So, I am playing around with a concept named "NewType" and I am taking inspiration from languages like F# and Scala.
My objective, for learning purposes mostly, is to build a macro that makes creating this abstraction something that takes no more than a single line of code.
I would like to create a macro that allows me to do something like this:
defmodule User do
require NewType # an absolutely original name for the macro :D
deftype Name, String.t() # Usage of said macro. Here I am defining a new type called "Name"
@enforce_keys [:name, :age]
defstruct [:name, :age]
@type t :: %__MODULE__{
name: Name.t,
age: integer()
}
@spec new(Name.t, integer) :: User.t
def new(name, age), do: %User{name: name, age, age}
end
And now, here is how I could create a User
:
defmodule Test do
alias User
import User.Name
@spec run :: User.t
def run do
name = Name("John")
User.new(name, 25)
end
end
This interface might remind you a little of the Record
interface. That's because I think its API has some good ideas I would like to explore.
So, as a starting point I tried reading the source code for Record, but I was not really able to pick it up and use it to create an implementation for my use case, mainly because I don't need/want to interface with Erlang records at all.
So, an implementation possibility would be to, under the hood, turn this into a tuple:
defmodule NewType do
defmacro new(name, val) do
quote do
NewType.to_tuple(unquote(name), unquote(val))
end
end
def to_tuple(name, val), do: {String.to_atom(name), val}
end
However, this is miles away from the interface I want to create ...
Name("John")
?After reading more about macros in Elixir, talking to the community and reading about NewType
, I have refined my ideas. While the exact implementation of my original idea is not possible, with some changes you can still get the core benefit of NewType
.
Name("John")
syntax. As explained in this post this syntax is not valid in Elixir.defguard
. Because the type is @opaque
it is not possible to have a guard that analyses the internal structure of the data without having dialyzer complaining. Since the main goal here is to have Dialyzer help me detect issues, and since the internal structure of the opaque data can only be analyzed by functions that belong to the module itself, this means this idea is not possible.new
. Originally I thought about having some verification mechanism, but this is not necessary, since dialyzer will let the user know if the user is invoking new
with an incorrect parameter.Age.age?
or Name.name?
I have opted for the more general NewType.is_type?/2
, which will accomplish the same and is more general.With these changes in mind, this is the macro I came up with:
defmodule NewType do
defmacro deftype(name, type) do
quote do
defmodule unquote(name) do
@opaque t :: {unquote(name), unquote(type)}
@spec new(value :: unquote(type)) :: t
def new(value), do: {unquote(name), value}
@spec extract(new_type :: t) :: unquote(type)
def extract({unquote(name), value}), do: value
end
end
end
@spec is_type?(data :: {atom, any}, new_type :: atom) :: boolean
def is_type?({type, _data}, new_type) when type == new_type, do: true
def is_type?(_data, _new_type), do: false
end
Which can be used like:
type.ex:
defmodule Type do
import NewType
deftype Name, String.t()
end
test.ex:
defmodule Test do
alias Type.Name
@spec print(Name.t()) :: binary
def print(name), do: Name.extract(name)
def run do
arg = 1
name = Name.new(arg) # dialyzer detects error !
{:ok, name}
end
end