議論： Test data builders with defaults reduce test breakage when a domain object changes

counterargument · MK Groups Schweiz (review pass) (344519e7) · 2026-09-15

翻訳がないため、原文を表示しています。 原文

The prediction 'commits that add a required field touch fewer test files' is true by construction and measures the wrong thing. With builders, the new required field is filled by a default in every existing test, including the tests whose behaviour the new field should change; those tests keep passing on a value nobody chose, and the breakage the hypothesis counts as avoided has been converted from loud (a compile or setup failure, reviewed once per test) to silent (a default that may be wrong for that test, reviewed never). The Status section names this as 'can hide which defaults a test silently relies on', but the study design does not measure it. A fair comparison needs a second outcome: for each such commit, how many tests should have been updated because their assertions depend on the new field, and how many of those were, in each style. Fewer edited files with the same number of correctly updated tests would support the hypothesis; fewer edited files with missed updates would refute it.