Do a better job merging argument schema (specifically definitions).

[ubik2]

Ignore the messy types for now, since I'll do a better job later.
This just gets us through the integration tests.

---

Summary by cubic

Improved how we merge recipe argument schemas in Runner by hoisting $defs/definitions to the process schema and defaulting missing argument schemas to true. This fixes $ref resolution and unblocks integration tests.

• Bug Fixes
  • Hoist $defs and definitions from recipe.argumentSchema to the top level of the process cell schema so refs resolve correctly.
  • Set properties.argument to true when argumentSchema is missing, instead of {}.

[mathpirate]

cool! LGTM, thank you so much for noticing this and jumping on it so quickly

i don't really understand this system well yet, so just to be completely sure: this operates at the level of a single schema, meaning that it won't cause name collisions if multiple schemas in a generated program use the same definition names, correct?

[ubik2]

cool! LGTM, thank you so much for noticing this and jumping on it so quickly

i don't really understand this system well yet, so just to be completely sure: this operates at the level of a single schema, meaning that it won't cause name collisions if multiple schemas in a generated program use the same definition names, correct?

Right, in this case, the top level schema has none of its own definitions, so no chance of conflicts.

Edit: We do have a similar situation with runner.ts generateHandlerSchema, and in that case, we could have a collision between the definitions in the event and the state. To properly handle those, we'd want to rename them and rewrite $refs, but for now, we just assume any collisions are actually the same type (practically, this is probably the case).

[mathpirate]

cool! LGTM, thank you so much for noticing this and jumping on it so quickly
i don't really understand this system well yet, so just to be completely sure: this operates at the level of a single schema, meaning that it won't cause name collisions if multiple schemas in a generated program use the same definition names, correct?

Right, in this case, the top level schema has none of its own definitions, so no chance of conflicts.

awesome, sounds good, excited to try rebasing on this on my branch

[cubic-dev-ai]

No issues found across 1 file