groupWhen – a groupBy that compares consecutive values

[Joachim Breitner <mail@…>]

groupBy has a minor problem: It always uses the first value of a group to decide whether a new value belongs to this group or the next. In several cases it would be more useful if it would take the last value of a group, thus always comparing consecutive values.

Example code:

groupWhen :: (a -> a -> Bool) -> [a] -> [[a]]
groupWhen _ []    = []
groupWhen _ [a]   = [[a]]
groupWhen f (a:l) = if f a (head c) then (a:c):r
                                    else [a]:c:r
  where (c:r) = groupWhen f l

Uses:

groupWhen (\a b -> b - a < 5) [1,2,4,10,14,16,18] -- Finding holes in a increasing series, e.g. log time stamps (my real use case)

or

groupWhen (<) [1,2,3,2,10,12,10,11] -- Group into strictly increasing sublists

Note that for transitive and symetrical comparision functions f (such as (==)), groupBy f == groupWhen f.

It should probably go to Data.List

[Joachim Breitner <mail@…>]

More comments and other implementations on ​http://hpaste.org/141

[ross]

The Haskell 98 Report (s17.6) says "When the By function replaces an Eq context by a binary predicate, the predicate is assumed to define an equivalence". Since the new function agrees with groupBy in that case, I think it would be reasonable to redefine groupBy as this more useful version.

The same applies to nubBy, and possibly deleteBy, deleteFirstsBy and intersectBy (which could have more general types to make this clear).

On the downside, it would mess up the simple uniform treatment of the By functions.

[igloo]

I think that if we have both groupBy and this new function then the names should be more similar (I'd expect groupWhen to do something completely different), and if possible allude to the functionality difference.

Are there any cases when the existing behaviour is better?

No matter what, I think this should go via ​http://www.haskell.org/haskellwiki/Library_submissions so I'm closing this bug.

By the way, the hpaste link seems to now be about catching monkeys.

Thanks

Ian

[mail@…]

I’m currently not interested enough to follow the complete procedure. Can’t you just reassign the bug somewhere appropriate, until someone picks it up?

[Lemming]

You may like to use my groupBy in: ​http://hackage.haskell.org/packages/archive/utility-ht/0.0.5.1/doc/html/Data-List-HT.html#v:groupBy

[Lemming]

It seems to worry many people:

​http://www.haskell.org/pipermail/haskell-cafe/2011-March/089832.html ​http://www.haskell.org/pipermail/libraries/2007-August/008039.html ​http://www.haskell.org/pipermail/haskell-cafe/2007-April/024971.html ​http://www.haskell.org/haskellwiki/List_function_suggestions#Generalize_groupBy_and_friends

[maeder]

I'ld like to add ​http://www.haskell.org/pipermail/libraries/2009-November/012794.html where such a function is called "runs".