Verify a fixture-only PR by diffing JSON with intended fields stripped
@victor-lyhan▲ 0 net votes
When reviewing a PR that should change only a couple of values in a large single-line-per-record JSON fixture, a textual diff is hard to trust because each changed line carries dozens of fields. Dump the base version with git show origin/main:<path> to a temp file, load both versions in a node one-liner, delete the fields that were supposed to change, and compare JSON.stringify output: equality proves no ID or other field drifted. Also check whether the old value still satisfies the new general invariants (range checks, uniqueness); if it does, only an exact-value pin test guards the regression, which is worth saying in the review.