Fixed a bug where the URL content negotiation "format" parameter values
were case sensitive and only lowercase values were accepted. For
example, URL query parameter format=json returned the appropriate JSON
response but format=JSON resulted in a
HttpMediaTypeNotAcceptableException and returned:
406 - The resource identified by this request is only capable of
generating responses with characteristics not acceptable according to
the request "accept" headers.
When the MappingMediaTypeFileExtensionResolver is constructed, it is
passed a map containing the media type key to MediaType mappings
defined in the ContentNegotiationConfigurer. In the constructor of
MappingMediaTypeFileExtensionResolver, the keys are converted to
lowercase and the mappings of keys to MediaTypes are added to the
ConcurrentMap<String, MediaType> mediaTypes using the lowercase
version of the keys. However, when retrieving the MediaType from a key
in the lookupMediaType method, no conversion to lowercase is performed
so any value for the URL "format" parameter other than the lowercase
version will not return the proper MediaType result.
On May 1st, 2014, a change was made to
ParameterContentNegotiationStrategy to handle cases where the content
negotiation format URL parameter does not result in a match for a
MediaType. If no match is found, a HttpMediaTypeNotAcceptableException
is thrown resulting in the 406 response above. Prior to this commit, a
null was returned instead of throwing an exception so this issue was
hidden and appeared to function correctly.
To make the media type lookup case insensitive, added a line to the
lookupMediaType method in MediaTypeFileExtensionResolver to first
convert the extension (media type key) to lowercase prior to attempting
to retrieve it from the mediaTypes map.
Issue: SPR-13747
This commit reworks the arrangement to a centralized cache, avoiding any extra reflection attempts if a cache entry is known already. In the course of this, it also enforces toXXX methods to be declared as non-static now (which is the only sensible arrangement anyway).
Issue: SPR-13703
Prior to this change, the resource handling FixedVersionStrategy would
be applied on all links that match the configured pattern. This is
problematic for relative links and can lead to rewritten links such as
"/fixedversion/../css/main.css" which breaks.
This commit prevents that Strategy from being applied to such links.
Of course, one should avoid to use that VersionStrategy with relative
links, but this change aims at not breaking existing links even if it
means not prefixing the version as expected.
Issue: SPR-13727
The recent commit 971f04 replaced the use of a NAMES_PATTERN regex in
favor of direct parsing in order to deal with nested curly braces.
The change also incorrectly replicated this logic which removes a
trailing slash after Pattern quoting (and not before):
cca037a74d/spring-web/src/main/java/org/springframework/web/util/UriTemplate.java (L207-L210)
After some more investigation there doesn't appear to be any scenario
where the quoted pattern would end with a trailing slash. It should
always end with \E (end of quote) or a ")" (end of group). Nor are
there any failing tests so this commit removes the logic altogether.
Issue: SPR-13705
This commit refines getMergedBeanDefinition's scope adaptation algorithm for inner beans, never leaving a custom scope within a containing bean of a different scope. The inner bean's scope will either be aligned with the containing bean's scope (matching the effective state) or be switched to prototype in case of an outer singleton (indicating that no singleton state management is desired).
Issue: SPR-13739
Also switches 4.2.4's new formatter implementations to package visibility, just in case they'll be superseded by another variant in the future.
Issue: SPR-13730
When the plus operator is used between strings in a SpEL
expression and that expression is compiled, it is
important to include a cast if computation of any of
the operands isn't obviously leaving strings on the
stack. Likewise if the stack contents are known to
be strings, a cast should not be included.
Issue: SPR-12426
Prior to this commit, range requests would be served by
ResourceHttpRequestHandler by partially reading the inputstream of
static resources. In case of resources contained in ZIP/JAR containers,
InputStreams may not fill the entire read buffer when calling
`inputStream.read(byte[])`. This was the case when using Spring Boot's
ZipInflaterInputStream - this would then not read the entire file
content and would close the response without writing the expected body
length indicated in the "Content-Length" header.
This commit makes sure that the whole resource is read.
Issue: SPR-13661
Prior to this commit, when adding a ShallowEtagHeaderFilter to an
application, the ServletResponse would be wrapped by a
ContentCachingResponseWrapper. When any part of the Spring
infrastructure calls `flushBuffer` on the wrapped response, the call is
delegated to the actual response, which is committed. It's not possible
to alter the response (headers, content) anymore - the ETag filter can't
act.
This change prevents the `flushBuffer` call to be delegated and only
commits the underlying response once the cached content is copied to the
actual response stream.
Issue: SPR-13717
With this change the MapAccessor now extends CompilablePropertyAccessor
rather than just PropertyAccessor. This means that any expression that
ends up using the MapAccessor is now compilable for fast performance.
Issue: SPR-13638
Juergen Hoeller
Rossen Stoyanchev
mhartsock Melissa Hartsock
Brian Clozel
Andy Clement
Stephane Nicoll