Protocols and Behaviours in Elixir are similar to interfaces in languages such as C#, Java etc.

Protocols can be thought of as interfaces for data whereas behaviours are like interfaces for modules, let’s see what this really means…

Protocols

A protocol is available for a data type, so let’s assuming we want a toString function on several data types but we obviously cannot cover all possible types that may be created in the future, i.e a Person struct or the likes. We can define a protocol which can be applied to data types, like this…

Let’s start by define the protocol

defprotocol Utils do
  @spec toString(t) ::String.t()
  def toString(value)
end

Basically we’re declaring the specification for the protocol using the @spec annotation. This defines the inputs and outputs, taking any params the after the :: is the return type. Next we define the function.

At this point we have now implementations, so let’s create a couple of implementations for a couple of the standard types, String and Integer

defimpl Utils, for: String  do
  def toString(value), do: "String: #{value}"
end

defimpl Utils, for: Integer  do
  def toString(value), do: "Integer: #{value}"
end

The for is followed by the data type supported by this implementation. So as you can see, we have a couple of simple implementation, but where protocols become more important is that we can now define the toString function on other types, let’s assume we have the Person struct from a previous post

defmodule Person do
  @enforce_keys [:firstName, :lastName]
  defstruct [:age, :firstName, :lastName]

  def create() do
    %Person{ firstName: "Scooby", lastName: "Doo", age: 30 }
  end
end

and we want to give it a toString function, we would simply define a new implementation of the protocol for the Person data type, like this

defimpl Utils, for: Person  do
  def toString(value), do: "Person: #{value.firstName} #{value.lastName}"
end

Now from iex or your code you can do sometihing like this

scooby = Parson.create()
Utils.toString(scooby)

and you’ve got toString working with the Person type.

Behaviours

Behaviours are again similar to interfaces but are used to define what a module is expected to implement. Let’s stick with the idea of a toString function which just outputs some information about the module that’s implementing it, but this time we’re expecting a module to implement this function, so we declare the behaviour as follows

defmodule UtilBehaviour do
  @callback toString() :: String.t()
end

We use the @callback annotation to declare the expected function(s) and @macrocallback for macros. As per the protocol we give the signature of the function followed by :: and the expected return type.

Now to implement this, let’s again go to our Person struct (remember this version of toString is just going to output some predefined string that represents the module)

defmodule Person do
  @behaviour UtilBehaviour

  @enforce_keys [:firstName, :lastName]
  defstruct [:age, :firstName, :lastName]

  def create() do
    %Person{ firstName: "Scooby", lastName: "Doo", age: 30 }
  end

  def toString() do
    "This is a Person module/struct"
  end
end

Now our module implements the behaviour and using Person.toString() outputs “This is a Person module/struct”.

We can also use the @impl annotation to ensure that you explicitly define the behaviour being implement like this

@impl UtilBehaviour
def toString() do
  "This is a Person module/struct"
end

This @impl annotation tells the compiler explicitly what you’re implementing, this is just an aid to development by making it clear what’s implementing what. If you use @impl once you have to use it on every behaviour.

Disclaimer: I’m going through some old posts that were in draft and publishing one’s which look relatively complete in case they’re of use: This post may not be 100% complete but does give a good overview of Elixir collections.

Lists in Elixir are implemented as linked lists which handle handle different types.

[3, "Three" :three]

Prepending to a list is faster than appending

list = [3, "Three" :three]
["pre" | list]

Appending

list = [3, "Three" :three]
list ++ ["post"]

List concat

[3, "Three" :three] ++ ["four", :4, 4]

List subtraction,

[2] -- [2.0]

Head and tail

hd [3, "Three" :three]
tl [3, "Three" :three]

Pattern matching

We can split the head an tail using Z

[head | tail] = [3.14, :pie, "Apple"]

The equivalent of a dictionary known as keyword lists in Elixir

[foo: "bar", hello: "world"]
[{:foo, "bar"}, {:hello, "world"}]

Keys can be atoms, keys are ordered and do not have to be unique

Maps

Unlike keyword lists they allows keys of any type and are unordered the syntax for a ,ap is %{}

map = %{:foo => "bar", "hello" => :world}

Defining structs is pretty simple in Elixir.

We create a struct within a module (like we do with functions), for example

defmodule Person do
  defstruct firstName: "", lastName: "", age: nil
end

As you can see, we’re explicitly settings the default values.

We’ve set age to nil, but we can also declare values with the default value of nil implicitly, but the fields that are to be implicitly set to nil must be at the start of the struct and we use [ ] to enclose the struct, for example

defstruct [:age, firstName: "", lastName: ""]

We can also mark fields/keys as required using the @enforce_keys attribute, for example

@enforce_keys [:firstName, :lastName]
defstruct [:age, :firstName, :lastName]

Now if we try to create and empty instance (i.e not setting firstName and lastName) we’ll get an error such as “(ArgumentError) the following keys must also be given when building struct Person: [:firstName, :lastName]”, but we’re jumping ahead of ourselves, let’s find out how we actually do create an instance of our struct first.

Structs take the name of the module, hence this struct is Person and we can create an instance of the struct using the following syntax, %{}, as shown below

def create() do
  %Person{ firstName: "Scooby", lastName: "Doo", age: 30 }
end

Now if you’ve already encountered maps, you’ll see that the struct syntax %{} is the same to the map syntax and that’s because structs are built on top of maps (but do not have all the capabilities of maps).

We can create an instance of a struct with it’s default values be simply using the following (assuming we’re not using @enforce_keys here)

defaultPerson = %Person{}

and this will simply bind to an instance with firstName and lastName of “” and age of nil.

We use standard “dot” syntax to, for example

iex(1)> scooby = Person.create()
%Person{firstName: "Scooby", lastName: "Doo", age: 30}
iex(2)> scooby.firstName
"Scooby"

We can create a new instance of a Person, where we change some fields using the | (update syntax), for example

iex(3)> scrappy = %{scooby | firstName: "Scrappy"}
%Person{firstName: "Scrappy", lastName: "Doo", age: 30}

We can also bind to a struct using pattern matching, i.e.

iex(4)> %Person{firstName: firstName} = scrappy
%Person{firstName: "Scrappy", lastName: "Doo", age: 30}
iex(5)> firstName
"Scrappy"

I said that structs were built on top of maps so let’s see if that’s true, try this

is_map(scooby)

and you’ll see true.

Structs can be made up of basic and more complicated types, i.e. structs can have fields which themselves are structs and so on.

I didn’t include much information in my post More modules in Elixer around the use macro and that’s because it really requires a post on it’s own. So here we go…

If you’ve uses Phoenix (partially covered in Elixir and Phoenix) you’ll have probably noticed that when we created the controller we gained access to functions such as json and when we setup the router we had a long list of except: atoms. This is because using use bought functions into the modules from Phoenix automatically, i.e. new, edit, create etc. functions.

The use macro is quite powerful, it essentially calls the __using__ macro within another module. The __using__ macro allows us to inject code from the other module.

Let’s see this in action…

We’ll start by creating a module which will include functionality that can be injected into another module

defmodule MyUse do
  defmacro __using__(_opts) do
    quote do
      def my_injected_fn() do
        "Hello World"
      end
    end
  end
end

In the above example, we create the macro with the my_injected_fn. The neat bit is the __using__ which injects this code into a module which uses the MyUse module, for example

defmodule MyModule do
  use MyUse

  def test_use() do
    my_injected_fn()
  end
end

This will inject all macros, but in the Phoenix example we want to inject only certain pieces from a module.

Before we move on let’s quickly address a couple of things in the code above. The quote macro tranforms the block of code into an AST (Abstract Syntax Tree), we can see an example of this by type the following into iex quote do: MyModule.testuse() and it will display something like {{:., [], [{:__aliases__, [alias: false], [:MyModule]}, :testuse]}, [], []}. It’s probably quite obvious that defmacro defines a macro and the __using__ calback macro is what allows us to extend other modules (as already mentioned).

Let’s change the MyUse module to allow us to inject specific functionality.

defmodule MyUse do
  defmacro __using__(which) when is_atom(which) do
    apply(__MODULE__, which, [])
  end

  def injector do
    quote do
      def my_injected_fn() do
        "Hello World"
      end
    end
  end
end

We’re now using the macro __using__ to select which bits of functionality we want to inject into another module. Admittedly in this example we just have a single piece of code to be injected, but bare with me.

So now to use this in our other modules we write the following

defmodule MyModule do
  use MyUse, :injector

  def test_use() do
    my_injected_fn()
  end
end

This will inject the injector defined code.

Let’s be honest this is not that useful with one function, so let’s extend the MyUse with a couple of functions (in my case, just to demonstrate things I’ve given them the same name but in most modules you’ll probably not be doing this

defmodule MyUse do
  defmacro __using__(which) when is_atom(which) do
    apply(__MODULE__, which, [])
  end

  def injector1 do
    quote do
      def my_injected_fn() do
        "Hello World 1"
      end
    end
  end

  def injector2 do
    quote do
      def my_injected_fn() do
        "Hello World 2"
      end
    end
  end
end

What we’re going to do, is in the calling module, we can select injector1 OR injector2 an without changing the calling function name (as it’s unchanged in the MyUse module)

defmodule MyModule do
  use MyUse, :injector1

  def test_use() do
    my_injected_fn()
  end
end

This will display “Hello World1” when evaluated. Switching the :injector2 will display “Hello World 2.

As mentioned it’s unlikely you’ve normally do this, it’s much more likely you might include different functions, maybe along these lines

defmodule MyUse do
  defmacro __using__(which) when is_atom(which) do
    apply(__MODULE__, which, [])
  end

  def injector1 do
    quote do
      def hello1() do
        "Hello World 1"
      end
    end
  end

  def injector2 do
    quote do
      def hello2() do
        "Hello World 2"
      end
    end
  end
end

Now we can inject these via use, like this

defmodule MyModule do
  use MyUse, :injector1
  use MyUse, :injector2

  def test_use() do
    IO.puts hello1()
    IO.puts hello2()
  end
end

Note: I’m still quite new to Elixir, the usage of two use clauses seems a little odd, there may be a better way to define such things.

I mentioned you could use import in much the same way, except use allows us to inject aliases, imports other use modules etc.

References

The ‘use’ Macro in Elixir.
Understanding Elixir’s Macros by Phoenix example
How to Use Macros in Elixir
Phoenix repo on GitHub

In the post A little more Elixir, we’re talking modules and functions we looked at a basic use of modules to allows us to declare functions (and macros, although we’ve not really touched these yet).

Modules can be nested, for example

defmodule Math do
  defmodule Fractions do
  # Nested module and functions 
  end

  # Top level module an functions
end

These don’t actually have a relationship with one another, instead Elixir essentially has them as separate modules, like this

defmodule Math do
  # functions
end

defmodule Math.Fractions do
  # functions
end

Importing modules

The import keyword (as the name suggests) imports a module’s functions and/or macros into the scope of the module or function they’re imported into. The scope is limited to that defined by the start and end of the module or function. Importing allows us to call functions from the module without having to use it’s module name. For example, we have a Math module with functions add and sub but we import it into our function like this

def f do 
  import Math
  add(1, 7)
end

Notice how the highlighted line does not need to prefix the module name to the function.

We can limit what’s imported using additional syntax where where we have only: or except: to reduce the scope of imports to the minimal, for example we do not need sub in our import so we could write

def f do 
  import Math, only [add: 2]
  add(1, 7)
end

In the above we import only the add function with arity of 2 (i.e. the 2 parameter function named add).

Aliasing modules

As the name suggests, alias allows us to create an alias to a module name, for example let’s say we have Math.Fractions whilst not a big deal to type if we’re typing it for every function it create a lot of “clutter” in our code, so instead we can alias the name like this

defmodule TestMod do
  alias Math.Fraction, as: F
  def f do
    F.some_function()
  end 
end

Require a module

The require keyword ensure the macro definitions of a module are compiled into the scope using the require. Or to put this another way require ensures that a required module is loaded before the module that’s calling into it. This will ensure any macros within it are available to the calling module and they’re scoped to that calling module.

Note: require is not like an alias, so you still need to prefix any macro/function calls with the module name (unless you also alias it ofcourse)

An example might be something like this

defmodule A do
  defmacro hello(arg) do
    quote do
      IO.puts "Hello #{unquote(arg)}"
    end
  end
end

defmodule B do
  def hello_world() do
    A.hello "World"
  end
end

In the above, if you compile this into iex using c(“require_sample.rx”) you’ll get warnings such as warning: you must require A before invoking the macro A.hello/1 if you try to execute the command B.hello_world(“Scooby”) you’ll get an error like this ** (UndefinedFunctionError) function A.hello/1 is undefined or private. However, there is a macro with the same name and arity. Be sure to require A if you intend to invoke this macro.

So as you can see, we need to use require, simply add the require A as below

defmodule A do
  defmacro hello(arg) do
    quote do
      IO.puts "Hello #{unquote(arg)}"
    end
  end
end

defmodule B do
  require A
  def hello_world() do
    A.hello "World"
  end
end

The use macro

The use macros allows us to “inject” any code into the current module. It’s used as an extension point.

I’ll dedicate a post of it’s own to the use keyword as it’s interesting what you can do with it.

Module Attributes

Attributes in Elixir are prefixed with the @ symbol. These add metadata to our module. An attribute is declared as a @name value pair. Whilst the name can be pretty much anything (within the allowable syntax), for example I might have a @my_ver 1, there are some reserved names

• @moduledoc is use for module documentation
• @doc is use for function or macro documentation
• @spec is use to supply a typespec for the function which follows
• @behaviour is used for OTP or under-defined behaviour (and yes it’s the UK spelling)

Here’s an example of creating our own attribute and we’re able to use it within our functions

defmodule Attributes do
  @some_name PutridParrot

  def attrib() do
    @some_name
  end
end

This will return the @some_name value.

You can set the attribute multiple times within the module scope. Elixir evaluates from top to bottom so functions after a change to the @some_name (above) will get the new value.

Guard clauses in Elixir allow us to extend the standard parameter pattern match to evaluating against a predicate.

For example

defmodule Guard do
  def check(value) when value == nil do
    IO.puts "Value is nil"
  end

  def check(value) when is_integer(value) or is_float(value) do
    IO.puts "#{value} is number"
  end

  def check(value) when is_atom(value) do
    IO.puts "#{value} is atom"
  end

end

We can see the use of the when keyword, also for multiple guards we use the or keyword. The right side of the when clauses is a predicate, hence should return true or false.

In these examples if we enter Guard.check(nil) the result is “Value is nil” if you check 123 or 123.4 (for example) the second guarded function is called etc.

Now one must remember that Elixir will evaluate functions from the top to the bottom, hence if the is_atom check is moved to the top of the functions, a nil will match as an atom and hence never reach the “Value is nil” returning function.

As we’ve seen, we supply a predicate to the function which means we can check the value for things like it’s type. We can also handle ranges, for example maybe we have one function that handle division but guards against a denominator of 0, then we might create two functions like this

def div(a, b) when b == 0 do
  if b == 0, do: raise("Divide by Zero")
end

def div(a, b) do
  a / b
end

I’m not saying this is better than writing something like the code below, but it’s another way of doing things

def div(a, b) do
  if b == 0, do: raise("Divide by Zero")

  a / b
end

As we saw, we use or not ||.

• The boolean operators allowed are or, and, not and !.
• Comparison operators include ==, !=, ===, >, <, >= and <=.
• Arithmetic operators +, –, *, / can be used.
• Join operators <> and ++ can be used.
• The in operator can be used.
• Type check functions such as is_integer etc.

For a more complete list which also include “guard-friendly functions” such as abs etc. can be viewed in the Guards documentation

One of the stranger bits of syntax I see in Elixir is for Atoms.

To quote the previous link “Atoms are constants whose values are their own name”.

What does this mean ?

Well imagine you were writing code such as

configuarion = "configuration"
ok = "ok"

Essentially the value is the same as the name, so basically instead if assigning a value of the same name we simply prefix the name with : to get the following

:configuration
:ok

In fact :true, :false and :nil are atoms but we would tend to use the alias of these, i.e. without the colon, true, false and nil.

The syntax for atoms is that they start with an alpha or underscore, can contain alphanumeric characters, can also contain @ and can end with and alphanumeric character or either ? or !. If you wish to create an atom with violates this syntax you can simply enclose in double quotes, i.e.

:"123atom"

and to prove it we can use

is_atom(:"123atom")

Atoms can also be declared starting with an uppercase alpha character such as

Atom
is_atom(Atom)

Atoms with the same content are always equivalent.

Modules are represented as atoms, for example :math is the module math and :”Elixir.String” is the same as a String.