JIRA: https://jira.spring.io/browse/INT-3876
INT-3876 track groups ids in AbstractCorrelatingMessageHandler and docs
INT-3876 Fix merge conflict and add missing link in doc
INT-3876 Move groupId check in message group processor and rework on docs
INT-3876 Remove extra space from the doc
INT-3876 Fix checkstyle refer instance variable groupid correctly
INT-3876 Update aggregator doc with more details
* Polishing `aggregator.adoc`
* Rework `AbstractCorrelatingMessageHandlerTests.testDontReapMessageOfOtherHandler`
do not use redundant options
JIRA: https://jira.spring.io/browse/INT-550
* We need verify schema version and register `IntegrationRegistrar`
only once, not on parsing each element
* `@Ignore` time-sensitive `GroovyScriptExecutingMessageProcessorTests`
* Upgrade to Gradle-4.3.1 to overcome NPE with caches
JIRA: https://jira.spring.io/browse/INT-4363
* Register `HandlerMethodArgumentResolver`s in the
`IntegrationRegistrar` as `BeanDefinition`s to let them be configured by
the `BeanFactory` callback.
The direct instance doesn't call callbacks
* Use `JacksonPresent` in the `IntegrationRegistrar` instead of local
property
* Some code style polishing for the `IntegrationRegistrar`
* Add `JsonPathTests.testJsonPathOnPayloadAnnotation()` to ensure
that `PayloadExpressionArgumentResolver` is properly configured by the
`BeanFactory` and `#jsonPath()` SpEL-function is properly evaluated on
the `@Payload` for method `@ServiceActivator` method argument
JIRA: https://jira.spring.io/browse/INT-4363
The JIRA requests making the `MessageHandlerMethodFactory` a bean.
This is not possible without a major rework of the `MessagingMethodInvokerHelper`.
The problem is that the `InvokerHandlerMethod` s are created before the factory
is initialized. In fact, it only works at all since the IHMs have a hard reference
to the factory's argument resolvers so, when they are initialized in `initialize()`
each handler sees the resolvers.
This really needs to be improved but, since we are so close to GA, the compromise
was to add a `HandlerMethodArgumentResolversHolder` which holds the standard resolvers
which are then wired into the factory (along with the message converter).
Delaying the creation of the `InvokerHandlerMethod` is not trivial; today they are
built in the CTOR; moving it to `initialize()` would prevent the fast failure for
a badly configured bean; the proper solution might be to make the MMIHs beans themselves.
Polishing - PR Comments
Resolves: /spring-projects/spring-integration#2268
* To indicate the source the Redis message in the `RedisInboundChannelAdapter`
populate the `RedisHeaders.MESSAGE_SOURCE` header to the messages to produce
* Fix the `SimpleMessageConverter` to populate the provided `MessageHeaders`
to the message to produce
* Add `toMessage(T object, @Nullable Map<String, Object> headers)`
to the `InboundMessageMapper` to propagate additional header
to the message to create
* Rework all the out-of-the-box `InboundMessageMapper` implementations
to properly propagate additional headers via `toMessage()`
from the `SimpleMessageConverter`
* Provide optimizations in the `InboundMessageMapper` implementations
do not re-create messages
* Refactor `MutableMessage.toString()` to align with the `GenericMessage`
* Add more `@Nullable` to method arguments
* Increase latch wait timeout in the `EndpointParserTests`
* Address `redis.adoc` PR comment
JIRA: https://jira.spring.io/browse/INT-4362
INT-4312 added a JSON conversion when there is a single method with a non-String parameter.
Add more tests for parameters of type `Message`.
Catch conversion exceptions and use the message as-is if the conversion fails.
* Since we already have `PriorityChannelSpec` that isn't
`QueueChannelSpec` responsibility to worry about `priority` option.
More over it has been done wrongly and `MessageGroupQueue.setPriority()`
has been called unconditionally by the provided `ChannelMessageStore` type
* Fix `PriorityChannelSpec` logic as well:
- set `this.messageGroupQueue.setPriority(true);` to check the state of
the provided `PriorityCapableChannelMessageStore`
- fix assert condition in the `doGet()`
* And simple compiled SpEL test-case for the JDK Proxy
JIRA: https://jira.spring.io/browse/INT-4361
* To get a gain of the method chain code flow and avoid extra annotation
parsing, add `.role(String)` hook to the `EndpointSpec`
* Delegate the provided `role` property to the `AbstractEndpoint`
* Register `AbstractEndpoint` as itself `SmartLifecycle` in the
`SmartLifecycleRoleController`
* Add `destroy()` to the `AbstractEndpoint` and remove it from the
`SmartLifecycleRoleController`
* Provide some Java 8 code style refactoring
* Rework XML parsers and Annotation processors to populate `role`
property on the `AbstractEndpoint`
* Wrap `roleController` bean extraction to the `NoSuchBeanDefinitionException` catch
* Fix several `AbstractEndpoint` implementation to properly call
`super.onInit()` which has been missed before
Since `MutableMessage` can be mutated, the `ID` and `TIMESTAMP` headers
can be added lately.
Or there might be some scenarios which don't care about those headers at
all.
Plus we get a performance gain when we skip their generation
Coerce the `null` cast to the proper target ctor
JIRA: https://jira.spring.io/browse/INT-4349
When headers come from the external system there is no guarantee
that special headers (e.g. `sequenceNumber`, `priority` etc.) in the
expected (`Integer`) type.
* Widen `int` headers setting value to the `Number` type
* Return primitive `int` for the `sequenceNumber` and `sequenceSize`
headers since for them `IntegrationMessageHeaderAccessor` never return
null for them
* Remove `SequenceNumberComparator` in favor of `MessageSequenceComparator`
since they are essentially duplicate each other
* Fix tests to deal with primitive `int` already
* Register `BeanDefinition`s for DSL components.
That way we get a gain when the `BeanDefinition` is involved,
e.g. Spring Cloud Function uses this approach to determine the generics
of the target `Function` class
* This programmatic `BeanDefinition` registration approach
(available since SF-5.0) allows us to avoid some manual lifecycle processes:
bean initialization, autowiring, event listener registration etc.
* Rework `IntegrationFlowContext` to register and remove `BeanDefinition`s as well
* Remove unused `registerComponents` flag from the `StandardIntegrationFlow`.
With the proper logic in the `IntegrationFlowContext` it does not make sense any more
* Remove redundant `AnnotationGatewayProxyFactoryBean` in the `IntegrationFlows`;
populate `defaultRequestChannel` directly to the `GatewayProxyFactoryBean` instance
* Add an overloaded `IntegrationFlows.from(Class, String)` to allow to specify an
explicit bean name for the target gateway proxy
* Extract and populate bean name from the `@MessagingGateway.name()` in the
`AnnotationGatewayProxyFactoryBean`
* Stop lifecycles in the `AmqpTests` after using for proper test suit shutdown
JIRA: https://jira.spring.io/browse/INT-4357
Updated based on code review.
Updated adocs
*Polishing code style, typos
* Add protection from NPE when `LockRegistryLeaderInitiator` isn't
supplied with the `leaderEventPublisher`
* Add `What's New` note
Fixes: spring-projects/spring-integration#2261
* Fix `ScatterGatherHandler.stop()` to call `GatherEndpoint.stop()` instead of `start()`
Adding unit test as per artembilan's suggestion
Replacing asterisk import with specific namespaces
Fixing formatting/codestyle errors shown by travis
Creating separate bean for new unit test
* Some simple polishing
* Merge the assertions for the fix into an existing test
**Cherry-pick to 4.3.x**
JIRA: https://jira.spring.io/browse/INT-3936
* Extract `smartMatch()` logic from the `IntegrationManagementConfigurer`
and `IntegrationMBeanExporter` into the `PatternMatchUtils` class
* Add negative (`!`) pattern matching configuration support to the
`GlobalChannelInterceptor` annotation and `<int:channel-interceptor>` component
* Code style, Docs and JavaDocs polishing
JIRA: https://jira.spring.io/browse/INT-4312
If there is a single eligible method, the message content is JSON
and the method parameter doesn't match the payload, perform JSON
conversion if there is an ObjectMapper available.
Fixesspring-projects/spring-integration#2213
JIRA: https://jira.spring.io/browse/INT-4123
Fully different `MessageStore`s can be configured for the same shared
Key-Value data-base.
Since the retrieval logic is based on the keys, that may cause the
unexpected messages expiration via `MessageGroupStoreReaper`.
* To distinguish store instances on the shared store add `prefix`
option to the `AbstractKeyValueMessageStore`
* Deprecate the `GemfireMessageStore` `Cache`-based configuration - `setIgnoreJta()` and `afterPropertiesSet()`.
The `GemfireMessageStore` relies only on an externally configured `Region`.
**Cherry-pick to 4.3.x**
Doc Polishing
JIRA: https://jira.spring.io/browse/INT-4255
- Support embedding headers for transports that don't support headers (TCP, Kafka, etc)
- Use the new message-aware Jackson ObjectMapper
- Provide a mechanism to more efficiently support byte[] payloads (avoid Base64 encoding)
- Support decoding "legacy" SCSt embedded headers
Polishing; PR Comments - Add Support to MQTT and TCP
Switch to simple patterns instead of regex
* Fix JavaDoc typo
* Upgrade to Jackson 2.9.1
JIRA: https://jira.spring.io/browse/INT-4342
Use similar code to Spring AMQP to add white list support for Integration's
use of the `DeserializingMessageConverter`; introduce the `WhiteListDeserializingMessageConverter`.
Polishing
Missed this change in PR.
Fix XSD attribute
The `PollableChannelPublisherAdapter` is based on the poll model of the
`FluxSink` and iterate and poll downstream `PollableChannel` until
there is an item or `n > 0`.
Having `take(6)` we end up with the cancel from the downstream
`Subscriber` after the batch is filled, but at the same time we continue
to poll the upstream source because `n` is like `Long.MAX_VALUE`.
* The proper way to interact is check for the `!sink.isCancelled()`
as well.
This way cancelled `sink` won't "steal" data from other subscribers
Fix compatibility with the latest dependencies
* Upgrade to Gradle 4.1
* Upgrade as much dependencies as possible
* Fix MongoDB module to resolve deprecation in the latest driver
* Increase receive timeout in the `ResequencerTests`
* Restore generic argument for the method reference in the `ReactiveStreamsTests`
* Fix `WebFluxInboundEndpoint` for the compatibility with the latest Reactor API
JIRA: https://jira.spring.io/browse/INT-4333
If the user has explicitly set management properties in a bean definition,
the configurer should not override those settings with defaults.
JIRA: https://jira.spring.io/browse/INT-4336
Don't add "hidden" unreachable methods to the map in the `GatewayProxyFactoryBean`.
The methods can never be executed and may not have all expected properties; they
should not appear in the graph at all since they are unreachable.
Also add defensive coding to the `IntegrationGraphServer` to avoid NPEs when unexpected
conditions arise.
JIRA: https://jira.spring.io/browse/INT-4326
To let to disable autoStartup for particular endpoint bean globally,
in one place, add `spring.integration.endpoints.noAutoStartup`
to the `spring.integration.properties`
So, now all the `AbstractEndpoint` checks that property for matching its
bean name to configure `autoStartup` property to `false`
**Cherry-pick to 4.3.x**
* Add `validateOnlyOverlapping = false` for `RequireThis` Checkstyle rule
* Add `this.` to caught places
* Use `CustomizableThreadFactory` instead of raw interface in the
`UnicastSendingMessageHandler`
JIRA: https://jira.spring.io/browse/INT-4328
Add support for sending `ErrorMessage`s to the return and nack channels.
**cherry-pick to 4.3.x, but change default EMS to null (will require minor adjustment to test - set the EMS in `adapterWithReturnsAndErrorMessageStrategy`)**
JIRA: https://jira.spring.io/browse/INT-4327
To avoid reflection and annotation processing overhead on the bean classes
when we deal with non-singleton beans and there is really no any
messaging on the bean class, store bean classes to the local cache to skip
them in the future when `BPP` is applied for request/prototype beans
* Simple code style polishing
**Cherry-pick to 4.3.x**
JIRA: https://jira.spring.io/browse/INT-4318
Do not `ObjectMApper.findAndRegisterModules()` unconditionally.
If there is no the target module library in classpath, we end up
with the `NoClassDefFoundError`
* Copy-paste the well-known modules registration logic from the
`org.springframework.http.converter.json.Jackson2ObjectMapperBuilder`
since we can't add `spring-web` dependency to the `SI-core`
Fixesspring-cloud/spring-cloud-stream#983
When the `AbstractMessageChannel.getFullChannelName()` is called
before the full bean initialization, we don't have the proper name any
more lately - that is only `unknown.channel.name`
* Reset `fullChannelName` property to `null` in the end of `onInit()`,
so the next `getFullChannelName()` will build the proper component name
**Cherry-pick to 4.3.x**
JIRA: https://jira.spring.io/browse/INT-4316
To allow to customize a JSON representation of the Object, provide a
`JsonObjectMapper`-based ctor for the `ObjectToMapTransformer`
When we use `IntegrationFlows.from(Class<?>)`, we don't need to require
present for `@MessagingGateway`, we can just synthesize it with the
default properties
This way we can simply make gateway proxies for the `java.util.function`
interfaces which should be useful for Spring Cloud Function
JIRA: https://jira.spring.io/browse/INT-4300
* Add `ReactiveHttpInboundEndpoint` based on the WebFlux foundation
* Extract `BaseHttpInboundEndpoint` for common
HTTP Inbound Channel Adapters options
* Make `spring-webmvc` and `spring-webflux` as `optional` dependencies
to let end-user to choose
* Refactor `HttpContextUtils` to include constants
for newly added WebFlux support
* Introduce `BaseHttpInboundEndpoint.setRequestPayloadTypeClass()`
for raw `Class<?>` and modify existing `setRequestPayloadType()`
for the `ResolvableType`
* Refactor existing MVC tests and XML components parsers to use
new `setRequestPayloadTypeClass()`
* Add `MessagingGatewaySupport.sendAndReceiveMessageReactive()` to get
a reply from downstream flow reactive back-pressure manner
* Add `IntegrationHandlerResultHandler` implementation to let WebFlux
infrastructure to handle the `Mono<Void>` from the `ReactiveHttpInboundEndpoint`
properly
* Fix `JdbcLockRegistryLeaderInitiatorTests` race condition to assert
the `initiator1` is elected eventually after yielding when the `initiator2`
is stopped
* Fix JavaDocs issue in the `HttpRequestHandlingMessagingGateway`
* Move all the "hard" logic in the `MessagingGatewaySupport#doSendAndReceiveMessageReactive`
to the `Mono` chain ensuring back-pressure when `sendAndReceiveMessageReactive()` is
called not from the Reactive Stream
Add test-case to demonstrate SSE
JIRA: https://jira.spring.io/browse/INT-3625
Some polishing and optimization for the
`MessagingGatewaySupport.doSendAndReceiveMessageReactive()`
More optimization for `MessagingGatewaySupport`
* Upgrade to Reactor 3.1 M3
* Document WebFlux-based components
Minor Doc Polishing
* Check for `null` package in the `JacksonJsonUtils`.
Also don't call `this.delegate.typeFromId()` the second time
* Increase wait timeout in the `QueueChannelTests`
* Check `ZookeeperMetadataStore.addListener()` for `null`
**Cherry-pick to 4.3.x**
According Reactor recommendation it would be better to `CoreSubscriber`
from Reactor instead of Reactive Streams `Subscriber` for chasing
some optimisation in case of Reactor flows
Increase `receive()` timeout in the `AttributePollingChannelAdapterParserTests`
JIRA: https://jira.spring.io/browse/INT-4301
To keep the context of the `failedMessage` and avoid extra wrapping
to the `MessagingException` make
`CircuitBreakerOpenException extends MessagingException` directly
JIRA: https://jira.spring.io/browse/INT-4280
To let end-user easily to inject custom `TaskScheduler` to the endpoint
add `ConsumerEndpointSpec.taskScheduler()` option
* Address PR comments
