Eclipse Scout: Technical Guide

To participate in the application startup or shutdown a platform listener can be created. For this a class implementing org.eclipse.scout.rt.platform.IPlatformListener must be created. The listener is automatically a bean and must therefore not be registered anywhere. See Section 2.3 to learn more about bean management in Scout and how the listener becomes a bean. As soon as the state of the platform changes the listener will be notified.

In its static initializer, the ClassInventory collects classes in projects containing a resource called META-INF/scout.xml.

Scanning all classes would be unnecessarily slow and consume too much memory. The file scout.xml is just an empty xml file. Scout itself also includes scout.xml files in all its projects.

The format XML was chosen to allow adding exclusions in large projects, but this feature is not implemented right now.

Scout uses Jandex ^[1] to build the class inventory. The meta data to find classes can be pre-computed during build time into an index file describing the contents of the jar file. See the jandex project for details.

Usually beans are registered during application startup. The application startup can be intercepted using platform listeners as described in Section 2.1.1.

There is also a predefined bean registration built into the Scout runtime. This automatically registers all classes having an org.eclipse.scout.rt.platform.@Bean annotation. Therefore it is usually sufficient to only annotate a class with @Bean to have it available in the bean manager as shown in listing Listing 4.

The most important meta data of a bean is the scope. It describes how many instances of a bean can exist in a single application. There are two different possibilities:

As like all bean meta data this characteristic can be provided in two different ways:

So the Java annotation org.eclipse.scout.rt.platform.@ApplicationScoped describes a bean having singleton characteristics.

It is not only possible to influence the number of instances to be created (see Section 2.3.2), but also to create beans eagerly, execute methods after creation (like constructors) or to delegate the bean creation completely. These topics are described in the next sections.

To retrieve a bean the class org.eclipse.scout.rt.platform.BEANS should be used. This class provides (amongst others) the following methods:

There are now two more annotations that have an effect on which beans are returned if multiple beans match a certain class. Consider the following example bean hierarchy:

In this situation 4 bean candidates are available: MyServiceImpl, MyServiceMod, MySpecialVersion and AnotherVersion. But which one is returned by BEANS.get(IMyService.class)? Or by BEANS.get(MySpecialVersion.class)? This can be influenced with the org.eclipse.scout.rt.platform.@Order and org.eclipse.scout.rt.platform.@Replace annotations. The next sections describe the idea behind these annotations and gives some examples.

The next examples use the bean situation as shown in figure Figure 1. In this situation the bean manager actually contains 3 beans:

The bean MySpecialVersion is not part of the bean manager because it has been replaced by AnotherVersion.

Bean decorations allow to wrap interfaces with a proxy to intercept each method call to the interface of a bean and apply some custom logic. For this a IBeanDecorationFactory has to be implemented. This is one single factory instance for the entire application. It decides which decorators are created for a bean request. The factory is asked for decorators on every bean retrieval. This allows to write bean decoration factories depending on dynamic conditions.

As bean decoration factories are beans themselves, it is sufficient to create an implementation of org.eclipse.scout.rt.platform.IBeanDecorationFactory and to ensure this implementation is used (see Section 2.3.4). This factory receives the bean to be decorated and the originally requested bean class to decide which decorators it should create. In case no decoration is required the factory may return null. Then the original bean is used without decorations.

The sample in listing Listing 7 wraps each call to the server with a profiler decorator that measures how long a server call takes.

Application scoped beans can declare methods annotated with javax.annotation.@PreDestroy. These methods will be called when the Scout platform is stopping. The methods may have any visibility modifier but must not be static and must not declare any parameters. If such a pre-destroy method throws an exception, the platform will continue to call all other pre-destroy methods (even methods on the same bean).

Please note that pre-destroy methods are only called for application-scoped beans that already have created their instance.

Pre-destroy methods inherited from super classes are always called after the ones from the class itself. Methods that are overridden are only called on the leaf class. Private methods are always called (because they cannot be overridden). The order in which multiple methods in the same declaring class are called is undefined.

Because the property classes are managed by the bean manager, you can use all the mechanisms to change the behavior (@Replace in particular).

Listing 10 demonstrates how you can use the replace annotation to change the existing ApplicationNameProperty class. The value is no longer fetched via the config mechanism, because the getValue(String) method is overriden. In this case a fixed value is returned.

The next example presented in Listing 11 uses the same idea. In this case, the getKey() method is overriden to read the value from an other key as demonstrated is the Listing 12.

The Desktop Layout can be configured using the IDesktop.setBenchLayoutData method. This property is observed and might be changed during the applications lifecycle. The desktop consists out of 9 view stacks (see Figure 2). Each form can be assigned to a single view stack using the property DisplayViewId (IForm.getConfiguredDisplayViewId). If multiple forms are assigned to the same view stack the views will be displayed as tabs where the top form is visible and the corresponding tab selected.

The east, center and west columns are separated with splitters which can be moved according to the layout data properties. Each column is split into a north, center and south part. Within a column the north, center and south parts can not differ in their width.

The modifications (splitter movements) are cached when a cache key (BenchLayoutData.withCacheKey) is set. In case the cache key is null the layout starts always with the initial values.

An example of a bench layout data configuration with a fixed north (N) view stack and an south (S) view stack with an minimal size. See org.eclipse.scout.rt.client.ui.desktop.bench.layout.FlexboxLayoutData API for the documentation of the properties.

By default the internationalization mechanism relies on .properties files using a reference implementation of the TextProviderServices:

Service extending the AbstractDynamicNlsTextProviderService class.

A Text Provider Service working with the default implementation need to define where the properties files are located. This is realized by overriding the getter getDynamicNlsBaseName(). Here an example:

If configured like this, it means that the .properties files will be located in the same plug-in at the location:

If you decide to store your translated texts in .properties files, you migth want to use the NLS Editor to edit them.

You need to respect the format defined by the Java Properties class. In particular the encoding of a .properties file is ISO-8859-1 (also known as Latin-1). All non-Latin-1 characters must be encoded. Examples:

The encoding is the "Unicode escape characters": \uHHHH where HHHH is a hexadecimal id of the character in the Unicode character table. Read more on the .properties File on wikipedia.

Delegation to the Lookup Service on server side.

They are not necessarily restricted to a fix number of records. Instead they should be favoured if the set of records is rather large.

Such a LookupCall is used if the data can be provided directly without the need to make a backend call.

An example of this approach is when a SmartField or a SmartColumn is configured to be used with a CodeType. The code types are cached so it is not necessary to fetch them using a lookup service. Instead a LocalLookupCall, in that case the CodeLookupCall, may be used to load the data. It creates the LookupRows corresponding to the codes in the CodeType.

Defined with getConfiguredXxxxxx() methods.

Using a LookupCall in a SmartField:

Accessing a LookupRow directly:

It is possible to access a LookupRow directly. In this example the input is a key (thisKey) and the method getDataByKey() is used. Before accessing the text, we ensure that a LookupRow has been retrieved.

The codes are organized in a tree. Therefore a CodeType can have one or more child codes at the root level, and each code can have other child codes. In a lot of cases a list of codes (meaning a tree containing only leaves at the first level) is sufficient to cover most of the need.

Child codes are ordered in their parent code. This is realized with the order annotation.

The type of the key is defined by its generic parameter <T>. It is very common to use a type from the java.lang.* package (like Integer or String), but any Java Object is suitable. It must:

There is no obligation to have the same type for the Id between the codes of a CodeType (meaning the same generic type parameter for the codes inner-class). However it is a good practice to have the same type between the codes of a CodeType, because the Id are used as value of SmartFields. Therefore the generic parameter describing the type of value of a SmartField must be compatible with the type of the codes contained in the CodeType.

CodeType in a SmartField (or SmartColumn).

If the SmartField (or SmartColumn) works with a CodeType, a specific LookupCall is instantiated to get the LookupRows based on the Codes contained in a CodeType.

Scout-runtime will handle the instantiation and the caching of CodeTypes.

This function returns the text corresponding to the key using a CodeType:

The common way to define a CodeType is to extend AbstractCodeType. Each code is an inner-class extending AbstractCode. Like usual the properties of Codes and CodeTypes can be set using the getConfiguredXxxxxx() methods.

See the Java Code of a simple YesOrNoCodeType having just two codes:

The SDK provides some help to generate CodeTypes and Codes. Use File → New → Scout → Scout Code Type to generate a new code.

Runtime Errors are used to indicate an error, that shouldn’t be catched/treated by business logic and therefore bubble up to the appropriate exception handler in the scout framework. Because those errors are handled by the framework internals, they should never be catched on the server (Services etc.) nor on the client side (Pages, Forms, etc.).

All Scout Runtime Errors extend PlatformError.

Use this translator to work with checked exceptions and runtime exceptions, but not with Throwable.

If given an Exception, or a RuntimeException, or if being a subclass thereof, that exception is returned as given. Otherwise, a PlatformException is returned which wraps the given Throwable.

Use this translator to work with runtime exceptions. When working with RunContext or IFuture, some methods optionally accept a translator. If not specified, this translator is used by default.

If given a RuntimeException, it is returned as given. For a checked exception, a PlatformException is returned which wraps the given checked exception.

Use this translator to work with PlatformExceptions.

If given a PlatformException, it is returned as given. For all other exceptions (checked or unchecked), a PlatformException is returned which wraps the given exception.

Typically, this translator is used if you require to add some context information via IThrowableWithContextInfo.withContextInfo(String, Object, Object).

Use this translator to work with Throwable as given.

Also, if given a wrapped exception like UndeclaredThrowableException, InvocationTargetException or ExecutionException, that exception is returned as given without unwrapping its cause.

For instance, this translator can be used if working with the Job API, e.g. to distinguish between a FutureCancelledError thrown by the job’s runnable, or because the job was effectively cancelled.

To optionally specify the name of the job, which is used to name the worker thread (only in development environment) and for logging purpose. Optionally, formatting anchors in the form of {} pairs can be used in the name, which will be replaced by the respective argument.

To optionally specify the RunContext to be installed during job execution. The RunMonitor associated with the RunContext will be used as the job’s monitor, meaning that cancellation requests to the job future or the context’s monitor are equivalent. If no context is given, the job manager ensures a monitor to be installed, so that executing code can always query its cancellation status via RunMonitor.CURRENT.get().isCancelled().

To optionally set the trigger to define the schedule upon which the job will commence execution. If not set, the job will commence execution immediately after being scheduled, and will execute exactly once.

The trigger mechanism is provided by Quartz Scheduler, meaning that you can profit from the powerful Quartz schedule capabilities.

For more information, see http://www.quartz-scheduler.org.

Use the static factory method Jobs.newExecutionTrigger() to get an instance:

Learn more about ExecutionTrigger.

To optionally control the maximal concurrently level among jobs assigned to the same semaphore.

With a semaphore in place, this job only commences execution, once a permit is free or gets available. If free, the job commences execution immediately at the next reasonable opportunity, unless no worker thread is available.

A semaphore initialized to one allows to run jobs in a mutually exclusive manner, and a semaphore initialized to zero to run no job at all. The number of total permits available can be changed at any time, which allows to adapt the maximal concurrency level to some dynamic criteria like time of day or system load. However, a semaphore can be sealed, meaning that the number of permits cannot be changed anymore, and any attempts will be rejected.

A new semaphore instance can be obtained via Jobs class.

Learn more about ExecutionSemaphore.

To associate the job with an execution hint. An execution hint is simply a marker to mark a job, and can be evaluated by filters to select jobs, e.g. to listen to job lifecycle events of some particular jobs, or to wait for some particular jobs to complete, or to cancel some particular jobs. A job may have multiple hints associated. Further, hints can be registered directly on the future via IFuture.addExecutionHint(hint), or removed via IFuture.removeExecutionHint(hint).

To control how to deal with uncaught exceptions. By default, an uncaught exception is handled by ExceptionHandler bean and then propagated to the submitter, unless the submitter is not waiting for the job to complete via IFuture.awaitDoneAndGet().

This method expects two arguments: an optional exception handler, and a boolean flag indicating whether to swallow exceptions. 'Swallow' is independent of the specified exception handler, and indicates whether an exception should be propagated to the submitter, or swallowed otherwise.

If running a repetitive job with swallowing set to true, the job will continue its repetitive execution upon an uncaught exception. If set to false, the execution would exit.

To set the thread name of the worker thread that will execute the job.

To set the maximal expiration time upon which the job must commence execution. If elapsed, the job is cancelled and does not commence execution. By default, a job never expires.

For a job that executes once, the expiration is evaluated just before it commences execution. For a job with a repeating schedule, it is evaluated before every single execution.

In contrast, the trigger’s end time specifies the time at which the trigger will no longer fire. However, if fired, the job may not be executed immediately at this time, which depends on whether having to compete for an execution permit first. So the end time may already have elapsed once commencing execution. In contrast, the expiration time is evaluated just before starting execution.

When awaiting futures, the definition of 'done' and 'finished' state should be understood - 'done' means that the future completed either normally, or was cancelled. But, if cancelled while running, the job may still continue its execution, whereas a job which not commenced execution yet, will never do so. The latter typically applies for jobs scheduled with a delay. However, 'finished' state differs from 'done' state insofar as a cancelled, currently running job enters 'finished' state only upon its actual completion. Otherwise, if not cancelled, or cancelled before executing, it is equivalent to 'done' state. In most situations, it is sufficient to await for the future’s done state, especially because a cancelled job cannot return a result to the submitter anyway.

Besides of some overloaded methods, IFuture basically provides two methods to wait for a future to enter 'done' state, namely awaitDone and awaitDoneAndGet, with the difference that the latter additionally returns the job’s result or exception. If the future is already done, those methods will return immediately. For both methods, there exists an overloaded version to wait for at most a given time, which once elapsed results in a TimedOutError thrown.

Further, awaitDoneAndGet allows to specify an IExceptionTranslator to control exception translation. By default, DefaultRuntimeExceptionTranslator is used, meaning that a RuntimeException is propagated as it is, whereas a checked exception would be wrapped into a PlatformException. If you require checked exceptions to be thrown as they are, use DefaultExceptionTranslator instead, or even NullExceptionTranslator to work with the raw ExecutionException as being thrown by Java Executor framework.

It is further possible to await asynchronously on a future to enter done state by registering a callback via whenDone method. The advantage over registering a listener is that the callback is invoked even if the future already entered done state upon registration.

Because invoked in another thread, this method optionally accepts a RunContext to be applied when being invoked.

Use the method awaitFinished to wait for the job to finish, meaning that the job either completed normally or by an exception, or that it will never commence execution due to a premature cancellation. To learn more about the difference between 'done' and 'finished' state, click here. Please note that this method does not return the job’s result, because by Java Future definition, a cancelled job cannot provide a result.

Job Manager allows to await for multiple futures at once. The filter to be provided limits the futures to await for. This method requires you to provide a maximal time to wait.

Filters can be plugged by using logical filters like AndFilter or OrFilter, or negated by enclosing a filter in NotFilter. Also see Section 9.7 to create a filter to match multiple criteria joined by logical 'AND' operation.

Use the method awaitFinished to wait for multiple jobs to finish, meaning that the jobs either completed normally or by an exception, or that they will never commence execution due to a premature cancellation. To learn more about the difference between 'done' and 'finished' state, click here.

You are running in a semaphore aware job and require to do some long running operation. During that time you do not require to be the permit owner. A simple but wrong approach would be the following:

The problem with this approach is, that you still are the permit owner while waiting, meaning that you possibly prevent other jobs from running. Instead, you could use a blocking condition for that to achieve:

Regardless of the schedule used, job manager guarantees no concurrent execution of the same job. That may happen, if using a repeatedly schedule with the job not terminated its last execution yet, but the schedule’s trigger would like to fire for the next execution already. Such a situation is called a misfiring. The action to be taken upon a misfiring is configurable via the schedule’s misfiring policy. A policy can be to run the job immediately upon termination of the previous execution, or to just ignore that missed firing. See the JavaDoc of the schedule for more information.

In order to deal with multiple ui-server nodes, the server holds a single notifications queue per ui-server node.

In this queues only the relevant notifications need to be kept: If a client notification is addressed to a session or user, that does not exist on a ui-server node, it is not added to the queue.

Sessions and corresponding users are registered on the server upon creation (and de-registered after destruction).

The ClientNotificationAddress determines which how the client notification needs to be dispatched and handled. A client notification can be addressed to

Client notifications can be published in a transactional or non-transactional way.

Generally, it makes sense to distribute the client notifications automatically to all other server cluster nodes (if available). This is achieved using ClusterNotifications. It is however also possible to publish client notifications without cluster distribution. E.g. in case of client notifications already received from other cluster nodes.

It is possible that a service generates a lot of client notifications that are obsolete once a newer notification is created. In this case a coalescer can be created to reduce the notifications:

To create a new client notification handler for a specific client notification, all you need to do is creating a class implementing org.eclipse.scout.rt.shared.notification.INotificationHandler<T>, where T is the type (or subtype) of the notification to handle.

The new handler does not need to be registered anywhere. It is available via jandex class inventory.

Sometimes it is necessary to start and stop handling notification dynamically, (e.g. when a form is opened) in this case AbstractObservableNotificationHandler can be used to add and remove listeners.

Dispatching is always done asynchronously. However, in case of transactional notifications, a service call blocks until all transactional notifications returned with the service response are handled.

This behavior was implemented to simplify for example the usage of shared caches:

In the example above, it is guaranteed, that the codetype is up-to-date as soon as reloadCodeType is finished.

Notification handlers are never called from a scout model thread. If the scout model needs to be updated when handling notifications, a model job needs to be created for that task.

The following example changes the initial value of a StringField called NameField:

Note: The type parameter of the extension (e.g. NameField) denotes the element which is extended.

The extension needs to be registeres when starting the application:

Extending a AbstractForm and one (or more) of its AbstractFormHandlers that can be achieved as follows:

There are a few things to note about this example:

The X-Frame-Options HTTP response header ^[3] can be used to indicate whether or not a user agent should be allowed to render a page in a <frame>, <iframe> or <object>. Sites can use this to avoid clickjacking ^[4] attacks, by ensuring that their content is not embedded into other sites. The X-Frame-Options header is described in RFC 7034 ^[5].

In Scout this header is set to SAMEORIGIN which allows the page to be displayed in a frame on the same origin (scheme, host and port) as the page itself only.

This header enables the XSS ^[6] filter built into most recent user agents. It’s usually enabled by default anyway, so the role of this header is to re-enable the filter for the website if it was disabled by the user. The X-XSS-Protection header is described in controlling-the-xss-filter.

In Scout this header is configured to enable XSS protections and instructs the user-agent to block a page from loading if reflected XSS is detected.

Content Security Policy is a HTTP response header that helps you reduce XSS risks on modern user agents by declaring what dynamic resources are allowed to load ^[7]. The CSP header is described in Level 1 and Level 2. There is also a working draft for a Level 3.

Scout makes use of Level 1 (and one directive from Level 2) and sets by default the following settings:

If a resource is blocked because it violates the CSP a report is created and logged on server side using level warning. This is done in the class ContentSecurityPolicyReportHandler. This enables admins to monitor the application and to react if a CSP violation is detected.

First the existence of the HttpOnly flag is checked. The servlet container will then add this flag to the Set-Cookie HTTP response header. If the user agent supports this flag, the cookie cannot be accessed through a client side script. As a result even if a cross-site scripting (XSS) flaw exists and a user accidentally accesses a link that exploits this flaw, the user agent will not reveal the cookie to a third party. For a list of user agents supporting this feature please refer to OWASP.

It is recommended to always enable this flag.

Since Java Servlet 3.0 specification this property can be set in the configuration in the deployment descriptor WEB-INF/web.xml:

Second the existence of the Secure flag is checked. The servlet container will then add this flag to the Set-Cookie HTTP response header. The purpose of the secure flag is to prevent cookies from being observed by unauthorized parties due to the transmission of a the cookie in clear text. Therefore setting this flag will prevent the user agent from transmitting the session id over an unencrypted channel.

Since Java Servlet 3.0 specification this property can be set in the configuration in the deployment descriptor WEB-INF/web.xml:

This of course only makes sense if the application is exposed to the end user using an encrypted channel like HTTPS (which is strongly recommended).

Unfortunately for the UI server it is not possible to detect if an application uses a secured channel. Consider the following example: The servlet container is protected by a reverse proxy. The communication between the user agent and the proxy is encrypted while the channel between the proxy and the servlet container is not. In this scenario the container cannot know that from a user agent point of view the channel is secured.

Because of this the validation assumes that the channel from the user agent to the entering node is secured and by default checks for the Secure flag. In case this assumption is not true and an unencrypted channel must be used this validation step can be disabled by setting the following property in the config.propertis file:

scout.auth.cookie.session.validate.secure=false

This skips the Secure flag check completely. In this scenario (not using https) it is also required to remove the secure tag from the cookie config in the WEB-INF/web.xml.

By default, all input in the Scout model is encoded. Examples are values/labels in value fields, cells in tables, message in message box. The reason behind this default choice is that developers do not have to think about output encoding in the standard case and are therefore less likely to forget output encoding and introduce a security vulnerability.

Example: In the following label field, the HTML <b> tag is encoded as <b>bold text</b>:

Sometimes developers may want to use HTML in the Scout model.

Examples are

Html input should only partially be encoded or not at all.

To disable the encoding of the whole value, the property HtmlEnabled can be used:

There are several ways to implement the use cases above. Some typical implementations are described in the following sections.

The JAX-WS Scout integration provides a thin layer on top of JAX-WS implementors to facilitate working with webservices. It depends on the JAX-WS 2.2.x API as specified in JSR 224. It is implementor neutral, and was tested with with the following implementations:

The integration does not require you to bundle the JAX-WS implementor with your application, which is a prerequisite for running in an EE container.

A servlet container like Apache Tomcat typically does not ship with a JAX-WS implementor. As the actual implementor, you can either use JAX-WS RI as shipped with the JRE, or provide a separate implementor like JAX-WS METRO or Apache CXF in the form of a Maven dependency. However, JAX-WS RI does not provide a servlet based entry point, because the Servlet API is not part of the Java SE specification.

When publishing webservices, it therefore is easiest to ship with a separate implementor: Declare a respective Maven dependency in your webbapp project - that is the Maven module typically containing the application’s web.xml.

When running in an EE container, the container typically ships with a JAX-WS implementor. It is highly recommended to use that implementor, primarily to avoid classloading issues, and to further profit from the container’s monitoring and authentication facility. Refer to the containers documentation for more information.

JAX-WS Scout integration is prepared to run with different implementors. Unfortunately, some implementors do not implement the JSR exactly, or some important functionality is missing in the JSR. To address this fact without loosing implementor independence, the delegate bean JaxWsImplementorSpecifics exists.

As of now, Scout ships with three such implementor specific classes, which are activated via config.properties by setting the property jaxws.implementor with its fully qualified class name. By default, JAX-WS METRO implementor is installed.

For instance, support for Apache CXF implementor is activated as following:

Of course, other implementors can be used as well. For that to work, install your own JaxWsImplementorSpecifics class, and reference its fully qualified name in config.properties.

The Maven plugin 'org.codehaus.mojo:jaxws-maven-plugin' with the goal 'wsimport' is used to generate a webservice stub from a WSDL file and its referenced XSD schema files. If your Maven module inherits from the Scout module 'maven_rt_plugin_config-master', the 'jaxws' profile is available, which activates automatically upon the presence of a 'WEB-INF/wsdl' folder. Instead of inheriting from that module, you can alternatively copy the 'jaxws' profile into your projects parent POM module.

This profile is for convenience purpose, and provides a ready-to-go configuration to generate webservice stubs and webservice provider artifacts. It configures the 'jaxws-maven-plugin' to look for WSDL and XSD files in the folder 'src/main/resources/WEB-INF/wsdl', and for binding files in the folder '/src/main/resources/WEB-INF/binding'. Upon generation, the stub will be put into the folder 'target/generated-sources/wsimport'.

The profiles requires the Scout runtime version to be specified, and which is used to refer to org.eclipse.scout.jaxws.apt module to generate webservice provider artifacts. However, this version is typically defined in pom.xml of the parent module, because also used to refer to other Scout runtime artifacts.

If your project design envisions a separate JAR module per WSDL, you simply have to set the property 'jaxws.wsdl.file' with the name of your WSDL file in the module’s pom.xml (example in Listing 73).

Otherwise, if having multiple WSDL files in your JAR module, some little more configuration is required, namely a respective execution section per WSDL file. Thereby, the 'id' of the execution section must be unique. Scout 'jaxws' profile already provides one such section, which is used to generate the stub for a single WSDL file (see such configuration in Listing 73), and names it 'wsimport-1'. It is simplest to name the subsequent execution sections 'wsimport-2', 'wsimport-3', and so on.
For each execution section, you must configure its unique id, the goal 'wsimport', and in the configuration section the respective wsdlLocation and wsdlFile. For 'wsimport' to work, wsdlLocation is not required. However, that location will be referenced in generated artifacts to set the wsdl location via @WebService and @WebServiceClient. The complete configuration is presented in Listing 74.

Further, you can overwrite any configuration as defined by 'jaxws-maven-plugin'. See http://www.mojohaus.org/jaxws-maven-plugin/ for supported configuration properties.

Also, it is good practice to create a separate folder for each WSDL file, which also contains all its referenced XSD schemas. Then, do not forget to change the properties wsdlLocation and wsdlFile accordingly.

By default, all XML files contained in folder 'WEB-INF/binding' are used as binding files. But, most often, you will have a global binding file, which applies to all your WSDL files, and some custom binding files different per WSDL file and XSD schema files. See how to explicitly configure binding files in Listing 75.

With binding files in place, you can customize almost every WSDL component and XSD element that can be mapped to Java, such as the service endpoint interface class, packages, method name, parameter name, exception class, etc.

The global binding file typically contains some customization for common data types like java.util.Date or java.util.Calendar, whereas the custom binding files are specific for a WSDL or XSD schema. See Section 16.7.

By default, generated artifacts are put into the package corresponding to the element’s namespace. Sometimes, you like to control the package names, but you want to do that on a per-namespace basis, and not put all the artifacts of a webservice into the very same package. That is mainly to omit collisions, and to have artifacts shared among webservices not duplicated.

Two separate binding files are required to customize WSDL components and XSD schema elements.
That is because WSDL component customization is to be done in 'jaxws' namespace http://java.sun.com/xml/ns/jaxws, whereas XSD schema element customization in 'jaxb' namespace http://java.sun.com/xml/ns/jaxb.

Annotation Processing (APT) is a tool which can be enabled to fire for annotated types during compilation. In JAX-WS Scout integration, it is used as a trigger to generate webservice port type implementations. Such an auto-generated port type implementation is called an entry point. It is to be published as the webservice’s endpoint, and acts as an interceptor for webservice requests. It optionally enforces for authentication, and makes the request to be executed in a RunContext. Then, it handles the web request to the effectively implementing port type bean for actual processing.

The entry point generated simplifies the actual port type implementation by removing lot of glue code to be written by hand otherwise. Of course, this entry point is just for convenience purpose, and it is up to you to make use of this artifact.

When using 'jaxws' Scout Maven profile, annotation processing is enabled for that module by default. But, an entry point for a webservice port type will only be generated if enabled for that port type, meaning that a class annotated with WebServiceEntryPoint pointing to that very endpoint interface is found in this module. Anyway, for a sole webservice consumer, it makes no sense to generate an entry point at all.

Simply run mvn clean compile on the project. If you are experiencing some problems, run with -X debug flag to get a more detailed error message.

In the Eclipse IDE, there are three ways to generate webservice stubs and APT artifacts.

If the webservice stub(s) or APT artifacts are not generated (anew or at all), delete the target folder manually, and continue according to procedure number three. A possible reason might be the presence of 'target\jaxws\wsartifact-hash'. Then, for each webservice, a 'hash file' is computed by 'wsimport', so that regeneration only occurs upon a change of WSDL or XSD files.

Stub and APT artifacts are derived resources, and should be excluded from version control. When working with Eclipse IDE, this is done automatically by eGit, because it adds derived resources to .gitignore (if configured to do so).

JaxWsAnnotationProcessor is an annotation processor provided by Scout JAX-WS integration to generate an entry point for an endpoint interface during compilation. The instructions how to generate the entry point is given via a Java class or Java interface annotated with WebServiceEntryPoint annotation.

An entry point implements the endpoint interface (or port type interface), and is published as the webservice endpoint for that endpoint interface. The entry point itself is auto generated by JaxWsAnnotationProcessor during compile time, based on instructions as given by the respective class/interface annotated with WebServiceEntryPoint annotation. The entry point is responsible to enforce authentication and to run the web request in a RunContext. In turn, the request is propagated to the bean implementing the endpoint interface.

Figure 13 illustrates the endpoint’s class hierarchy and the message flow for a web request.

As you can see, both, entry point and port type implement the endpoint interface. But it is the entry point which is actually installed as the webservice endpoint, and which receives web requests. However, the webservice itself is implemented in the implementing bean, which typically is located in server module. See Section 16.3 for more information. Upon a web request, the entry point simply intercepts the web request, and then invokes the web method on the implementing bean for further processing.

See an example of an implementing port type bean, which is invoked by entry point.

This section describes the steps required to generate an entry point. For demonstration purposes, a simple ping webservice is used, which provides a single method 'ping' to accept and return a String object.

See the WSDL file of ping webservice: Section 16.8.1
See the endpoint interface of ping webservice: Section 16.8.2

To generate an entry point for the webservice’s endpoint interface, create an interface as following in your jaxws project.

It is important to understand, that the interface PingWebServiceEntryPointDefinition solely acts as the anchor for the WebServiceEntryPoint annotation. This class or interface has no special meaning, except that it declares the annotation to be interpreted by annotation processor. Typically, this class is called Entry Point Definition.

If running mvn clean compile, an entry point is generated for that endpoint interface. See the entry point as generated for ping webservice: Section 16.8.3

If you should experience some problems in the entry point generation, refer to Build webservice stubs and APT artifacts from within Eclipse IDE, or Build webservice stubs and APT artifacts from console.

This section gives an overview on how to configure the entry point to be generated.

Besides the instructions which can be set via WebServiceEntryPoint annotation, it is further possible to contribute other annotations to the entry point. Simply declare the annotation of your choice as a sibling annotation to WebServiceEntryPoint annotation. In turn, this annotation will be added to the entry point as well. This may be useful to enable some vendor specific features, or e.g. to enable MTOM to efficiently send binary data to a client.
That also applies for WebService annotation to overwrite values as declared in the WSDL file.

Further, you can also provide your own handler chain binding file. However, handlers and authentication as declared via WebServiceEntryPoint annotation are ignored then.

This configuration generates the following artifacts:

All artifacts are generated into the package 'org.eclipse.scout.docs.ws.ping', as specified by the definition. The entry point itself is generated into 'PingWebServiceEntryPoint.java'. Further, for each handler, a respective handler delegate is generated. That allows handlers to be looked up via bean manager, and to run the handlers on behalf of a RunContext. Also, an AuthHandler is generated to authenticate web requests as configured.

The handler-chain XML file generated looks as following. As specified, the authentication handler is installed as the third handler.

The following listing shows the beginning of the entry point generated. As you can see, the handler-chain XML file is referenced via HandlerChain annotation, and the MTOM annotation was added as well.

See listing for an example of how to configure JAX-WS handlers.

JAX-WS handlers are configured directly on the entry point definition via the array field handlerChain. In turn, JaxWsAnnotationProcessor generates a 'handler XML file' with the handler’s order preserved, and which is registered in entry point via annotation handlerChain.

A handler can be initialized with static 'init parameters', which will be injected into the handler instance. For the injection to work, declare a member of the type Map in the handler class, and annotate it with javax.annotation.Resource annotation.

Because handlers are looked up via bean manager, a handler must be annotated with ApplicationScoped annotation.

If a handler requires to be run in a RunContext, annotate the handler with RunWithRunContext annotation, and optionally specify a RunContextProducer. If the web request is authenticated upon entering the handler, the RunContext is run on behalf of the authenticated user. Otherwise, if not authenticated yet, it is invoked with the Subject as configured in jaxws.provider.user.handler config property.

Sometimes it is useful to share state among handlers, and even with the port type. This can be done via javax.xml.ws.handler.MessageContext. By default, a property put onto message context is only available in the handler chain. To make it available to the port type as well, set its scope to 'APPLICATION' accordingly.

The following listings gives an example of how to propagate state among handlers and port type.

Scout’s JAX-WS integration already provides complete support for reading a correlation ID from the HTTP header named X-Scout-Correlation-Id of the incoming web service request and propagates it to the RunContext that executes the actual service operation. A new correlation ID is created if the HTTP header is empty or missing.

The registration of webservice endpoints depends on the implementor you use, and whether you are running in an EE container with webservice auto discovery enabled.

When running in an EE container, webservice providers are typically found by their presence. In order to be found, such webservice providers must be annotated with WebService annotation, and must have the coordinates 'serviceName' and 'portName' set. Still, most application servers allow for manual registration as well. E.g. if using Oracle WebLogic Server (WLS), endpoints to be published can be registered directly in 'web.xml' as a Servlet. However, this is vendor specific. Refer to the container’s documentation for more information.

If not running in an EE container, the registration is implementor specific. In the following, an example for JAX-WS METRO and Apache CXF is given.

A webservice operation is invoked on behalf of an invocation context, which is associated with a dedicated port, and which specifies the data to be included in the web request. Upon a webservice call, the invocation context should be discarded.

Invoking newInvocationContext() returns a new context and port instance. The context returned inherits all properties as configured for the WS-Client (endpoint URL, credentials, timeouts, …​), but which can be overwritten for the scope of this context.

The following listing illustrates how to set/overwrite properties.

The WS-Client provides port instances via a preemptive port cache. This cache improves performance because port creation may be an expensive operation due to WSDL/schema validation. The cache is based on a 'corePoolSize', meaning that that number of ports is created on a preemptively basis. If more ports than that number are required, they are created on demand, and additionally added to the cache until expired, which is useful at a high load.

The WS-Client supports for cancellation of webservice requests. Internally, every web request is run in another thread, which the calling thread waits for to complete. Upon cancellation, that other thread is interrupted, and the calling thread released with a WebServiceRequestCancelledException. However, depending on the JAX-WS implementor, the web request may still be running, because JAX-WS API does not support the cancellation of a web request.

The invocation context allows you to access HTTP status code and HTTP headers of the last web request.

An invocation context can be associated with request context properties, which are propagated to handlers and JAX-WS implementor.

Learn more how to access context properties from within a handler in Listing 84.

To install a handler, overwrite execInstallHandlers and add the handler to the given List. The handlers are invoked in the order as added to the handler-chain. By default, there is no handler installed.

The method execInstallHandlers is invoked upon preemptive creation of the port. Consequently, you cannot do any assumption about the calling thread.

If a handler requires to run in another RunContext than the calling context, annotate it with RunWithRunContext annotation, e.g. to start a new transaction to log into database.

If the endpoint requires to authenticate requests, an authentication handler is typically added to the list, e.g. BasicAuthenticationHandler for 'Basic authentication', or WsseUsernameTokenAuthenticationHandler for 'Message Level WS-Security authentication', or some other handler to provide credentials.

The current context’s correlation ID can be forwarded to the consumed web service. Scout provides a handler that sets the X-Scout-Correlation-Id HTTP header on the outgoing request.

The following properties can be set globally for all WS-Clients. However, a WS-Client can overwrite any of this values.