We turned Example and ExampleMatcher into interfaces with default implementations created via the static of methods. The default ExampleMatcher has strongly typed default implementation. This change allows store implementations to create their very own flavor of Example and ExampleMatcher allowing eg. non typed ones.
Original pull request: #238.
Path expressions are now evaluated one by one with the first one actually available in the payload being used for value lookup. Also tweaked the default behavior to return null in case of an single, invalid path.
Methods for domain event registration and access are now not following bean semantics anymore to avoid those accidentally leaking into formats that use the bean spec (e.g. Jackson to produce JSON).
We're now favoring the generic TypeInformation over trying to resolve the property type via field or PropertyDescriptor as only the former does proper generic type resolution.
Type specialization - i.e. enrichment of a raw type with a current generic context - is now only done if the current type is not yet resolved completely. This allows wildcarded target references to just fall back to the type to specialize, which will then by definition carry more generics information than the one to be specialized.
Revert the merging of a parent type's type variable map and only apply the locally available declared generics in case of parameterized types. Moved that augmentation into ParameterizedTypeInformation.
We now use CGLibs ReflectUtils.defineClass(…) to inject classes into the entity's class loader instead of our own code. ReflectUtils itself discovers whether the class loader is encapsulated and falls back to Unsafe for class definition. We no longer require our own code so the evil class was removed.
Original pull request: #234.
RxJava 1 development is phased out over time (Feature freeze in June 2017, EOL in March 2018) and we don't want to ship newly developed first-class support for libraries known to be no longer maintained. RxJava 1 type adoption support is not affected. Single and Observable can be still used in user-space code declaration, also copies of RxJava1CrudRepository (in user-space code) remain working.
XmlBeam 1.4.11 accidentally removed ProjectionFactory as public interface. Adapted our HttpMessageConverter implementation to make sure we still work on those versions.
Related ticket: https://github.com/SvenEwald/xmlbeam/issues/45
Marked all packages with Spring Frameworks @NonNullApi. Added Spring's @Nullable to methods, parameters and fields that take or produce null values. Adapted using code to make sure the IDE can evaluate the null flow properly. Fixed Javadoc in places where an invalid null handling policy was advertised. Strengthened null requirements for types that expose null-instances.
Removed null handling from converters for JodaTime and ThreeTenBP. Introduced factory methods Page.empty() and Page.empty(Pageable). Introduced default methods getRequiredGetter(), …Setter() and …Field() on PersistentProperty to allow non-nullable lookups of members. The same for TypeInformation.getrequiredActualType(), …SuperTypeInformation().
Tweaked PersistentPropertyCreator.addPropertiesForRemainingDescriptors() to filter unsuitable PropertyDescriptors before actually trying to create a Property instance from them as the new stronger nullability requirements would cause exceptions downstream.
Lazy.get() now expects a non-null return value. Clients being able to cope with null need to call ….orElse(…).
Original pull request: #232.
If you previously asked ProxyProjectionFactory for a proxy of an interface which the target instance already implements, it created a proxy although it could've returned the instance as is. It's now actually doing that avoiding superfluous proxy creation.
RepositoryFactoryBeanSupport.getRepositoryInformation() previously didn't consider any store-specific compositions (incl. store-specific fragments like the Querydsl support) as it just handed a fragment with the custom implementation to RepositoryFactorySupport.
This is now fixed by improving the API in RepositoryFactorySupport which now exposes a getRepositoryInformation(RepositoryMetadata, RepositoryFragments) which creates a fresh RepositoryComposition from the given RepositoryMetadata. Intermediate setup methods are still kept around but private as RepositoryFactorySupport.getRepository(…) still needs to forward the base RepositoryComposition to method interceptors for query result mapping.
As the calculation of a PersistentPropertyPath instances is rather expensive and they're heavily used in downstream mapping operations (e.g. query and fields mapping in MongoDB) we now cache the instances created per base type and property path.
Previously, RepositoryBeanNameGenerator applied the custom bean name lookup if the BeanDefinition given was not a ScannedGenericBeanDefinition. That in turn had been the case for custom implementation classes that were obtained through classpath scanning. With Spring 5 an index file can be used by the scanner, which in turn will cause AnnotatedGenericBeanDefinition instances being returned. That caused the code path to lookup the first constructor argument to kick in (usually used to obtain the repository interface from repository factory beans) and cause a NullPointerException.
We now forward AnnotatedBeanDefinitions as is and only apply the custom lookup for everything else, i.e. the bean definitions used for the factories.
Deprecated the additional method in favor if a lookup via MappingContext.getPersistentPropertyPath(PropertyPath) using the resolvable PropertyPath exposed by InvalidPersistentPropertyPath.
Re-enabled test cases for handling of different date/time APIs in auditing. Tweaked the lookup of the last modified value to try to convert to LocalDateTime as well in case it's a subtype of the type requested. In combination with two newly introduced converters for local date type times in JodaTime and ThreeTenBP, this allows us to still convert legacy JodaTime and ThreeTenBP types to be used as well.
We now keep instances of ReturnedType around to avoid the repeated reflection inspection for constructor and parameter name detection for class types and projection information for interfaces.
This is an overload for the already existing method taking a plain Iterator so that we only apply the close callback to the Stream in the newly introduced one.
To avoid the repeated scanning of projection interfaces for property descriptors we now cache the ProjectionInformation instances created via ProjectionFactory.getProjectionInformation(…). Adapted SpelAwareProxyProjectionFactory to now apply its customizations in newly introduced createProjectionInformation(…).
Removed the deprecated implementation of ResourceLoaderAware in favor of BeanClassLoaderAware. Removed deprecated ProjectionFactory.getInputProperties() in favor of ….getProjectionInformation(). Removed the Java conditional to add a MethodInterceptor for default methods. Adapted test cases.
Updated Sonargraph architecture description. Polished imports in domain package to avoid unnecessary imports (mostly due to Javadoc references).
Removed manually declared constructor in ExampleMatcherAccessor in favor of Lombok's @RequiredArgsConstructor. Removed deprecations in type information subsystem. Moved to different way of class loading to avoid null pointer warnings in ReflectionUtils.
Introduced new ExampleMatcherAccessor in data.support in favor of the now deprecated one in data.repository.core to avoid a dependency into the repositories subsystem to work with examples in general.
Added Degraph based test in order to avoid future cyclic package dependency violations.
Original pull request: #230.
Related ticket: DATAMONGO-1721.
We now use a marker interface to store cache hits for conversion targets without requesting a conversion target in the conversion query. The dedicated marker type has an individual cache view and does not interfere with other types.
Previously we used Object as cache key which caused false positives due to assignability checks. If an earlier conversion query with a requested target type yielded a hit (e.g. negative hit), subsequent queries without a target type received the same answer. This might happen although the non-cached answer would return a different result.
