We now override canRead(Type, Class<?>, MediaType) and tweak some of the delegate to reflect the patch in Spring HATEOAS spring-hateoas/#360 but without having to upgrade to the new version.
The preparation of the JsonNode that we map onto the domain object now correctly looks up fields by their mapped names. Previously we accidentally checked the incoming JSON for fields with raw unmapped property names with caused fields invalid detected or not detected if a custom Jackson naming strategy was in place.
Previously, RepositoryRestMvcConfiguration exposed its two ResourceSupportHttpMessageConverter beans as MappingJackson2HttpMessageConverters. This made it impossible for Spring Boot support for conditional beans to identify that they were type-constrained and, therefore, were not a suitable replacement for a general-purpose Mapping Jackson2HttpMessageConverter bean.
This commit updates RepositoryRestMvcConfiguration to expose both of its Jackson message converters as TypeConstrinaedMappingJackson2HttpMessageConverter instances. This will also Spring Boot to identify that their converters are type-constrained so that it can continue to auto-configure its general purpose converter. This is important as it allows Boot to provide a converter that honors the user’s configuration rather than Spring MVC’s default converter which does not. See spring-projects/spring-boot#2914 for details.
Original pull request: #174.
Before replacing the remove operation with a custom one that nulls out the object, we check whether the JSON Pointer points to an array element. If that's the case we skip the replacement and proceed.
We now treat embedded projections as Resource instances and equip them with a self link. This should also allow to register ResourceProcessor instances for projection types to even add projection specific links if necessary.
Also removed the explicit removal of the association link in case a projection is rendered for an association.
We now explicitly activate the resolution of lazy-loading proxies on Hibernate 4 to make sure not already populated lazy-loading proxies are actively resolved.
IncomingRequest.isJsonPatchRequest()/….isJsonMergePatchRequest() previously did a hard equals comparison to the media type discovered in the request. This fails if the request's media type contains parameters (e.g. the charset).
We now rather use MediaType.isCompatibleWith(…) to compare the media types more leniently.
Renamed test case to match conventions so that the Maven build actually executes them.
During our custom JSON handling on PUT requests, we previously piped all JSON objects through code that expected that one to be mapped to a PersistentEntity. This fails for JSON objects that basically represent maps.
We now skip that code paths for non-entity properties.
The test case introduced for DATAREST-482 failed when running against Spring 4.1 generation as the handling of empty request payload changed slightly in 4.1. We now explicitly use required = false to get null handed into the controller and manually default the Resources<Object> in it.
Related tickets: DATAREST-482, SPR-12778.
Scrubbed old format. Lightened up paging section and focus more on "next"/"prev" hypermedia controls and their benefit. Moved several sections under one top level section called "Customizing Spring Data REST". Cleaned up index.adoc to match the style of Spring Data JPA. Removed JSONP section.
Remove embedded entity operations, based on old spring-data-rest-compact media type. Utilized relative leveloffset for each include. Corrected some ALPS mistakes.
Original pull request: #162.
Collapsed if clauses leading to the very same return in RootResourceinformationToAlpsDescriptorConverter. Renamed isExportableProperty(…) to isExportable(…) in JacksonMetadata. Removed commented out code in AssociationLinks. Cleanups in test code.
Original pull request: #160.
If @JsonIgnore is applied to either a field or a getter method for an association, then do not expose that property's metadata down from ALPS.
Original pull request: #160.
We now make sure that unmapped properties (e.g. by using @JsonIgnore) are ignored when applying PUT request payloads. To do so we now basically use the merging algorithm used for PATCH requests but make sure we fill up the request payload with explicit null values for all mapped but not contained properties.
To be sure to pick up domain type lookup customizations we now consistently use RepositoryMetadata instead of trying to be clever and re-detecting it via RepositoriesUtils.
