spec/cycles/Structural_cycles: optional field bug?

[cueckoo]

Originally opened by @rudolph9 in cuelang/cue#145

spec/cycles/Structural_cycles

Arguably the following falls under the category of

It is allowed for a value to define an infinite set of possibilities without evaluating to an infinite structure itself.

Foo: {
  foo?: Foo
  bar: string
}

foo: Foo & {
  foo:{ 
    bar: "barNested"
  }
  bar: "barParent"
}

but in the current version of cue it doesn't return (probably eventually a stack overflow error but killed it after like 5 minutes).

$ cue eval optional_structural_cycle.cue
# waiting
# waiting
# waiting
^C

I could also see this fall under the category of an infinite cycle so could use some clarity around if it's a bug with the evaluation or if it just need some clarity in the doc and maybe make sure it's detected in the cycle check that will eventually be rolled cuelang/cue#29

[cueckoo]

Original reply by @rudolph9 in cuelang/cue#145 (comment)

This feature is relevant to a little testing package I'm working on refining, if supported it would allow me to infinitely nest describe scopes https://gist.github.com/rudolph9/eccc978e71fd9c32b0ae182ce6854c16

[cueckoo]

Original reply by @jlongtine in cuelang/cue#145 (comment)

You might try with the latest master. Marcel made some updates that I think might be relevant.

[cueckoo]

Original reply by @mpvl in cuelang/cue#145 (comment)

I tested this does not work with the new version. This is indeed another case. The current implementation explicitly tests for disjunctions that allow such infinite structure. But optional fields implicitly such a disjunction. And as @rudolph9 correctly notes, the spec allows for such cases.

Note that none of this is related to evaluation, but merely to printing. The engine is happy to compute on infinite structures. It is just the representation of these infinite structures that is causing the trouble.