minitest validations - is there a cleaner way?

rubyonrails-talk
1

i am trying out minitest, but need some advice on a clean way to test validations.

I setup my testing to use minitest by following the following railscast plus i added miniskirt for Factories. http://railscasts.com/episodes/327-minitest-with-rails

everything works well, but there has to be a cleaner way of testing failed validations. would i be better off using something besides “must_raise”? if not, is there a way to make a helper function or something to clean this up so it is more readable?

2

I usually test my validations like this:

3

Carlos,

I love the way you are testing this. I have a question tho. When I tried implementing something similar in my tests, it doesn't seem to be running the validations I set in the def validate function of my model. Can you think of any reason why this might be happening or how I should alternatively test to make sure it includes the validations in validate?

Also, do you test associations with minitest? An example of that would be incredibly helpful as well.

Thanks for any help!

4

Oh fun, just found out def validate doesn't quite work like that anymore in rails3, so never mind on the first question, I need to change my def validate method.

An example of an association test would still be crazy helpful tho. Here's what I used to do with spec and shoulda:

it { should have_many(:products).through(:buying_guides_products) }

Any ideas?

5

I’m glad you found my snippet helpful!

Testing associations is trickier, and I must admit that I don’t test associations as thoroughly as I probably should. I usually just test that the record responds to the association (e.g.: record.must_respond_to :association).

I took a look at the “shoulda-matchers” gem’s source code. Those association matchers test for a whole bunch of stuff (foreign keys, join tables, correct class names, etc.), some of which are probably re-testing ActiveRecord functionality which we should be able to assume works correctly. A good test for associations should fall somewhere in between these two extremes.

I’d love to hear if anyone else has suggestions.