Fields filters should be underscored

[zlatio]

Hello,

First of all thank you for this library :) I think it's great. However I have one issue. When passing fields parameter like fields[users]=first-name,last-name I get argument_error

** (ArgumentError) argument error
     stacktrace:
       :erlang.binary_to_existing_atom("first-name", :utf8)
       (elixir) lib/string.ex:1792: String.to_existing_atom/1
       (elixir) lib/enum.ex:1184: Enum."-map/2-lists^map/1-0-"/2
       lib/ja_serializer/builder/attribute.ex:29: JaSerializer.Builder.Attribute.do_filter/2

Attributes filters needs to be underscored because otherwise they can't be converted to existing atoms. So in the attribute builder on line 31 we need something like this:

  defp safe_atom_list(field_str) do
    field_str |> String.replace("-", "_") |> String.split(",") |> Enum.map(&String.to_existing_atom/1)
  end

What do you think?

[yordis]

@alanpeabody there is no default naming as far as I know http://jsonapi.org/format/#document-member-names so I would like to see by default use underscore instead of dashes.

Specially when most of the programming languages are easier to read from underscore and probably most of the people that use this come from a background where underscores are normally the default name. I barely see dashes been use.

Could we default the attributes name to underscore please 🙏 . It's normal for me to go and add

config :ja_serializer,
    key_format: :underscored

in my configuration in every single project that use this

[alanpeabody]

@yordis the default naming is what jsonapi.org recommends: http://jsonapi.org/recommendations/#naming

[yordis]

😭 I guess I will continue using the code snippet

[alanpeabody]

sorry. :-/

[zlatio]

@alanpeabody @yordis I am not sure you got the exact issue I have with this. As per JSONAPI spec

Member names SHOULD contain only the characters “a-z” (U+0061 to U+007A), “0-9” (U+0030 to U+0039), and the hyphen minus (U+002D HYPHEN-MINUS, “-“) as separator between multiple words.

Thing is when I pass multiple words separated by - like fields[users]=first-name,last-name I get

** (ArgumentError) argument error
     stacktrace:
       :erlang.binary_to_existing_atom("first-name", :utf8)
       (elixir) lib/string.ex:1792: String.to_existing_atom/1
       (elixir) lib/enum.ex:1184: Enum."-map/2-lists^map/1-0-"/2
       lib/ja_serializer/builder/attribute.ex:29: JaSerializer.Builder.Attribute.do_filter/2

because actually there is no first-name but first_name and it can't be converted to existing atom. Right now I am using a plug to convert those but I see it easy to patch like:

  defp safe_atom_list(field_str) do
    field_str |> String.replace("-", "_") |> String.split(",") |> Enum.map(&String.to_existing_atom/1)
  end

I am just not sure if this is the best solution :)

[alanpeabody]

@zlatio yeah, I understand your issue. To be frank I don't know the best solution. Currently the formatting of the keys all happens here:

https://github.com/vt-elixir/ja_serializer/blob/master/lib/ja_serializer/param_parser.ex

However it is impractical/incorrect to format values there. The piece your solution is missing is a way to customize it... If we hooked into Application.get_env(:ja_serializer, :key_format) and did it conditionally I would be happy to accept a PR.

[zlatio]

True, there is no way to customize it. I will try to come up with a PR for this :)

[alanpeabody]

Thanks a bunch!

[habeel]

I think the handling is inconsistent through out if by default inserted_at is being serialized to "inserted-at" then the generated templates should also handle the input the same way.

I believe ParamParser should be called inside Params.to_attributes because Params.to_attributes is taking care of parsing relationships which is a json-api spec so then it should also take care of translating the keys.

I will open up a new issue for this.