskell functions. -deech On Mon, Jul 26, 2010 at 3:42 PM, Nils wrote:

...

On 26.07.2010 08:33, David Virebayre wrote:

...
listeEtagTot = concatMap (`listeEtagArm` cfgTypesTringle) listeArmOrd

You can use flip as a "wildcard" aswell:

...
listeEtagTot = concatMap (listeEtagArm `flip` cfgTypesTringle) listeArmOrd

Makes it even more readable in my opinion, since this really "shows" you where the value belongs to.

_______________________________________________ Haskell-Cafe mailing list Haskell-Cafe@haskell.org http://www.haskell.org/mailman/listinfo/haskell-cafe

Max Rabkin

5:25 p.m.

On Mon, Jul 26, 2010 at 10:42 PM, Nils wrote:

...

On 26.07.2010 08:33, David Virebayre wrote:

...
listeEtagTot = concatMap (`listeEtagArm` cfgTypesTringle) listeArmOrd

You can use flip as a "wildcard" aswell:

...
listeEtagTot = concatMap (listeEtagArm `flip` cfgTypesTringle) listeArmOrd

Makes it even more readable in my opinion, since this really "shows" you where the value belongs to.

It took me a fair while (I'm talking on the order of half a minute) to figure out what that meant, but it's pretty cool. Maybe a different name would be better? How about (??) or "it"?

...

listeEtagTot = concatMap (listeEtagArm ?? cfgTypesTringle) listeArmOrd listeEtagTot = concatMap (listeEtagArm `it` cfgTypesTringle) listeArmOrd

--Max

Daniel Fischer

5:36 p.m.

On Monday 26 July 2010 23:25:27, Max Rabkin wrote:

...

It took me a fair while (I'm talking on the order of half a minute) to figure out what that meant, but it's pretty cool.

Yeah, really neat.

...

Maybe a different name would be better? How about (??) or "it"?

...
listeEtagTot = concatMap (listeEtagArm ?? cfgTypesTringle) listeArmOrd listeEtagTot = concatMap (listeEtagArm `it` cfgTypesTringle) listeArmOrd

I think (??) is far better. Additionally, it doesn't run into problems with ghci's magical `it'.

...

--Max

Vo Minh Thu

5:45 p.m.

2010/7/26 Daniel Fischer :

...

On Monday 26 July 2010 23:25:27, Max Rabkin wrote:

...
It took me a fair while (I'm talking on the order of half a minute) to figure out what that meant, but it's pretty cool.

Yeah, really neat.

...
Maybe a different name would be better? How about (??) or "it"?

...
listeEtagTot = concatMap (listeEtagArm ?? cfgTypesTringle) listeArmOrd listeEtagTot = concatMap (listeEtagArm `it` cfgTypesTringle) listeArmOrd

I think (??) is far better. Additionally, it doesn't run into problems with ghci's magical `it'.

(__) is quite good :)

Vo Minh Thu

5:47 p.m.

2010/7/26 Vo Minh Thu :

...

2010/7/26 Daniel Fischer :

...
On Monday 26 July 2010 23:25:27, Max Rabkin wrote:

...
It took me a fair while (I'm talking on the order of half a minute) to figure out what that meant, but it's pretty cool.

Yeah, really neat.

...
Maybe a different name would be better? How about (??) or "it"?

...
listeEtagTot = concatMap (listeEtagArm ?? cfgTypesTringle) listeArmOrd listeEtagTot = concatMap (listeEtagArm `it` cfgTypesTringle) listeArmOrd

I think (??) is far better. Additionally, it doesn't run into problems with ghci's magical `it'.

(__) is quite good :)

Well, I meant `__` Cheers, Thu

aditya siram

5:53 p.m.

It seems confusing to alias a function without adding any functionality just to make things slightly easier to read. Instead wouldn't it be better if this idiom were documented on haskell.org? -deech On Mon, Jul 26, 2010 at 4:47 PM, Vo Minh Thu wrote:

...

...
2010/7/26 Daniel Fischer :

...
On Monday 26 July 2010 23:25:27, Max Rabkin wrote:

...
It took me a fair while (I'm talking on the order of half a minute) to figure out what that meant, but it's pretty cool.

Yeah, really neat.

...
Maybe a different name would be better? How about (??) or "it"?

...
listeEtagTot = concatMap (listeEtagArm ?? cfgTypesTringle)

2010/7/26 Vo Minh Thu : listeArmOrd

...
...
...
...
listeEtagTot = concatMap (listeEtagArm `it` cfgTypesTringle) listeArmOrd

I think (??) is far better. Additionally, it doesn't run into problems with ghci's magical `it'.

(__) is quite good :)

Well, I meant `__`

Cheers, Thu _______________________________________________ Haskell-Cafe mailing list Haskell-Cafe@haskell.org http://www.haskell.org/mailman/listinfo/haskell-cafe

Brandon S Allbery KF8NH

9:21 p.m.

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 On 7/26/10 17:53 , aditya siram wrote:

...

It seems confusing to alias a function without adding any functionality just to make things slightly easier to read. Instead wouldn't it be better if this idiom were documented on haskell.org http://haskell.org?

And yet much of Applicative is exactly that. (It even started out being called "idioms".) - -- brandon s. allbery [linux,solaris,freebsd,perl] allbery@kf8nh.com system administrator [openafs,heimdal,too many hats] allbery@ece.cmu.edu electrical and computer engineering, carnegie mellon university KF8NH -----BEGIN PGP SIGNATURE----- Version: GnuPG v2.0.10 (Darwin) Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/ iEYEARECAAYFAkxONH0ACgkQIn7hlCsL25XLsgCcDLhDWBfkHcAGgc7145NuY08/ N4IAmgPHachdFS28mOGHeYD7yy1A3px2 =NLVp -----END PGP SIGNATURE-----

Ozgur Akgun

5:55 p.m.

I think it is pretty cool as well. But I think there is a problem with viewing it as a wildcard. let's say we define the following: (??) = flip foo :: a -> b -> c foo ?? x :: a -> c Perfect! But saying ?? can be used as a wildcard might in the following wrong perception: foo x ?? :: b -> c -- WRONG Just a small concern. Other than that, very neat! On 26 July 2010 21:42, Nils wrote:

...

On 26.07.2010 08:33, David Virebayre wrote:

...
listeEtagTot = concatMap (`listeEtagArm` cfgTypesTringle) listeArmOrd

You can use flip as a "wildcard" aswell:

...
listeEtagTot = concatMap (listeEtagArm `flip` cfgTypesTringle) listeArmOrd

Makes it even more readable in my opinion, since this really "shows" you where the value belongs to.

_______________________________________________ Haskell-Cafe mailing list Haskell-Cafe@haskell.org http://www.haskell.org/mailman/listinfo/haskell-cafe

-- Ozgur Akgun

Edward Z. Yang

6:24 p.m.

IMO, if you really want a wildcard, just write a lambda... \x -> foo 1 x 3 Cheers, Edward

Bill Atkins

9:27 p.m.

Seconded. On Monday Jul 26, 2010, at 6:24 PM, Edward Z. Yang wrote:

...

IMO, if you really want a wildcard, just write a lambda...

\x -> foo 1 x 3

Cheers, Edward _______________________________________________ Haskell-Cafe mailing list Haskell-Cafe@haskell.org http://www.haskell.org/mailman/listinfo/haskell-cafe

Yves Parès

27 Jul 27 Jul

3:55 a.m.

(??) is misleading, some may be tempted to write things like: func ?? 45 ?? x , forgetting that ?? is just a mere operator, not a syntactic convenience. Unfortunately, Haskell doesn't provide Scala's underscore for partially applied functions: func(56, _, "foo", _)

Nils Schweinsberg

26 Jul 26 Jul

7:09 p.m.

On 26.07.2010 23:55, Ozgur Akgun wrote:

...

I think it is pretty cool as well. But I think there is a problem with viewing it as a wildcard.

let's say we define the following:

(??) = flip

foo :: a -> b -> c foo ?? x :: a -> c

Perfect!

But saying ?? can be used as a wildcard might in the following wrong perception:

foo x ?? :: b -> c -- WRONG

This looks interesting. I played around with this for a bit: {-# LANGUAGE MultiParamTypeClasses , FunctionalDependencies , FlexibleInstances #-} class Wildcard f v r | f -> v r where (??) :: f -> v -> r instance Wildcard (a -> b -> c) b (a -> c) where (??) = flip instance Wildcard (b -> c) b c where (??) = id f :: String -> Int -> String f s i = s ++ show i a :: String -> String a = (f ?? 5) b :: Int -> String b = (f "Int: " ??) Sadly, this won't typecheck: pattern.hs:19:0: Couldn't match expected type `Int' against inferred type `[Char]' Expected type: Int Inferred type: String When using functional dependencies to combine Wildcard (b -> c) b c, arising from the dependency `f -> a r' in the instance declaration at pattern.hs:12:9 Wildcard (String -> Int -> String) Int (String -> String), arising from a use of `??' at pattern.hs:19:5-10 When generalising the type(s) for `a' Ideas anyone? :)

wren ng thornton

27 Jul 27 Jul

11:45 a.m.

aditya siram wrote:

...

Eta-reducing is nice, and sometimes it makes code more readable. But 'flip' is one of those functions that always seems to hinder rather than help readability, conversely factoring out flip always makes code easier to comprehend. I don't see a need for its existence - maybe I'm missing something and more experienced Haskellers can comment.

Not quite Haskell but, in combinator calculi 'flip' is unexpectedly powerful. It'd be odd not to have/use it in Haskell. Though in general I agree that it should be used sparingly, for legibility's sake. -- Live well, ~wren

5412

Age (days ago)

5414

Last active (days ago)

List overview

Download

19 comments

15 participants

participants (15)

aditya siram
Antoine Latter
Bill Atkins
Brandon S Allbery KF8NH
Daniel Fischer
David Virebayre
Edward Z. Yang
Ivan Miljenovic
Max Rabkin
Nils
Nils Schweinsberg
Ozgur Akgun
Vo Minh Thu
wren ng thornton
Yves Parès