Structs: enforce_keys is not a good name

By defining a struct one guarantees that the instances’ keys are exactly those provided in the definition. By including a key within the enforce_keys attribute one ensures that the corresponding values cannot be nil.

In my view, the attribute name enforce_keys suggests that the keys listed there will be always present in the struct (which is actually the case for all keys provided in the definition) rather than that the corresponding values cannot be nil.

This is indeed the case of how enforce_keys behaves. Where are you seeing otherwise? Perhaps the docs could be clearer but the current behavior is exactly as described :slight_smile:

In my view, the keys are always in the struct as long as they are listed in definition:

iex(1)> defmodule A do defstruct [:a, :b] end
iex(2)> %A{}
%A{a: nil, b: nil}

In my view, one could say that “all the keys listed in the definition are enforced to be present in the struct instance”. And by looking at code itself one could expect that “enforce_keys” does exactly that (meaning the presence of keys not listed under “enforce_keys” within the struct is optional).

What enforce_keys does is that it makes sure the corresponding values cannot be nil. Hence these are not the keys that are “enforced” by enforce_keys. It is rather the non-nilness of values that is enforced. Hence the name enforce_keys is somewhat misleading and non-transparent (IMO).

Docs are fine. IMO, guess clear docs are not a substitute for clear unambiguous names (as they obviate the need to look in the docs in the first place).

EDIT: hopefully better explanations

Sorry, I was saying that your ideal view is the current behavior. For example:

iex(1)> defmodule A do
...(1)>   @enforce_keys [:name]
...(1)>   defstruct [:name, :age]
...(1)> end
{:module, A,
 <<70, 79, 82, 49, 0, 0, 8, 128, 66, 69, 65, 77, 65, 116, 85, 56, 0, 0, 1, 14,
   0, 0, 0, 27, 8, 69, 108, 105, 120, 105, 114, 46, 65, 8, 95, 95, 105, 110,
   102, 111, 95, 95, 9, 102, 117, 110, 99, ...>>, %A{age: nil, name: nil}}
iex> %A{age: 1}
** (ArgumentError) the following keys must also be given when building struct A: [:name]
    expanding struct: A.__struct__/1
    iex:3: (file)

iex> %A{name: nil}
%A{age: nil, name: nil}

Enforce keys doesn’t care about nil values, only keys. Does that clear things up?

4 Likes

Oh I see. So keys listed under enforce_keys must be present in the constructor and that’s all. They are, in a sense, like positional args to functions as they need to be always present.

1 Like