Fluency for notAccessed

[abeyecon]

Using the following rule I can ensure that no impl class gets accessed directly (they will instead be injected via spring @component injection against their interface). However it doesn't read as well as it could, and the violation message doesn't read well either. Suggest adding .notBeAccessed()

@archtest
public static final ArchRule implOnlyUsedByInjection = classes()
.that().resideInAPackage(IMPL)
.should().onlyBeAccessed().byAnyPackage();

Violation Message:
Architecture Violation [Priority: MEDIUM] - Rule 'classes that reside in a package 'com.eyecon.xyz.impl..' should only be accessed by any package ['']' was violated (51 times):

[hankem]

Even though not fluent, you can easily use a custom condition:

    @ArchTest
    public static final ArchRule implOnlyUsedByInjection = classes()
            .that().resideInAPackage(IMPL)
            .should(notBeAccessedAtAll());

    static ArchCondition<JavaClass> notBeAccessedAtAll() {
        return new ArchCondition<JavaClass>("not be accessed at all") {
            @Override
            public void check(JavaClass clazz, ConditionEvents events) {
                Set<JavaAccess<?>> accessesToSelf = clazz.getAccessesToSelf();
                StringBuilder message = new StringBuilder(clazz.getName()).append(" is accessed ").append(accessesToSelf.size()).append(" times");
                if (!accessesToSelf.isEmpty()) {
                    message.append(":");
                }
                for (JavaAccess<?> javaAccess : accessesToSelf) {
                    message.append("\n").append(javaAccess.getDescription());
                }
                events.add(new SimpleConditionEvent(clazz, accessesToSelf.isEmpty(), message.toString()));
            }
        };
    }

Violation Message:

java.lang.AssertionError: Architecture Violation [Priority: MEDIUM] -
 Rule 'classes that reside in a package 'com.eyecon.xyz.impl..' should not be accessed at all'
 was violated (51 times):

[abeyecon]

Thanks @hankem that is a very tidy example

[codecholeric]

Just to mention, if it does what you want, and you only don't like the message, you could also just overwrite that:

@ArchTest
public static final ArchRule implOnlyUsedByInjection = classes()
    .that().resideInAPackage(IMPL)
    .should().onlyBeAccessed().byAnyPackage()

    .as("Spring bean implementations should not be directly accessed")
    .because("we want to use Spring DI and only to refer to them through their interface");

That's also a documentation for anyone reading your rule, what was the intention.

[abeyecon]

        return new ArchCondition<JavaClass>("not be accessed at all") {
            @Override
            public void check(final JavaClass clazz, final ConditionEvents events) {
                final Set<JavaAccess<?>> accessesToSelf = clazz.getAccessesToSelf();
                final Set<JavaAccess<?>> externalAccessesToSelf = accessesToSelf.stream()
                        .filter(access -> access.getOriginOwner() != clazz)
                        .collect(Collectors.toSet());
                final StringBuilder message = new StringBuilder(clazz.getName()).append(" is accessed ")
                        .append(externalAccessesToSelf.size()).append(" times");
                if (!externalAccessesToSelf.isEmpty()) {
                    message.append(":");
                }
                for (final JavaAccess<?> javaAccess : externalAccessesToSelf) {
                    message.append("\n").append(javaAccess.getDescription());
                }
                events.add(new SimpleConditionEvent(clazz, externalAccessesToSelf.isEmpty(), message.toString()));
            }
        };
    }

Modified code to allow class to access itself internally

[abeyecon]

Thanks for the help. Love ArchUnit. Looking forward to 1.0 so that our large business friends can use it without fighting with the corporate policies.

[codecholeric]

Yes, we should really release 1.0
There is no real reason for keeping it 0.x, it's stable and has been out there for a long time, it's mainly just me thinking "why should this be 1.0, it's not different from the last release" 🙄
I really feel #171 should be in there, but since I'm gonna merge that into the next release, maybe then it's time 😉

[codecholeric]

PS: If your question is solved, do you want to close the issue?