So what is a function anyway?

Let us start with a set of entities, for example, the cars parked in a given block of a street. One can identify each car by its license plate, and then build a table with two columns: the first one with the license plate, and the second one with the corresponding the color of each car:

That’s a function that we can call carColor, that associates each car of the block with its color. The table is a representation of that function. A function from a set of entities A to a set of entities B is then a relation that associates elements between the sets A and B, with the property that each and every element of A has a one and only one corresponding element of B.

There are two things to note. First, a function is defined from one set to another set, in our example, from the set of license plates to the set of colors. To get the color of a particular car, you go to the table representation of the function, look up the plate in the column of license plates, and get the color from the second column. With this in mind, one can identify the from set as the input of the function, while the to set is the output.

Second, all elements of the input must relate to some element in the output set. In our example, every car in the block has a color assigned in the table. This means that there cannot be empty cells in the second column of our table.

❓ Can you come up with more examples of functions in the real world?

In math, functions are ubiquitous. Let us assume that $x$ is a real number, and define the function

that is, the function that returns the value $x$ plus one. The input set (which is called the domain of the function in math) is the set of all real values, $\mathbb{R}$, while the output set (the codomain in math) is also $\mathbb{R}$, because adding one to any real number is also another real number. Using the arrow notation, the complete definition of the function would be

that can be read as following: $f$ is a function from $\mathbb{R}$ to $\mathbb{R}$ such that $f$ of $x$ is $x + 1$”.

🔔 A slight detour around the codomain. The codomain is the set of entities where the function can possibly map input values into. For example, in the case of the color of the cars, the codomain is simply the set of all the possible colors. In many cases that information is too general, and it is convenient to define the range of the function, which is the set of actual values of outputs the function maps inputs into. The range in the cars example is the set {black, red, white}.

Multiple inputs and outputs

Let us take the example of a vending machine. In a vending machine, products are arranged in shelves, where each shelf is named by a letter. In each shelf, the products are aligned and identified by a number. Then, in A1 you have a bag of chips, in A2 a chocolate, in B1 a soda, and so on. The machine also has a keyboard with letters and numbers for you to choose the product. To buy something, you need to give the machine some money (coins, bills, credit card, etc), select the product by clicking the letter and the number on the keyboard. The machine returns the product and some cashback, if any. The inputs of our vendingMachine function are the money, the letter and the number you selected, and the outputs are the product and the cashback (if any).

An example from math could be a translation function, where given a point with coordinates (x) and $y$ in the plane, it returns a point with coordinates $x+1$ and $y+1$:

or more succintly

Partial application

When we feed a function of several input elements, we can obtain the proper output(s). But having many inputs opens a new possibility: what happens when one decides not to complete all the inputs? Let us find out. Let us assume that we entered a bill into the vending machine. It is clear that we will not get any product, because the machine still needs two more inputs: the letter of the shelf and the product number that we want. After we complete these two inputs, we will get our treat (and cashback, if any). So, entering money only in the vending machine leads to a state where two inputs are needed and two outputs will be returned. But, this is another function!!!. Let us call it vendingMachineAfterInsertBill that receives the letter of the shelf and the product number that we want and returns the product (and cashback, if any).

Going back to the math example, let us feed the function with just the $x = 3$ value,

Again, the result of feeding the function with one value is another function:

or

This property of functions is called partial application. Whenever you do not complete all the inputs of a function, you get another function.

Composition

Finally, we look here how to work with several functions at once. Let us assume that we have a function getFirstName that given the full name of a person, it returns the first name (it does not matter at this point the specifics of the implementation, not even the language). For example, when applied to ‘David Gilmour’, it returns ‘David’, or when applied to ‘Annie Lennox’, it returns, of course ‘Annie’. We also have a function getInitial that for a given name, it returns the initial. In the previous cases, ‘D’ for ‘David’ and ‘A’ for ‘Annie’.

Now we want to build a function that gives us the initial of the first name, given the full name. For ‘Paul McCartney’. We feed ‘Paul McCartney’ as input to the function getFirstName, which gives us the output ‘Paul’. Now, ‘Paul’ is the input of the function getInitial and returns ‘P’ as the final output. This plumbing where the output of one function is the input of another is called composition. Note that it is absolutely necessary that the output of the first called function (getFirstName) and the input of the second one (getInitial) are the same kind of entity, in our case, both are first names. You can see this pictures for a graphical explanation.

Let us now look at a math example. We defined before the function $f(x)$ that adds one to $x$, for example

What if we apply the $f$ function again? It means to compute

In general,

If we look carefully to the last expression, it looks like composing the function is like passing the function as the input itself ($f(f(x))$). This means that if we are going to have a programming language that implements function composition, in some way functions should be able to be passed as input to another functions.

❗️ The fact that we use the same function to compose with itself is not relevant to this discussion, one can compose as many different functions as one wants, provided that inputs and outputs are compatible in each composition step.

🔔 However, composing that particular function with itself is interesting. Imaging having only the zero and this function. You can create all the natural numbers {1, 2, …} just by composing this function with itself again and again. For example, $4 = f(f(f(f(0))))$, and so on. Therefore, given the number 0, $f(x) = x + 1$ and function composition, one can get all the natural numbers. Looks like there is something going on here. More on this, hopefully, in a future episode.

Functions in F#

The F# language implements functions in such a way that they satisfy the properties mentioned above. To define a function, the language also uses the keyword let:

let next x =
    x + 1 

We defined the function named next that receives an argument x. Notice that there are no other symbols or parentheses in the function definition. The body of the function should be indented, and there is no return keyword at the end. The function simply returns the last expression found in its body. Clean, isn’t it? Using the function is easy as well:

let one = next 0 
let two = next (next 0)

printfn "one: %A" one 
printfn "two: %A" two 

one: 1
two: 2

Notice that there is no need to use parentheses around the argument when using a function. However, you need to use them when passing a more complex expression as the argument to the function, such as in the case of two.

There is another way to write the computation of two, by using the pipe operator |>:

let anotherTwo =
    0 
    |> next
    |> next 
    
printfn "anotherTwo: %A" anotherTwo

anotherTwo: 2

This operator takes care of the plumbing when calling functions one after another. In the example above, the first |> receives 0 as the input, passes it to the next function, the second |> receives the output of the first next function and feeds it as input to the second next.

Another example. Let us assume that we have the functions getInitial and getFirstName defined as:

let getInitial name = 
    .... //Implementation not important right now

and

let getFirstName fullName = 
    .... //Implementation not important right now

and we defined the value

let paul = "Paul McCartney"

Then,

paul
|> getFirstName
|> getInitial 

Here the string value paul is fed into the getFirstName function as the input by the first pipe |>, and returns ‘Paul’ as output. Then, the string ‘Paul’ is passed as the input of the function getInitial that gives us the ‘P’.

Composition is so important in functional languages, that it has its own symbol in F#, >> :

let add2 = next >> next 
let two' = add2 0 
printfn "%A" two'

2

Yes, you can use the ' symbol in any identifier! (provided it is not the first character). Note also that we defined a function add2 by using the composition operator (no argument needed). This is equivalent to:

let add2' x = 
    x
    |> next
    |> next 

Remember that there is no return at the end of the function, just the last expression of the function is the return value.

Back to the names example, to clarify the order in which functions are composed.

let getInitialFromFirstName fullName =
    fullName
        |> getFirstName 
        |> getInitial 

and

let getInitialFromFirstName' =
        getFirstName >> getInitial 

are equivalent.

❓ Think about routines, procedures or functions that maybe you have written in your language of preference. Do they behave as F# functions? What are the main differences you see?

🏋🏽 We have a function mult2 that given a number x doubles that number. Without coding, can you determine what the next composite functions return when applied to 3? :

let f = mult2 >> next 
let g = next >> mult2 

Code the function mult2 and see the result by yourself.

Some final remarks for now on functions. First, note that the language use the same keyword let to bind simple values and functions to a name or identifier. This emphasizes the fact that in F# functions are ‘just’ values, and can be treated in the same way as, say, a simpler binding of an expression to an identifier. Second, the properties of functions that were discussed above match perfectly inmutability. In fact, functions receive immutable inputs and return an immutable value.