Do we need behaviours?

[SimonLab]

The example is using behaviours to define the structure of the API interface:

phoenix-ecto-append-only-log-example/lib/append/append_only_log.ex

Lines 9 to 14 in e8eee00

	@callback insert(struct) :: {:ok, Ecto.Schema.t()} | {:error, Ecto.Changeset.t()}
	@callback get(integer) :: Ecto.Schema.t() | nil | no_return()
	@callback all() :: [Ecto.Schema.t()]
	@callback update(Ecto.Schema.t(), struct) ::
	{:ok, Ecto.Schema.t()} | {:error, Ecto.Changeset.t()}
	@callback get_history(Ecto.Schema.t()) :: [Ecto.Schema.t()] | no_return()

The implementation of the API is then done in the before_compile macro:

phoenix-ecto-append-only-log-example/lib/append/append_only_log.ex

Lines 25 to 48 in e8eee00

	defmacro __before_compile__(_env) do
	quote do
	import Ecto.Query
	def insert(attrs) do
	%__MODULE__{}
	|> __MODULE__.changeset(attrs)
	|> Repo.insert()
	end
	def get(entry_id) do
	sub =
	from(
	m in __MODULE__,
	where: m.entry_id == ^entry_id,
	order_by: [desc: :inserted_at],
	limit: 1,
	select: m
	)
	query = from(m in subquery(sub), where: not m.deleted, select: m)
	Repo.one(query)
	end

...
However I'm not sure behaviours are needed for the append only example.
From my understanding it makes sense to use behaviours when multiple modules need to provide the same functions but with a different logic of implemention. In our example only the AppendOnlyLog module is implementing the behaviours.

For a learning purpose it makes sense to see how behaviours works 👍 but I'm not sure if they provide a lot of values for this repository?

Should we still keep them to demonstrate how they can be used?

ref: https://elixir-lang.org/getting-started/typespecs-and-behaviours.html#behaviours

[nelsonic]

Related to: #8