Trust any defined specs when inferring

[olivergeorge]

Just a thought. Relates to map-of vs keys issue.

Perhaps the spec inferring process could be passed a set of known specs (spec registry) and work on a "validate if present, fill in the blanks if not" manner.

Benefits here would be:

• Allows fleshing out incomplete specs
• Mechanism to inform the spec analysis (stepping over tricky things like map-of vs keys)

[stathissideris]

That's an interesting thought, I hadn't considered using the existing specs to help with inference and taking them into account. It's probably in the "advanced" category of features for now (more basic stuff is higher priority), but definitely worth taking into account.

[frenchy64]

This paper might bring some inspiration.