The benchmarks module does not produce a JAR by default which let's our Maven Central deployment fail as a module has to produce one according to their rules. We're now only including the benchmark module when the benchmarks profile is active.
We now make sure that the SimpleTypeHolder produced by MongoCustomConversions is used to set up default MongoMappingContext instances in (Reactive)MongoTemplate and unit tests.
Use lombok's Value for ObjectPathItem. Make methods accessible in DefaultDbRefResolver before calling. Use class.cast to avoid warnings. Update Javadoc.
Original pull request: #478.
We now check if already resolved DBRef's are assignable to the target property type. If not, we perform conversion again to prevent ClassCastException when trying to assign non matching types.
Remove non applicable public modifiers in ObjectPath.
Original pull request: #478.
Enhance benchmark statistics with Git/working tree details. Specify byte encoding for JSON to byte encoder.
Add status code check to HttpResultsWriter to verify that the results were accepted. Convert spaces to tabs in pom.xml.
Original pull request: #483.
Run the benchmark via the maven profile "benchmarks":
mvn -P benchmarks clean test
Or run them customized:
mvn -P benchmarks -DwarmupIterations=2 -DmeasurementIterations=5 -Dforks=1 clean test
Original pull request: #483.
Use Lombok's Value annotation for immutable value objects. Use IllegalArgumentException for NonNull validation exceptions. Trim whitespaces, formatting.
Original pull request: #472.
We now make sure to run any query / update object through the Query- / UpdateMapper. This ensures @Field annotations and potential custom conversions get processed correctly for update / remove operations.
Original pull request: #472.
We now explicitly mention mapping/support limitations for API variants like count(Query, String) not having domain type specific information that allows field specific mapping.
We log CyclicPropertyReferenceException with its message only and removed the stack trace from the log. The stacktrace points to a verifier location and is not particularly useful in finding the offending code. This change creates consistency over how CyclicPropertyReferenceException is logged.
Original pull request: #397.
We now consider the property's collection type after bulk-fetching DBRefs before returning the actual result value. The issue got only visible if bulk fetching is possible and constructor creation is used. Setting the property value on through an property accessor works fine because the property accessor checks all values for assignability and potentially converts values to their target type. That's different for constructor creation.
Original Pull Request: #457
We now make sure to safely serialize JSON output for mapped documents. This prevents the logger from rendering false exception messages to log appender.
Original pull request: #450.
Use Integer.MIN_VALUE as placeholder for unset values to allow setting zero (immediately) and -1 (indefinite wait) server selection timeouts. Fix test method name. Add JavaDoc.
Original pull request: #449.
We now make sure to safely serialize the criteria object used for creating the error message when raising an `InvalidMongoDbApiUsageException` in cases where `addCriteria` is used to add multiple entries for the same property.
Original pull request: #448.
The changes for DATAMONGO-1617 subtley changed the behavior for entity updates in terms of the version value they see for entities using optimistic locking. Previously the updates already saw the new version value, where after the fix for DATAMONGO-1617 it saw the old one. That caused BeforeConvertEvent listeners not being able to distinguish between an original insert and the first update anymore.
This change is now rolled back and we introduced a test case that encodes this expectation explicitly.
Oliver Gierke
Mark Paluch
Christoph Strobl
Borislav Rangelov
Martin Macko
Michael J. Simons