This is an alternative site for discovering Elm packages. You may be looking for the official Elm package site instead.
Self balancing trees in elm.
version 1.0.0
license BSD3
native-modules False
elm-version 0.18.0 <= v < 0.19.0
Tag 1.0.0
Committed At 2017-02-08 18:48:03 UTC
elm-lang/keyboard 1.0.1 <= v < 2.0.0 1.0.1
elm-lang/html 2.0.0 <= v < 3.0.0 2.0.0
elm-lang/core 5.1.1 <= v < 6.0.0 5.1.1
Fresheyeball/elm-function-extra 3.0.0 <= v < 4.0.0 3.0.0


Self-balancing binary search trees, in Elm.

Self-balancing binary search trees are a node-based data structure that allow amortized worst case asymptotic O (log n) lookup, insertion and removal, as well as O (n) enumeration.

Basic tree operations

Type signatures in this section assume a Tree type has been imported and exposed.

There are only a few operation that need to be implemented on a tree to allow the construction of a more complete set of operations.


First of all, we'll need some constructors.

Tree.empty : Tree comparable
Tree.singleton : comparable -> Tree comparable

empty allows creation of the empty, null tree. This is useful as a target for folding or comparison to other nodes.

Using the singleton constructor, we can create a tree for one single comparable value.

Manipulation and inspection

Other than construction, we'll also need some operations to manipulate and inspect the tree.

Tree.insert : comparable -> Tree comparable -> Tree comparable
Tree.remove : comparable -> Tree comparable -> Tree comparable
Tree.member : comparable -> Tree comparable -> Bool

insert and remove speak for themselves: allowing the insertion into and removal of nodes from a tree, means we can actually store the exact data we want in a tree, and remove it once it is no longer required or once it needs to be removed. Of course, without a way of knowing whether an element is in the tree, we can't do much with it. So, we also need a member function that will check if a given comparable appears in a given Tree.


Tree.foldl : (comparable -> a -> a) -> a -> Tree comparable -> a
Tree.foldr : (comparable -> a -> a) -> a -> Tree comparable -> a

Finally, there is the pair of folding functions foldl and foldr that allow enumerating the values in a tree, passing them into an accumulator and finally, returning this accumulated value. This accumulator could be any type of value, like a single value (for example, to sum all the elements of a tree), a list or even a tree. Note that this enumeration happens in O (n).

Using this last case -- folding a tree into another tree -- allows arbitrary types of manipulation. For example, removal of a node could be implemented as folding all values except the node to be removed into a new tree. Of course, that means the performance for removal would drop to O (n) rather than O (log n), so that's not how it's implemented.

However, a whole host of other operations could be defined -- in a rather performant manner -- in terms of folding. map, filter, toList, size, union, intersect, partition and difference are a few standard operations that may be implemented this way, with example implementations given below.

Fold-based operations

This section is a quick overview of operations that can efficiently and often succinctly be implemented in terms of a fold.


Convert tree to list in ascending order, using foldl.

Example implementation:

toList : Tree comparable -> List comparable
toList =
    foldl (::) []


Create tree from list by folding over the list and inserting into an initially empty tree. Folds over the list, rather than the tree.

Example implementation:

fromList : List comparable -> Tree comparable
fromList =
    List.foldl insert empty


Foldl over the list and incrementing an accumulator by one for each value that passes through the accumulator operation.

Example implementation:

size : Tree comparable -> Int
size =
    foldl (\_ acc -> acc + 1) 0


Fold over the tree, executing the specified operation on each value, and accumulating these values into a new tree.

Example implementation:

map : (comparable -> comparable2) -> Tree comparable -> Tree comparable2
map operator =
        (insert << operator)


Create a new set with elements that match the predicate.

Example implementation:

filter : (comparable -> Bool) -> Tree comparable -> Tree comparable
filter predicate =
        (\item ->
            if predicate item then
                insert item


Union is implemented by folding over the second tree and inserting it into the first tree.

Example implementation:

union : Tree comparable -> Tree comparable -> Tree comparable
union =
    foldl insert


Tree intersection creates a new Tree containing only those values found in both trees. This is implemented by filtering the right-hand set, only keeping values found in the left-hand set.

Example implementation:

intersect : Tree comparable -> Tree comparable -> Tree comparable
intersect left =
    filter (flip member left)


The differences between two trees is, in Elm land, defined as the elements of the left tree that do not exists in the right tree. As such, this is implemented by filtering the left tree for values that do not exist in the right set.

diff : Tree comparable -> Tree comparable -> Tree comparable
diff left right =
    filter (not << flip member right) left


Similar to filtering, this does not throw away the values that do not match the predicate, but creating a second tree from those values. The resulting trees are then returned as a tuple.

Example implementation:

partition : (comparable -> Bool) -> Tree comparable -> ( Tree comparable, Tree comparable )
partition predicate =
        (\item ->
            if predicate item then
                Tuple.mapFirst <| insert item
                Tuple.mapSecond <| insert item
        ( empty, empty )