This commit introduce a change of behaviour when component scan is used
with conditions. Previously, any condition in the REGISTER_BEAN phase
were ignored and the scan was applied regardless of the outcome of
those conditions. This is because REGISTER_BEAN condition evaluation
happens later in the bean factory preparation.
Rather than ignoring those conditions, this commit fails fast when it
detects such use case. Code will have to be adapted accordingly.
Closes gh-23206
This commit fixes an issue where a Cacheable method which returns a
Flux (or multi-value publisher) will be invoked once, but the returned
publisher is actually subscribed twice.
The previous fix 988f3630c would cause the cached elements to depend on
the first usage pattern / request pattern, which is likely to be too
confusing to users. This fix reintroduces the notion of exhausting the
original Flux by having a second subscriber dedicated to that, but uses
`refCount(2)` to ensure that the original `Flux` returned by the cached
method is still only subscribed once.
Closes gh-32370
This commit fixes an issue where a Cacheable method which returns a
Flux (or multi-value publisher) will be invoked once, but the returned
publisher is actually subscribed twice.
By using the Reactor `tap` operator, we ensure that we can emit values
downstream AND accumulate emitted values into the List with a single
subscription.
The SignalListener additionally handles scenarios involving cancel,
for instance in case of a `take(1)` in the chain. In that case values
emitted up until that point will have been stored into the List buffer,
so we can still put it in the cache. In case of error, no caching occurs
and the internal buffer is cleared. This implementation also protects
against competing onComplete/onError signals and cancel signals.
Closes gh-32370
This commit removes the previous implementation in favor of the new
PlaceholderParser. The only noticeable side effect is that the exception
is no longer an IllegalArgumentException, but rather the dedicated
PlaceholderResolutionException.
See gh-9628
To improve consistency and avoid confusion regarding primitive types
and their wrapper types, this commit ensures that we always use class
literals for primitive types.
For example, instead of using the `Void.TYPE` constant, we now
consistently use `void.class`.
