At this point let's review the process followed in creating the program. First an algorithm was conceived and refined, moving from the original problem to the subtasks into which it was broken down: the four tasks for operating on the data abstractions, plus the task of determining stable. The refinement for this last task was specified, treating the pairings and preferences as data abstractions. Next the functions embodying the operations on the data abstractions were specified. In doing so, the trade-off between storage (adding the priority arrays) and execution time (requiring a traversal of rows of mpref or wpref if there were no priority arrays) came into play. This led to the "right" information to be stored and the "right" choice of data structure in which to store it?/FONT>the priority arrays.
Finally, the solution was packaged into a main program that (1) created and input data to the data structures chosen; (2) called stabilitycheck to check the entire pairing. The functions main and stabilitycheck were written, treating the pairs and preferences as data abstractions. As a result, should their implementations need to be changed, the changes are localized to the data structure definitions and the functions operating on them.