« Home

Master Elm pt. 2

Introduction to data structures

Ossi Hanhinen @ohanhi

Read this to

I expect you to know

Most people come to Elm from JavaScript, and traditionally JavaScript has really only had two data structures, object and array. Today I will go through all of the basic data structures in Elm, telling when to use them and comparing them to JavaScript as applicable.

Note: the double dash -- starts a comment in Elm for the rest of the line, like // in JavaScript.

I suggest you to read these in order, but here’s a quick navigation just in case:


Records are the closest thing to the object in JavaScript, but they are statically typed:

dog =
    { name = "Fluffy"
    , mother = "Muffin"

The type of dog here is now defined, even though I didn’t specify anything. This record has two String fields: name and mother. There’s no way of changing it anymore. The order of the fields is not relevant to the type.

Usually we write out the type definition, too. This might be done like so:

type alias Dog =
    { mother : String
    , name : String

Note that the record was using =, and the type alias is using :. In Elm, the colon : reads as “has the type”. Another thing to note is that type alias is literally an alias for a type. These are always interchangeable:

-- type annotation says:
-- "dogA has the type Dog"
dogA : Dog
dogA =
    { name = "Fluffy"
    , mother = "Muffin"

-- type annotation says:
-- "dogB has the type (record with two String fields, mother and name)"
dogB : { mother : String, name : String }
dogB =
    { name = "Fluffy"
    , mother = "Muffin"

sameDogs =
    dogA == dogB -- True

The two parts of syntax related to records are getting and setting:

-- Getting
dogA.name -- "Fluffy"
-- or
.name dogA -- "Fluffy"

-- Setting (always returns a new record)
{ dogA | name = "Fluuuffy" } -- { name = "Fluuuffy", mother = "Muffin" }

Remember that you can also pattern match on records!

-- type annotation says:
-- dogToString is a function from Dog to String
dogToString : Dog -> String
dogToString ({ name } as dog) =
    "The name of this dog is "
        ++ name
        ++ " and all of the data is: "
        ++ toString dog


Tuples are basically records without field names. Like records, the fields can be of varying types: one piece of a tuple can be a number while the other is a string. They are quite common in Elm, since they are very handy to create on the fly.

Tuples can look a little confusing at first, since they are constructed with parentheses () and the items are separated with commas ,.

-- info has the type "tuple of Int, String and String"
info : ( Int, String, String )
info =
    ( 42, "Life", "everything" )

You can pattern match on tuples too:

getComment : ( Bool, String ) -> String
getComment pair =
    case pair of
        ( True, word ) ->
            "Oh yes, " ++ word ++ " it is!"

        ( False, word ) ->
            "Nope, " ++ word ++ " isn't right."

Note: If you see the empty tuple () somewhere, it’s because () holds no information whatsoever, and therefore serves as a type (and value) for certain cases where the contents would be ignored anyway.


Lists are the only collection data structure that has any special syntax in Elm. They are quite similar to the JavaScript arrays. However, Elm lists don’t have indices for the items. They are simply sequences of items of a single type.

-- dogs has the type "a list containing Dogs"
dogs : List Dog
dogs =
    [ fluffy
    , scruffy

Lists are written with brackets [] and items within are separated with commas , just like JavaScript arrays.

Adding an item in front of a list is very efficient, because it doesn’t affect the rest of the list in any way! For the CS folks, Elm lists are immutable linked lists. In the drawing below, the blue one is the existing list with items always linking to the next one. The yellow is the new list. As you can see, only the new item "E" had to be added, linking into the previous first item "D".

Drawing of a List

Drawing of a List

You can combine (concatenate) two lists with ++, and add a single item to the front of a list with :: (pronounced cons).

moreDogs =
    sparky :: [ lassie, fifi ] ++ [ muffin ] -- [ sparky, lassie, fifi, muffin ]

Lists are the most used collection in Elm. They are very versatile, but if you need to access, and especially update certain items often, you should look at Array and Dict.

Oh and also, lists are the only collection you can pattern match on! This is very useful for some algorithms.

getFirstTwo : List Int -> Maybe ( Int, Int )
getFirstTwo list =
    case list of
        -- cons (::) can be used to "deconstruct" a list when pattern matching
        first :: second :: rest ->
            Just ( first, second )

        _ ->
            Nothing -- there weren't enough items in the list


Sets are an often overlooked data structure, even though they are very useful in some cases. (Ironically, I missed this section in the original edition of the article, too.) A set is an orderless collection of items, where each item can only appear once. You can also use Set to get only unique values in a List with Set.toList (Set.fromList myList). Keep in mind the resulting list won’t necessarily be in the same order, however.

As a practical example, in the lottery game Keno each number can only occur once.

kenoRound : Set Int
kenoRound =
        [ 1, 10, 43, 39, -- ...

Trying to Set.insert 10 kenoRound would return the same set as before, because 10 was already in the set.

Set has a whole bunch of interesting functions, such as union, intersect and diff, which are what Venn diagrams were made to depict:

Intersection of two sets

Intersection of two sets

In the figure above, the highlighted part is the intersection of the two sets. Union would be all of the yellow and blue set, and difference would be everything in the two sets that is not highlighted.

Note: The core Set can only have comparables as keys: numbers and strings, really. In a pinch, this might help you: a Set is really just a Dict that uses only the keys, and uses () for all the values. Thus, eeue56/elm-all-dict can be used for making a Set implementation of your own.


Dict is short for dictionary: the data structure has keywords and values. You can only find a value by its keyword, not the other way around. The keywords also need to be unique within the Dict. This data structure is incredibly useful for storing things like translations, but also collections of records that are referenced by their IDs.

Here’s a depiction of how data is organized in a Dict:

Drawing of a Dict

Drawing of a Dict

This is what it looks like in code:

type alias Id = Int

authors : Dict Id Author
authors =
        [ ( 0, { name = "Margaret Atwood" } )
        , ( 1, { name = "John Irving" } )

Note that getting a value from a Dict is always an unknown, and because Elm guarantees no runtime exceptions, Dict.get returns a Maybe value.

Note: The core Dict can only have comparables as keys: numbers and strings, really. If you need other kinds of keys, check out eeue56/elm-all-dict!


First off, if you are looking to use Arrays, use Skinney/elm-array-exploration, which will soon replace the current core implementation.

Array occupies a fixed chunk of the computer’s memory. If you add an item so that the array gets bigger, a new bigger chunk will be allocated somewhere else and then all of the contents are copied over. This makes adding items inefficient. But unlike List, you can efficiently retrieve or update e.g. the 50th item in an Array.

This is how arrays look like in my mind:

Drawing of an Array

Drawing of an Array

And this is how it looks in code:

words : Array String
words =
    Array.fromList [ "Hello", "world" ]

newWords : Array String
newWords =
    Array.set 1 "you!" words -- == Array.fromList [ "Hello", "you!" ]
Maybe your followers would be interested in this post? Get tweety with it!