[Typescript] Generate oneOf schemas as type unions

[joscha]

#19027 + latest master + revert of #19481, see #19027 (comment)

reverts #19481
closes #19027

[simon-abbott]

This change breaks allOf unions, they don't generate discriminator information at all.

EDIT: Opened an issue for this: #19868.

[joscha]

@ksvirkou-hubspot would you mind having a look at #19868 by any chance, please?

[joscha]

@simon-abbott can you produce a pull request with a fixture that shows the breakage or link me to some code showing what it should be vs. what it is right now? Maybe we can work together to try and solve this one?

[ksvirkou-hubspot]

@ksvirkou-hubspot would you mind having a look at #19868 by any chance, please?

This issue occurs because the schema described doesn't include a discriminator. Without a discriminator, it's challenging to find out which object the API returns. I'm uncertain about the best course of action in this scenario. I could either throw an exception or return a JSON object.
Currently, I'm reviewing other generators to understand their approach to this situation.
Do you have any idea?

[simon-abbott]

@simon-abbott can you produce a pull request with a fixture that shows the breakage or link me to some code showing what it should be vs. what it is right now? Maybe we can work together to try and solve this one?

Unfortunately I don't have time to make a full MRE at the moment, but If you generate the example schema in #19868 in 7.7.0 and again in 7.9.0 (I didn't test in 7.8, but I predict it will behave like 7.7) you can see the difference in the main Fruit model. The main cause is that previously the Fruit model would be generated as a franken-class containing the properties (and thus the getAttributeTypeMap) of the constituent parts. After this PR the main Fruit model is a super-thin wrapper.

This issue occurs because the schema described doesn't include a discriminator. [...] I'm uncertain about the best course of action in this scenario. [...]

See my reply in #19868. The spec says what to do.

[ksvirkou-hubspot]

@joscha let's discuss it in the issue.