SCA Java Development Guide : Apache Tuscany

This document is the development guideline for SCA Java 2.x project.

General Guide
Getting Source code
Setting up your development environment
Importing SCA modules into your Development IDE
Coding Guidelines
Testing
Client Applications
Maven Build Structure
Development Hints

OSGi Overview and Approach

General Guide

Welcome to the Tuscany SCA Java subproject project. We look forward to your participation and try to help you get on board. Feel free to ask your questions on the mailing list.

Here are some general guidelines we use in this project.

Java SCA sub-project aims to provide enterprise-grade service infrastructure based on SCA.
Tuscany SCA is not just a reference implementation. We encourage innovation based on the tenets of SCA. A lot of work we do provides feedback to the specifications.
The Java SCA infrastructure should provide flexibility and choice. It should not dictate programming models but support many.
The Java SCA infrastructure is very modularized and is designed to be highly extensible so users can customize it to fit their needs.

Prerequisites

Java SCA requires the following:

Checking out code from Subversion

Use the command as follows:

svn checkout https://svn.apache.org/repos/asf/tuscany/sca-java-2.x/trunk/

Setting up your Development Environment

top-down build (recommended approach)

Check out all of the java source code.

svn checkout https://svn.apache.org/repos/asf/tuscany/sca-java-2.x/trunk/

Building the SCA source code :

cd trunk
mvn -fae clean install

It should work even if you start with an empty Maven local repository, and it should always work, however when you are building for Tuscany for the first time there are a lot of dependencies which must be downloaded so the first build can take a long time and it may fail with problems retrieving the dependencies.

There can be occasional problems downloading artifacts from remote Maven repositories so if mvn fails with network related sounding messages sometimes just trying again can fix the problem.

The trunk code sometimes has SNAPSHOT dependencies which can get out of date in your local repository so if you see odd build failures try updating the SNAPSHOT jars by using the "-U" parameter in the mvn command.

Once you have done a top-down build, and your local maven repository is populated, you can start using the maven off line option to speed up the build process by using the "-o" parameter in the mvn command.

The SCA build consumes a good amount of memory, in case you are seeing issues during the build, set a MAVEN_OPTS environment variable to allocate more memory for the build process.

Windows : SET MAVEN_OPTS="-Xmx512m -XX:MaxPermSize=384m"
Unix : export MAVEN_OPTS="-Xmx512m -XX:MaxPermSize=384m"

If you are using MAC OS, please see 'Special instructions for MAC OS users' section below

Importing SCA modules into your Development IDE

Using Eclipse

If this is the first time you are using your workspace with maven m2 local repository, you will need to tell your Eclipse workspace the location of the directory, and you can do this with the following command :

mvn -Declipse.workspace=[path-to-eclipse-workspace] eclipse:add-maven-repo

In order to generate the necessary project files to import the SCA modules to Eclipse, you can use the maven eclipse plugin

mvn eclipse:eclipse

Now launch your Eclipse IDE, select File->Import->Existing projects into Workplace, and then import the project from SCA Modules into your Eclipse Workspace.

Using Eclipse PDE

To help with development in an OSGi environment Tuscany also has a build profile to setup the Eclipse Plugin Development Environment:

mvn -fae -Peclipse

Now set the "Target Platform" in your Eclipse IDE by following the steps below :

Inside eclipse, select File->Open File" and open "distribution/all/target/features/tuscany.target"
- click "Set as target platform" on the upper-right side of the overview window that opened

Then as when using eclipse:eclipse launch your Eclipse IDE, select File->Import->Existing projects into Workplace, and then import the project from SCA Modules into your Eclipse Workspace.

There are also some Tuscany Eclipse code templates available:
Eclipse Style Formatter
Eclipse Templates

Eclipse Build Troubleshooting

Sometimes a Maven build will work from the command line, yet the same component will not build in the Eclipse environment. Of course, developers try to prevent this from happening, but it does happen and makes for a valid Jira. In the meantime, here are some steps that might help correct build issues in the Eclipse environment:

Missing classpath variables - Sometimes Java build classpath variables are incorrect or missing, for instance "Unbound classpath variable: 'M2_REPO/com/sun/xml/bind/jaxb-impl/2.1.9/jaxb-impl-2.1.9.jar'". Add the classpath variable 'M2_REPOS' to your workspace. Window > Preferences > Java > Build Path > Classpath variables. Create variables for any missing variables and point them to the appropriate place in your file system. Do a complete rebuild.

Missing projects - Sometimes developers add or remove project dependencies. These show up with error message "The project cannot be built until build path errors are resolved". This is often a sign that your Eclipse workspace is out of sync with your local repository or the svn repository. The full remedy is to refresh your local repos (svn update), rebuild your local repos (mvn clean install -U), regenerate Eclipse projects (mvn -Declipse), reimport projects in Eclipse, and clean/rebuild your projects. You may be able to do less than this full refresh with simple changes, but the full refresh is often needed for major changes.

Project won't build - Sometimes a particular project will not build despite repeated efforts. Consider if this project is necessary to the task at hand. You may be able to delete the problematic Eclipse project and continue with other work.

Special instructions for MAC OS users

Some plugins used in the Tuscany build requires a explicit dependency on some classes from JDK tools.jar, which is in a different place in the MAC OS environment.

We have created duplicate profiles in Tuscany to accommodate the most used user tasks

Running a build : there is a default profile for the mac environment properly configuring the tools.jar property
```
   mvn clean install
```

Creating Eclipse IDE project files : there is a "eclipse-mac" profile
```
   mvn -Peclipse-mac
```

Update your maven settings.xml (%user_home%/.m2/settings.xml)

Updating your settings.xml with proper property configuration will make all profiles work in a MAC OS environment.

<settings>

  <profiles>
     <profile>
        <id>mac-os-configuration</id>
        <properties>
           <tools.jar>${java.home}/../Classes/classes.jar</tools.jar>
        </properties>
     </profile>
  </profiles>
  
  <activeProfiles>
    <activeProfile>mac-os-configuration</activeProfile>
  </activeProfiles>

</settings>

Coding Guidelines

There are a few simple guidelines when developing for JAVA SCA:

The basic coding style used is the described at Sun Java coding standards but the main thing is to be consistent with the existing code you're updating, so for example, if you're updating a method that uses the braces on the same line style don't add code with the hanging braces style.

Always include the Apache License Headers on all files (both source code files and resource files such as xml documents)

Include a descriptive log message for checkins, for example "fixed such and such problem".

While Tuscany does not yet have an official style or template, here are some templates that folks have been using and have been checked into the build which are stored at https://svn.apache.org/repos/asf/tuscany/java/etc/

Naming conventions to increase consistency

Folder Names: Please use all lowercases and dashes in folder names (like in the jar names)

Maven artifact id = tuscany-<folder name>

Package names: Package names within modules should include the module name so that source code can be located in the source tree easily. So, for example, java/sca/module/implementation-java would be in package structure org.apache.tuscany.implementation.java.*

Testing

Tuscany uses plain junit test cases to perform unit and integration testing, below is an example that can also be used as a template for writing new test cases; it demonstrates how to bootstrap the Tuscany SCA runtime in your test case, and because they are based on junit, you can run it from your IDE of choice or from Maven.

Note that we are using JUnit 4.2 code style in OSGI development stream

/**
 * Description of your test case and necessary details you find necessary
 */
@Scope("COMPOSITE") @EagerInit
public class CalculatorTestCase {

    private static CalculatorService calculatorService;
    private static NodeLauncher launcher;
    private static Node node;

    @Reference
    public void setCalculatorService(CalculatorService calculatorService) {
        CalculatorTestCase.calculatorService = calculatorService;
    }
    
    
    @BeforeClass
    public static void setUpBeforeClass() throws Exception {
        launcher = NodeLauncher.newInstance();
        String location = ContributionLocationHelper.getContributionLocation(CalculatorClient.class);
        node = launcher.createNode("Calculator.composite", new Contribution("test", location));
        System.out.println("SCA Node API ClassLoader: " + node.getClass().getClassLoader());
        node.start();
    }

    @AfterClass
    public static void tearDownAfterClass() throws Exception {
        if (node != null) {
            node.stop();
            node.destroy();
        }
        if (launcher != null) {
            launcher.destroy();
        }
    }

    @Test
    public void testCalculator() throws Exception {
        // Calculate
        assertEquals(calculatorService.add(3, 2), 5.0);
        assertEquals(calculatorService.subtract(3, 2), 1.0);
        assertEquals(calculatorService.multiply(3, 2), 6.0);
        assertEquals(calculatorService.divide(3, 2), 1.5);
    }
}

Note that we use surefire maven plugin to run the unit and integration tests, and in most cases, they are configured to match a **/*TestCase.java file name pattern. Because of this, if your test case has a different file name pattern, you might execute it from your IDE of choice, but the maven build won't execute the test.

Client Applications as SCA Components

Below is how you can build client applications as an SCA component.

@Scope("COMPOSITE") @EagerInit
public class CalculatorClient {
    
    private CalculatorService calculatorService;

    @Reference
    public void setCalculatorService(CalculatorService calculatorService) {
        this.calculatorService = calculatorService;
    }
    
    @Init
    public void calculate() {

        // Calculate
        System.out.println("SCA API ClassLoader: " + print(Reference.class.getClassLoader()));
        System.out.println("3 + 2=" + calculatorService.add(3, 2));
        System.out.println("3 - 2=" + calculatorService.subtract(3, 2));
        System.out.println("3 * 2=" + calculatorService.multiply(3, 2));
        System.out.println("3 / 2=" + calculatorService.divide(3, 2));
    }
    
    private static String print(ClassLoader cl) {
        StringBuffer buf = new StringBuffer();
        for (; cl != null;) {
            buf.append(cl.toString());
            buf.append(' ');
            cl = cl.getParent();
        }
        return buf.toString();
    }

}

Maven Build Structure

We use the term Module to refer to the leaf of maven tree.

sca/pom.xml's parent will be pom/parent/pom.xml
Other poms will use the pom from the parent folder as parent pom
Group id: org.apache.tuscany.sca
Version of our modules will be specified once in java/sca/pom.xml, child poms don't need specify a version as they get it from their parent
pom names begin Apache Tuscany SCA
Eclipse projects are generated for all built modules using mvn eclipse:eclipse

Adding a new module and not ready to integrate?

'work-in-progress' modules can be worked on in the same source tree and yet not break the top-down build. You can do this by not listing your module(s) in java/sca/modules/pom.xml.

Development Hints

Importing existing Tuscany SCA projects into Eclipse

This section has talked about how to get set up ready to develop Tuscany. If you need to import existing samples into Eclipse to work on there are some instructions here. These are instructions for 1.x but should work OK on 2.x.

Generating Eclipse WTP Web Projects for Webapp samples

If you're using Eclipse WTP and want to get WTP Web Projects generated
for our Webapp samples you can simply pass a -Dwtpversion=1.5 option to
the usual mvn eclipse:eclipse command, like this:
mvn -Dwtpversion=1.5 -Peclipse eclipse:eclipse

The magic -Dwtpversion=1.5 option will add the WTP Web project nature to
all the Eclipse projects with <packaging>war</packaging> in their Maven
pom.xml. You'll then be able to add these projects to a WTP Tomcat or
Geronimo Server configuration, to publish and run them straight from
your Eclipse workspace.

Generating Dependencies for Ant in Samples

Figuring out the package dependency to include in Ant build.xml can be a pain. Here is a quick
script which works in Linux environment for war files.

jar tvf sample-feed-aggregator-webapp.war | grep .jar | awk '{ printf "%s\n", $8 }' |
   sed -e "s/WEB-INF\/lib\///" | awk '{ printf "<include name=\"%s\"/>\n", $1 }' | grep -v tuscany