Generate better schema types

[dani-garcia]

Type of change

- [ ] Bug fix
- [ ] New feature development
- [x] Tech debt (refactoring, code cleanup, dependency upgrades, etc)
- [ ] Build/deploy pipeline (DevOps)
- [ ] Other

Objective

The quicktype schema generation seems to have trouble with types in schema files being referenced from other types in separate schema files, which created duplicate classes, with a lot of strange names.

Instead of generating one individual file for each response, we create a big struct with all the options. This seems to solve the issue of references between schema files, and so no duplicate classes are created.

Note that the way that the file is loaded in quicktype is different. By default, quicktype generates JSON conversion helper functions only for the root type, which would be the big struct SchemaTypes. By importing the file with #/definitions/ appended, we only bring the sub schemas as root types, and ignore SchemaTypes, which we don't care about.

Examples of the schema changes, using C# as an example:

• ProjectsResponse previously contained an array of DatumElement, which were identical to ProjectResponse. Now it properly contains an array of ProjectResponse. This is also applicable to SecretsResponse which contained an array of DatumClass
• Previously it would generate multiple variants for Argon2ID, named: FluffyArgon2Id and PurpleArgon2Id, now there is only one Argon2Id. This is applicable to the two factor response types, which change in the same way

This is working correctly with the Java, Go, C++, PHP and Ruby bindings.

[bitwarden-bot]

Checkmarx One – Scan Summary & Details – 22f67793-e167-46df-8332-f6e62451b4fd

No New Or Fixed Issues Found

[Hinton]

Did you verify the TS types to?

[dani-garcia]

I did now and they work correctly, I also created a PR with some of the changes to get the TS bindings up to par with the rest of them #398