Module name is incorrect when the name as acronym word

[wingyplus]

My module has acronym word like PymServerClient.PagerOfGPDailySummary. The library at revision 377c923c5597e76445124cb6af749cdf3456b4b8 working correctly. But when bumping to the lastest main (11ed134aa3b8f6ebc469496ba79fe6c2d3585dd9 at this point), the module name renders to PymServerClient.PagerOfGpdailySummary.

I assume it is probably because of the word GP that produces incorrect results.

[aj-foster]

Hello there! I did make some changes to the way module names are handled in https://github.com/aj-foster/open-api-generator/commit/8052f69108694baac656185acc0003a32f32f8b2 (the commit after the one that worked for you), however I haven't figured out what went wrong yet. Could you share an example OpenAPI spec that generated the issue?

Thanks!

[xadhoom]

Same here, before 1.0.0-rc* modules names was reflecting openapi spec, now they're lowercased+camelized.

A minimal schema that exposes this issue can be:

components:
  schemas:
    TOTPCreate:
      additionalProperties: false
      description: |
        Object for a TOTP token config, with confirmation OTP.
      properties:
        otp:
          description: The OTP as obtained from the authenticator app/device
          type: string
        uri:
          description: The OTP Auth URI obtained from `TOTPRequest`
          type: string
        verification_code:
          description: The opaque signature obtained from `TOTPRequest`
          type: string
      required:
        - uri
        - otp
        - verification_code
      title: TOTPCreate
      type: object
info:
  title: Acme API
  version: '1'
openapi: 3.0.0
paths:
  /api/v1/me/totp:
    post:
      callbacks: {}
      description: |
        Saves a config obtained with `TOTPRequest`, along with a device generated
        OTP to confirm it's correctness.
      operationId: Controllers.ApiV1.TOTP.totp_save
      parameters: []
      requestBody:
        content:
          application/json:
            schema:
              $ref: '#/components/schemas/TOTPCreate'
        description: The TOTP config to save
        required: false
      responses:
        '201':
          description: TOTP request config saved
        '429':
          description: Too many requests, retry later
      summary: Saves a TOTP config
      tags: []
servers:
  - url: '{endpoint_url}'
    variables:
      endpoint_url:
        default: https://localhost:5443/

TOTPCreate is rendered as Totpcreate after library update.

[xadhoom]

I've tried playing a bit with the fun normalize_identifier/1 to allow sequences of uppercases like the following:

  def normalize_identifier(input) do
    input
    |> String.split(~r/(([^A-Za-z0-9]+)?([A-Z0-9]+(?![a-z0-9])))|(([^A-Za-z0-9]+)?([A-Z]+)?([a-z0-9]+)?)/, include_captures: true, trim: true)
    |> Enum.map_join("_", fn segment ->
      segment
      |> String.replace(~r/^[^A-Za-z]+/, "")
      |> String.replace(~r/[^A-Za-z0-9]+$/, "")
      |> maybe_downcase_segment()
    end)
  end

  defp maybe_downcase_segment(segment) do
    case String.upcase(segment)  do
      ^segment -> segment
      _ -> String.downcase(segment)
    end
  end

Which translates "PymServerClient.PagerOfGPDailySummary" in:

iex(1)> normalize_identifier("PymServerClient.PagerOfGPDailySummary")|>Macro.camelize()
"PymServerClientPagerOf_GPDailySummary"

and "TOTPRequest" in:

iex(1)> normalize_identifier("TOTPRequest")|>Macro.camelize()
"TOTPRequest"

But there're corner cases where uppers are at the end of the name:

iex(1)> normalize_identifier("RequestTOTP")|>Macro.camelize()
"Request_TOTP"

iex(2)> normalize_identifier("RequestTOTPCreate")|>Macro.camelize
"Request_TOTPCreate"

It's a bit funny, given the interaction with Macro.camelize()

So I don't have a real solution, except maybe to avoid the same normalization for all tokens? and use different approaches for different parts of the schema?

[aj-foster]

Thank you for the nice test case, and thank you for exploring some variations in the normalization. PR #43 has a possible solution, if you'd like to try it out. Let me know how it works for you.

A lot of information was being lost thanks to the String.downcase call in the normalization. Now, instead of always normalizing to lowercase snake_case, callers now have their choice of casing. The default is snake case for compatibility.

[wingyplus]

I forgot to reply to you. I'm sorry. 🙇

And thank you for a quick fix. Will try it in 1 or 2 days.

[flaviogrossi]

I had the same problem after the update, and i can confirm that the linked PR creates names compatible with the old version in my case

[aj-foster]

The fix from #43 is now available in version 0.1.0. ❤️