Answer evaluation in UnlockProtocol

[jathak]

This implements @papajohn's proposal in #59.

[albert12132]

The way unlocking works is the following:

1. Verify if the student's guess is correct (this is the part you modified).
2. If it is deemed correct, we take the student input (literally just the string they input) to be the answer for the test case.
3. When GradingProtocol runs, it will check if the repr of the prompt expression (>>> x + 4) is equal to the unlocked answer (again, literally just the string they input).

Suppose we have a prompt:

>>> x
?

If a student enters 2 + 3 and this is deemed correct, the unlocked test case will look like

>>> x
2 + 3

Then GradingProtocol will check if repr(x) (which is the string '5') is equal to the string '2 + 3', which it isn't.

Getting around this requires either rewriting interact or rewriting GradingProtocol; changing _verify won't be enough.

[jathak]

Okay. I get it now. I rewrote this to leave _verify as-is, but change interact to call _verify again with evaluated input if exact string matching fails. If evaluated input matches, it then uses that as the answer for the test case.

[albert12132]

The code looks good to me. Can you write one or two tests for this in tests/protocols/unlock_test.py?

[jathak]

Done.

[albert12132]

LGTM.