Table of Contents
This section introduces the JAX-RS Client API, which is a fluent Java based API for communication with RESTful Web services. This standard API that is also part of Jakarta EE 9 is designed to make it very easy to consume a Web service exposed via HTTP protocol and enables developers to concisely and efficiently implement portable client-side solutions that leverage existing and well established client-side HTTP connector implementations.
The JAX-RS client API can be utilized to consume any Web service exposed on top of a HTTP protocol or it's extension (e.g. WebDAV), and is not restricted to services implemented using JAX-RS. Yet, developers familiar with JAX-RS should find the client API complementary to their services, especially if the client API is utilized by those services themselves, or to test those services. The JAX-RS client API finds inspiration in the proprietary Jersey 1.x Client API and developers familiar with the Jersey 1.x Client API should find it easy to understand all the concepts introduced in the new JAX-RS Client API.
The goals of the client API are threefold:
Encapsulate a key constraint of the REST architectural style, namely the Uniform Interface Constraint and associated data elements, as client-side Java artifacts;
Make it as easy to consume RESTful Web services exposed over HTTP, same as the JAX-RS server-side API makes it easy to develop RESTful Web services; and
Share common concepts and extensibility points of the JAX-RS API between the server and the client side programming models.
As an extension to the standard JAX-RS Client API, the Jersey Client API supports a pluggable architecture to enable the
use of different underlying HTTP client Connector implementations. Several such implementations are
currently provided with Jersey. We have a default client connector using Http(s)URLConnection
supplied
with the JDK as well as connector implementations based on Apache HTTP Client, Jetty HTTP client and Grizzly Asynchronous Client.
The uniform interface constraint bounds the architecture of RESTful Web services so that a client, such as a browser, can utilize the same interface to communicate with any service. This is a very powerful concept in software engineering that makes Web-based search engines and service mash-ups possible. It induces properties such as:
simplicity, the architecture is easier to understand and maintain; and
evolvability or loose coupling, clients and services can evolve over time perhaps in new and unexpected ways, while retaining backwards compatibility.
Further constraints are required:
every resource is identified by a URI;
a client interacts with the resource via HTTP requests and responses using a fixed set of HTTP methods;
one or more representations can be returned and are identified by media types; and
the contents of which can link to further resources.
The above process repeated over and again should be familiar to anyone who has used a browser to fill in HTML forms and follow links. That same process is applicable to non-browser based clients.
Many existing Java-based client APIs, such as the Apache HTTP client API or HttpUrlConnection
supplied with the JDK place too much focus on the Client-Server constraint for the exchanges of request and
responses rather than a resource, identified by a URI, and the use of a fixed set of HTTP methods.
A resource in the JAX-RS client API is an instance of the Java class
WebTarget.
and encapsulates an URI. The fixed set of HTTP methods can be invoked based on the
WebTarget
.
The representations are Java types, instances of which, may contain links that new instances of
WebTarget
may be created from.
Since a JAX-RS component is represented as an annotated Java type, it makes it easy to configure, pass around and inject in ways that are not so intuitive or possible with other client-side APIs. The Jersey Client API reuses many aspects of the JAX-RS and the Jersey implementation such as:
URI building using UriBuilder and UriTemplate to safely build URIs;
Built-in support for Java types of representations such as
byte[]
,
String
,
Number
,
Boolean
,
Character
,
InputStream
,
java.io.Reader
,
File
,
DataSource
,
JAXB beans as well as additional Jersey-specific JSON and Multi Part support.
Using the fluent builder-style API pattern to make it easier to construct requests.
Some APIs, like the Apache HTTP Client or HttpURLConnection
can be rather hard to use and/or require too much code to do something relatively simple, especially when
the client needs to understand different payload representations.
This is why the Jersey implementation of JAX-RS Client API provides support for wrapping HttpUrlConnection
and the Apache HTTP client. Thus it is possible to get the benefits of the established JAX-RS implementations and
features while getting the ease of use benefit of the simple design of the JAX-RS client API.
For example, with a low-level HTTP client library, sending a POST request with a bunch of typed HTML form parameters
and receiving a response de-serialized into a JAXB bean is not straightforward at all. With the new JAX-RS Client API
supported by Jersey this task is very easy:
Example 5.1. POST request with form parameters
Client client = ClientBuilder.newClient(); WebTarget target = client.target("http://localhost:9998").path("resource"); Form form = new Form(); form.param("x", "foo"); form.param("y", "bar"); MyJAXBBean bean = target.request(MediaType.APPLICATION_JSON_TYPE) .post(Entity.entity(form,MediaType.APPLICATION_FORM_URLENCODED_TYPE), MyJAXBBean.class);
In the Example 5.1, “POST request with form parameters” a new WebTarget
instance is created using a new
Client instance first, next a Form instance is created with two form parameters.
Once ready, the Form
instance is POST
ed to the target resource.
First, the acceptable media type is specified in the request(...)
method. Then in the
post(...)
method, a call to a static method on JAX-RS Entity is made to construct
the request entity instance and attach the proper content media type to the form entity that is being sent. The
second parameter in the post(...)
method specifies the Java type of the response entity that should
be returned from the method in case of a successful response. In this case an instance of JAXB bean is requested to
be returned on success. The Jersey client API takes care of selecting the proper MessageBodyWriter<T> for
the serialization of the Form
instance, invoking the POST
request and producing and
de-serialization of the response message payload into an instance of a JAXB bean using a proper
MessageBodyReader<T>.
If the code above had to be written using HttpUrlConnection
, the developer would have to write custom
code to serialize the form data that are sent within the POST request and de-serialize the response input stream
into a JAXB bean. Additionally, more code would have to be written to make it easy to reuse the logic when
communicating with the same resource “http://localhost:8080/resource”
that is represented by
the JAX-RS WebTarget
instance in our example.
Refer to the dependencies for details on the dependencies when using the Jersey JAX-RS Client support.
You may also want to use a custom Connector implementation. In such case you would need to include
additional dependencies on the module(s) containing the custom client connector that you want to use. See section
"Configuring custom Connectors" about how to use and configure a custom
Jersey client transport Connector
.
JAX-RS Client API is designed to allow fluent programming model. This means, a construction of a
Client
instance, from which a WebTarget
is created, from which a
request Invocation is built and invoked can be chained in a single "flow" of invocations.
The individual steps of the flow will be shown in the following sections.
To utilize the client API it is first necessary to build an instance of a
Client using one of the static ClientBuilder factory methods. Here's the most
simple example:
Client client = ClientBuilder.newClient();
The ClientBuilder
is a JAX-RS API used to create new instances of Client
.
In a slightly more advanced scenarios, ClientBuilder
can be used to configure additional
client instance properties, such as a SSL transport settings, if needed (see ???
below).
A Client
instance can be configured during creation by passing a ClientConfig
to the newClient(Configurable)
ClientBuilder
factory method.
ClientConfig implements Configurable and therefore it offers methods to register
providers (e.g. features or individual entity providers, filters or interceptors) and setup properties.
The following code shows a registration of custom client filters:
ClientConfig clientConfig = new ClientConfig(); clientConfig.register(MyClientResponseFilter.class); clientConfig.register(new AnotherClientFilter()); Client client = ClientBuilder.newClient(clientConfig);
In the example, filters are registered using the ClientConfig.register(...)
method. There are
multiple overloaded versions of the method that support registration of feature and provider classes or instances.
Once a ClientConfig
instance is configured, it can be passed to the
ClientBuilder
to create a pre-configured Client
instance.
Note that the Jersey ClientConfig
supports the fluent API model of Configurable.
With that the code that configures a new client instance can be also written using a more compact style as shown
below.
Client client = ClientBuilder.newClient(new ClientConfig() .register(MyClientResponseFilter.class) .register(new AnotherClientFilter());
The ability to leverage this compact pattern is inherent to all JAX-RS and Jersey Client API components.
Since Client
implements Configurable interface too, it can be configured further
even after it has been created. Important is to mention that any configuration change done on a
Client
instance will not influence the ClientConfig instance that was used to
provide the initial Client
instance configuration at the instance creation time.
The next piece of code shows a configuration of an existing Client
instance.
client.register(ThirdClientFilter.class);
Similarly to earlier examples, since Client.register(...)
method supports the fluent API style,
multiple client instance configuration calls can be chained:
client.register(FilterA.class) .register(new FilterB()) .property("my-property", true);
To get the current configuration of the Client
instance a getConfiguration()
method can be used.
ClientConfig clientConfig = new ClientConfig(); clientConfig.register(MyClientResponseFilter.class); clientConfig.register(new AnotherClientFilter()); Client client = ClientBuilder.newClient(clientConfig); client.register(ThirdClientFilter.class); Configuration newConfiguration = client.getConfiguration();
In the code, an additional MyClientResponseFilter
class and
AnotherClientFilter
instance are registered in the clientConfig
. The
clientConfig
is then used to construct a new Client
instance. The
ThirdClientFilter
is added separately to the constructed Client
instance.
This does not influence the configuration represented by the original clientConfig
.
In the last step a newConfiguration
is retrieved from the client
. This
configuration contains all three registered filters while the original clientConfig
instance
still contains only two filters. Unlike clientConfig
created separately, the
newConfiguration
retrieved from the client
instance represents a live
client configuration view. Any additional configuration changes made to the client
instance
are also reflected in the newConfiguration
. So, newConfiguration
is really
a view of the client
configuration and not a configuration state copy. These principles are
important in the client API and will be used in the following sections too. For example, you can construct a
common base configuration for all clients (in our case it would be clientConfig
) and
then reuse this common configuration instance to configure multiple client
instances that can
be further specialized. Similarly, you can use an existing client
instance configuration to
configure another client instance without having to worry about any side effects in the original
client
instance.
Once you have a Client
instance you can create a WebTarget
from it.
WebTarget webTarget = client.target("http://example.com/rest");
A Client
contains several target(...)
methods that allow for creation of
WebTarget
instance. In this case we're using target(String uri)
version.
The uri
passed to the method as a String
is the URI of the targeted
web resource. In more complex scenarios it could be the context root URI of the whole RESTful application, from
which WebTarget
instances representing individual resource targets can be derived and
individually configured. This is possible, because JAX-RS WebTarget
also implements
Configurable
:
WebTarget webTarget = client.target("http://example.com/rest"); webTarget.register(FilterForExampleCom.class);
The configuration principles used in JAX-RS client API apply to WebTarget
as well. Each
WebTarget
instance inherits a configuration from its parent (either a client or another
web target) and can be further custom-configured without affecting the configuration of the parent component.
In this case, the FilterForExampleCom
will be registered only in the
webTarget
and not in client
. So, the client
can still be used to create new WebTarget
instances pointing at other URIs using just the
common client configuration, which FilterForExampleCom
filter is not part of.
Let's assume we have a webTarget
pointing at "http://example.com/rest"
URI
that represents a context root of a RESTful application and there is a resource exposed on the URI
"http://example.com/rest/resource"
. As already mentioned, a WebTarget
instance can be used to derive other web targets. Use the following code to define a path to the resource.
WebTarget resourceWebTarget = webTarget.path("resource");
The resourceWebTarget
now points to the resource on URI
"http://example.com/rest/resource"
. Again if we configure the
resourceWebTarget
with a filter specific to the resource
,
it will not influence the original webTarget
instance. However, the filter
FilterForExampleCom
registration will still be inherited by the
resourceWebTarget
as it has been created from webTarget
. This mechanism
allows you to share the common configuration of related resources (typically hosted under the same URI root,
in our case represented by the webTarget
instance), while allowing for further configuration
specialization based on the specific requirements of each individual resource. The same configuration principles
of inheritance (to allow common config propagation) and decoupling (to allow individual config customization)
applies to all components in JAX-RS Client API discussed below.
Let's say there is a sub resource on the path "http://example.com/rest/resource/helloworld"
.
You can derive a WebTarget
for this resource simply by:
WebTarget helloworldWebTarget = resourceWebTarget.path("helloworld");
Let's assume that the helloworld
resource accepts a query param for GET
requests which defines the greeting message. The next code snippet shows a code that creates
a new WebTarget
with the query param defined.
WebTarget helloworldWebTargetWithQueryParam = helloworldWebTarget.queryParam("greeting", "Hi World!");
Please note that apart from methods that can derive new WebTarget
instance based on a URI path
or query parameters, the JAX-RS WebTarget
API contains also methods for working with matrix
parameters too.
Let's now focus on invoking a GET
HTTP request on the created web targets. To start building a new
HTTP request invocation, we need to create a new Invocation.Builder.
Invocation.Builder invocationBuilder = helloworldWebTargetWithQueryParam.request(MediaType.TEXT_PLAIN_TYPE); invocationBuilder.header("some-header", "true");
A new invocation builder instance is created using one of the request(...)
methods that are
available on WebTarget
. A couple of these methods accept parameters that let you define
the media type of the representation requested to be returned from the resource. Here we are saying that we
request a "text/plain"
type. This tells Jersey to add a Accept: text/plain
HTTP header to our request.
The invocationBuilder
is used to setup request specific parameters. Here we can setup headers
for the request or for example cookie parameters. In our example we set up a "some-header"
header to value true
.
Once finished with request customizations, it's time to invoke the request. We have two options now.
We can use the Invocation.Builder
to build a generic Invocation instance
that will be invoked some time later. Using Invocation
we will be able to e.g. set additional
request properties which are properties in a batch of several requests and use the generic JAX-RS
Invocation
API to invoke the batch of requests without actually knowing all the details
(such as request HTTP method, configuration etc.). Any properties set on an invocation instance can be read
during the request processing. For example, in a custom ClientRequestFilter you can call
getProperty()
method on the supplied ClientRequestContext to read a request
property. Note that these request properties are different from the configuration properties set on
Configurable
. As mentioned earlier, an Invocation
instance provides generic
invocation API to invoke the HTTP request it represents either synchronously or asynchronously. See
the Chapter 11, Asynchronous Services and Clients for more information on asynchronous invocations.
In case you do not want to do any batch processing on your HTTP request invocations prior to invoking them, there
is another, more convenient approach that you can use to invoke your requests directly from an
Invocation.Builder
instance. This approach is demonstrated in the next Java code listing.
Response response = invocationBuilder.get();
While short, the code in the example performs multiple actions. First, it will build the the request from the
invocationBuilder
. The URI of request will be
http://example.com/rest/resource/helloworld?greeting="Hi%20World!"
and the request will contain
some-header: true
and Accept: text/plain
headers. The request will then pass
trough all configured request filters ( AnotherClientFilter
,
ThirdClientFilter
and
FilterForExampleCom
). Once processed by the filters, the request will be sent to the remote
resource. Let's say the resource then returns an HTTP 200 message with a plain text response content that contains
the value sent in the request greeting
query parameter. Now we can observe the returned
response:
System.out.println(response.getStatus()); System.out.println(response.readEntity(String.class));
which will produce the following output to the console:
200 Hi World!
As we can see, the request was successfully processed (code 200) and returned an entity (representation) is
"Hi World!"
. Note that since we have configured a MyClientResponseFilter
in the resource target, when response.readEntity(String.class)
gets called, the response
returned from the remote endpoint is passed through the response filter chain (including the
MyClientResponseFilter
) and entity interceptor chain and at last a proper
MessageBodyReader<T> is located to read the response content bytes from the response stream into a
Java String
instance. Check Chapter 10, Filters and Interceptors to lear more about
request and response filters and entity interceptors.
Imagine now that you would like to invoke a POST
request but without any query parameters. You would
just use the helloworldWebTarget
instance created earlier and call the
post()
instead of get()
.
Response postResponse = helloworldWebTarget.request(MediaType.TEXT_PLAIN_TYPE) .post(Entity.entity("A string entity to be POSTed", MediaType.TEXT_PLAIN));
The following code puts together the pieces used in the earlier examples.
Example 5.2. Using JAX-RS Client API
ClientConfig clientConfig = new ClientConfig(); clientConfig.register(MyClientResponseFilter.class); clientConfig.register(new AnotherClientFilter()); Client client = ClientBuilder.newClient(clientConfig); client.register(ThirdClientFilter.class); WebTarget webTarget = client.target("http://example.com/rest"); webTarget.register(FilterForExampleCom.class); WebTarget resourceWebTarget = webTarget.path("resource"); WebTarget helloworldWebTarget = resourceWebTarget.path("helloworld"); WebTarget helloworldWebTargetWithQueryParam = helloworldWebTarget.queryParam("greeting", "Hi World!"); Invocation.Builder invocationBuilder = helloworldWebTargetWithQueryParam.request(MediaType.TEXT_PLAIN_TYPE); invocationBuilder.header("some-header", "true"); Response response = invocationBuilder.get(); System.out.println(response.getStatus()); System.out.println(response.readEntity(String.class));
Now we can try to leverage the fluent API style to write this code in a more compact way.
Example 5.3. Using JAX-RS Client API fluently
Client client = ClientBuilder.newClient(new ClientConfig() .register(MyClientResponseFilter.class) .register(new AnotherClientFilter())); String entity = client.target("http://example.com/rest") .register(FilterForExampleCom.class) .path("resource/helloworld") .queryParam("greeting", "Hi World!") .request(MediaType.TEXT_PLAIN_TYPE) .header("some-header", "true") .get(String.class);
The code above does the same thing except it skips the generic Response
processing and directly
requests an entity in the last get(String.class)
method call. This shortcut method let's you
specify that (in case the response was returned successfully with a HTTP 2xx status code) the response entity
should be returned as Java String
type. This compact example demonstrates another advantage of
the JAX-RS client API. The fluency of JAX-RS Client API is convenient especially with simple use cases.
Here is another a very simple GET request returning a String representation (entity):
String responseEntity = ClientBuilder.newClient() .target("http://example.com").path("resource/rest") .request().get(String.class);
Some client invocations, like asynchronous or reactive, could lead to a need to start a new thread. This is
being done on provided ExecutorService or ScheduledExecutorService. ClientBuilder
has two
methods, which can be used to define them: executorService(ExecutorService)
and
scheduledExecutorService(ScheduledExecutorService)
. When specified, all invocations which
do require running on another thread, should be executed using provided services.
Default values do depend on the environment - in Java/Jakarta EE container, it has to be ManagedExecutorService
and ManagedScheduledExecutorService, for Java SE it would be
ForkJoinPool.commonPool
for Executor service and something undefined for Scheduled
executor service.
Example 5.4. Setting JAX-RS Client ExecutorService
ExecutorService myExecutorService = Executors.newCachedThreadPool(); Client client = ClientBuilder.newBuilder().executorService(myExecutorService).build();
All the Java types and representations supported by default on the Jersey server side for requests and responses are also supported on the client side. For example, to process a response entity (or representation) as a stream of bytes use InputStream as follows:
InputStream in = response.readEntity(InputStream.class); ... // Read from the stream in.close();
Note that it is important to close the stream after processing so that resources are freed up.
To POST
a file use a File
instance as follows:
File f = ... ... webTarget.request().post(Entity.entity(f, MediaType.TEXT_PLAIN_TYPE));
The support for new application-defined representations as Java types requires the implementation of the same JAX-RS entity provider extension interfaces as for the server side JAX-RS API, namely MessageBodyReader<T> and MessageBodyWriter<T> respectively, for request and response entities (or inbound and outbound representations).
Classes or implementations of the provider-based interfaces need to be registered as providers within the
JAX-RS or Jersey Client API components that implement Configurable
contract
(ClientBuilder
, Client
, WebTarget
or
ClientConfig
), as was shown in the earlier sections.
Some media types are provided in the form of JAX-RS Feature a concept that allows the extension
providers to group together multiple different extension providers and/or configuration properties in order
to simplify the registration and configuration of the provided feature by the end users. For example,
MoxyJsonFeature can be register to enable and configure JSON binding support via MOXy
library.
By default, the transport layer in Jersey is provided by HttpUrlConnection
. This transport is implemented
in Jersey via HttpUrlConnectorProvider that implements Jersey-specific Connector SPI.
You can implement and/or register your own Connector
instance to the Jersey
Client
implementation, that will replace the default HttpUrlConnection
-based
transport layer. Jersey provides several alternative client transport connector implementations that are ready-to-use.
Table 5.1. List of Jersey Connectors
Transport framework | Jersey Connector implementation | Maven dependency |
---|---|---|
Grizzly NIO framework | GrizzlyConnectorProvider | org.glassfish.jersey.connectors:jersey-grizzly-connector |
Apache HTTP client | ApacheConnectorProvider | org.glassfish.jersey.connectors:jersey-apache-connector |
Apache 5 HTTP client | Apache5ConnectorProvider | org.glassfish.jersey.connectors:jersey-apache5-connector |
Helidon HTTP client | HelidonConnectorProvider | org.glassfish.jersey.connectors:jersey-helidon-connector |
Jetty HTTP client (JDK 17+) | JettyConnectorProvider | org.glassfish.jersey.connectors:jersey-jetty-connector |
Jetty 11.x HTTP client | Jetty11ConnectorProvider | org.glassfish.jersey.connectors:jersey-jetty11-connector |
Jetty 11.x HTTP/2 client | Jetty11Http2ConnectorProvider | org.glassfish.jersey.connectors:jersey-jetty11-http2-connector |
Netty NIO framework | NettyConnectorProvider | org.glassfish.jersey.connectors:jersey-netty-connector |
JDK NIO client | JdkConnectorProvider | org.glassfish.jersey.connectors:jersey-jdk-connector |
Java java.net.http client | JavaNetHttpConnectorProvider | org.glassfish.jersey.connectors:jersey-jnh-connector |
As indicated earlier, Connector and ConnectorProvider contracts are Jersey-specific
extension APIs that would only work with Jersey and as such are not part of JAX-RS. Following example shows how to
setup the custom Grizzly Asynchronous HTTP Client based ConnectorProvider
in a Jersey client
instance:
ClientConfig clientConfig = new ClientConfig(); clientConfig.connectorProvider(new GrizzlyConnectorProvider()); Client client = ClientBuilder.newClient(clientConfig);
Client
accepts as a constructor argument a Configurable
instance. Jersey
implementation of the Configurable
provider for the client is ClientConfig
.
By using the Jersey ClientConfig
you can configure the custom
ConnectorProvider
into the ClientConfig
. The GrizzlyConnectorProvider
is used as a custom
connector provider in the example above. Please note that the connector provider cannot be registered as a provider
using Configurable
.register(...)
. Also, please note that in this API has changed
since Jersey 2.5, where the ConnectorProvider
SPI has been introduced in order to decouple client
initialization from the connector instantiation. Starting with Jersey 2.5 it is therefore not possible to directly
register Connector
instances in the Jersey ClientConfig. The new
ConnectorProvider
SPI must be used instead to configure a custom client-side transport connector.
A ConnectorProvider can also be set by a property on a ClientBuilder
starting with
Jersey 2.40. The following example shows how to setup the custom Grizzly Asynchronous HTTP Client based
ConnectorProvider
in a Jersey client instance:
Client client = ClientBuilder.newBuilder() .property(ClientProperties.CONNECTOR_PROVIDER, "org.glassfish.jersey.grizzly.connector.GrizzlyConnectorProvider") .build();
For more information about the property see Appendix A, Configuration Properties.
Be aware of using other than default Connector
implementation.
There is an issue handling HTTP headers in
WriterInterceptor
or MessageBodyWriter<T>
.
If you need to change header fields do not use neither
ApacheConnectorProvider
nor GrizzlyConnectorProvider
nor JettyConnectorProvider
(for asynchronous requests).
Other older version connectors can be affected by this issue as well.
The issue for example applies to Jersey Multipart
feature that also modifies HTTP headers.
The default connector is the most advanced connector, and it supports the most features Jersey has to offer.
However, there are a few limitations coming from the HttpUrlConnection
One limitation is in the variety of HTTP methods supported by the HttpUrlConnection
, since only the
original HTTP/1.1 methods are supported. For instance, HTTP Patch method is not supported. See
property HttpUrlConnector.SET_METHOD_WORKAROUND in the Appendix
Section A.7, “The default HttpUrlConnector properties” for a possible workaround.
Also, in the default transport connector, there are some restrictions on the headers, that
can be sent in the default configuration.
HttpUrlConnectorProvider
uses HttpUrlConnection
as an underlying connection
implementation. This JDK class by default restricts the use of following headers:
Access-Control-Request-Headers
Access-Control-Request-Method
Connection
(with one exception - Connection
header with
value Closed
is allowed by default)Content-Length
Content-Transfer-Encoding
-Host
Keep-Alive
Origin
Trailer
Transfer-Encoding
Upgrade
Via
Sec-
The underlying connection can be configured to permit all headers to be sent,
however this behaviour can be changed only by setting the system property
sun.net.http.allowRestrictedHeaders
.
Example 5.5. Sending restricted headers with HttpUrlConnector
Client client = ClientBuilder.newClient(); System.setProperty("sun.net.http.allowRestrictedHeaders", "true"); Response response = client.target(yourUri).path(yourPath).request(). header("Origin", "http://example.com"). header("Access-Control-Request-Method", "POST"). get();
Internally, the HttpUrlConnection
instances are pooled, so (un)setting the
property after already creating a target typically does not have any effect.
The property influences all the connections created after the property has been
(un)set, but there is no guarantee, that your request will use a connection
created after the property change.
In a simple environment, setting the property before creating the first target is sufficient, but in complex environments (such as application servers), where some poolable connections might exist before your application even bootstraps, this approach is not 100% reliable and we recommend using a different client transport connector, such as Apache Connector. These limitations have to be considered especially when invoking CORS (Cross Origin Resource Sharing) requests.
The limited configurability of the HttpUrlConnection
is another aspect to consider. For details, see
Java Networking Properties
, for instance property http.maxConnections
.
For each Connector a property file defining properties tweaking the actual
Connector
exists. There are ClientProperties properties that apply to the
default HttpUrlConnectorProvider. Many of the properties are also understood by
various Connectors
. Each of the ConnectorProvider
defines properties that apply to it.
Moreover, each Connector supports a list of properties that are unique for the Connector. The list of the client connector properties can be found in the Appendix A, Configuration Properties.
It is not possible to provide all the settings the underlying HTTP Clients support. For Apache HTTP Client, and for Jetty HTTP Client, it is possible to access directly the HTTP Client classes and invoke setter methods there.
For Apache Connector, an ApacheHttpClientBuilderConfigurator SPI allows for invoking methods
of org.apache.http.impl.client.HttpClientBuilder
, such as setDefaultCredentialsProvider
:
org.apache.http.client.CredentialsProvider credentialsProvider = new org.apache.http.impl.client.BasicCredentialsProvider(); credentialsProvider.setCredentials( org.apache.http.auth.AuthScope.ANY, new org.apache.http.auth.UsernamePasswordCredentials("name", "password") ); ApacheHttpClientBuilderConfigurator apacheHttpClientBuilderConfigurator = (httpClientBuilder) -> { return httpClientBuilder.setDefaultCredentialsProvider(credentialsProvider); }; ClientConfig cc = new ClientConfig(); cc.register(apacheHttpClientBuilderConfigurator); cc.connectorProvider(new ApacheConnectorProvider()); Client client = ClientBuilder.newClient(cc); ...
For Apache 5 Connector, an Apache5HttpClientBuilderConfigurator SPI allows for invoking methods
of org.apache.hc.client5.http.impl.classic.HttpClientBuilder
, such as setDefaultCredentialsProvider
:
org.apache.hc.client5.http.auth.CredentialsStore credentialsProvider = new org.apache.hc.client5.http.impl.auth.BasicCredentialsProvider(); credentialsProvider.setCredentials( new org.apache.hc.client5.http.auth.AuthScope("localhost", getPort()), new org.apache.hc.client5.http.auth.UsernamePasswordCredentials("name", "password".toCharArray()) ); Apache5HttpClientBuilderConfigurator apache5HttpClientBuilderConfigurator = (httpClientBuilder) -> { return httpClientBuilder.setDefaultCredentialsProvider(credentialsProvider); }; ClientConfig cc = new ClientConfig(); cc.register(apache5HttpClientBuilderConfigurator); cc.connectorProvider(new Apache5ConnectorProvider()); Client client = ClientBuilder.newClient(cc); ...
For Jetty Connector, an JettyHttpClientSupplier SPI allows for providing a configured instance
of org.eclipse.jetty.client.HttpClient
:
HttpClient httpClient = new HttpClient(...); ClientConfig clientConfig = new ClientConfig() .connectorProvider(new JettyConnectorProvider()) .register(new JettyHttpClientSupplier(httpClient)); Client client = ClientBuilder.newClient(clientConfig); ...
For Jetty Connector, an Jetty11HttpClientSupplier SPI allows for providing a configured instance
of org.eclipse.jetty.client.HttpClient
:
HttpClient httpClient = new HttpClient(...); ClientConfig clientConfig = new ClientConfig() .connectorProvider(new Jetty11ConnectorProvider()) .register(new Jetty11HttpClientSupplier(httpClient)); Client client = ClientBuilder.newClient(clientConfig); ...
Filtering requests and responses can provide useful lower-level concept focused on a certain independent aspect or domain that is decoupled from the application layer of building and sending requests, and processing responses. Filters can read/modify the request URI, headers and entity or read/modify the response status, headers and entity.
Jersey contains the following useful client-side filters (and features registering filters) that you may want to use in your applications:
CsrfProtectionFilter: Cross-site request forgery protection filter (adds
X-Requested-By to each state changing request). |
EncodingFeature: Feature that registers encoding filter which use registered ContentEncoders to encode and decode the communication. The encoding/decoding is performed in interceptor (you don't need to register this interceptor). Check the javadoc of the EncodingFeature in order to use it. |
HttpAuthenticationFeature: HTTP Authentication Feature (see authentication below). |
Note that these features are provided by Jersey, but since they use and implement JAX-RS API, the features should be portable and run in any JAX-RS implementation, not just Jersey. See Chapter 10, Filters and Interceptors chapter for more information on filters and interceptors.
The underlying connections are opened for each request and closed after the response is received and entity is processed (entity is read). See the following example:
Example 5.6. Closing connections
final WebTarget target = ... some web target Response response = target.path("resource").request().get(); System.out.println("Connection is still open."); System.out.println("string response: " + response.readEntity(String.class)); System.out.println("Now the connection is closed.");
If you don't read the entity, then you need to close the response manually by
response.close()
. Also if the entity is read into an InputStream
(by response.readEntity(InputStream.class)
), the connection stays open until
you finish reading from the InputStream
. In that case, the InputStream
or the Response should be closed manually at the end of reading from InputStream.
In some cases you might need to inject some custom types into your client provider instance. JAX-RS types do not need to be injected as they are passed as arguments into API methods. Injections into client providers (filters, interceptor) are possible as long as the provider is registered as a class. If the provider is registered as an instance then runtime will not inject the provider. The reason is that this provider instance might be registered into multiple client configurations. For example one instance of ClientRequestFilter can be registered to two Clients.
To solve injection of a custom type into a client provider instance
use InjectionManagerClientProvider to
extract InjectionManager
which can return the required injection. The following example shows how to utilize
InjectionManagerClientProvider
:
Example 5.7. InjectionManagerClientProvider example
public static class MyRequestFilter implements ClientRequestFilter { // this injection does not work as filter is registered as an instance: // @Inject // private MyInjectedService service; @Override public void filter(ClientRequestContext requestContext) throws IOException { // use InjectionManagerClientProvider to extract InjectionManager from request final InjectionManager injectionManager = InjectionManagerClientProvider.getInjectionManager(requestContext); // and ask for MyInjectedService: final MyInjectedService service = injectionManager.getInstance(MyInjectedService.class); final String name = service.getName(); ... } }
For more information see javadoc of InjectionManagerClientProvider
(and javadoc of InjectionManagerProvider
which supports common Jakarta REST components).
This section describes how to setup SSL configuration on Jersey client (using JAX-RS API). The SSL configuration is setup in ClientBuilder. The client builder contains methods for definition of KeyStore, TrustStore or entire SslContext. See the following example:
SSLContext ssl = ... your configured SSL context; Client client = ClientBuilder.newBuilder().sslContext(ssl).build(); Response response = client.target("https://example.com/resource").request().get();
The example above shows how to setup a custom SslContext
to the ClientBuilder
.
Creating a SslContext
can be more difficult as you might need to init instance properly with the protocol,
KeyStore
, TrustStore
, etc. Jersey offers a utility SslConfigurator class that
can be used to setup the SslContext
. The SslConfigurator
can be configured based on
standardized system properties for SSL configuration, so for example you can configure the KeyStore
file
name using a environment variable javax.net.ssl.keyStore
and SslConfigurator
will use such a variable to setup the SslContext
. See javadoc of SslConfigurator for more
details. The following code shows how a SslConfigurator
can be used to create a custom SSL
context.
SslConfigurator sslConfig = SslConfigurator.newInstance() .trustStoreFile("./truststore_client") .trustStorePassword("secret-password-for-truststore") .keyStoreFile("./keystore_client") .keyPassword("secret-password-for-keystore"); SSLContext sslContext = sslConfig.createSSLContext(); Client client = ClientBuilder.newBuilder().sslContext(sslContext).build();
Note that you can also setup KeyStore
and TrustStore
directly on a
ClientBuilder
instance without wrapping them into the SslContext
. However, if you setup
a SslContext
it will override any previously defined KeyStore
and TrustStore
settings.
ClientBuilder
also offers a method for defining a custom HostnameVerifier implementation.
HostnameVerifier
implementations are invoked when default host URL verification fails.
A behaviour of HostnameVerifier is dependent on an http client implementation.
HttpUrlConnectorProvider
and ApacheConnectorProvider
work properly, that means that after
the unsuccessful URL verification HostnameVerifier
is called and by means of it is possible to
revalidate URL using a custom implementation of HostnameVerifier
and go on in a handshake processing.
JettyConnectorProvider
and GrizzlyConnectorProvider
provide only host URL verification
and throw a CertificateException
without any possibility to use custom HostnameVerifier
.
Moreover, in case of JettyConnectorProvider
and Jetty11ConnectorProvider
there are the properties
JettyClientProperties.ENABLE_SSL_HOSTNAME_VERIFICATION and Jetty11ClientProperties.ENABLE_SSL_HOSTNAME_VERIFICATION
to disable an entire host URL verification mechanism in a handshake.
Note that to utilize HTTP with SSL it is necessary to utilize the “https”
scheme.
Currently the default connector provider HttpUrlConnectorProvider provides connectors based on
HttpUrlConnection
which implement support for SSL defined by JAX-RS configuration discussed in this
example.
Jersey supports Basic and Digest HTTP Authentication.
In version of Jersey 3.x both authentication methods are provided by single Feature
HttpAuthenticationFeature.
For migration of older applications: org.glassfish.jersey.client.filter.HttpBasicAuthFilter
and org.glassfish.jersey.client.filter.HttpDigestAuthFilter
shall be replaced by
those two authentication methods.
In order to enable http authentication support in Jersey client register the HttpAuthenticationFeature. This feature can provide both authentication methods, digest and basic. Feature can work in the following modes:
BASIC: Basic preemptive authentication. In preemptive mode the authentication information is send always with each HTTP request. This mode is more usual than the following non-preemptive mode (if you require BASIC authentication you will probably use this preemptive mode). This mode must be combined with usage of SSL/TLS as the password is send only BASE64 encoded.
BASIC NON-PREEMPTIVE:Basic non-preemptive authentication. In non-preemptive mode the
authentication information is added only when server refuses the request with 401
status code and
then the request is repeated with authentication information. This mode has negative impact on the performance.
The advantage is that it does not send credentials when they are not needed. This mode must
be combined with usage of SSL/TLS as the password is send only BASE64 encoded.
DIGEST: Http digest authentication. Does not require usage of SSL/TLS.
UNIVERSAL: Combination of basic and digest authentication. The feature works in non-preemptive
mode which means that it sends requests without authentication information. If 401
status
code is returned, the request is repeated and an appropriate authentication is used based on the
authentication requested in the response (defined in WWW-Authenticate
HTTP header). The feature
remembers which authentication requests were successful for given URI and next time tries to preemptively
authenticate against this URI with latest successful authentication method.
To initialize the feature use static methods and builder of this feature. Example of building the feature in Basic authentication mode:
HttpAuthenticationFeature feature = HttpAuthenticationFeature.basic("user", "superSecretPassword");
Example of building the feature in basic non-preemptive mode:
HttpAuthenticationFeature feature = HttpAuthenticationFeature.basicBuilder() .nonPreemptive().credentials("user", "superSecretPassword").build();
You can also build the feature without any default credentials:
HttpAuthenticationFeature feature = HttpAuthenticationFeature.basicBuilder().build();
In this case you need to supply username and password for each request using request properties:
Response response = client.target("http://localhost:8080/rest/homer/contact").request() .property(HTTP_AUTHENTICATION_BASIC_USERNAME, "homer") .property(HTTP_AUTHENTICATION_BASIC_PASSWORD, "p1swd745").get();
This allows you to reuse the same client for authenticating with many different credentials.
See javadoc of the HttpAuthenticationFeature for more details.
When using SSL/TLS protocols for the connection, SNIHostName
is set automatically
based on the host name in the HTTPS request.
There might be use-cases where the SNIHostName
is required to be set for other host
than the host specified in the HTTPS request. For those cases, when the HTTP header Host
is set, the SNIHostName
is set for the host specified in the Host
header.
Note that only Apache Connector, JDK Connector, Netty connector
, and the default
HttpUrlConnector
do support this feature.
Sometimes, it may be required the SNIHostName
is not set, or it differs from the
HTTP Host
header. In that case, the ClientProperties.SNI_HOST_NAME
property can be utilized. The property sets the host name to be used for calculating the
javax.net.ssl.SNIHostName
. The property takes precedence over the HTTP
Host
header.
When the host name in the the ClientProperties.SNI_HOST_NAME property matches the HTTP request
host, the SNIHostName
is not set, and the HTTP Host
header is not used
for setting the SNIHostName
.
Turning the SNI off allows for Domain Fronting.
Suppose a case that the start of the request is to be logged and even measured.
This can be done by ClientRequestFilter
, which is usually invoked before the request is wired on the network.
However, the filter may be called as a last of the filters in the chain. Sure, it can have the highest priority,
but the other filters can have the very same priority! Some long-running operations can be performed before the
measuring can actually start. Even worse, the filter may even be skipped from the chain by the previous
#abortWith
!
For this, PreInvocationInterceptor
, the code that executes before the ClientRequestFilters
are invoked, has been added to the client request chain. Jersey ensures all the interceptors are invoked with each request.
The interceptor contains a single #beforeRequest
method, which corresponds to ClientRequestFilter
:
/** * The method invoked before the request starts. * @param requestContext the request context shared with * ClientRequestFilter. */ void beforeRequest(ClientRequestContext requestContext);
Note that only a single #abortWith
is allowed in all PreInvocationInterceptors
,
otherwise an IllegalStateException
is thrown.
All the exceptions accumulated in PreInvocationInterceptors
are thrown in a single Exception,
available through #getSuppressed()
.
Similarly, ClientResponseFilter
seems to be a good place where the total time of the HTTP request can be measured,
but similarly to ClientRequestFilter
, the response filter may not be invoked at all.
For this, PostInvocationInterceptor
has been introduced. Jersey runtime ensures that every
PostInvocationInterceptor
is called. Since an exception can occur during the HTTP request,
PostInvocationInterceptor
comes with two methods:
/** * The method is invoked after a request when no * is thrown, or the Throwables are resolved * by previous PostInvocationInterceptor. * * @param requestContext the request context. * @param responseContext the response context * of the original Response or response context * defined by the new resolving Response. */ void afterRequest(ClientRequestContext requestContext, ClientResponseContext responseContext); /** * The method is invoked after a Throwable is caught * during the client request chain processing. * * @param requestContext the request context. * @param exceptionContext the context available to handle the * caught Throwables. */ void onException(ClientRequestContext requestContext, ExceptionContext exceptionContext);
The #afterRequest
method is executed when no exception has been thrown during the HTTP request,
#onException
method is executed if the exception has been thrown during the request.
It is possible to set a response in #onException
, and the consecutive PostInvocationInterceptor
will
execute its #afterRequest
method.
The measuring example can looks as follows, then:
String response = ClientBuilder.newClient().target("path") .register(new PreInvocationInterceptor() { @Override public void beforeRequest(ClientRequestContext requestContext) { startTime = System.currentTimeMillis(); } }) .register(new PostInvocationInterceptor() { @Override public void afterRequest(ClientRequestContext requestContext, ClientResponseContext responseContext) { logDuration(System.currentTimeMillis() - startTime); } @Override public void onException(ClientRequestContext requestContext, ExceptionContext exceptionContext) { logDuration(System.currentTimeMillis() - startTime); } }) .request().get().readEntity(String.class);
InvocationBuilderListener is an interface that is inspired by Microprofile REST Client RestClientBuilderListener
and it contains a single method:
/** * Whenever an Invocation.Builder is created, (i.e. when * WebTarget#request() is called, this method would be invoked. * * @param context the updated InvocationBuilderContext. */ void onNewBuilder(InvocationBuilderContext context);
InvocationBuilderContext
a subset of methods of the Invocation.Builder
. It can be used to call the default
values of the Invocation.Builder
. Since it is invoked at the time Invocation.Builder
is instantiated, any consequent
calls of the Invocation.Builder
‘s methods will replace the defaults set by the InvocationBuilderListener
.
For instance, if all the HTTP requests should contain a custom HTTP header,
there can be created a feature that would be registered on the client:
public static class MyFeature implements Feature { @Override public boolean configure(FeatureContext context) { context.register( (InvocationBuilderListener)(l)-> l.getHeaders().add("MY_HEADER", "MY_VALUE") ); return true; } }
This section describes support of Expect:100-continue in Jersey client using Expect100Continue feature. Jersey client supports given header for default JDK HTTP connector only.
Jersey client Expect100Continue feature
Since Jersey 2.32 it is possible to send Expect:100-continue header from Jersey client. Feature shall be registered in client using (for example)
target(RESOURCE_PATH).register(Expect100ContinueFeature.basic());
Note that registration can be done in several ways: with basic settings, and with custom settings:
target(RESOURCE_PATH).register(Expect100ContinueFeature.withCustomThreshold(100L));
Basic registration means that default sending threshold will be used. Value of the default threshold is
DEFAULT_EXPECT_100_CONTINUE_THRESHOLD_SIZE = 65536L;
Threshold is used to determine allowed size of request after which 100-continue header shall be sent before sending request itself.
Environment properties configuration
Previous paragraph described programmatic way of configuration. However the Expect100Continue feature can be configured using environment variables as well.
Since Jersey client can be influenced through environment variables, there are two variables which come since Jersey 2.32:
-Djersey.config.client.request.expect.100.continue.processing=true/false -Djersey.config.client.request.expect.100.continue.threshold.size=12345
First variable can be used to forbid the Expect (100-continue) header be sent at all even though it is registered as described in the previous paragraph. If this property is not provided (or true) and the Expect100Continue feature is registered, sending of the Expect header is enabled.
The second property defines (or modifies) threshold size. So, if the Expect100Continue feature is registered
using basic (default threshold size) parameters, value of the threshold can be modified using this property.
This is valid for custom threshold as well - when the Expect100Continue feature is registered using
withCustomThreshold method its value can be modified anyway by the environment property
jersey.config.client.request.expect.100.continue.threshold.size
.