Cloak.Ecto.SHA256.hash_string/1 break when the value is nil

[SimonLab]

First of all thank you for the recent release!

I've noticed some of my tests are failing due to:

cloak_ecto/lib/cloak_ecto/types/sha_256.ex

Lines 78 to 86 in 05ea920

	@doc false
	@impl Ecto.Type
	def equal?(value1, value2) do
	hash_string(value1) == hash_string(value2)
	end
	defp hash_string(string) do
	if String.valid?(string), do: hash(string), else: string
	end

I have a schema similar to

  schema "tokens" do
    field(:token_hash, Cloak.Ecto.SHA256)
   ...
   end

I can create a new changeset from this schema, however the initial value will be nil for :token_hash.
If I then call the following: put_change(changeset, :token_hash, "my_token") the String.valid?(nil) in hash_string(string) will fail.

I'm not entirely sure if a check needs to be added in the hash_string function to make sure first the value are not nil, or if I need to make sure first the changeset contains a valid string for the hash first.

One way I can resolve this is to add a default empty string value to field(:token_hash, Cloak.Ecto.SHA256, default: "").
Let me know if this is a non issue and feel free to close this.
Thanks

[vitalis]

This is my solution for the issue:

    changeset
    |> get_field(:contact_email)
    |> case do
      nil ->
        changeset

      email ->
        put_change(changeset, :contact_email_hash, email)
    end

[SimonLab]

Thanks @vitalis this resolves the issue when the contact_email is nil but I don't think it does when the current contact_email_hash is nil.

If I'm not mistaken put_change(changeset, :contact_email_hash, email) will call the equal?/2 which then calls the private function has_string/1 which contains the call to String.valid?, breaking with the current nil value

[vitalis]

@SimonLab
It did resolve the issue for me, together with your suggestion: "Cloak.Ecto.SHA256, default: "")"
I hope the PR will solve the issue without "workarounds" :)