Hello,

I’m building a small app for learning purposes where I have users and those users can create projects. Projects can have different statuses and when the project status changes, embeded meta field is updated. Projects can also have multiple users so in the meta field I save who updated the project status.

# projects.ex
def update_project(%User{} = user, %Project{} = project, %{"status" => status} = attrs) do
    attrs = prepare_status_attrs(user, project, attrs)
    changeset = Ecto.Changeset.change(project)

    Repo.update(changeset)
end

defp prepare_status_attrs(%User{} = user, %Project{} = project, attrs) do
    cond do
        attrs["status"] == "draft" ->
            Map.put_new(attrs, "meta", %{"created_by" => user.id, "created_at" => NaiveDateTime.utc_now()})
            # other
        attrs["status"] == "active" ->
            Map.put_new(attrs, "meta", %{"activated_by" => user.id, "activated_at" => NaiveDateTime.utc_now()})
            # other
        attrs["status"] == "deleted" ->
            Map.put_new(attrs, "meta", %{"deleted_by" => user.id, "deleted_at" => NaiveDateTime.utc_now()})
            # other
    end
end

# meta_embed.ex
embedded_schema do
    field :created_at, :naive_datetime
    field :created_by, :string
    field :activated_at, :naive_datetime
    field :activated_by, :string
    field :deleted_at, :naive_datetime
    field :deleted_by, :string
    # other fields
  end

def changeset(%MetaEmbed{} = meta, attrs) do
    meta
    |> cast(attrs, [:created_at, :created_by, :activated_at, :activated_by, :deleted_at, :deleted_by])
  end

The code above is a little bit simplified but what I have is a function that prepares the attrs in my projects.ex file before it goes to changeset and there are no changeset validations. The reason why I’m doing it is simply because here I have access to the user data etc. Is this a good practice or should I change the logic, pass all the user data to changeset and then do all the validations with changeset?