When traversing nested property paths, AbstractMappingContext.getPersistentPropertyPath(…) previously used the raw actual property type. If the property path contains a reference to a generically typed property, this causes the deeper paths not being resolved correctly.
We now explicitly use the TypeInformation of the property to retain generics information while traversing the path.
In case the PageableHandlerMethodArgumentResolver was configured to use one-indexed parameters, it wasn't defaulting the lower bounds for the page number. This caused indexes out of the allowed bound submitted causing an invalid index handed tor PageRequest. We now apply better range shifting before the bounds are applied.
Instead of manually using Introspector.getBeanInfo(…) we now use Spring's BeanUtils.getPropertyDescriptors(…) to benefit from some caching of descriptor instances as well as advanced support for fluent setters, default methods etc.
Original pull request: #122.
AnnotationBasedPersistentProperty now also caches the absence of properties that are expressed through accessors only. Previously the absence of a field caused us to skip the registration of the absence in the cache.
MappingContextTypeInformationMapper now caches detected type aliases.
PreferredConstructor.isEnclosingClassParameter(…) now eagerly returns if the parameter itself is not an enclosing one and thus avoids a collection lookup and equals check. Moved equals check for the type to the very end of the equals check to increase the chances that other inequality guards kick in earlier.
AbstractMappingContext now leaves the non-null-check for getPersistentEntity(…) to the factory method of ClassTypeInformation.
We now pre-calculate the hash codes for TypeInformation implementations as far as possible as the instances are used as cache keys quite a lot. The same applies to AbstractPersistentProperty.
BasicPersistentEntity now uses an ArrayList we sort during the verify() phase to mimic the previous behavior wich was implemented using a TreeSet as ArrayLists are way more performant when iterating over all elements which doWithProperties(…) is doing which is used quite a lot.
BeanWrapper now avoids the getter lookup if field access is used.
SimpleTypeHolder now uses a CopyOnWriteArrySet to leniently add types detected to be simple to the set of simple types to avoid ongoing checks against the inheritance hierarchy.
In case the repository lookup for a given domain type fails we traverse the given types super-types and try to detect a repository for those.
Original pull request: #110.
Simplified implementation of Path conversion. Inlined helper domain types to not to pollute the packages with types that are only used within that very one test class.
Original pull request: #111.
We now make the field detected by the callback accessible so that the value lookup doesn't fail for private fields. Added a few more unit tests to verify behavior.
PageImpl now makes sure that the total given to the constructor is never less than then number of items given to make sure we do not mask broken count calculation by creating an actually invalid instance.
Related ticket: DATAMONGO-1120.
The initial setup of the type variable map now unfolds the detected types in the map to make sure we detect all type variables present in the current scope.
Added caching of component and map value TypeInformation instances to avoid repeated creation.
So far, the lookup of the type variable map was preferring the type variable maps detected on parent types in nested structures. This caused the concrete type variables in nested types not being considered correctly which caused the type resolution to fall back to the generic bounds.
We now accumulate the type variable maps to avoid having to lookup a certain map in the nesting hierarchy. The core fix is in ParentTypeAwareTypeInformation's constructor and mergeMaps(…) respectively. Simplified the handling of type variable maps and made proper use of generics throughout the class hierarchy.
DomainClassConverter now returns input when the source type is identical to the target type. Previously we erroneously tried to convert the sourceType to the idType of the targetType's backing repository. As a side effect we also avoid performing some unnecessary converter hops.
Original pull request: #101.
Clarified return values if predicates don't match or return multiple results (for the findOne(…) method). Copied summary to the @return tag.
Original pull request: #96.
Previously we only allowed to sort by QueryDsl path expressions. With this change we now also support ordering by operator expressions, e.g. yearMonth() on a date property.
Original pull request: #94.
The type detection for entity candidates of PersistentProperty instances now consistently handles collection types even if they're used als collection or map values.
Annotations used on both fields and accessors are now only rejected if they're Spring Data annotations (i.e. from the org.springframework.data namespace) and the configurations of these annotations differ.
PropertyPath now exposes the complete property named the resolution failed for in case it can't resolve a source String. It will expose the part of the Path from the deepest successful resolution result.
Changed the resolution algorithm to use a List instead of a Stack and a fresh instance of the former so that the nested recursion steps keep track of the correct resolution depths.
RepositoryConfigurationDelegate now registers the store-specific RepositoryConfigurationExtension as lazy-init BeanDefinition so that other parties can detect that definition.
We're now firing the component registration event manually in RepositoryBeanDefinitionParser as ParserContext.registerBeanComponent(…) registers the given BeanComponentDefinition with the registry. As we do that manually before this previously resulted in the BeanDefinition overriding the very same one and container warnings in turn.
We now synchronize the (seldom) writes to the isPropertyParameterCache HashMap in PreferredConstructor via a ReadWriteLock. We could as well have used a ConcurrentHashMap here without the need for manual locking but this would potentially waste memory for a mostly read-only data structure. We also anticipate potential multiple writes for the same property.
Original pull request: #86.
Parameters erroneously rejected @Param annotated query method parameters that were preceded by a non-bindable type (like Pageable or Sort) as it checked for a 0 index of the parameter. We now manually keep an index in the check itself and have removed the Parameter.isFirst() method as it's not used anywhere else.
We're now using the newly introduced method on ControllerLinkBuilderFactory that takes both a type and a method to forward the method the invocation is happening on independently from the method being invoked.
The property "location" should actually be "locations" since this attribute is read by org.springframework.data.repository.config.ResourceReaderRepositoryPopulatorBeanDefinitionParser.doParse(Element, BeanDefinitionBuilder).
