Concepts, in plain words

You do not need to know property testing or domain modelling to use Nullean.Entities. You need about ten words. Here they are.

The types that describe your business: an Order, a Ledger, a Docset. In C#, records. A domain model says what things are.

A statement about the domain that must be true: "an order total is never negative", "you can only ship a confirmed order". Nullean.Entities calls a rule a constraint, and writes it as a named value next to the type. A type cannot say these things, so they usually live in if statements and validators.

A method that takes a value and gives you the next one: order.Confirm(), ledger.MoveToSavings(10). The library finds the operations of your type by itself (every public method that returns the type).

An example test says: for this input, expect this output. It covers the case its author thought of.

A property test says: for every input of this kind, this is true. A tool makes up many inputs, runs your check on each, and reports the first one that breaks it. Nullean.Entities writes the property tests for you, from the rules.

The inputs a property test makes up: many orders, with different lines, in different states. Nullean.Entities builds them from the shape of your type, and mixes in nasty values on purpose (empty strings, negative numbers, duplicates) because that is where bugs are.

When a generated input breaks a rule, it is usually large and noisy. The tool then tries smaller and smaller versions of it and keeps the smallest one that still breaks the rule. A failure that says "an order with no lines" is far easier to understand than one with five lines and a made-up address. The library calls this "smallest case found".

Generated data is random, but the randomness is repeatable. A failure prints a short code, the seed. Pass it back with .Seed("…") and the test makes exactly the same inputs and fails the same way, so you can fix the bug and re-run the same case.

To prove a rule can fail, you need one value that breaks it. That value is its counter-example: "an order with a negative price" for "the total is never negative". If a rule has no counter-example, you cannot tell a rule that holds from a rule that is never checked. The library finds most counter-examples itself; see counter-examples.

A kind of question you can ask of a rule. Nullean.Entities tests nine of them: the nine axes. Each is a different kind of bug.

A test is vacuous when it passes without proving anything, because the data never reached the rule it was meant to check. A smoke alarm test where nobody lit a match. Nullean.Entities fails a vacuous test instead of letting it look green.

For rules about time ("ships within a minute"), the test does not wait. It gives your service a clock it controls and moves that clock forward by chosen amounts, so a test of a 60-second deadline takes milliseconds.