Effect reasoning

Purpose

Decide where to put tests by reasoning about the effects of a change rather than guessing. Use effect sketches — a bubble per affected variable or return value with arrows to what it can change — to find where to test (P004), then place characterization tests where they will actually sense the change, ideally at pinch points, narrowings where tests on one or two methods detect changes across many (P043).

When to use

• You must choose the smallest set of test points that will sense a change rippling across several methods or classes.

Do not apply when the change is local and its single test point is obvious — just test there.

Procedure

1. Mark the change point(s). Identify what the change will modify.
2. Sketch the effects (P004). From each change point, draw a bubble per affected variable or return value with arrows to what it can change, building an effect sketch of everything the change can affect.
3. Trace the three ways effects propagate (P061): used return values, mutated objects passed as parameters, and mutated static or global data. Reason forward about the effect chain, which propagates to callers up to a system boundary while unrelated code keeps its behaviour (P041).
4. Find interception points (P062). Trace effects outward from the change points and prefer interception points close to the change — fewer steps make a stronger safety assurance.
5. Look for a pinch point (P043). Find a narrowing where one or two methods sit on the path of changes that fan out across many; tests there detect a wide set of changes with few tests. A pinch point is also a natural encapsulation boundary and can reveal hidden classes worth extracting later (P091).
6. Place characterization tests at the chosen interception/pinch points (see characterization-testing).

Inputs

• The change point(s) and enough of the call/data structure to trace effects.

Output

An effect sketch for the change, the identified interception/pinch point(s), and the recommended test points — the minimal set that senses the change — plus any pinch point flagged as a natural encapsulation boundary.

References

• characterization-testing — how to write the tests placed at the chosen points.
• legacy-code-change-algorithm — this supports step 2 (find test points).

Provenance

Derived from principles P004 (effect sketches), P041 (forward effect-chain reasoning), P061 (three propagation mechanisms), P043 (pinch points), P062 (interception points close to the change), and P091 (feature sketches reveal hidden classes / encapsulation boundaries). Source is distillation-only; paraphrased, not quoted.