Step 1 of 6 · about 40 min

Equivalence partitioning and boundary values

Bugs are easier to find on purpose than by luck. Two techniques do most of the heavy lifting, and neither requires maths beyond counting.

Equivalence partitioning is the idea that inputs fall into groups that all behave the same way, so testing one value from each group tells you what testing all of them would, at a fraction of the effort. Boundary value analysis adds the observation that bugs cluster at the edges of those groups, so you test the values right at and just past each edge.

Here is a worked example. Take a requirement: an age field accepts whole numbers from 18 to 65 inclusive, and rejects everything else.

Partitions (equivalence classes):
- below range:   anything less than 18   -> rejected
- in range:      18 to 65                -> accepted
- above range:   anything more than 65   -> rejected

One representative from each partition:
10   -> rejected
40   -> accepted
80   -> rejected

Boundary values (the edges, and one step past each):
17   -> rejected   (just below the lower edge)
18   -> accepted   (the lower edge)
65   -> accepted   (the upper edge)
66   -> rejected   (just past the upper edge)

Three partition tests confirm the broad behaviour. Four boundary tests catch the off-by-one errors that live at the edges, the classic "should it be less-than or less-than-or-equal" bug. Seven precise tests replace the hundred vague ones you might have written.