wttj / ecto_anon Goto Github PK

Currently it is only possible to anonymise a field with a value different from the default value providing a custom function for each field.

For example:

defmodule User do
  use Ecto.Schema
  use EctoAnon.Schema

  anon_schema [
    name: &__MODULE__.custom_string/3
    email: &__MODULE__.custom_string/3
  ]

  schema "users" do
    field :name, :string
    field :age, :integer
    field :email, :string

    anonymized()
  end

  def custom_string(:string, %__MODULE__{} = _user, _opts \\ []), do: "[REDACTED]"
end

user = Repo.get(User, id)
%User{name: "jane", age: 24, email: "[email protected]"}

EctoAnon.run(user, Repo)
{:ok, %User{name: "[REDACTED]", age: 24, email: "[REDACTED]"}}

My suggestion is to have an alternative to provider a different value per type.

For example:

defmodule User do
  use Ecto.Schema
  use EctoAnon.Schema

   anon_schema(
     [:name, :email],
     anonymized_values: [string: "[REDACTED]"]
   )

  schema "users" do
    field :name, :string
    field :age, :integer
    field :email, :string

    anonymized()
  end
end

user = Repo.get(User, id)
%User{name: "jane", age: 24, email: "[email protected]"}

EctoAnon.run(user, Repo)
{:ok, %User{name: "[REDACTED]", age: 24, email: "[REDACTED]"}}

I am not thinking about the implementation details now, only the public interface and its usage for the sake of the discussion of the new feature. 🙂

What do you think?

Would be great in migration & schema for field anonymized to have an option for a custom name as anonymized could be already taken !

Implement the anon_with: option in anon_field

Being able to specify an option for the runner like cascade: true and let it anonymize associations of our object could be awesome ! 🚀

Based on https://github.com/revelrylabs/ecto_soft_delete
We could add a anonymized_columns() for migrations and without_anonymized/with_anonymized query helpers

Right now, you have to precise email: [:anonymized_email, options: [partial: true]] to enable the option instead of something like email: [:anonymized_email, options: [:partial]]

As for now, we only have default anonymizing functions. It would be great to provide functions that are more business aware to keep anonymous data more readable.

Examples:

• Anonymous phone (ie XX-XX-XX-XX-02 where locale is configurable)
• Anonymous email (ie [email protected], where domain and tld are configurable)

TODO

• Provide basic migrations see Oban
• Make logging optional
• Provide (hooks | behaviour) to decorate the log changeset with specific needs (anonymizing done in an async queue? add a status, requested_at and finished_at fields to the table structure and fill them with your own data)
  • may we return the log id in order to find it later from the async queue processing?
  • the {source_id, source_type} identifier tuple may serve as well to find it later

Table structure

serialized_changeset can be created with:

:io_lib.format("~p", [email: "redacted"]) |> :lists.flatten()