Fri, Aug 7, 2020

Understanding type aliases

I recently wrote a single case discriminated union which is what I wanted but was also confused why it didn’t behave like a type alias and then learnt that these two are different things.

type CustomerId = int

type CustomerId = CustomerId of int

I was aware of both syntaxes and from a quick scan they look the same however they behave differently and rightly so. As I travel the F# road there is more emphasis on creating types for your functions. I have used this approach in C# to enforce type discrimination but it seems less prevalent in the mainstream from my experience. For example think of this:

public string DoSomething(string name, int age, string address)

This would become:

public string DoSomething(Name name, Age age, Address address)

This goes someway for example to disallow passing any random string or int value to methods throughout your domain. If everything is a int or string or any other primitive type there’s nothing stopping you calling the above method like so:

DoSomething("10 Downing St", 21, "Jon for PM").

This would compile fine and only at runtime are you likely to spot the issue.

So with this design in mind I wanted to create a type and away I went and created type CustomerId = CustomerId of int. This type was passed into a function and I needed to get the underlying value to convert to SQL. For some reason I had type CustomerId = int in my mind so I called ToString() on my type assuming that would get the underlying value. In fact what it returned was "CustomerId 2" ie/ a string-ified .NET object. This didn’t raise it’s head until runtime however as my SQL statement failed. Now if I had created my type as an alias and not a type eg. type CustomerId = int it would have worked fine. However, a type alias is just that, an alias and does not give you the design I described above. For example,

type Name = string
type Age = int
type Address = string

let doSomething (name:Name) (age:Age) (address:Address) =

I can call the function like so doSomething "Jon for PM" 21 "10 Downing St" and also doSomething "10 Downing St" 21 "Jon for PM" and both are valid at compile time but you haven’t achieved what you set out to achieve. What you actually want is:

type Name = Name of string
type Age = Age of int
type Address = Address of string

let doSomething (name:Name) (age:Age) (address:Address) =

To call this function you have to be much more explicit:

let res = doSomething (Name("Jon for PM")) (Age(21)) (Address("10 Downing St"))
printfn "%s" res

There’s no way to mix up the arguments however we are still at the original issue I faced, the return value will be a string-ified .NET object. In F# to get the inner value out you have to create a module to take your type and extract the value:

module Age =
       let value (Age input) = input

We can then amend our function to call it like so:

let doSomething (name:Name) (age:Age) (address:Address) =
    (Age.value age).ToString()

Now this might seem a bit of a pain but that is because we have type safety and have to be very explicit if we want to expose that value which is no bad thing really.

comments powered by Disqus