Skip to content

fixturemonkey.md

Latest commit

 

History

History
364 lines (267 loc) · 12 KB

fixturemonkey.md

File metadata and controls

364 lines (267 loc) · 12 KB

Fixture Monkey

Dependencies

  • Core
<dependency>
	<groupId>com.navercorp.fixturemonkey</groupId>
	<artifactId>fixture-monkey-starter</artifactId>
	<version>0.5.0</version>
</dependency>
  • Bean Validations support
<dependency>
	<groupId>com.navercorp.fixturemonkey</groupId>
	<artifactId>fixture-monkey-javax-validation</artifactId>
	<version>0.5.0</version>
</dependency>

Pre-requisite (For Java)

Immutable Type

Method 1: Through Constructor

  • Requires each object to be mocked to be either:

    • A record class
    • Any constructors with @ConstructorProperties
      • If using lombok:
        • Can annotate class with @Data or @ConstructorProperties
        • Can also just add lombok.anyConstructor.addConstructorProperties=true in lombok.config

  • Create Fixture Monkey with objectIntrospector option

FixtureMonkey fixtureMonkey=FixtureMonkey.builder()
        .objectIntrospector(ConstructorPropertiesArbitraryIntrospector.INSTANCE)
        .build();

Method 2: Through Factory

  • Requires each object to be mocked to have a static factory method
  • Create Fixture Monkey with objectIntrospector option
FixtureMonkey fixtureMonkey=FixtureMonkey.builder()
        .objectIntrospector(FactoryMethodArbitraryIntrospector.INSTANCE)
        .build();

Method 3: Through Jackson

  • Requires each object to be mocked to be serializable/deserializable by Jackson
  • Requires additional dependency: com.navercorp.fixturemonkey:fixture-monkey-jackson
  • Create Fixture Monkey with plugin option
FixtureMonkey fixtureMonkey=FixtureMonkey.builder()
        .plugin(new JacksonPlugin(objectMapper))
        .build();

Mutable Type

Method 1: Through Constructor (Same as the immutable type above)

Method 2: Through FieldReflection

  • Requires each object to be mocked to have:
    • No args constructor
    • Getter / Setter
  • Create Fixture Monkey with objectIntrospector option
FixtureMonkey fixtureMonkey=FixtureMonkey.builder()
        .objectIntrospector(FieldReflectionArbitraryIntrospector.INSTANCE)
        .build();

Method 3: Through Java Bean (Default objectIntrospector)

  • Requires each object to be mocked to have:
    • No args constructor
    • Setter

Useful features from Fixture Monkey

See Fixture Monkey Component for the full list of features

Validation

  • Read more in Fixture Monkey Arbitrary Validator
  • Invalid instance could not be generated
    • If an object is invalid which sampled over 10k times from Arbitrary, TooManyFilterMissesException will be thrown
    • See below in "@Pattern" bug section for a sample log
  • By default, there is no validator included

Bean Validation by Javax

  1. To enable bean validations by Javax, add fixture-monkey-javax-validation dependency
  2. Add plugin option
FixtureMonkey fixtureMonkey=FixtureMonkey.builder()
        .plugin(new JavaxValidationPlugin())
        .build();

Bean Validation by Jakarta

  1. To enable bean validations by Javax, add fixture-monkey-jakarta-validation dependency
  2. Add plugin option
FixtureMonkey fixtureMonkey=FixtureMonkey.builder()
        .plugin(new JakartaValidationPlugin())
        .build();

Arbitrary Validation

  • Current version only have DefaultArbitraryValidator which javax.validation.Validator would validate instance
  • Can create own custom Arbitrary Validator by implementing ArbitraryValidator
  • Can combine Arbitrary Validators with CompositeFixtureValidator
  • Create FixtureMonkey with arbitraryValidator option

Null instance

  • Determines when a null instance is created with defaultNullInjectGenerator option

InnerSpec

  • This is an alternative from using Arbitrary to create custom logic for generation of mock data
  • However, to date, there are no documentations on the usage of InnerSpec

Jqwik - Arbitrary

  • Arbitrary is a fixture in jqwik

    • It is used in fixture monkey to generate random data for the different types of data in the properties of Java object

  • To use Arbitrary in Fixture Monkey:

    1. Create a Fixture Monkey generator for the object that we want to generate
    FixtureMonkey fixtureMonkey = FixtureMonkey.builder().build();
    1. Create Arbitrary for whatever fields we want to have custom generation logic
    • E.g.:
    Arbitrary<String> genderArbitrary = Arbitraries.of(gender);
    1. Use ArbitraryBuilder to set the fields using Arbitrary created
      ArbitraryBuilder<Person> personArbitraryBuilder = fixtureMonkey.giveMeBuilder(Person.class)
              .set("gender", Arbitraries.of(genderArbitrary))
              ...
    1. Create the object
     Person person = personArbitraryBuilder.build().sample();

  • See Arbitrary Component for more details

Other useful properties for Arbitrary

  • Filtering
    • Include only part of the values generated by Arbitrary
    • Arbitrary.filter(predicate)
  • Mapping
    • Start with existing arbitrary and use its generated values to build other objects from them
    • Arbitrary.map(function)

Jqwik - Constraints

See Jqwik Constraining Default Generation for more details

  • Default parameter generation can be constrained with these additional annotations

Commonly used

  • Allow null values with probability
    • @WithNull(value): Inject null with probability of value
      • Works for all generated types
  • Set string length
    • @StringLength(value, min, max): Set fixed length / range length
    • @NotEmpty
  • Set string not blank
    • @NotBlank
  • Characters
    • @LowerChars
    • @NumericChars
  • Set Collection (List, Set, Stream, Iterator, Map, Array) size
    • @Size(value, min, max): Set fixed length / range length
  • Set uniqueness in container object (List, Set, Stream, Iterator, Array)
    • @UniqueElements(by): Ensure elements are unique or unique in relation to a certain feature (by)

Jqwik - Parameterized Type

  • Jqwik can handle type variables and wildcard types as well

See Handling of parameterized Types for more details

Jqwik - Custom meta-annotations

  • Can also create own annotations apart from using Jqwik built-in annotations
  • See Self-Made Annotations for more details

How-to

Multi-valued Fields

  • Create ListArbitrary
  • For specific list of values, pass in through Arbitraries.of(<list of values>)
List<String> countries=["Singapore","Malaysia",...];
        ListArbitrary<String> countryArbitrary=Arbitraries.of(countries).list();

Abstract classes with concrete subclasses

  • Unable to simply just create superclass when its abstract type, i.e. cannot simply just fixtureMonkey.giveMeOne(CommsInfo.class) where CommsInfo is an abstract class
  • One possible approach - creating minimal Arbitrary for each concrete class:
Arbitrary<OnlineAccount> onlineAccountArbitrary=
        fixtureMonkey.giveMeBuilder(OnlineAccount.class).build();
        Arbitrary<InternetSubscriptionAccount> internetSubscriptionArbitrary=
        fixtureMonkey.giveMeBuilder(InternetSubscriptionAccount.class).build();

        Arbitrary<CommsInfo> commsInfoArbitrary=
        Arbitraries.oneOf(onlineAccountArbitrary,internetSubscriptionArbitrary);

Nested fields

  • Cannot simply do this to have custom logic for nested fields:
fixtureMonkey.giveMeBuilder(Person.class)
        .set("hobbies.hobby",hobbyArbitrary) // where hobbies is a nested field

Common formatted data

Email

  • Can use Web Module (jqwik-web) to generate data of email format from
    • Create EmailArbitrary using Web.emails()

Time

  • Can use Time Module(jqwik-time) to generate date/time-related java types

Considerations

Bean validation

  • Bean validations will be overridden with Arbitraries
    • E.g.: Cannot use @Email on a field that we are going to overrride with Arbitraries (such as Arbitraries.strings().ofMaxLength(20))), it will not be possible to generate mock data the fits the @Email constraint yet also abiding to the Arbitraries attributes

Field name type safety

  • To override / custom generator for different fields, we have to pass in the fieldName in string, like
 ArbitraryBuilder<Person> personArbitraryBuilder=fixtureMonkey.giveMeBuilder(Person.class)
        .set("gender",Arbitraries.of(genderArbitrary))
        ...
- There are thus risks that the field name may not be valid / present in the object to be mocked
  • Create Fixture Monkey with useExpressionStrictMode = true option to throw exception if the property does not exist

Leftover

@Pattern does not seem to work

@Pattern(regexp = "^[FM]?$") // javax.validation.constraints.Pattern
private String gender;
  • Threw exception, which indicates mocked data was not created based off the @Pattern constraint
22:25: 48.792 [main] ERROR c.n.f.r.ArbitraryValue$MonkeyRandomGenerator - Fail to create valid arbitrary.

Fixture factory Constraint Violation messages.

- violation: must match "^[FM]?$", type: class com.lqr.javafixture.domain.Person, property: updatedBy, invalidValue: ^M$

javax.validation.ConstraintViolationException: DefaultArbitrayValidator ConstraintViolations. type: class com.lqr.javafixture.domain.Person
...
net.jqwik.api.TooManyFilterMissesException: Filtering [net.jqwik.api.RandomGenerator$$Lambda$634/0x0000000800df30e0@9a9aa68] missed more than 10000 times.

Thoughts

  • Lack of documentations for fixture monkey components:
    • Some of the useful components such as InnerSpec, ArbitraryExpression, ExpressionGenerator may be powerful, but the lack of documentations make it difficult to implement and make use of
  • Fixture Monkey is heavily reliant on Jqwik in the creation of Arbitrary
    • Jqwik is a property-based testing framework for java
    • Here is a good article that summarize the key features of jqwik: https://www.baeldung.com/java-jqwik-property-based-testing
    • This means that apart from generating of mock data (with Arbitrary + Fixture Monkey), we are also able to make use of jqwik to perform property-based tests which is a complement to what we already set up for mocking of data
  • Complexities in simple use cases
    • In order to override a particular
  • Flexibility in independent field-level constraints
    • The support for Bean Validations (javax, jakarta), jqwik as well as self-made annotations provide much flexibility for having custom logic in generating data in fields
    • However, this can only be done on the class itself, where cases if we do not want to "dirty" the POJO (e.g.: to segregate mock data constraints and actual domain constraints), we will need to duplicate the class itself OR to go for another approach for custom generation
      • Another approach would be through the use of Arbitrary which can result in too much boilerplate codes & steeper learning curve

Resources