Usually the data is a combination of externally received information and internal calculations/annotations. I just use copious asserts, and one line: assert

Usually the data is a combination of externally received information and internal calculations/annotations. I just use copious asserts, and one line: assert bubble_sort(data) == sort(data) Since this always runs during development it will automatically pick up internal and external changes. What you laid out is explicitly testing everything against everything else and is 100% comprehensive. Bubble sort is less than 100% comprehensive but still gives good coverage, while optimised sorts just mean you get lucky in the face of buggy comparisons.