Client: Guarantee case-insensitive name uniqueness?

[kubukoz]

Hi!

I tried to regenerate the sources for the GitLab API again (for https://github.com/kubukoz/caliban-gitlab/), but noticed I'm getting compilation warnings:

[E]      Generated class io.pg.gitlab.graphql$IssueSort$UPDATED_ASC$ differs only in case from io.pg.gitlab.graphql$IssueSort$updated_asc$.
[E]        Such classes will overwrite one another on case-insensitive filesystems.
[E]      L961:     case object UPDATED_ASC extends IssueSort
[E]                ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

and some more.

Context:

object IssueSort {
    case object updated_desc extends IssueSort
    case object updated_asc extends IssueSort //here
    case object created_desc extends IssueSort
    case object created_asc extends IssueSort
    case object UPDATED_DESC extends IssueSort
    case object UPDATED_ASC extends IssueSort //here
    case object CREATED_DESC extends IssueSort
    case object CREATED_ASC extends IssueSort
    case object PRIORITY_ASC extends IssueSort
    case object PRIORITY_DESC extends IssueSort
    case object LABEL_PRIORITY_ASC extends IssueSort
    case object LABEL_PRIORITY_DESC extends IssueSort
    case object MILESTONE_DUE_ASC extends IssueSort
    case object MILESTONE_DUE_DESC extends IssueSort
    case object DUE_DATE_ASC extends IssueSort
    case object DUE_DATE_DESC extends IssueSort
    case object RELATIVE_POSITION_ASC extends IssueSort
    case object SEVERITY_ASC extends IssueSort
    case object SEVERITY_DESC extends IssueSort
    case object WEIGHT_ASC extends IssueSort
    case object WEIGHT_DESC extends IssueSort
    case object PUBLISHED_ASC extends IssueSort
    case object PUBLISHED_DESC extends IssueSort
    case object SLA_DUE_AT_ASC extends IssueSort
    case object SLA_DUE_AT_DESC extends IssueSort
// ...
}

The API in question: https://gitlab.com/api/graphql

As you can see in [https://docs.gitlab.com/ee/api/graphql/reference/index.html#issuesort](the documentation for this type), the lowercase values have been deprecated, so in theory shouldn't appear in the responses.

Still, it would be great to avoid generating classes / objects with the same name. Do you think this is something the Caliban generator should take care of?

Reproduction at https://github.com/kubukoz/caliban-gitlab/tree/da346a5440cb62b93f09d12c72e0718f3e34c626

[ghostdogpr]

Interesting, I haven't considered that before. This is valid as per the spec so the codegen should take care of that and add a suffix to avoid any clash.

It's a little tricky because the type name can be referenced somewhere else (e.g. an interface) but it surely can be done. Probably the best way is to add an entry to a set whenever we write a new type and deal with the conflicts there.

[javimartinez]

I would like to work on this one

[ghostdogpr]

@kubukoz in addition to fixing this, I added an issue for the option to exclude deprecated fields #645