groupByWhile seems unintuitive

[Dave Rapin]

I was expecting to pass a list and a grouping function and to get a list of lists back. E.g.
groupWhile (\a b -> a.id == b.id) records -> List (List Record)

Instead we get a list of tuples with the item passed as the first and not included in the list of records. So I have to do a map to remove unwrap the tuple and add the Tuple.first into the list. List.map (\g -> ( optionalLabelFunction (Tuple.first g), Tuple.first g :: Tuple.second g )) groups

When grouping by even simple types like a list of string or ints it's awkward that the item being matched on is removed from the list and made to be the first item in the tuple. I can see why you opted to return a tuple to show what was grouped on, but not why it's removed from the group itself.

I think it would be valuable for records and cover all use cases if you could pass a function for the group label (first in tuple) and still include the record in the list, or simply make it a List of Lists.

(I don't mean to sound unappreciative. List.Extra has been a boon for me!)

[Palmer Paul]

The reason each group is a ( a, List a ) rather than a List a is so that the fact that a group cannot be empty is encoded in the return type. ( a, List a ) is a fairly common pattern in Elm for expressing a non-empty list. In order to convert from a non-empty list to a normal list, you can define a helper function in your code like

fromNonempty : ( a, List a ) -> List a
fromNonempty ( x, xs ) =
    x :: xs

groupWhile used to return a List (List a) up until July 2018. To see the full discussion that led to the decision to change it to how it is now, see #49 and #105.

[Dave Rapin]

Understood and thanks for the comprehensive background behind the decision! While I would definitely prefer List (List a) as I find that more intuitive / obvious, I can also see that there was an intelligent (and interesting) discussion about it.

As an aside, the pattern matching helper you provided is definitely prettier than what I was doing. Cheers!