I would appreciate a ELI5. I skimmed through the paper and from what I saw they took the entire domain of time, mapped it into a couple categories (weekday/weekend, working hours/non-working hours) and then created a test matrix based off that. What I’m not clear on is if this special matrix was smaller than what you’d expect by just doing a combinatorial test of all categories. They mention generating a test with something “pairwise”, but I think I missed a piece about how this approach is better than “write a list of possible states”.

The first array is special. If you just generated every combination of the 5 different equivalence classes, you’d have 48 combinations to test.

They don’t explain how they arrived at that array, but it looks like they used some knowledge of the logic inside the system under test. They may have drawn a Karnaugh map to find equivalence classes that don’t have any mutual effect.

It’s not clear to me why they generate the results into a second array, other than for compact representation. Verification only consists of looking across a row to see if the values disagree. The paper didn’t talk about any column-wise verification in that second array. So in implementation terms, you could generate the output as a list of lists.

I was unable to access this at the provided url; the Google DNS did not have an entry for the domain name. Googling found me an alternate link.

I would appreciate a ELI5. I skimmed through the paper and from what I saw they took the entire domain of time, mapped it into a couple categories (weekday/weekend, working hours/non-working hours) and then created a test matrix based off that. What I’m not clear on is if this special matrix was smaller than what you’d expect by just doing a combinatorial test of all categories. They mention generating a test with something “pairwise”, but I think I missed a piece about how this approach is better than “write a list of possible states”.

The first array is special. If you just generated every combination of the 5 different equivalence classes, you’d have 48 combinations to test.

They don’t explain how they arrived at that array, but it looks like they used some knowledge of the logic inside the system under test. They may have drawn a Karnaugh map to find equivalence classes that don’t have any mutual effect.

It’s not clear to me why they generate the results into a second array, other than for compact representation. Verification only consists of looking across a row to see if the values disagree. The paper didn’t talk about any column-wise verification in that second array. So in implementation terms, you could generate the output as a list of lists.