We are currently investigating an issue with the editor of some pages. Please save your work and avoid to create new pages until this banner is gone.
Heiko Sommer (email@example.com)
|Current Version (v. 5)||Apr 13, 2019 22:51||Jorge Avarias|
|v. 4||Apr 13, 2019 22:33||Jorge Avarias|
|v. 3||Apr 13, 2019 22:03||Jorge Avarias|
|v. 2||Apr 13, 2019 22:00||Jorge Avarias|
|v. 1||Apr 13, 2019 22:00||Jorge Avarias|
The purpose of this document is to give a practical introduction to writing Java components for ALMA software, using the ACS container/component framework. You should have read the section on Technical Architecture in  or have learned otherwise about the concepts of container/component and XML binding classes to be used for ALMA software. You should also be familiar with the ALMA build environment (see ), even though a few things are explained redundantly in this tutorial.
While describing the steps involved in developing two sample components, I will try to provide information beyond the scope of the demo components in order to help using the framework for concrete ALMA subsystem development. However, this is not a concept or design document for the respective parts of ACS. It should help you to get started nonetheless.
The Java container/component model is fully integrated in ACS. It is meant to be used by ALMA subsystems or the parts of them that don’t have real-time requirements and don’t directly control hardware devices.
This document no longer maintains a separate section for abbreviations. Please use the ALMA software glossary .
Readers should feel very much encouraged to follow at their computers the steps described in the chapters below, switching between this tutorial and the real software.
The Java parts of the ACS framework were developed using the Eclipse IDE. Eclipse is not one of the “official” ALMA standards, but a majority of ALMA developers finds it so helpful that I like to recommend getting it. You can download Eclipse for free from . The latest milestone build is usually fine, but special plugins may require to use a released version.
We have started a description on how to configure Eclipse to work well together with the ALMA build system . It’s expected to grow, feedback is welcome. As background information, notes from a presentation of Eclipse given at ESO are also available on that page.
The ACS release (currently ACS 4.1.1) contains source code inside the delivered jar files, both for the implementation of the container, compiled IDL files etc, as well as for the examples used in this tutorial.
Automatically generated documentation and the source code for the examples can also be found online at
A live web-based CVS view for all relevant ACS modules is at https://bitbucket.sco.alma.cl/projects/ASW/repos/acs/browse/LGPL/CommonSoftware
The modules that play a major role in the Java container/component part of ACS are listed below in compile order, together with the jar files they produce.
Misc. utility classes, test support
Java binding classes for XML (schema compiler, runtime) based on Castor
Java code generator for component helper classes, see 6
ACS IDL compiler (produces XML binding class aware component interfaces)
IDL and XML schema files that are part of the framework;
xmlentity.jar, archive_xmlstore_if.jar, systementities.jar
IDL and XML schema files used for demos and testing; forked from ObsPrep
Java container, component interfaces
Examples used in this tutorial, other examples (e.g. used for testing)
HelloDemo.jar, XmlComponent.jar, BrightLamp.jar, BrightUser.jar, jcontexmpl.jar
Here's an overview of the directories and files in the module jcontexmpl:
Nothing too exciting since the API documentation is generated by the Makefile and is not supposed to be checked in.
definition of the HelloDemo component used in this tutorial
definition of the XmlComponent component used in this tutorial
other demo components not considered in this tutorial
definition of errors with certain types and codes, see 
not used, but required by build system
compilation of /src, produced by running the Makefile in /src
compilation of /idl/HelloDemo.idl
compilation of /idl/XmlComponent.idl
other .jar, .so, .a files
output of Java/C++ IDL compilation of the IDL files
not used, but required by build system
demo components and test clients
tests internal to ACS that use the demo components
To compile and run the software, you need a current installation of ACS. The basic concepts for Java components have not changed since ACS 3.0 though.
The /src and /test directories each contain a Makefile. The command
make clean all
will compile IDL and Java code and pack the results into jar files.
will copy IDL, schema, and jar files to the respective sub-directories of your INTROOT directory so that they are available for other modules.
To learn about using the make system, refer to "man acsMakefile" and "man javaMakefile".
To run the container, you must first start the ACS services & manager. Either use the graphical ACS Command Center (see ), or the command
Make sure your ACS_CDBvariable points to the CDB installation that you want to use. You find a CDB suitable for the examples under jcontexmpl/test in CVS. You could either use that CDB directly, or merge the entries from Components.xml into your $ACS_CDB/CDB/MACI/Components/Components.xml.
To run the container on the same machine where the Manager runs, use the command
acsStartContainer –java <containername>
For cases where the manager runs on a different machine, please refer to  (“FAQGeneralCompContainerOptions”).
To terminate a Java container, you can use the command line
or the brute-force approach with Ctrl-C on the container console; in this case the Manager will attempt to reload components in that container next time the container runs.
The rest of ACS you terminate with
acsStop (or killACSif all else fails).
The recommended method is Java remote debugging, which is described in the ACS FAQs  under “How do I debug my components?”.
Windows aficionados can run the Java container with its components (or just a client application that accesses components) on a Windows machine that has a network connection to another machine where the rest of ACS is running. Different subversions of JDK 1.4 have worked fine on Windows.
The build process (make) is not available under Windows. Assuming you have ACS installed on a Linux machine, you should keep your source code there, and mount the directories on Windows, e.g. using Samba The ACS FAQ will describe the setup in more detail.. The initial compilation must be done on Linux to compile IDL etc. Subsequent plain Java compilations can then be done from Windows.
There is an issue with CR/LF line termination: if you get ASCII sources from CVS on Windows, they come with Windows line breaks; the build process runs on Linux though, and a few tools will fail without telling you why. The Makefile itself and the IDL files are affected. In case of doubt, run dos2unix on them, e.g.
~/CVSPRJ/ACS/jcontexmpl/src 1001 > dos2unix Makefile
By default, log messages are sent to the local console and to the central ACS logger. There they simply get absorbed in the void if you don't register for them. For example,
acsStartLoggingClient | tee mylog.xml
will send all logs messages to the file mylog.xml and also to the console where you typed the command.
To adjust the log levels that get through, you may provide your own Java logging properties file. It must be compliant to the JDK 1.4 logging spec. Refer to the FAQ for the details. As a template for the properties file, you could use almalogging.properties from acsjlog/src.
There is also a graphical application "LoggingClient" that you can start with the command
During the design of your ALMA subsystem, you should partition your software into one or preferably several components The "subsystem master component" concept may be included in this tutorial in the future. A component will be the smallest chunk of software that can be deployed separately, e.g. in its own process or even on its own machine. Guidelines are that a component should be small enough to be functionally cohesive (having exactly one functional interface with related methods), but large enough to keep the number of components sufficiently small. Typically, a component will be made up of several Java classes, and a subsystem might consist of a handful of components.
There can be components that will be used by other subsystems, while other components are only used internally by the subsystem. In either case, the functional interface must be specified as CORBA IDL.
The example components that we'll look at in the tutorial are not jewels of OO design, but they are meant to illustrate the steps involved to get to an implementation. The detailed motivation for all methods used in the interfaces is described in the next section about IDL.
We will work with
The components and their containers can be pictured as follows, with the “antennas” sticking out on top being the functional interfaces.
The two Java components, HelloDemo and XmlComponent, can either run collocated inside one Java container as shown, or run in separate Java containers. The Java container is drawn as a closed box to indicate that it works as a “tight container” (cf. ), intercepting all functional calls to any of its components, in addition to starting/stopping the components and providing explicit services to them.
The Java components can talk with, say, a C++ component “otherComp”, which must run inside a C++ container (“other Container”). That container is drawn open to indicate that it is a “porous” kind of container which only takes care of starting/stopping the components, but does not intercept functional calls on the running components.
The ACS containers use CORBA and the ACS services, the ACS Manager, and the Configuration Database. The Manager itself also uses the CDB.
A component's functional interface must be specified as an interface in CORBA's Interface Definition Language (IDL). An IDL file may contain more than one interface. The Java container framework does not require further restrictions Currently there is a problem with IDL attributes which will be fixed later. on the usage of the more exotic IDL constructs beyond what ACS already mandates for the definition of C++ components.
Using the ALMA directory structure, you should put IDL files in the <xxxx>/idl directory of your module. Our HelloDemo component is defined in jcontexmpl/idl/HelloDemo.idl.
Let's look at the listing of HelloDemo.idl with line-by-line explanations.
1, 2, 16
Standard include guard to avoid multiple inclusions of this IDL file
Mandatory include of acscomponent.idl (see line 9)
Standard CORBA directive that affects generated Java classes At least for what we care about here; primarily the prefix controls the CORBA repository ID.. The prefix pragma must come after the include statements (this restriction might be lifted in the future, but for now the TAO IFR would have a problem otherwise.)
Declaration of the CORBA module "demo". A module may contain many interfaces.
Interface declaration for the HelloDemo component. Every ALMA component's IDL interface must inherit from ACS::ACSComponent. This ensures that all components have a unique instance name and a state visible to other components.
Silly methods using CORBA types string, double, and long
After the IDL file HelloDemo.idl has been written, a couple of Java classes must be generated from it.
The ALMA Makefile will take care of this, given the line (see jcontexmpl/src/Makefile)
IDL_FILES = HelloDemo
The generated Java classes are then automatically compiled and put into /lib/HelloDemo.jar. They belong to the package alma.demo and can be categorized as
Helper & holder classes
Warning: don't choose the same name for IDL_FILES as for (one of the values of) JARFILES in your Makefile! Otherwise the jar files produced from the IDL and from your module's code have the same name and one gets overwritten by the other.
The implementation of a Java component is a Java class that must implement
The implementation of a Java component is a Java class that must implement
The implementation class may inherit from any base class (This is possible because for Java components the framework uses the CORBA tie delegation model; C++ components can use multiple inheritance and extend the CORBA skeleton directly, thus saving one runtime delegation step.); no such base class is required by the framework though, which saves the single inheritance option offered by Java.
For components that don’t need inheritance for other purposes, there is a generic base class (alma.acs.component.ComponentImplBase)that provides standard implementations of the required methods from ComponentLifecycle and ACSComponent, just to keep the component code simpler.
In real life we would usually first create the component implementation class with just dummy implementations for all mandatory methods, e.g. using code generation features of an IDE like Eclipse, and then move on to the next steps described in sections 6and below. This document will ignore the iterative process and look at the final implementations right away.
The functional interface for the HelloDemo component is HelloDemoOperations, a Java interface generated by the CORBA IDL compiler.
By convention, we call the implementation class “HelloDemoImpl” and create it in the package “alma.demo.HelloDemoImpl”. It can be found under jcontexmpl/src in CVS.
The package declaration in accord with the "subpackage-per-component-implementation" convention
Import for logging (standard JDK logger configured by ACS)
Imports for CORBA Holder classes for the OUT parameters in the sayHelloWithParameters method (as defined in )
Import of the ComponentStates interface, needed to access (and optionally modify) the component state
Import for the mandatory Lifecycle interface
Import for the ContainerServices interface which the container provides to the component. From this interface, the component gets everything the container can do for it on request (that is, not transparently w/o the component noticing)
Import for the HelloDemoOperations interface, which defines the functional methods that a client of our component would use.
The component implementation class implements the functional and the lifecycle interface; it does not inherit from any base class (if so, there would be even less here to look at…)
Reference to the ContainerServices object that the container will give us (see line 47)
Ref to the logger object which we'll get from the container (see line 49)
The initialize method (from ComponentLifecycle) is called by the container after creating an instance of our component. It provides us with the ContainerServices object. We don't use it for much besides getting a logger and storing that reference in a separate variable for easy access.
Implementation of the execute method (from ComponentLifecycle). This method will be called by the container after initialize() has returned. Only few components are foreseen to use both initialize and execute; here we just log a message.
Implementation of the cleanUp method (from ComponentLifecycle). This method will be called by the container when the component is getting dismissed by the ACS Manager, but only after the last call to any of the methods from HelloDemoOperations has returned.
Implementation of the aboutToAbort method (from ComponentLifecycle). This method may be called asynchronously when the container or the entire ALMA system must shut down without waiting for proper termination of all components. It is meant as a notification to the component to perform the most urgent actions before being forcefully removed at any time.
Implementation of the componentState() method (from ACSComponent).
Implementation of the name() method (from ACSComponent). The name is the instance name of our component, e.g. something we can't know at compile time. We get the name from the container, using the ContainerServices interface.
Implementation of the sayHello method which is declared in HelloDemoOperations.
Implementation of the sayHelloWithParameters method: we set the OUT parameter and leave the INOUT parameter untouched.
Note the use of DoubleHolder and IntHolder for outgoing parameters, a concept that is not possible in Java without using these helper classes; they are therefore defined in the CORBA IDL-to-Java mapping spec. 
Note that the implementation of the HelloDemo component is done with just one Java class. A real component would rather have one main implementation class that implements the component interfaces and uses other classes to perform the functional tasks.
Unlike the previous HelloDemo, the component XmlComponent uses entity objects that are transported over the network as serialized XML. We will focus on this feature and not repeat things which have already been demonstrated before.
The term “entity” refers to non-primitive, non-binary data with identity such as scheduling blocks, system users, correlator configurations etc., c.f. . These data types are defined in XML schemas. Their instances are XML “documents” that get passed around by value and can be stored in the ALMA Archive.
From the XML schemas, we produce Java classes at build-time, based on the Castor framework . These binding classes represent the hierarchical data structure in memory. They have type-safe methods to navigate the tree and to access data items. The binding classes can be automatically instantiated from XML documents that correspond to the same schema, and can serialize their data into stringified XML.
The idea of “transparent XML entities” is to have the entity objects transported by CORBA as language-neutral XML strings. However, the Java component implementations (both client and server involved in the remote call) don’t have to worry about the serialization and parsing of the XML. The main advantage of this is that the framework ensures type safety, from the IDL/schema definitions, all the way down to the Java implementation classes of a component; it’s not possible to be surprised at runtime that the received XML differs in type or version from what the component expected. To achieve this, the implementation classes use modified component interfaces, in which Java binding classes are substituted for the structs with serialized XML. An example with detailed explanation will be given in 5.3.
A Java component that acts either as a client or as a server with respect to some other component, is completely independent in its decision whether to parse XML entity objects “by hand” or to enjoy transparent support for binding classes. The other component may be written in Java, C++, or whatever other language, and it may or may not use binding classes (If this other component is written in Java, the developer has a real choice on this; for C++ at least at the moment there is no support for XML binding classes, so the C++ XML parser (expat) that comes with ACS should be used.).
For example, a Java client that exchanges XML entity objects with a C++ server can work with Java binding classes, whereas the C++ server parses the XML using a SAX parser.
We define complex data types in XML schema and use these types in CORBA IDL similar to how CORBA structs would be used.
To keep this tutorial focused on component development, we only touch on the schema development issue. Rather than showing how to develop a new schema, we use the existing schemas from the module acstestentities.
For XML schemas describing ALMA entities, there are the following conventions:
Schema files must be put into the /idl directory of your module, or into the ICD/<yourSubsystem>/ws/idl directory if Like IDL files are meant to be used by other subsystems. Please check with ALMA software engineering.
You must add information to your module's Makefile so that the XML schemas can be compiled into Java binding classes. Let's look at how acstestentities/src/Makefile has it:
XSDBIND = acsTestEntities
tells the build process to look for a configuration file ../idl/acsTestEntities.xml which must provide information about the schema files to be compiled (see below). The resulting Java classes will then be packed into ../lib/acsTestEntities.jar.
Schema files can include or import other schema files. In our case, CommonEntity.xsd from the module ACS/define, systementities.jar, is imported. In our Makefile, the line
XSDBIND_INCLUDE = systementities
is required so that the build process can set the include path for the schema compiler.
Here's a reduced listing of the configuration file /idl/acsTestEntities.xml
We see that schema files are listed together with their xml namespaces (<EntitySchema>). Then the namespaces are mapped to Java packages (<XmlNamespace2JPackage>). Files and packages are separated here because several schema files can use the same xml namespace, but the Java package of the resulting classes must depend on the namespace, not the schema file.
Note that the compilation of xml schemas is specified differently in the Makefile, compared with the compilation of IDL files. IDL files are listed individually and are compiled into one jar file each, whereas schema files are only referenced indirectly through their XML configuration file, and are compiled into only one jar file per module.
The diagram below illustrates the relationships among the xml schema files, IDL files, and Makefiles from our module "acsjexmpl" and the modules "define" and "acstestentities" from where resources are used by the example module.
Using XML schema types in IDL – general idea
The diagram below shows how to use the XML schema definition of a SchedBlock in the IDL definition of a component interface.
Here's a thinned-out listing.
xmlentity.idl comes from define/idl and declares the CORBA struct XmlEntityStruct which is used to transport serialized XML together with some meta data. It must be included whenever xml entity objects are used (see I-K). The developer does usually not have to know the details of that struct.
The prefix pragma must come after the include statements; this restriction might be lifted in the future, but for now the TAO IFR would have a problem otherwise.
We use the same module as for the HelloDemo component
Typedefs for xml entity classes. The interface methods that use entity classes as parameters or return types (see H-K) must use these named typedefs like 'ObsProposal' instead of 'XmlEntityStruct'. This not only makes the interface more readable, but is also required for the automatic use of Java binding classes.
Declaration of an exception. (How to use the ACS Error System  instead of plain CORBA exceptions may be demonstrated in a future version of this document.)
A struct that contains XML entities.
Interface declaration for the XmlComponent
Just to have something dumb in here
Method that returns an xml entity object (as a struct that contains the xml data as a string if the component is accessed as a plain CORBA object, see below.)
Demonstrates the use of a sequence (~array) of xml entity objects
Demonstrates the use of xml entity objects as an OUT parameter
Uses the ObsProjectTree struct, with entities inside
Uses an exception (again, later the mechanism described in  will be used.)
We have to add the new IDL file to jcontexmpl/src/Makefile:
IDL_FILES = HelloDemo XmlComponent
In addition to running this file through the standard CORBA IDL compiler, it is also fed into the "ACS IDL compiler". This tool gets started by the build process, so you never have to see it; the interested reader can find it in the directory ACS/…/XmlIdl. The ACS IDL compiler creates additional Java classes which the CORBA IDL compiler would not create. They are used for working with Java XML binding classes instead of plain XML strings, see 5.1.1.
In the Makefile, the line
XML_IDL = "ObsProposal=alma.xmljbind.test.obsproposal.ObsProposal; \
provides the mapping from IDL typedefs (see lines 6-8) to Java binding classes (conceptually: to the schema that defines the entities).
The Java files that are produced from XmlComponent.idl are
Helper & holder classes
ObsProposalHelper.class, ObsProposalHolder.class, SchedBlockHelper.class, SchedBlockHolder.class, SchedBlockSeqHelper.class, SchedBlockSeqHolder.class, SchedBlockSeqJHolder.class, XmlComponentHelper.class, XmlComponentHolder.class
The classes shown in normal print are mandated by CORBA standards and generated by the CORBA IDL compiler, while the italicized classes are required by the ACS framework and are generated by the ACS IDL compiler.
In section 5.2, we saw the IDL file with the typedefs for the entity classes, and their use in the methods of the XmlComponent interface.
The following two functional interfaces have been generated for XmlComponent. The component implementation must implement one of them:
For each component class, a corresponding helper Naming convention: For the component of an interface "xxx", the helper file is named as "xxxComponentHelper". This should not be so easily confused with the IDL-generated "xxxHelper". class is needed. In most cases this helper class will be used only by the framework, which means that the component developer will not have to look at it much.
The framework uses the helper class to
We will choose the convenience and type-safety offered when implementing XmlComponentJ. The interface generated by the ACS IDL compiler is listed below (Javadoc lines omitted).
Comparing this with the IDL definition, we see that binding classes like
alma.xmljbind.test.schedblock.SchedBlockare used whenever the IDL contained a typedef’d entity struct. In addition to this direct substitution, the ACS IDL compiler created the class
ObsProjectTreeJwhich contains binding classes, thus substituting the IDL struct
ObsProjectTreeas the return type of the method getEntireTreeInAStruct();
Let’s look at the implementation of the method createObsProposal().
We see that the Java class ObsProposal has type-safe methods like “setPerformanceGoals” or “setObsProposalEntity”.
The latter takes an object of type ObsProposalEntityT, also generated by the binding framework. It is used to store administrational information about the ObsProposal object, most important the object ID.
For a new entity object, a valid and unique object ID can be obtained conveniently through a call to ContainerServices#assignUniqueEntityId, for which the container collaborates with the archive. If the archive is not present, it defaults to using random numbers which should be almost always unique (2^64).
Our method implementation simply returns the top-level Java object for the ObsProposal. The container framework will then automatically serialize it to XML, embed the string in the XmlEntity CORBA struct, and send it to the client using CORBA. If the client is another Java component, its container will automatically create an ObsProposal class and fill it with the data from the serialized XML.
To keep the size of this tutorial small, the implementations of the other methods are not listed here. They are meant to serve as more detailed examples, for which the source code of the module jcontexmpl should be retrieved anyway.
Alternatively, the code listing can be found online at the ACS documentation site:
To see how the code would look like if the component would implement XmlComponentOperations, take a look at the inner class XmlComponentHelper::IFTranslator.
For each component class, a corresponding helperclass is needed. In most cases this helper class will be used only by the framework, which means that the component developer will not have to look at it much.
Naming convention: For the component of an interface "xxx", the helper file is named as "xxxComponentHelper". This should not be so easily confused with the IDL-generated "xxxHelper".
The framework uses the helper class to
A template for the helper class gets generated automatically by the build system if the Makefile contains the line
The template has a file ending .java.tpl to ensure that no edited .java helper class gets overwritten. It is meant as a convenience for the component developer who will normally just have to remove the .tpl ending and then treat that Java file like his/her own, which requires checking it into GIT.
For our HelloDemo component, the helper class does not need to be changed (some Javadoc lines stripped from the listing):
Only in special situations, when a Java component chooses to sometimes use Java binding classes, but other times send or receive serialized XML instead, then the helper class must be modified. This goes beyond the scope of this tutorial and will be described in a more specialized document.
An example for this can be found in alma.demo.XmlComponentImpl.XmlComponentComponentHelper from jcontexmpl/src/.
The ACS configuration database contains information about which components are deployed on the different machines. It assists the ACS Manager to locate components at runtime. All the runtime action is hidden from the programmer by the container.
There are two different types of components:
Let’s make HelloDemo a static component. We create a new line in the file CDB/MACI/Components/ Components.xml with
Note that the generated component helper class contains these values in a comment (code listing above, line 42). This facilitates adding the CDB entry using copy and paste.
The CDB file Components.xml then looks similar to
ACS ships with a default CDB that contains the entries for the demo components.
You’ll have to add entries for your own components, or create a new CDB structure and reset the ACS_CDB environment variable.
A component can have different kinds of clients which we'll discuss separately:
We can use the ACS tool ObjectExplorer as a generic client for all components. Run it with
and you should find your Java components (e.g. XmlComponent) and their instances (e.g. XMLCOMP1) listed on the dialog, together with C++ components (e.g. Lamp).
At first, you see only a visualization of the information in the CDB, as opposed to actually loaded components. When you click on a component instance, the Manager will tell the Container to activate the component with the given instance name.
The screenshot shows the output after XMLCOMP1 has been loaded and the method getBestSchedBlock has been called (with View → Expand result data checked in ObjectExplorer.)
We’ve seen the example of a Java component accessing a C++ Component in 4.2, and a Java component accessing another Java component without transparent XML binding support referenced in 5.1.
To use support for xml binding classes on the client side of a call, we first obtain the other component (here: XMLCOMP1)
and then ask the container to wrap it with the more convenient interface that uses XML binding classes, provided in the ContainerServices interface
Now we can keep working with xmlCompJ, and the framework will delegate all calls to xmlComp, translating between XML binding classes and serialized XML presentation in between.
To facilitate the writing of JUnit test clients, ACS (module jcont) offers the class
which is a subclass of JUnit’s TestCase that does the talking with ACS Manager. It provides most of the functionality that ContainerServices provides for components.
A JUnit test that gets a reference to the XmlComponent component and runs some tests on it is
The constructor must call the constructor of the base class with a name to be used for this client (here: “XmlComponentClient”). If you override the setup() method, it’s necessary there to call super.setUp() first.
JUnit will call all methods that start with “test” in their name, such as testException().
In the Eclipse Java Perspective, you can run such a class with JUnit. Open the class in the editor, then choose
Menu Run → Run As → JUnit Test. It will display the results.