The previous implementation uses ArrayList for storing resolved
resources and ArrayList has O(n) time complexity for the contains
operation. By switching to the HashSet for storing resolved
resources we improve the time complexity of this operation to be O(1).
See gh-25927
There are more locations which could benefit from not using a
toCharArray on a String, but rather use the charAt method from
the String itself. This to prevent an additional copy of the
char[] being created.
This commit introduces `and()` default methods in the MethodFilter and
FieldFilter functional interfaces in ReflectionUtils in order to simplify
uses cases that need to compose filter logic.
Closes gh-26063
Prior to this commit, MergedAnnotationCollectors.toAnnotationSet()
created an intermediate ArrayList for storing the results prior to
creating a LinkedHashSet in the finishing step.
Since the creation of the intermediate list is unnecessary, this commit
simplifies the implementation of toAnnotationSet() by using the
Collector.of() factory method that does not accept a `finisher` argument.
The resulting Collector internally uses a `castingIdentity()` function
as the `finisher`.
Closes gh-26031
The migration from JUnit 4 assertions to AssertJ assertions resulted in
several unnecessary casts from int to long that actually cause
assertions to pass when they should otherwise fail.
This commit fixes all such bugs for the pattern `.isNotEqualTo((long)`.
The existing implementation was exposed to very poor performance when matching
with multiple delimiters against a large buffer with many delimiters. In that
case all matchers are invoked many times (as many as the number of delimiters)
even though some of them found no match at all on the first pass.
The revised implementation uses a single index and advances all matchers
together, checking one byte a time, and not letting any one of them search to
the end of the entire buffer on a single pass.
Closes gh-25915
Simplify and optimize the processing of the input stream.
The existing implementation was using bufferUntil and creating a List
for every line along with an EndFrameBuffer inserted for the
bufferUntil predicate. So the larger the input buffer and the more
lines it contained, the greater the overhead.
The new implementation avoids bufferUntil for all lines and
instead uses concatMapIterable to aggregate the lines from a buffer
into a single list. So the larger the input buffer and the more
lines it contains, the better the throughput. The only buffering
used then is for partial chunks and those are accumulated in a list.
See gh-25915
Before this commit, ReadCompletionHandler delayed closing the channel
to allow an ongoing read to complete/fail so we can get a hold of the
associated DataBuffer and release it. This can be problematic since the
read take time to complete but even more importantly there was a race
condition where we didn't check if we've been disposed concurrently
while releasing the read flag.
This commit removes the delay and closes the channel immediately from
cancel() and that should in turn fail any ongoing read operation,
according to AsynchronousChannel, and the DataBuffer is released.
Further improvements include:
- combining the "reading" and "disposed" AtomicBoolean's into a
single "state" AtomicReference.
- an optimistic check to remain in READING mode and avoid state
switches when there is demand to continue reading.
Closes gh-25831
This commit raises the minimum Coroutines version supported
to 1.4.0-M1 and above, and changes usages of awaitFirst() or
awaitFirstOrNull() to awaitSingle() or awaitSingleOrNull()
to fix gh-25007.
Closes gh-25914
Closes gh-25007
This commit adds support for Kotlin Coroutines suspending functions to
Spring MVC, by converting those to a Mono that can then be handled by
the asynchronous request processing feature.
It also optimizes Coroutines detection with the introduction of an
optimized KotlinDetector.isSuspendingFunction() method that does not
require kotlin-reflect.
Closes gh-23611
Now that there's a new JMH infrastructure for benchmarks and that
performance tests have been rewritten to use it, we should remove the
`PERFORMANCE` `TestGroup` to avoid introducing such tests in the future.
Closes gh-24830
We are passing custom values to the constructor of PooledByteBufAllocator in
tests in order to turn of caching. This is based on:
https://github.com/netty/netty/issues/5275#issuecomment-220547057
Netty 4.1.52 has significant changes in PooledByteBufAllocator:
https://github.com/netty/netty/pull/10267
After the changes, our current value for maxOrder=2, which results in
chunkSize=16K, causes an assert failure in PoolChunk where the runSize
exceeds the chunkSize.
Prior to this commit, the subtype suffix of a MIME type (see RFC 6839)
was not properly taken into account when checking compatibility between
MIME types.
For example, `"application/*"` was not considered as compatible with
`"application/vnd.io.spring+json"`.
This commit adds a new `MimeType#getSubtypeSuffix()` method to easily
extract the subtype suffix information. This method is then reused in
the `isCompatibleWith` implementation to better handle these cases.
Fixes gh-25350
Renaming the `tags` method to `getTags` for overall consistency, and
fixing the Javadoc to mention that this returns an immutable collection.
Closes gh-25678
This commit declares each of the following public interfaces as a
@FunctionalInterface.
- org.springframework.context.ApplicationContextInitializer
- org.springframework.test.web.servlet.DispatcherServletCustomizer
- org.springframework.validation.MessageCodeFormatter
- org.springframework.util.IdGenerator
- org.springframework.beans.factory.config.YamlProcessor.MatchCallback
- org.springframework.beans.factory.config.YamlProcessor.DocumentMatcher
Closes gh-25580
Prior to this commit, if null values were supplied for the
RepeatableContainers or AnnotationFilter arguments to `from()` factory
methods in MergedAnnotations, certain operations would eventually
result in a NullPointerException. This is to be expected; however, the
NullPointerException is often swallowed and only logged at INFO level
with an exception message similar to the following.
> Failed to introspect annotations on org.example.MyClass: NullPointerException
In such cases, the INFO log message is not helpful in diagnosing the
problem. Furthermore, since the exception is swallowed, the desired
operation (e.g., MergedAnnotations.stream(...)) simply returns no
results.
This commit improves the user experience by eagerly asserting non-null
preconditions for required arguments in MergedAnnotations.from()
factory methods.
Closes gh-25568
PR gh-25429 brought it to our attention that there was a bug in
AnnotationScanner when using a non-null class filter that filtered out
classes; however, it turns out that there is no production code that
utilizes the package-private class filtering support.
This commit therefore removes all class filtering support from
AnnotationScanner since that functionality is effectively unused.
Closes gh-25477
Prior to this commit, scanning for annotations resulted in an infinite
loop when using the INHERITED_ANNOTATIONS search strategy and a class
filter that filters out visited classes.
This commit avoids an infinite loop in AnnotationsScanner's
processClassInheritedAnnotations(...) method by skipping the current
level of the class hierarchy when the current source class has been
filtered out.
Closes gh-25429
Prior to this commit, a Profiles instance created via Profiles.of() was
not considered equivalent to another Profiles instance created via
Profiles.of() with the exact same expressions. This makes it difficult
to mock invocations of Environment#acceptsProfiles(Profiles) -- for
example, when using a mocking library such as Mockito.
This commit makes Profiles instances created via Profiles.of()
"comparable" by implementing equals() and hashCode() in ParsedProfiles.
Note, however, that equivalence is only guaranteed if the exact same
profile expression strings are supplied to Profiles.of(). In other
words, Profiles.of("A & B", "C | D") is equivalent to
Profiles.of("A & B", "C | D") and Profiles.of("C | D", "A & B"), but
Profiles.of("X & Y") is not equivalent to Profiles.of("X&Y") or
Profiles.of("Y & X").
Closes gh-25340
Prior to this commit, SocketUtils used System.currentTimeMillis() for
the seed for the java.util.Random instance used internally. The use of
the milliseconds value returned by currentTimeMillis() can lead to
collisions for randomly selected free ports for tests executing in
parallel on the same computer.
This commit therefore switches to System.nanoTime() for the Random seed
used in SocketUtils in an attempt to avoid such collisions for tests
executing in parallel in different JVMs on the same computer.
Closes gh-25321
This commit removes load time weaving, CGLIB and Objenesis support
from native images.
GraalDetector has been removed for now because of
https://github.com/oracle/graal/issues/2594. It should be reintroduced
when this bug will be fixed with NativeImageDetector class name.
Closes gh-25179
This commit introduces a spring.xml.ignore system property
which when set to true avoid initializing XML infrastructure.
A typical use case is optimizing GraalVM native image footprint
for applications not using XML. In order to be effective, those
classes should be initialized at build time:
- org.springframework.util.DefaultPropertiesPersister
- org.springframework.core.io.support.PropertiesLoaderUtils
- org.springframework.web.servlet.function.support.RouterFunctionMapping
- org.springframework.web.client.RestTemplate
- org.springframework.web.servlet.mvc.method.annotation.ExceptionHandlerExceptionResolver
- org.springframework.web.servlet.config.annotation.WebMvcConfigurationSupport
- org.springframework.web.servlet.mvc.method.annotation.RequestMappingHandlerAdapter
- org.springframework.http.converter.support.AllEncompassingFormHttpMessageConverter
- org.springframework.http.codec.support.BaseDefaultCodecs
- org.springframework.beans.PropertyEditorRegistrySupport
Closes gh-25151
Prior to this commit, a change to Javadoc in any class in spring-core
would result in ALL tests in the entire test suite being re-run via the
Gradle build.
Thanks to a tip from @melix, this commit aims to ensure that the
spring-core JAR is "reproducible".
Prior to this commit, a change to Javadoc in any class in spring-core
would result in ALL tests in the entire test suite being re-run via the
Gradle build.
Thanks to a tip from @melix, this commit ensures that the spring-core
JAR is "reproducible".
Commit 4000b244ff introduced new variants for noNullElements in
org.springframework.util.Assert.
This commit adds the corresponding tests to AssertTests.
Closes gh-25239
This commit avoids duplicate addition of void.class to the temporary
primitiveTypes Set in the static initialization block in ClassUtils.
Closes gh-25128
Prior to this commit (and the previous one), the `AntPathStringMatcher`
(inner class of `AntPathmatcher`) would compile `Pattern` instances and
use regex matching even for static patterns such as `"/home"`.
This change introduces a shortcut in the string matcher algorithm to
skip the `Pattern` creation and uses `String` equality instead.
Static patterns are quite common in applications and this change can
bring performance improvements, depending on the mix of patterns
configured in the web application.
In benchmarks (added with this commit), we're seeing +20% throughput
and -40% allocation. This of course can vary depending on the number of
static patterns configured in the application.
Closes gh-24887
Prior to this commit, `MediaType.parseMediaType` would already rely on
the internal LRU cache in `MimeTypeUtils` for better performance. With
that optimization, the parsing of raw media types is skipped for cached
elements.
But still, `MediaType.parseMediaType` would first get a cached
`MimeType` instance from that cache and then instantiate a
`new MediaType(type, subtype, parameters)`. This constructor not only
replays the `MimeType` checks on type/subtyme tokens and parameters, but
it also performs `MediaType`-specific checks on parameters.
Such checks are not required, as we're using an existing `MimeType`
instance in the first place.
This commit adds a new protected copy constructor (skipping checks) in
`MimeType` and uses it in `MediaType.parseMediaType` as a result.
This yields interesting performance improvements, with +400% throughput
and -40% allocation/call in benchmarks. This commit also introduces a
new JMH benchmark for future optimization work.
Closes gh-24769
Prior to this commit, the loadFactoryNames() and loadFactories() methods
in SpringFactoriesLoader effectively used a different default
ClassLoader.
This commit ensures that the ClassLoader for SpringFactoriesLoader.class
is now consistently used as the default ClassLoader.
Closes gh-24992
This commit ensures that the static cache in SpringFactoriesLoader
contains unmodifiable lists of unique names of factory implementations.
See gh-24985
Prior to this commit, SpringFactoriesLoader discovered all registered
factory implementations for a given factory type even if duplicates
were registered within a single META-INF/spring.factories file or in
multiple such files in the classpath.
This commit updates the internals of SpringFactoriesLoader so that
duplicate registrations are ignored, thereby aligning with the
well-known semantics for java.util.ServiceLoader in this regard.
Closes gh-24985
- Build Scan plugin is now Gradle Enterprise plugin applied in settings
- Compile task dependencies are now defined through classpath
- Test fixture publication can be disabled through public API
Closes gh-24384
This commit moves the toString() implementation for synthesized
annotations from TypeMappedAnnotation to
SynthesizedMergedAnnotationInvocationHandler in order to take advantage
of the synthesized annotation attribute value cache introduced in
72b1abd226.
Closes gh-24970
Prior to the introduction of the MergedAnnotation API in Spring Framework
5.2, our SynthesizedAnnotationInvocationHandler utilized a cache for
annotation attribute values; whereas, the new
SynthesizedMergedAnnotationInvocationHandler has no such caching in
place.
Issues such as gh-24961 indicate a regression in performance caused by
the lack of such an attribute value cache. For example, the required
attribute in @RequestParam is looked up using the internal meta-model
in the MergedAnnotation API twice per request for each @RequestParam in
a given controller handler method.
This commit reintroduces the attribute value cache to avoid the
unnecessary performance overhead associated with multiple lookups of
the same attribute in a synthesized annotation. This applies not only
to direct attribute method invocations but also to invocations of
equals() and hashCode() on a synthesized annotation.
Note, however, that this commit does NOT address multiple lookups of
annotation attribute values for invocations of toString(). That behavior
currently remains unchanged in the implementation of
org.springframework.core.annotation.TypeMappedAnnotation.toString().
Closes gh-24970
Commit 31fa1569c5 introduced initial support for avoiding unnecessary
annotation synthesis in the MergedAnnotation API; however, it only
avoided synthesis for annotations that do not declare any attributes.
This commit reworks this support to avoid unnecessary annotation
synthesis for annotations that declare attributes.
Specifically, this commit introduces a new `isSynthesizable()` method
in AnnotationTypeMapping that allows the "synthesizable" flag to be
computed once and cached along with the other metadata already cached
in AnnotationTypeMapping instances. TypeMappedAnnotation now delegates
to this new method when determining whether it should synthesize an
annotation.
Closes gh-24861
Prior to Spring Framework 5.2, some of our annotation utilities would
not synthesize an annotation if it was already synthesized or not
synthesizable (i.e., did not declare local aliases via @AliasFor and
did not declare attributes that could override attributes in the
meta-annotation hierarchy above the given annotation); however, we lost
most of this functionality with the introduction of the new
MergedAnnotations API.
This commit revises the implementation of createSynthesized() in
TypeMappedAnnotation so that, for invocations of
MergedAnnotation.synthesize() and indirectly for invocations of
AnnotatedElementUtils.findMergedAnnotation(), etc.:
1. An annotation that was previously synthesized will not be
synthesized again.
2. An annotation that is not "synthesizable" will not be synthesized.
For both of the above use cases, the original annotation is now
returned from createSynthesized().
Closes gh-24861
Prior to this commit, the `MimeTypeUtils` LRUCache would maintain a
queue to track least recently used cached values. In some cases,
concurrent access could create more entries in that unbounded queue than
expected and spend a significant amount of time removing entries in that
queue (i.e. iterating over a long list of elements).
This commit ensures that recently used entries are only added to the
queue if they've been removed by the current thread, in case of
concurrent access.
Fixes gh-24886
The following modification will also recursively traverse the directory that the symbolic link points to.
This modification is not proposed for deleteRecursively, since it can be harmful if not intended.
This modification relates to the following commit:
Consistent use of NIO.2 for file read/write interactions
Issue: SPR-15748
12114a9
Prior to this commmit, "mutipart/*" MIME types would be cached by the
`MimeTypeUtils` LRU cache. Since those MIME types are likely to have
random boundaries in them, they can waste space in the LRU cache.
This is not improving things since we're parsing them anyway.
This commit skips the caching step for all "multipart" MIME types.
Fixes gh-24767
This commit improves the performance of the `ConcurrentLruCache` and
applies a consistent style to the code:
* separating read/write locks into different variables does not help
performance, so this change is reverted
* apply a consistent style for read/write locks and try/cactch calls
* the reordering of recently used keys is only done when the cache is
full
Fixes gh-24671
- manage collection size manually
- check cache hit first before size check
- reduce read-lock scope
- use `map.get` to test cache instead of `queue.remove`
Closes gh-24469
See gh-24671
Prior to this commit, DefaultParameterNameDiscoverer did not register
any discovers when compiling or running as a GraalVM native image.
This commit relaxes the GraalVM check so that it is only applied to the
registration of KotlinReflectionParameterNameDiscoverer.
Consequently, StandardReflectionParameterNameDiscoverer and
LocalVariableTableParameterNameDiscoverer are once again always
registered by DefaultParameterNameDiscoverer.
Closes gh-24600
Rossen Stoyanchev
shevtsiv
Juergen Hoeller
Sam Brannen
Сергей Цыпанов
Marten Deinum
Sébastien Deleuze
Philippe Marschall
Josh Cummings
Brian Clozel
Shenker93
Yanming Zhou
Stephane Nicoll
yilianhuaixiao
Arjen Poutsma
XenoAmess
May
Rossen Stoyanchev
izeye
Roland Weisleder
GardenLee
Juergen Hoeller
Vlad Kisel
Brian Clozel
Sam Brannen
Sam Brannen
Jendrik Johannes
Sébastien Deleuze
Mikael
Moran Avigdor
Qimiao Chen
Christoph Dreis
Kwangyong Kim