The indicator of whether custom construction is needed for a class based return type now includes whether input properties are available in the first place.
In case a DTO type exposes multiple no-argument constructors we now leniently abort the parameter name detection and let downstream users decide what to do in this particular case.
The problematic case can't be outright rejected here already as there's valid scenarios for that, e.g. in JPA the type could be used for multiple query methods, using explict constructor expressions listing arguments explicitly and thereby select one of the multiple constructors.
Made the query-by-example.adoc a top level document. Clients including the documents have to tweak the indentation on the inclusion side then. Included query-by-example.adoc into index.adoc so that it can be double checked in test renderings easily.
Renamed ExampleSpecification to ExampleMatcher and introduced a matching() factory method, dropping the dedicated override mechanism for types for now. Updated documentation accordingly.
Used Lombok for constructor creation, field defaults, toString() as well as equals(…) and hashCode() implementations.
Tweaked imports in unit tests. Renamed methods to express expected behavior.
Original pull request: #153.
Split Example and ExampleSpec to create reusable components. Refactor builder pattern to a fluent API that creates immutable instances. Split user and framework API, Example and ExampleSpec are user API, created ExampleSpecAccessor for modules to access example spec configuration. Create static methods in GenericPropertyMatchers to ease creation of matchers in a readable style. Convert PropertySpecifier to inner class and move PropertySpecifiers to ExampleSpec.
Related tickets: DATAJPA-218, DATAMONGO-1245.
Original pull request: #153.
Added required types for Query by Example which should be used by the individual store implementations.
Generally the Example type captures a sample object and allow various settings concerning the mapping into an actual query. So there’s configuration options for handling null values, string matching, property paths to ignore,
Related tickets: DATAJPA-218, DATAMONGO-1245.
Original pull request: #153.
ResultProcessor.processResult(…) now explicitly handles null values to prevent IllegalArgumentExceptions being provoked in the ProjectionFactory which occurred if null values were handed down to it.
We now resolve composed annotation values using @AliasFor within AnnotationBasedPersistentProperty and BasicPersistentEntity. Nevertheless it is up to the individual store implementation to make use of this.
Original pull request: #156.
Previously, we registered an InstantiationAwareBeanPostProcessor to predict the type to be created by RepositoryFactoryBeanSupport by inspecting a particular property value of the registered BeanDefinition.
This has now been elevated to a more generic mechanism that can get a FactoryBean type configured with a set of properties to inspect for a configured type. That new infrastructure now replaces the explicit configuration for RepositoryFactoryBeanSupport with one that's set up via configuration.
Fixed indentation to use tabs instead of spaces. Tweaked structure of if-clauses in PropertyAccessingMethodInterceptor to reduce nesting. Made helper method static.
Restructured test cases slightly and used ExpectedException rule to verify exceptions. Moved newly introduced test methods more to the end of the file (new tests last).
Added author and copyright year extensions where necessary.
Original pull request: #155.
We now use a non-static cache for TypeInformation instances in SimpleTypeInformationMapper to make sure Spring Boot's development tools can correctly reload classes. Removed the static singleton instance of SimpleTypeInformationMapper in favor instantiating it directly.
Introduced a template method to override the creation of a PagedResource to allow subclasses to create a more concrete instance than PagedResource itself.
