命令调度器
Command Dispatchers
最后更新于
The pages provide the background on how to handle command messages in your application. The dispatching process is the starting point of such a command message. Axon provides two interfaces you can use to send the commands to your command handlers, being:
The Command Bus, and
The Command Gateway
This page will show how and when to use the command gateway and bus. How to configure and specifics on the command gateway and bus implementations are discussed in the section.
The 'Command Bus' is the mechanism that dispatches commands to their respective command handlers. As such it is the infrastructure component that is aware which component can handle which command.
Each command is always sent to exactly one command handler. If no command handler is available for the dispatched command, a NoHandlerForCommandException
exception is thrown.
The CommandBus
provides two methods to dispatch commands to their respective handler, being the dispatch(CommandMessage)
and dispatch(CommandMessage, CommandCallback)
methods:
The CommandDispatcher
described above exemplifies a couple of important aspects and capabilities of the dispatching commands:
The CommandBus
interface providing the functionality to dispatch command messages.
The aggregate identifier is, per best practice, initialized as the String of a random unique identifier.
Typed identifier objects are also possible, as long as the object implements a sensible toString()
function.
The GenericCommandMessage#asCommandMessage(Object)
method is used to create a CommandMessage
.
To be able to dispatch a command on the CommandBus
, you are required to wrap your own command object (e.g. the 'command message payload') in a CommandMessage
.
The CommandBus#dispatch(CommandMessage)
function will dispatch the provided CommandMessage
on the bus, for delivery to a command handler.
If an application isn't directly interested in the outcome of a command, this method can be used.
If the outcome of command handling is relevant for your application, the optional second parameter can be provided, the CommandCallback
.
The CommandCallback
allows the dispatching component to be notified when command handling is completed.
The Command Callback has one function, onResult(CommandMessage, CommandResultMessage)
, which is called when command handling has finished.
The first parameter is the dispatched command, whilst the second is execution result of the dispatched command.
Lastly, the CommandCallback
is a 'functional interface' due to onResult
being its only method.
As such, commandBus.dispatch(commandMessage, (cmdMsg, commandResultMessage) -> { /* ... */ })
would also be possible.
The CommandResultMessage
provides the API to verify whether command execution was exceptional or successful.
If CommandResultMessage#isExceptional
returns true, you can assume that the CommandResultMessage#exceptionResult
will return a Throwable
instance containing the actual exception.
Otherwise, the CommandResultMessage#getPayload
method may provide you with an actual result or null
, as further specified here.
Command Callback consideration
In the case that
dispatch(CommandMessage, CommandCallback)
is used, the calling component may not assume that the callback is invoked in the same thread that dispatched the command. If the calling thread depends on the result before continuing, you can use theFutureCallback
. TheFutureCallback
is a combination of aFuture
(as defined in the java.concurrent package) and Axon'sCommandCallback
. Alternatively, consider using aCommandGateway
.
The 'Command Gateway' is a convenience approach towards dispatching commands. It does so by abstracting certain aspects for you when dispatching a command on the CommandBus
. It this uses the CommandBus
underneath to perform the actual dispatching of the message.
While you are not required to use a gateway to dispatch commands, it is generally the easiest option to do so.
The CommandGateway
interface can be separated in two sets of methods, namely send
and sendAndWait
:
The send
API as shown above introduces a couple of concepts, marked with numbered comments:
The CommandGateway
interface providing the functionality to dispatch command messages.
The aggregate identifier is, per best practice, initialized as the String of a random unique identifier.
Typed identifier objects are also possible, as long as the object implements a sensible toString()
function.
The send(Object)
function requires a single parameter, the command object.
This is an asynchronous approach to dispatching commands.
As such the response of the send
method is a CompletableFuture
.
Callback when using
send(Object)
The
CommandGateway#send(Object)
method uses theFutureCallback
under the hood to unblock the command dispatching thread from the command handling thread.
A synchronous approach to sending messages can also be achieved, by using the sendAndWait
methods:
The CommandGateway#sendAndWait(Object)
function takes in a single parameter, your command object.
It will wait indefinitely until the command dispatching and handling process has been resolved.
If waiting indefinitely is not desirable, a 'timeout' paired with the 'time unit' can be provided along side the command object.
Doing so will ensure that the command dispatching thread will not wait longer than specified.
If command dispatching/handling was interrupted or the timeout was reached whilst using this approach, the command result will be null
.
Dispatching commands will, generally speaking, have two possible outcomes:
Command handled successfully, and
command handled exceptionally
The outcome to some extent depends on the dispatching process, but more so on the implementation of the command handler. Thus if the @CommandHandler
annotated function throws an exception due to some business logic, it will be that exception which will be the result of dispatching the command.
The successful resolution of command handling intentionally should not provide any return objects. Thus, if the CommandBus
/CommandGateway
provides a response (either directly or through the CommandResultMessage)
, then you should assume the result of successful command handling to return null
.
The CommandMessage
also allows the addition of to the Command Message.
It does so by internally leveraging the CommandBus
interface .
This allows for chaining of follow up operations after the command has been returned.
The result returned by this method can either be successful or exceptional, as will be explained .
In all other scenarios, the result follows the approach.
While it is possible to return results from command handlers, this should be used sparsely. The intent of the Command should never be to retrieve a value, as that would be an indication that the message should be designed as a . Exceptions to this would be the identifier of the Aggregate Root, or identifiers of entities the Aggregate Root has instantiated. The framework has one such exception build in, on the @CommandHandler
annotated constructor of an Aggregate. In case the 'command handling constructor' has executed successfully, instead of the Aggregate itself, the value of the @AggregateIdentifier
annotated field will be returned.