how do I write test methods

[whoburg]

What's the best practice for writing test methods for models in the model library? I'd like the general philosophy to be:

• models don't have default values placed directly in the variables or constraints
• test methods substitute some default values and choose a reasonable objective, and then solve

I don't want to use .substitutions.update in test(self, ...) because then calling test mutates the substitutions of the model, which is unexpected.

Should models also expose some sort of defaults() method that gives default substitutions?

Can we get an example of a simple model with a test method?

[bqpd]

How about test as a class method? On Mar 31, 2016 10:20, "Warren Hoburg" notifications@github.com wrote:

What's the best practice for writing test methods for models in the model library? I'd like the general philosophy to be:

• models don't have default values placed directly in the variables or constraints
• test methods substitute some default values and choose a reasonable objective, and then solve

I don't want to use .substitutions.update in test(self, ...) because then calling test mutates the substitutions of the model, which is unexpected.

Should models also expose some sort of defaults() method that gives default substitutions?

Can we get an example of a simple model with a test method?

— You are receiving this because you are subscribed to this thread. Reply to this email directly or view it on GitHub https://github.com/hoburg/gpkit-models/issues/59

[bqpd]

example:

class Atmosphere(ConstraintSet):
    # ...
    @classmethod
    def test(cls):
        import numpy.testing as npt
        from gpkit import Model
        cs = cls()
        substitutions = {}  # ...
        m = Model(cs["\\rho"], cs, substitutions)
        sol = m.solve()
        # test the solution

Atmosphere.test()

[whoburg]

sounds good.

[pgkirsch]

@bqpd: @whoburg and I discussed using this same structure for making different variants on models e.g.

@classmethod
def standalone_737(cls):
     objective = ...
     substitutions = {...}
    m  = Model(....)
    return m

@classmethod
def aircraft_737(cls): # for full aircraft model
    substitutions = {....} # shorter substitutions dict
    m  = Model(....)
    return m

How do you feel about this? I'm currently in the middle of implementing it for my GPjet models as a test before porting over to gpkit models and (hopefully) ending GPjet for good.

[pgkirsch]

Where the original

class VerticalTail(Model):
...

becomes

class VerticalTail(ConstraintSet):
...

[bqpd]

Ooh, I like that!

[pgkirsch]

Sweet. One thing I can't figure out: does this mean ConstraintSet will need a __getitem__ method to be able to do something like objective = cs['W_{vt}']?

[bqpd]

Yeah, I'm going to rename constraintset to something else so it can have user-facing things like that.

[bqpd]

But for now you could just use strings.

[pgkirsch]

what do you mean use strings?

[pgkirsch]

One thing I have figure out is how to do it with CostedConstraintSet, which also works pretty well I think

[bqpd]

Oh, derp. I was thinking for substitutions. For cost there's no nice way to do it right now.

[bqpd]

Yeah, costedconstraintset has getitem