Create temporary end-to-end suite of test cases about aospy HOT 11 CLOSED

How do you feel about this in light of #61?

from aospy.

With regard to potentially closing this issue? Should we wait until we have more coverage, or do you think #61 is good enough for now?

On the bright side, with #61 it shouldn't be too hard to check a lot of these things off the list. I can do my part in adding some more tests on AM2 and idealized model output, but I don't have any AM3, HiRAM, or CMIP5 runs handy. In principle, though, I could just lift some from your object library.

from aospy.

Ya, just wanted to see if this still has merits, which it does. I agree, it would be useful (and low-hanging fruit) to copy the other model objects here, since often in the past a change would break things for one model and not the other.

(Doing these does not obviate the need for more unit testing everywhere, of course.)

from aospy.

👍 I'll try and get to this soon.

from aospy.

from aospy.

I think the best way of extending #64 to other models will be to abstractify all of these test methods you've created for AM2 by moving them into a separate class that accepts a generic model object, and then creating classes for each model that inherits from that class and provides the needed model object; the tests will then automatically run for each model. See infinite-diff's tests for something of an example: the TestBwdDiff class inherits from TestFwdDiff and just supplies a different self.method, which the tests then use. (And the latter inherits from a generic class that supplies misc. arrays and other attributes to use as the test data.)

(FYI I'm in the midst of refactoring that package, so that link might not work after the next few days, but the concept is relatively straightforward.)

from aospy.

That's a really good idea. I'll do that in my next pull request.

from aospy.

👍 I am starting to get a better feel for OOP and polymorphism...it is really powerful.

from aospy.

So am I correct that after #66 we are all set on this except for CMIP5 (which I need to figure out)?

from aospy.

Yes, that's correct, and for what it's worth the CMIP5 tests are implemented (albeit they fail), but perhaps that's a separate issue (cf. spencerahill/aospy-obj-lib#10).

from aospy.

Right. In that case, let's call this done! I'll re-open if, once I get the CMIP5 ones sorted out, these tests don't work.

Thanks for all your work on this!

from aospy.