Stop truncating Flux results to first element with @⁠CacheEvict

Prior to this commit, ReactiveCachingHandler.processCacheEvicts()
adapted every reactive return value via Mono.from(), which subscribes
for only the first element and cancels the upstream Publisher. For a
@⁠CacheEvict method that returns a Flux, this silently truncated the
returned sequence to its first element. In addition, the `#result`
variable in `condition` SpEL expressions was bound to only that first
emitted element.

To address that, this commit mirrors the existing multi-value handling
in processPutRequest(). When the adapter reports isMultiValue(), a side
Subscriber is subscribed via publish().refCount(2) that exhausts the
Flux and collects its values into a List for eviction, while the
original, unmodified Flux is returned to the caller. Consequently, the
`#result` variable in `condition` SpEL expressions for a Flux-returning
@⁠CacheEvict method is now the full List of emitted elements rather
than just the first element, making it consistent with @⁠Cacheable and
@⁠CachePut.

This commit also improves spr14235AdaptsToReactorFlux() in
CacheReproTests. Previously it exercised @⁠CacheEvict only with a
single-element Flux and never asserted on the returned sequence. Now it
uses a multi-element Flux and verifies that all elements are both
returned to the caller and visible to the `condition` expression.

Last but not least, this commit documents the aforementioned
`#result`/Flux semantics in @⁠CacheEvict's Javadoc and in the reference
manual, since both were previously invalid or incomplete for this
scenario.

Closes gh-37309
This commit is contained in:
Sam Brannen
2026-09-21 17:16:34 +02:00
parent 9637d12785
commit 26de340102
4 changed files with 66 additions and 11 deletions
@@ -381,10 +381,12 @@ available to the context so that you can use them for key and conditional comput
| `result`
| Evaluation context
| The result of the method call (the value to be cached). Only available in `unless`
expressions, `cache put` expressions (to compute the `key`), or `cache evict`
expressions (when `beforeInvocation` is `false`). For supported wrappers (such as
`Optional`), `#result` refers to the actual object, not the wrapper.
| The result of the method call (the value to be cached, or evaluated for eviction). Only
available in `unless` expressions, `cache put` expressions (`key`, `condition`, or
`unless`), or `cache evict` expressions (`key` or `condition`, when `beforeInvocation`
is `false`). For supported wrappers (such as `Optional`), `#result` refers to the
actual object, not the wrapper. For a method returning a `Flux`, `#result` refers to a
`List` containing all values collected from the `Flux`.
| `#result`
|===
@@ -464,7 +466,12 @@ not the case with `@Cacheable` which adds data to the cache or updates data in t
and, thus, requires a result.
As of 6.1, `@CacheEvict` takes `CompletableFuture` and reactive return types into account,
performing an after-invocation evict operation whenever processing has completed.
performing an after-invocation evict operation whenever processing has completed. As with
`@Cacheable` and `@CachePut`, for a method returning a `Flux`, all emitted elements are
collected into a `List` before the evict operation is performed. When `beforeInvocation`
is `false`, that same `List` is what the `condition` SpEL expression sees as the
`#result`. Either way, the `Flux` returned to the caller is unaffected by this collection
process.
TIP: When `@CacheEvict` is combined with `@Retryable`, the retry advice is applied
outermost, so eviction runs again on every retry attempt -- and, with
@@ -77,7 +77,8 @@ public @interface CacheEvict {
* <li>{@code #result} for a reference to the result of the method invocation, which
* can only be used if {@link #beforeInvocation()} is {@code false}. For supported
* wrappers such as {@code Optional}, {@code #result} refers to the actual object,
* not the wrapper</li>
* not the wrapper. For a method that returns a {@code Flux}, {@code #result} refers
* to a {@code List} containing all values collected from the {@code Flux}.</li>
* <li>{@code #root.method}, {@code #root.target}, and {@code #root.caches} for
* references to the {@link java.lang.reflect.Method method}, target object, and
* affected cache(s) respectively.</li>
@@ -123,6 +124,11 @@ public @interface CacheEvict {
* <p>The SpEL expression evaluates against a dedicated context that provides the
* following meta-data:
* <ul>
* <li>{@code #result} for a reference to the result of the method invocation, which
* can only be used if {@link #beforeInvocation()} is {@code false}. For supported
* wrappers such as {@code Optional}, {@code #result} refers to the actual object,
* not the wrapper. For a method that returns a {@code Flux}, {@code #result} refers
* to a {@code List} containing all values collected from the {@code Flux}.</li>
* <li>{@code #root.method}, {@code #root.target}, and {@code #root.caches} for
* references to the {@link java.lang.reflect.Method method}, target object, and
* affected cache(s) respectively.</li>
@@ -1105,6 +1105,39 @@ public abstract class CacheAspectSupport extends AbstractCacheInvoker
}
/**
* Reactive Streams Subscriber for exhausting the Flux and collecting a List
* to evaluate for eviction.
*/
private final class CacheEvictListSubscriber implements Subscriber<Object> {
private final List<CacheOperationContext> contexts;
private final List<Object> cacheValue = new ArrayList<>();
public CacheEvictListSubscriber(List<CacheOperationContext> contexts) {
this.contexts = contexts;
}
@Override
public void onSubscribe(Subscription s) {
s.request(Integer.MAX_VALUE);
}
@Override
public void onNext(Object o) {
this.cacheValue.add(o);
}
@Override
public void onError(Throwable t) {
this.cacheValue.clear();
}
@Override
public void onComplete() {
performCacheEvicts(this.contexts, this.cacheValue);
}
}
/**
* Inner class to avoid a hard dependency on the Reactive Streams API at runtime.
*/
@@ -1186,8 +1219,16 @@ public abstract class CacheAspectSupport extends AbstractCacheInvoker
public @Nullable Object processCacheEvicts(List<CacheOperationContext> contexts, @Nullable Object result) {
ReactiveAdapter adapter = (result != null ? this.registry.getAdapter(result.getClass()) : null);
if (adapter != null) {
return adapter.fromPublisher(Mono.from(adapter.toPublisher(result))
.doOnSuccess(value -> performCacheEvicts(contexts, value)));
if (adapter.isMultiValue()) {
Flux<?> source = Flux.from(adapter.toPublisher(result))
.publish().refCount(2);
source.subscribe(new CacheEvictListSubscriber(contexts));
return adapter.fromPublisher(source);
}
else {
return adapter.fromPublisher(Mono.from(adapter.toPublisher(result))
.doOnSuccess(value -> performCacheEvicts(contexts, value)));
}
}
return NOT_HANDLED;
}
@@ -56,6 +56,7 @@ import static org.mockito.Mockito.verify;
* @author Phillip Webb
* @author Juergen Hoeller
* @author Stephane Nicoll
* @author Sam Brannen
*/
class CacheReproTests {
@@ -312,7 +313,7 @@ class CacheReproTests {
assertThat(bean.findById("tb1").collectList().block()).isEqualTo(tb);
assertThat(cache.get("tb1").get()).isEqualTo(tb);
bean.clear().blockLast();
assertThat(bean.clear().collectList().block()).containsExactly(1, 2, 3);
List<TestBean> tb2 = bean.findById("tb1").collectList().block();
assertThat(tb2).isNotEmpty();
assertThat(tb2).isNotEqualTo(tb);
@@ -705,9 +706,9 @@ class CacheReproTests {
return Flux.fromIterable(item);
}
@CacheEvict(cacheNames = "itemCache", allEntries = true, condition = "#result > 0")
@CacheEvict(cacheNames = "itemCache", allEntries = true, condition = "#result == {1, 2, 3}")
public Flux<Integer> clear() {
return Flux.just(1);
return Flux.just(1, 2, 3);
}
}