Skip to content

Instantly share code, notes, and snippets.

@arwagner
Created October 29, 2011 01:13
Show Gist options
  • Star 1 You must be signed in to star a gist
  • Fork 1 You must be signed in to fork a gist
  • Save arwagner/1323952 to your computer and use it in GitHub Desktop.
Save arwagner/1323952 to your computer and use it in GitHub Desktop.
Combinatorial explosion and testing
I found http://groups.google.com/group/growing-object-oriented-software/browse_thread/thread/47695af2c6b5adda fascinating, so I decided to make it a little more concrete. It's also in ruby using rspec, sorry about that. Here's the code under test:
class EligibleForDiscountPolicy
def initialize transaction, user
@transaction = transaction
@user = user
end
def decide
return true if @user.is_gold_member?
return false if @transaction.coupon_was_used?
return false if @transaction.contains_new_release?
return true
end
end
The idea here is that you're writing software for a video rental store. This particular class is responsible for determining whether a particular user is eligible for a discount on a particular transaction. If they have a gold membership, they always are. If not, two other conditions apply. Here are two approaches for testing this:
describe EligibleForDiscountPolicy do
[
[true,true,true, true],
[true,true,false, true],
[true,false,true, true],
[true,false,false, true],
[false,true,true, false],
[false,true,false, false],
[false,false,true, false],
[false,false,false, true]
].each do |gold, coupon, new_release, result|
scenario = "When the user #{gold ? 'IS' : 'IS NOT'} a gold member"
scenario += ", a coupon #{coupon ? 'IS' : 'IS NOT'} used"
scenario += ", and the transaction #{new_release ? 'DOES' : 'DOES NOT'} contain a new release"
scenario += ", the user #{result ? 'IS' : 'IS NOT'} eligible for a discount"
specify scenario do
user = stub :is_gold_member? => gold
transaction = stub :coupon_was_used? => coupon, :contains_new_release? => new_release
EligibleForDiscountPolicy.new(transaction,user).decide.should == result
end
end
end
This generates a test (spec) for each of the 8 scenarios in the truth table you could build from the above code. The output looks something like this:
/Users/andrewwagner/ruby> rspec -f nested combinatorial2_spec.rb
EligibleForDiscountPolicy
When the user IS a gold member, a coupon IS used, and the transaction DOES contain a new release, the user IS eligible for a discount
When the user IS a gold member, a coupon IS used, and the transaction DOES NOT contain a new release, the user IS eligible for a discount
When the user IS a gold member, a coupon IS NOT used, and the transaction DOES contain a new release, the user IS eligible for a discount
When the user IS a gold member, a coupon IS NOT used, and the transaction DOES NOT contain a new release, the user IS eligible for a discount
When the user IS NOT a gold member, a coupon IS used, and the transaction DOES contain a new release, the user IS NOT eligible for a discount
When the user IS NOT a gold member, a coupon IS used, and the transaction DOES NOT contain a new release, the user IS NOT eligible for a discount
When the user IS NOT a gold member, a coupon IS NOT used, and the transaction DOES contain a new release, the user IS NOT eligible for a discount
When the user IS NOT a gold member, a coupon IS NOT used, and the transaction DOES NOT contain a new release, the user IS eligible for a discount
Finished in 0.20627 seconds
8 examples, 0 failures
This is a very data-driven approach, which has its advantages and disadvantages. Here is another approach:
require './combinatorial'
describe EligibleForDiscountPolicy do
let (:user) { Object.new }
let (:transaction) { Object.new }
context "when the user is a gold member" do
before { user.stub :is_gold_member? => true }
it "doesn't matter whether a coupon was used" do
transaction.should_not_receive :coupon_was_used?
EligibleForDiscountPolicy.new(transaction,user).decide()
end
it "doesn't matter whether there was a new release" do
transaction.should_not_receive :contains_new_release?
EligibleForDiscountPolicy.new(transaction,user).decide()
end
it "the user is always eligible for a discount" do
EligibleForDiscountPolicy.new(transaction,user).decide().should == true
end
end
context "when the user is not a gold member" do
before { user.stub :is_gold_member? => false }
context "when a coupon was used" do
before { transaction.stub :coupon_was_used? => true }
it "doesn't matter whether there was a new release" do
transaction.should_not_receive :contains_new_release?
EligibleForDiscountPolicy.new(transaction,user).decide()
end
it "the user is not eligible for a discount" do
EligibleForDiscountPolicy.new(transaction,user).decide().should == false
end
end
context "when a coupon was not used" do
before { transaction.stub :coupon_was_used? => false }
context "when renting a new release" do
before { transaction.stub :contains_new_release? => true }
it "the user is not eligible for a discount" do
EligibleForDiscountPolicy.new(transaction,user).decide().should == false
end
end
context "when not renting a new release" do
before { transaction.stub :contains_new_release? => true }
it "the user is eligible for a discount" do
EligibleForDiscountPolicy.new(transaction,user).decide().should == false
end
end
end
end
end
And the resulting output:
/Users/andrewwagner/ruby> rspec -f nested combinatorial_spec.rb
EligibleForDiscountPolicy
when the user is a gold member
doesn't matter whether a coupon was used
doesn't matter whether there was a new release
the user is always eligible for a discount
when the user is not a gold member
when a coupon was used
doesn't matter whether there was a new release
the user is not eligible for a discount
when a coupon was not used
when renting a new release
the user is not eligible for a discount
when not renting a new release
the user is eligible for a discount
Finished in 0.20714 seconds
7 examples, 0 failures
I do feel like the second approach is more expressive, somehow, both in the actual spec code and the output. In both cases, though, I feel like if there were more conditions to check, the number of tests would increase exponentially.
Thoughts?
@jbrains
Copy link

jbrains commented Oct 29, 2011

I like the code in the second one, but I still want to see a summary of the behavior as a table.

@dyokomizo
Copy link

I prefer second approach. Also we can collect the stubs, contexts and tests as we run the spec and create a summary table, outputting something like this:

| is_gold_member? | coupon_was_used? | contains_new_release? | decide |
| true | nil | nil | true |
| false | true | nil | false |
| false | false | true | false |
| false | false | false | true |

Also I think the last test is wrong in the second approach, the stub should be false and the should should be true :)

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment