This allows re-use of existing MethodParameter instances from controller
methods with cached metadata, and also ensures additional capabilities
such as looking up parameter annotations on interfaces.
See gh-29825
This commit ensures that any storage used for multipart handling only
gets cleaned up if multipart data is actually retrieved via
ServerWebExchange::getMultipartData.
Closes gh-30590
Prior to this commit, the Observation instrumentation for Reactive
server applications was implemented with a `WebFilter`. This allowed to
record observations and set up a tracing context for the controller
handlers.
The limitation of this approach is that all processing happening at a
lower level is not aware of any observation. Here, the
`HttpWebHandlerAdapter` handles several interesting aspects:
* logging of HTTP requests and responses at the TRACE level
* logging of client disconnect errors
* handling of unresolved errors
With the current instrumentation, these logging statements will miss the
tracing context information. As a result, this commit deprecates the
`ServerHttpObservationFilter` in favor of a more direct instrumentation
of the `HttpWebHandlerAdapter`. This enables a more precise
instrumentattion and allows to set up the current observation earlier in
the reactor context: log statements will now contain the relevant
information.
Fixes gh-30013
Prior to this commit, some FreeMarker tests would fail when involving
property lookup in a template file, where this bean property is linked
with a method implemented as a default method on an interface.
While I did not manage to reproduce this behavior in an independent test
case, this is most likely related to a change in JDK 21:
https://bugs.openjdk.org/browse/JDK-8071693
This commit changes the template expression to using the default method
directly.
* use forEach and putIfAbsent to copy headers in DefaultClientRequestBuilder
* use forEach in ReactorClientHttpRequest and ReactorNetty2ClientHttpRequest
* circumvent ReadOnlyHttpHeaders.entrySet()
* ensure the fast path to LinkedCaseInsensitiveMap for forEach and putIfAbsent exists
Closes gh-29972
Prior to this commit, `WebClient` observations would be recorded as
aborted (with tags "outcome":"UNKNOWN", "status":"CLIENT_ERROR")
for use cases like this:
```
Flux<String> result = client.get()
.uri("/path")
.retrieve()
.bodyToFlux(String.class)
.take(1);
```
This is due to operators like `take` or `next` that consume *some*
`onNext` signals and then cancels the subscription before completion.
This means the subscriber is only partially interested in the response
and we should not count this as a client error.
This commit ensures that observations are only recorded as aborted if
the response was not published at the time the CANCEL signal was
received.
The code snippet above will now publish observations with
"outcome":"SUCCESS" and "status":"200" tags, for example.
Closes gh-30070
This commit refactors some AssertJ assertions into more idiomatic and
readable ones. Using the dedicated assertion instead of a generic one
will produce more meaningful error messages.
For instance, consider collection size:
```
// expected: 5 but was: 2
assertThat(collection.size()).equals(5);
// Expected size: 5 but was: 2 in: [1, 2]
assertThat(collection).hasSize(5);
```
Closes gh-30104
This commit documents the fact that default status handlers configured
on the `WebClient` are not applied to `exchangeTo*` methods as those
variants give full access to the client response.
Applying them here would restrict the ability to adapt the behavior
depending on the HTTP response status.
Closes gh-30059
Prior to this commit, an error thrown by a `ExchangeFilterFunction`
configured on a `WebClient` instance would be recorded as such by the
client observation, but the response details would be missing from the
observation.
All filter functions and the exchange function (performing the HTTP
call) would be merged into a single `ExchangeFunction`; this instance
was instrumented and osberved. As a result, the instrumentation would
only get the error signal returned by the filter function and would not
see the HTTP response even if it was received. This means that the
recorded observation would not have the relevant information for the
HTTP status.
This commit ensures that between the configured `ExchangeFilterFunction`
and the `ExchangeFunction`, an instrumentation `ExchangeFilterFunction`
is inserted. This allows to set the client response to the observation
context, even if a later error signal is thrown by a filter function.
Note that with this change, an error signal sent by a filter function
will be still recorded in the observation.
See gh-30059
This commit updates AbstractMessageWriterResultHandler#writeBody in
order to use the declared bodyParameter instead of
ResolvableType.forInstance(body) when the former has unresolvable
generics.
Closes gh-30214
This commit turns some stream-based iterations back into simpler
enhanced for loops.
For simple use cases like these, where the stream API is merely used to
map/filter + collect to a List, a for loop is more efficient.
This is especially true for small collections like the ones we deal
with in BodyInserters/BodyExtractors here (in the order of 50ns/op vs
5ns/op). These cases are also simple enough that they don't lose in
readability after the conversion.
Closes gh-30136
Prior to this commit, the `RequestedContentTypeResolverBuilder` would
create a `RequestedContentTypeResolver` that internally delegates to a
list of resolvers. Each resolver would either return the list of
requested media types, or a singleton list with the "*/*" media type; in
this case this signals that the resolver cannot find a specific media
type requested and that we should continue with the next resolver in the
list.
Media Types returned by resolvers can contain parameters, such as the
quality factor. If the HTTP client requests "*/*;q=0.8", the
`HeaderContentTypeResolver` will return this as a singleton list. While
this has been resolved from the request, such a media type should not be
selected over other media types that could be returned by other
resolvers.
This commit changes the `RequestedContentTypeResolverBuilder` so that it
does not select "*/*;q=0.8" as the requested media type, but instead
continues delegating to other resolvers in the list. This means we need
to remove the quality factor before comparing it to the "*/*" for
equality check.
Fixes gh-29915
Juergen Hoeller
rstoyanchev
Arjen Poutsma
Sam Brannen
Sébastien Deleuze
galingerv
Brian Clozel
James Yuzawa
Krzysztof Krasoń
Vatsa
Harry Yang