Data validation two ways

[caufieldjh]

Describe the desired behavior

Would like to verify transforms in at least the following two manners:

• Check for changes above a certain threshold, e.g., too many changes to new version
• Do unit tests to verify certain axiom-dependent patterns are being represented as expected (but will require pre-processing to relax first)

[caufieldjh]

This may be easier if coupled with #30

[caufieldjh]

Doing a full diff for every update can be slow, but we can run sanity checks:

• Is the new version empty, particularly after ROBOT processing? (already do this)
• Does the new version involve a different set of imports from the last?
• Is the new version >10% different from the previous in the total number of lines?

[caufieldjh]

The first point is covered in #125