Concepts

A short glossary of the vocabulary used across this site, in the order you tend to meet the terms.

Property   A law about your code, expressed as a test method under #[Property]: "for any input matching this shape, this assertion holds." See Getting started.

Generator / arbitrary   Something that knows how to produce random values of a given shape and how to minimize them. Gen::intBetween(0, 100) is a generator; the object it returns implements ArbitraryInterface. See Generators.

Run   One execution of the property body with one set of generated arguments. #[Property(runs: 200)] asks for 200 successful runs — see "discard" below for what doesn't count.

Discard   A run abandoned via Assume::that(false) because a precondition didn't hold. Discards don't consume runs; they're retried until maxDiscards is exceeded.

Seed   The integer that deterministically drives a property's random phase. Two runs with the same seed generate the identical sequence of values — this is what makes a reported failure reproducible: #[Property(seed: 7382910)].

Falsify   What happens when a generated (or explicit) input makes the property's assertion fail. The runner catches the failure and hands the input to the shrinker.

Shrink / shrinking   Minimizing a falsifying input by descending through its value's tree of smaller candidates, looking for the simplest one that still fails. See Shrinking for how the tree itself is built and why it's guaranteed to terminate.

Counterexample   The shrunk (minimal) input that falsified the property, plus the original input, the failure, and run/shrink statistics — carried by CounterExample and reported inside PropertyViolationException.

Classify / coverage   Recording which "kind" of input each run happened to be (even, empty, boundary, …), to check that your generators actually reach the interesting cases instead of passing vacuously. See Distribution.

Regression corpus   Optional on-disk memory (PROPERTY_DB) of past failures, replayed before the random phase on every run, so a fixed bug can't silently come back unnoticed. See Regression corpus.

Command / state machine   The vocabulary for testing a sequence of operations rather than a single input — a generated list of Commands run against both a real system and a simplified in-memory model. See State machine: concepts.