Added debug logging in RepositoryFactorySupport and RepositoryConfigurationDelegate so that the scanning and the instantiation of repositories can be a lot easier identified in the logs.
Use weak references in annotation and property annotation cache to retain references until the last GC root is cleared. Remove trailing whitespaces. Reformat.
Original pull request: #304.
We now use HashMap to store persistent properties of a PersistentEntity. An entity is built in a single thread so no concurrent modification happens. Concurrent reads may happen during entity usage which is fine as the PersistentEntity is not changed anymore. Previously, we used ConcurrentReferenceHashMap defaulting to soft references. Soft references can be cleared at the discretion of the GC in response to memory demand. So a default ConcurrentReferenceHashMap is memory-sensitive and acts like a cache with memory-based eviction rules.
Persistent properties are not subject to be cached but elements of a PersistentEntity and cannot be recovered once cleared.
Original pull request: #304.
We now eagerly look up the aggregate root type of a repository in the MappingContext. Some implementations might not have pre-populated the context with all entities and we need to make sure it knows about the aggregate root as other clients (e.g. the auditing subsystem) might only defensively access the entities via PersistentEntities which is not adding new entities to avoid store clashes.
ProxyProjectionFactory and ResultProcessor require a DefaultConversionService to convert values for projection proxies and to post-process query method results. Creating instances of these types requires an instance of the conversion service which previously created a DefaultConversionService instance (along all converter registrations) upon ProxyProjection/ResultProcessor instantiation.
We now use the shared instance that is initialized once and shared across all ProxyProjection/ResultProcessor instances to reduce garbage and improve the CPU profile.
We now calculate the number of expected default masks to filter synthetic constructors that do not match the expected parameter count. A Kotlin constructor with argument defaulting generates a synthetic integer argument for every 32 constructor arguments. This is independent of the number of actual optional arguments.
Previously, we used an non-exact check to consider constructors as default ones if they had at least two additional arguments. This caused the wrong constructor being used if the non-synthetic types matched the types of the default constructor.
Removed the explicit registration for JodaTime and ThreeTenBP to JSR-310 converters (originally introduced to support the unifying lookup of the last modified date in the auditing subsystem) as reading converters. This avoids the warning reporting that the source types (JodaTime and ThreeTenBP LocalDateTime) not being store-native types (which usually indicates a superfluous converter registration).
Updated the test cases to make sure these warnings aren't trigger due to test setups causing the same issue.
Moved test cases into PropertyPathUnit test so that they're closer to the implementation. Switched to Introspector.decapitalize(…) to follow the Java Beans Specification regarding the handling of all-uppercase properties.
Original pull request: #289.
Introduced ProxyUtils.getUserClass(…) that by default is basically a facade for Spring's ClassUtils.getUserClass(…) but allows the registration of ProxyDetector implementations via Spring's SpringFactoriesLoader mechanism.
Moved all existing usages of ClassUtils.getUserClass(…) to ProxyUtils.
I added a cover page for the epub output, now that the Spring data reference guides will have epub output (through a commit against spring-data-build).
Original pull request: #290.
We now explicitly disable entity expansion in the DocumentBuilderFactory used by XMLBeam.
Introduced constructor in XmlBeamHttpMessageConverter to allow dedicated configuration of an XBProjector instance in case the defaults need tweaking and augmented the web configuration setup to automatically pick up a custom XmlBeamHttpMessageConverter bean instead of our own default if present.
After considering JSR-310 types to be simple we map these types primarily to java.util.Date as the majority of stores does not natively support JSR-310 types. Converters referencing JSR-310 types are now properly annotated with Reading/WritingConverter annotations to distinguish between reading and writing intents.
Othwerise, converters between JSR-310/java.util types and Joda/ThreeTenBackport to JSR-310 types interfere with conversion as regular java.util.Date types would convert to e.g. LocalDateTime.
We now consider all types in the java.time package as simple types to prevent deep reflective access. We are already excluding java.lang types for the same reason.
Original pull request: #286.
Previously a custom Iterable implementation would've caused QueryMethod.isCollectionQuery() to return true. We now solely rely on TypeInformation.isCollectionLike() (which checks for exact Iterable, Collection assignability and arrays) after handling potential wrapper types.
Instead of a simple check for assignability from Iterable, we now properly use TypeInformation.isCollectionLike(), which checks for Iterable equality or assignability of collections or arrays as well as an explicit check for Slice as that is needed to properly unwrap Page instances and Slices themselves. That prevents custom domain types implementing Iterable from being unwrapped into their element types.
