If the free function is really useful, I think I'd rather have a separate free function (that might return a boolean rather than raise) and a test-specific assert* function that uses it, rather than overloading too much into the parameters of the latter. We should keep all the assert* functions as similar as possible.

Some of this was discussed in RFC-28. Two functions seems reasonable now that we are clearly using the function more widely.

Based on the comments I have added a free predicate function wcsNearlyEqualOverBBox that uses short circuit evalutation. That and the existing related assert are now implemented in terms of a private function _compareWcsOverBBox.

I also updated the documentation for testCase to explicitly list its requirements, making it safe to mock if desired. In one case (the bbox test) a trivial modification was required to make it true that only testCase.fail was called.

I did not attempt to add predicates for any other asserts. I'd rather leave that for another ticket.

This work is in afw on tickets/DM-3347.

I did not attempt to allow testCase=None, as that was rightly pointed out as less than ideal. For the record, it would probably be easy to change our decorator such that testCase=None was turned into an instance of TestCase. That would provide consistency, but I think it would make the code for the asserts too difficult to understand ("how does None support fail(msgStr)?"). Best to just drop the whole idea.

Sorry I'm a bit late with this, but neither GitHub nor JIRA seems to be able to see that branch. Did you push?

K-T you are right: I had not pushed. My apologies. I just did so now.

Two small comments on the PR, one on the code and one on the documentation.

Thank you for the very helpful corrections. I have corrected both issues and will merge as soon as it passes Jenkins.

I found and fixed a bug in the short circuit code. The new code should also be somewhat more efficient, since it only records measured error if the error is too large. Merged to master.