[
https://issues.apache.org/jira/browse/CAMEL-13287?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Claus Ibsen resolved CAMEL-13287.
---------------------------------
Resolution: Fixed
> AggregationStrategy - Access original exchange in aggregate method
> -------------------------------------------------------------------
>
> Key: CAMEL-13287
> URL: https://issues.apache.org/jira/browse/CAMEL-13287
> Project: Camel
> Issue Type: New Feature
> Components: camel-core
> Reporter: Balazs Szeti
> Assignee: Claus Ibsen
> Priority: Major
> Fix For: 3.0.0, 3.0.0.M5
>
>
> For aggregation after multicast/splitter the original exchange should
> optionally be available in the aggregate method.
> There are several use cases when we would like to go-on processing the
> original exchange after multicast, but we'd like to enrich it with the
> outcome of the called routes. For example:
> {code:java}
> rest()
> .get("orders/{orderId}")
> .route()
> .to("direct:getOrderDetails") //get UserId, ItemId
> .setBody(method(this,"createResponsePojo"))
> .multicast(new MyAggregationStrategy())
> .to("direct:getUserDetails")
> .to("direct:getDeliveryAddress")
> .to("direct:getItemDetails")
> .end()
> ;
> {code}
> If any of the called routes fail, we still would like return a partial
> response in our service (this is a common requirement in case of
> microservices). The MyAggregationStrategy should simply enrich the
> ResponePojo object from the original exchange somehow with the new exchanges
> coming from the sub-routes. See this example:
> {code:java}
> public class MyAggregationStrategy implements AggregationStrategy {
> public Exchange aggregateWithOriginal(Exchange oldExchange, Exchange
> newExchange, Exchange originalExchange) {
> Exchange exchange = oldExchange != null ? oldExchange : originalExchange;
> ResponsePojo response = exchange.getMessage().getBody(ResponsePojo.class);
> if (! newExchange.isFailed()) {
> // ... Add newExchange body somehow to ResponsePojo object...
> }
> return exchange;
> }
> ...
> }
> {code}
> Currently only the exchanges from the "sub-routes" are available during
> aggregation, so the exchange after the aggregate will be one (the first) of
> those. This comes with multiple problems:
> * Though the exchanges in the sub-routes are copies of the original
> exchange, sub routes make modifications: modify headers, modify properties,
> etc. Usually we don't want to see all these set by a sub-route on the final
> aggregated exchange. It's only noise. "Whatever happens in the sub-route,
> should stay in the sub-route." - We only want to see on the aggregated
> exchange what we "took" intentionally from the sub-route exchanges.
> * If we use stopOnException(true) our life is simple because we usually
> don't have to worry about exceptions in aggregate, we will stop anyway. The
> aggregation logic can become complicated if we want to go on with processing
> in case of errors. The first time aggregate() is called the oldExchange is
> null, so we usually take the newExchange as the return value. If this
> exchange has an Exception, we need to "clean" it first, otherwise the error
> handler will kick in after aggregation. This is non-trivial.
> h3. Suggested approach
> Let's extend the AggregationStrategy interface with a new method that takes
> three exchanges. This should be called after Multicast EIP (Enrich EIP is
> simple, it only has two exchanges).
> With a default implementations we can keep the interface compatible:
> {code:java}
> public interface AggregationStrategy {
> /**
> * Aggregates an old, a new and the original exchange together to create a
> single combined exchange.
> *
> * @param oldExchange the oldest exchange, which is the returned value of the
> previous aggregation on null.
> * @param newExchange the newest exchange
> * @param originalExchange the original exchange before Multicast or Splitter
> EIP. Null in case of Enrich EIP.
> * @return a combined exchange, favor returning the oldExchange
> */
> default Exchange aggregateWithOriginal(Exchange oldExchange, Exchange
> newExchange, Exchange originalExchange) {
> return aggregate(oldExchange, newExchange);
> };
> //Maybe we should have a default implementation here too so one can only
> implement aggregateWithOriginal()
> Exchange aggregate(Exchange oldExchange, Exchange newExchange);
> ...
> }
> {code}
> {code:java}
> {code}
--
This message was sent by Atlassian JIRA
(v7.6.14#76016)
