DATACMNS-783 introduced code to specialize a raw type containing generics with the type information from it's context so that it can be established if only the raw type has been e.g. stored as type hint in a store (SomeType<T> stored as SomeType and T needing to be reestablished from the context). That application of the context unfortunately did not consider that the raw type might carry generics information itself, like `SomeType extends Foo<String, Integer>`.
We've now changed the specialization algorithm so that the the target type is being looked at in the context of the type used for the assignment (e.g. `Foo<T, ID> foo`). If that assignment fully resolves the declared type, we just outright use the given type as it by definition carries all generics information plus better type information in the first place.
If the type to specialize on is only partially expanding the generics we now create a synthetic type to merge the generics information. I.e. a `SomeType<T> extends Foo<T, Integer>` stored as SomeType would still resolve ID to Integer but capture T from the context including all potentially declared bounds.
Related tickets: DATACMNS-783, DATACMNS-1138.
To resolve XSD files properly from the classpath, their HTTPS reference must be present in the spring.schemas to avoid internet interaction for resolving an XSD file.
A side effect of the fix for DATACMNS-1482 trigged source objects that already perfectly match the desired return type would still be subject to detailed inspection of potential conversions, which – by definition – will not kick in.
The code has now been changed to only trigger detailed inspection for collection values.
Related tickets: DATACMNS-1482.
If plain dots were submitted as elements in a Sort expression to be parsed by SortHandlerMethodArgumentResolver, those dots would be considered a property of the sort expression, which is of course wrong. We now drop property candidates solely consisting of dots and whitespace.
We now cache the BeanFactory lookup for EvaluationContextExtension within the ExtensionAwareEvalutationContextProvider to avoid (expensive) repeated context scans when creating the actual EvaluationContext.
Original pull request: #395.
Before this commit we haven't properly resolved methods on a root object provided by an EvaluationContextExtension that was using varargs. With a vararg method, the number of parameters handed into the method is not necessary equal to the number of parameters. We previously simply skipped methods with a different number of arguments. We now try direct matches first but calculate valid varargs alternatives in case that initial lookup fails and try to match those alternatives.
This lookup is implemented in ….util.ParameterTypes now and used by ….spel.spi.Function. The latter now also handles the actual invocation of those methods properly by collecting the trailing arguments into an array.
Various fast returns and the use of Class instead of TypeDescriptor led to e.g. List<BigDecimal> not getting properly converted to List<Integer> leading to unexpected ClassCastExceptions when the collection elements where accessed.
We now consider more than 16 immutable and nullable Kotlin properties per bucket in generated PropertyAccessors.
Previously only the first 16 properties were considered due to truncation of the defaulting bitmap. We used SIPUSH to render the defaulting mask in bytecode which is intended for 16 bit integers (short). Migrating to LDC (runtime constants) preserves the actual constant value of 32 bits and so we're considering now full buckets.
Oliver Drotbohm
Christoph Strobl
Jens Schauder
Stas
Mark Paluch
Sangwoo Lee
Michael Simons
David Schneider
Jerzy Kwiatkowski
Spring Operator
Keegan Witt
Oliver Drotbohm
Sebastien Deleuze