Creating an application with EclipseRT Virgo Web Server

GreenPages: a demonstration

Christopher Frost

Ben Hale

Rob Harrop

Glyn Normington

Steve Powell

Andy Wilkinson

Web Server

2.1.0.RELEASE

Abstract

Spring application programmers are introduced to Virgo Web Server by installing the Web Server and examining a small application called GreenPages. Despite its simplicity, GreenPages is designed to demonstrate many different Web Server features and to act as a template from which other modular applications can be built.

This version of the guide is based on the following software versions:

Web Server2.1.0.RELEASE
GreenPages2.3.0.RELEASE
SpringSource Tool Suite2.5.0.RELEASE
Apache Maven2.2.0

Copyright © 2009, 2010 VMware Inc.

All rights reserved. This document is made available under the terms of the Eclipse Public License v1.0 which is available at http://www.eclipse.org/legal/epl-v10.html.

Java, Sun, and Sun Microsystems are trademarks or registered trademarks of Sun Microsystems, Inc. in the United States and other countries.

OSGi is a trademark or a registered trademark of the OSGi Alliance in the United States, other countries, or both.

Eclipse is a trademark of Eclipse Foundation, Inc.

UNIX is a registered trademark of The Open Group.

Windows is a registered trademark of Microsoft Corporation in the United States and other countries.

Mac and Mac OS are trademarks of Apple Inc., registered in the U.S. and other countries.


Table of Contents

Preface
1. Concepts
1.1. OSGi concepts
1.2. Spring DM concepts
1.3. Grouping bundles in Web Server
2. Installation
2.1. Pre-requisites
2.2. Installing Web Server
2.3. Installing the SpringSource Tool Suite
2.4. Installing Apache Maven
3. Installing GreenPages
3.1. Introduction
3.2. Obtaining GreenPages
3.3. Building and installing GreenPages
3.4. Browsing the GreenPages application
3.5. Running GreenPages from Eclipse
4. The Web Module
4.1. Introduction
4.2. GreenPages set up
4.3. The controller
4.4. Deploying a bundle
4.5. Creating a PAR
4.6. Referencing an OSGi Service
4.7. Publishing an OSGi Service
5. The Middle Tier
5.1. Introduction
5.2. Creating the DataSource project
5.3. Building the JPA module
5.4. Trying out the JPA middle tier
5.5. Applying best practices to the middle tier
6. Testing GreenPages
6.1. Introduction
6.2. Single bundle integration testing
6.3. Contributing OSGi sourced dependencies
6.4. Multi bundle integration testing
7. Automated Build
7.1. Introduction
7.2. Setting up for Automated Build
7.3. Create POM
7.4. Adding the par plugin
7.5. Adding the dependency plugin
7.6. Automatically running the tests
7.7. Deploying the application
A. Further Resources
A.1. Projects
A.2. Documentation

Preface

This Guide introduces some basic concepts, explains how to install Virgo Web Server and the associated Eclipse™ tools, and examines a sample web application in detail, including how the application is built and tested.

It is intended for Spring application programmers and assumes little or no understanding of OSGi™ or Virgo Web Server.

Questions about Web Server and SpringSource Tool Suite (or this Guide) may be posted to the Web Server Community Forums (http://www.eclipse.org/forums).

1. Concepts

Virgo Web Server is a Java application server composed of a collection of modules which supports applications which are also composed of a collection of modules. These may be traditional Java web applications packaged as Web ARchive (.war) files as well as other modular applications. Modules may be shared between applications and multiple versions of modules can co-exist.

This chapter introduces concepts necessary for developing Web Server applications. These concepts will become clearer as the GreenPages application is explored in later chapters.

1.1 OSGi concepts

Modules in Virgo are represented using a standard Java module system known as OSGi. Modules consist of programs and resources organised by Java package together with metadata which declares imported and exported packages. A module exports a package to make the corresponding programs and resources available for use by other modules. A module imports a package to use the corresponding programs and resources of another module.

Representing a program as a collection of modules makes it easier for the programmer to manage it and modify it and for teams of programmers to divide responsibilities between themselves. A module is similar to a Java class in this respect. Design principles similar to those for organising data and programs into classes can be applied to organising applications into modules.

An industry consortium known as the OSGi Alliance (see the appendix Projects) develops OSGi specifications, reference implementations, and compliance tests. Virgo Web Server is built on the Equinox OSGi framework which is also the reference implementation for the OSGi framework specification.

Bundles

Modules in OSGi are known as bundles. Each bundle is stored in a file which conforms to the JAR file format and can contain Java classes, a manifest (in META-INF/MANIFEST.MF), and further resource files.

The OSGi framework enables bundles to be installed and run.

OSGi identifies bundles “by name” or “by identifier” (id).

The symbolic name and version of a bundle is an attribute of the bundle itself and uniquely identifies that bundle (by name) in an OSGi framework. A bundle usually declares its symbolic name and version in its manifest (a file called MANIFEST.MF) like this:

Bundle-SymbolicName: org.foo.bundle
Bundle-Version: 1.2.3.BUILD-2009-06-04

Additionally, the OSGi framework assigns a distinct number, known as a bundle id, to each bundle as it is installed. Bundles may be referred to “by identifier” using this number. The OSGi framework itself resides in a bundle with bundle id 0.

The dependencies between bundles are expressed statically in terms of packages and dynamically in terms of services. A package is familiar to Java programmers. For example, a Java program may depend on a class org.foo.X, from package org.foo, and a bundle containing that program would either need to contain org.foo.X or depend on the package org.foo. Package dependencies are specified in the bundle manifest, for example:

Import-Package: org.foo

A bundle which provides a package for use by other bundles must export the package in its manifest. For example:

Export-Package: org.foo

The OSGi framework ensures that a given bundle’s package dependencies can be satisfied before the bundle runs. This process is known as resolution.

After a bundle is resolved, its classes and resources are available for loading. In OSGi, bundles and their packages do not appear on the application classpath. Instead, each bundle has a class loader which loads its own classes and loads classes belonging to each of its imported packages by deferring to the bundle class loader that exports the package.

Life cycle

The OSGi framework manages the life cycle of each bundle. A bundle is first of all installed and will be in the INSTALLED state. If a request is made to start the bundle, the OSGi framework resolves the bundle and, if resolution was successful, will subsequently move the bundle to the ACTIVE state. If a request is made to stop the bundle, the OSGi framework will move the bundle back to the RESOLVED state. A request may then be made to uninstall the bundle.

While the bundle is INSTALLED, ACTIVE or RESOLVED, it may be updated to pick up some changes. These changes are not detected by bundles which were depending on the bundle before it was updated. A “refresh packages” operation may be performed to ripple the changes out to those bundles. (See Services concepts.)

The life cycle of a bundle can be summarised by a state transition diagram. This diagram shows some more of the intermediate states of a bundle not described in the overview above:

Figure 1.1. Bundle life cycle

Bundle life cycle


Services

Bundles may publish Java objects, known as services, to a registry managed by the OSGi framework. Other bundles running in the same OSGi framework can then find and use those services. Services are typically instances of some shared Java interface. A bundle which provides a service need not then export the package containing the implementation class of the service.

For example, a bundle could export a package containing the interface org.bar.SomeInterface, thus:

Export-Package: org.bar

…implement the interface with a class SomeImpl:

package org.bar.impl;

class SomeImpl implements SomeInterface {
	…
}

…create an instance of SomeImpl and then publish this instance (as an instance of the interface SomeInterface).

An OSGi framework publishes a number of standard services. For example, the Package Admin service provides the “refresh packages” life cycle operation mentioned above.

OSGi provides an API which can be used to publish and find services, but it is much simpler to use Spring DM to accomplish this. (See Spring DM concepts.)

Versioning

OSGi allows different versions of bundles, packages, and several other entities, to co-exist in the same framework and provides some mechanisms for managing these versions.

Version numbers

An OSGi version number consists of up to three numeric components, or exactly three numeric components followed by a string component. These components are separated by a period (“.”) and are called the major, minor, micro, and qualifier components, respectively.

For example, the version 2.4.1.ga has major component 2, minor component 4, micro component 1, and a qualifier component ga. (There are restrictions on the characters that can appear in a qualifier. For example: letters, digits, underscores and hyphens are allowed; periods and commas are not.)

Trailing components may be omitted along with their period (.). So, for example, the version numbers 2, 2.0, and 2.0.0 all denote the same version. This example demonstrates that 0 is assumed if a numeric component is omitted, and the empty string is assumed for an omitted qualifier.

Version ranges

Dependencies on bundles and packages have an associated version range which is specified using an interval notation: a square bracket “[” or “]” denotes an inclusive end of the range and a round bracket “(” or “)” denotes an exclusive end of the range. Where one end of the range is to be included and the other excluded, it is permitted to pair a round bracket with a square bracket. The examples below make this clear.

If a single version number is used where a version range is required this does not indicate a single version, but the range starting from that version and including all higher versions.

There are three common cases:

  • A “strict” version range, such as [1.2,1.2], which denotes that version and only that version.

  • A “half-open” range, such as [1.2,2), which has an inclusive lower limit and an exclusive upper limit, denoting version 1.2.0 and any version after this, up to, but not including, version 2.0.0.

  • An “unbounded” version range, such as 1.2, which denotes version 1.2 and all later versions.

Versioning policies

A versioning policy is a way of using version numbers to indicate compatible and incompatible changes. OSGi does not mandate a particular versioning policy. Instead, a specific versioning policy may be implemented using version ranges.

Strict and half-open version ranges are most useful in representing versioning policies. Unbounded version ranges can lead to problems as they (unrealistically) assume that compatibility will be preserved indefinitely.

For example, a conservative versioning policy might assume that any change, other than in the qualifier component of a version, implies an incompatible change to the object. Such a policy would employ version ranges such as [1.2.1.beta,1.2.2) which accept any version from 1.2.1.beta (inclusive) up to but not including 1.2.2 (exclusive).

Alternatively, a relaxed versioning policy might assume that only changes in the major component of a version denote an incompatible change. Such a policy would employ version ranges such as [1.2,2) to capture this.

Bundle version

Each bundle has a version. The bundle’s version may be specified in the manifest using a Bundle-Version header:

Bundle-Version: 1.4.3.BUILD-20090302

If not specified the bundle version is assumed to be 0.

Package version

Each exported package has a version. The exported package’s version may be specified on the Export-Package manifest header. For example

Export-Package: org.foo;version="2.9",org.bar;version="1"

exports two packages: org.foo, at version 2.9.0 and org.bar, at version 1.0.0.

If the version attribute is omitted the version is assumed to be 0.

Each package import has a version range. The package import version range may be specified on the Import-Package manifest header. If interval notation is used, the version range must be enclosed in double quotes, for example:

Import-Package: org.foo;version="[2,3)",org.bar;version="[1,1]"

seeks to import a package org.foo in the range [2.0.0,3.0.0) and a package org.bar with the (exact) version 1.0.0.

If a version range is not specified on an import, the range 0 is assumed, meaning that any version of this package would satisfy the import.

Bundle manifest version

Bundle manifests have a version which is 1 by default, indicating OSGi Release 3 semantics. Web Server is based on OSGi Release 4 and therefore expects bundle manifests to be at version 2, indicating OSGi Release 4 semantics. (See the appendix Projects.) The bundle manifest’s version should be specified on the Bundle-ManifestVersion manifest header, exactly as follows:

Bundle-ManifestVersion: 2

Manifest version

Manifests themselves also have a version which must be specified as 1.0. This is not an OSGi definition but part of the JAR file specification (http://java.sun.com/javase/6/docs/technotes/guides/jar/jar.html).

Manifest-Version: 1.0

1.2 Spring DM concepts

Spring DM is a project which enables services to be published and consumed using descriptions written in XML. Web Server has Spring DM built-in.

The XML descriptions reside in files with extension .xml in the bundle’s META-INF/spring sub-directory.

To publish a service, an <osgi:service> tag is used, specifying the implementation class of the service and the interface class to be used. Spring DM constructs an instance of the implementation class and publishes that instance in the OSGi service registry under the interface when the bundle is started.

To consume a service, an <osgi:reference> tag is used and the service may be passed into other Spring beans using Spring’s dependency injection facilities.

Spring DM automatically creates proxies for OSGi services so that the actual service object may come and go at runtime. If a service disappears, any proxies to the service will wait for the service to re-appear. This effect is known as damping.

When a bundle is started, Spring DM builds the application contexts specified by the XML descriptions, creates proxies for the specified services, and publishes the specified services to the OSGi service registry.

When a bundle is stopped, Spring DM retracts any services it published on behalf of the bundle and closes the bundle’s application contexts. Web Server turns off damping of a service proxy while the proxy’s application context is being closed.

(Spring DM has been contributed to Eclipse as the Gemini Blueprint project.)

1.3 Grouping bundles in Web Server

Web Server provides a way of grouping together a collection of OSGi bundles which comprise a single application. These bundles are placed in a JAR file with extension “.par”. This is called a PAR file.

All the bundles in a PAR file are resolved together and so mutual dependencies are permitted.

At runtime a PAR file provides a scope in the sense that bundles inside the PAR file may depend on packages and services outside the PAR file, but bundles outside the PAR file may not depend on packages and services provided by the PAR file.

Virgo also provides the plan artifact as another way of grouping bundles and other artifacts into an application. A plan is a file (in XML format) listing a collection of artifacts. This Guide makes no further reference to plans. See Section A.2, “Documentation” for a link to more Virgo documentation.

PAR files (or individual bundles) are deployed into Web Server by dropping them into a “pickup” directory or using the Administration Console web application provided with Web Server. During deployment, the bundles in the PAR file are installed into OSGi, resolved together, and then started together.

2. Installation

Before developing an application with Web Server, it is essential to install Web Server, an Integrated Development Environment (IDE), and a build system integrated with Eclipse. The IDE used here is the Eclipse-based SpringSource Tool Suite (STS), and the build system used here is Apache Maven.

STS is supplied as a fully configured Eclipse IDE, with Virgo Web Server and Maven plugins built-in.

2.1 Pre-requisites

Before proceeding, ensure that a Java™ Standard Edition Development Kit (JDK) for Java 6 or later is installed and that the JAVA_HOME environment variable is set to the root directory of the JDK. (A Java Runtime Environment (JRE) alone is not sufficient, a development kit is necessary to use the facilities in STS.)

To verify this, issue the command "%JAVA_HOME%"\bin\java -version from a command prompt on Windows (or $JAVA_HOME/bin/java -version from a terminal window on UNIX) and ensure that the command completes successfully and reports a Java version 1.6.x (denoting Java 6) or greater.

Also issue the command "%JAVA_HOME%"\bin\jar to ensure that there is a means of extracting files from zip archives. If the jar command is unavailable, download and install a suitable zip program such as 7zip, gzip, or WinZip. This is most relevant for Windows operating systems where the inbuilt zip extraction utility may not handle long pathnames correctly.

2.2 Installing Web Server

Although the steps are similar, the details of installing the Virgo Web Server depend on the operating system.

Obtain Web Server from the download site (http://www.eclipse.org/virgo/download). This guide is consistent with version 2.1.0.RELEASE of Web Server.

Installing Web Server on Windows™ operating systems

Unzip the download of Web Server to the root directory of a drive (this will avoid possible problems with long pathnames). Set an environment variable %VWS_HOME% to refer to the unzipped folder…

prompt> cd C:\
prompt> "%JAVA_HOME%"\bin\jar xf \path\to\virgo-web-server-2.1.0.RELEASE.zip
prompt> set VWS_HOME=C:\virgo-web-server-2.1.0.RELEASE

To verify the installation, issue the command: "%VWS_HOME%"\bin\startup.bat and ensure a message numbered UR0001I is displayed. Many other messages about starting and installing other required artifacts are produced, but the UR0001I message indicates that the user region is ready for use. (Timestamps have been removed and thread names and other details may vary with different installations and versions.)

system-artifacts             <TC0000I> Starting Tomcat. 
system-artifacts             <TC0010I> Creating HTTP/1.1 connector with scheme http on port 8080. 
system-artifacts             <TC0010I> Creating HTTP/1.1 connector with scheme https on port 8443. 
system-artifacts             <TC0010I> Creating AJP/1.3 connector with scheme http on port 8009.
system-artifacts             <TC0001I> Started Tomcat. 
system-artifacts             <DE0004I> Starting bundle 'org.eclipse.virgo.web.core' version '2.1.0.RELEASE'. 
system-artifacts             <DE0004I> Starting bundle 'org.eclipse.virgo.web.dm' version '2.1.0.RELEASE'. 
start-signalling-1           <DE0005I> Started bundle 'org.eclipse.virgo.web.dm' version '2.1.0.RELEASE'. 
system-artifacts             <DE0005I> Started bundle 'org.eclipse.virgo.web.tomcat' version '2.1.0.RELEASE'. 
start-signalling-1           <DE0005I> Started bundle 'org.eclipse.gemini.web.tomcat' version '1.1.0.RELEASE'. 
start-signalling-2           <DE0005I> Started bundle 'org.eclipse.virgo.web.core' version '2.1.0.RELEASE'. 
start-signalling-2           <DE0005I> Started plan 'org.eclipse.virgo.web' version '2.1.0'. 
Thread-2                     <UR0001I> User region ready. 

Shut down the server by pressing Ctrl-C.

Installing Web Server on UNIX™ operating systems

Unzip the download of Web Server to a suitable location on the file system, such as the home directory. (If the download was automatically unzipped by the operating system, simply move the unzipped directory to the chosen location.) Set an environment variable $VWS_HOME to refer to the unzipped folder…

prompt$ mkdir /path/to/home/springsource
prompt$ cd /path/to/home/springsource
prompt$ unzip /path/to/virgo-web-server-2.1.0.RELEASE.zip
prompt$ export VWS_HOME=/path/to/home/springsource/virgo-web-server-2.1.0.RELEASE

To verify the installation, use a terminal window to issue the command: $VWS_HOME/bin/startup.sh and ensure a message numbered UR0001I is displayed. Many other messages about starting and installing other required artifacts are produced, but the UR0001I message indicates that the user region is ready for use. (Timestamps have been removed and thread names and other details may vary with different installations and versions.)

system-artifacts             <TC0000I> Starting Tomcat. 
system-artifacts             <TC0010I> Creating HTTP/1.1 connector with scheme http on port 8080. 
system-artifacts             <TC0010I> Creating HTTP/1.1 connector with scheme https on port 8443. 
system-artifacts             <TC0010I> Creating AJP/1.3 connector with scheme http on port 8009.
system-artifacts             <TC0001I> Started Tomcat. 
system-artifacts             <DE0004I> Starting bundle 'org.eclipse.virgo.web.core' version '2.1.0.RELEASE'. 
system-artifacts             <DE0004I> Starting bundle 'org.eclipse.virgo.web.dm' version '2.1.0.RELEASE'. 
start-signalling-1           <DE0005I> Started bundle 'org.eclipse.virgo.web.dm' version '2.1.0.RELEASE'. 
system-artifacts             <DE0005I> Started bundle 'org.eclipse.virgo.web.tomcat' version '2.1.0.RELEASE'. 
start-signalling-1           <DE0005I> Started bundle 'org.eclipse.gemini.web.tomcat' version '1.1.0.RELEASE'. 
start-signalling-2           <DE0005I> Started bundle 'org.eclipse.virgo.web.core' version '2.1.0.RELEASE'. 
start-signalling-2           <DE0005I> Started plan 'org.eclipse.virgo.web' version '2.1.0'. 
Thread-2                     <UR0001I> User region ready. 

Shut down the server by pressing Ctrl-C.

2.3 Installing the SpringSource Tool Suite

The SpringSource Tool Suite (STS) is a development environment based on Eclipse that is already configured with the plugins needed to work with Web Server and OSGi. Although the steps are similar, the details of installing STS depend on the operating system.

Go to the STS download site (http://www.springsource.com/products/springsource-tool-suite-download) and download the variant appropriate to the operating system being used. This guide is consistent with STS version 2.5.0.RELEASE. Previous versions may not work properly with the latest revision of GreenPages, currently 2.3.0.RELEASE.

It is possible to use vanilla Eclipse, and add the SpringSource Tools to it. Please refer to the installation instructions in the Virgo Programmer Guide.

Installing STS on Windows™ operating systems

Unzip the download of STS to the root directory of a drive (this will avoid possible problems with long pathnames).

prompt> cd C:\
prompt> "%JAVA_HOME%"\bin\jar xf \full…path…to\springsource-tool-suite-2.5.0.RELEASE-e3.6.1-win32.zip

To verify the installation, run the eclipse.exe (or sts.exe) executable in the unzipped directory and check that STS displays a welcome panel. The first time there may be a short delay due to the initial set-up of indexes.

Installing STS on UNIX™ operating systems

Unpack the download of STS to a suitable location on the file system, such as /opt or, if root access is not available, the home directory. (If the download was automatically unpacked by the operating system, simply move the unpacked directory to the chosen location.)

To verify the installation, run the STS executable (STS.app on Mac OS X) in the unpacked directory and check that STS displays a welcome panel. The first time there may be a short delay due to the initial set-up of indexes.

Note about Java versions in STS

SpringSource Tool Suite runs on Eclipse using Java Version 1.6, and Web Server requires Java Version 1.6. The GreenPages application built here also requires Java Version 1.6. The default Java compiler settings in STS should not need adjusting, but should be checked.

2.4 Installing Apache Maven

Apache Maven, or Maven for short, is a software project management and comprehension tool which uses a central Project Object Model (POM) to manage a project’s build, reporting and documentation generation. POM files (pom.xml) are included in the projects for GreenPages.

To install Maven, visit the Maven website (http://maven.apache.org) and follow the download instructions from there. This document has been written and tested with Maven version 2.2.0. The rest of the document assumes that Maven commands (mvn …) are available from the command line.

3. Installing and exploring GreenPages

3.1 Introduction

GreenPages is a simple application that allows users to search an online email address directory. Each listing in the directory details the relevant email addresses and the name of the owner. GreenPages has only three screens: the search screen, the results screen and the listing detail screen.

In the search screen, users can enter search criteria to be matched against the listings in the directory. The result screen displays any listings that match the criteria entered by the user. The listing detail screen shows all the data known about a given listing.

Despite its simplicity, GreenPages is designed to demonstrate many different Web Server features and to act as a template from which other modular applications can be built. In particular, GreenPages demonstrates:

  • module dependencies with Import-Package,

  • load-time weaving with JPA and AspectJ,

  • bundle classpath scanning, and

  • service export, lookup and injection.

In addition to demonstrating common Web Server features, GreenPages demonstrates integration with:

  • Spring Framework 3.0;

  • FreeMarker 2.3;

  • EclipseLink 1.0.0;

  • H2 1.0.71; and

  • Commons DBCP 1.2.2.

The GreenPages application is packaged as a PAR file containing four modules.

The greenpages.db module provides access to an external database and publishes a javax.sql.DataSource service.

The greenpages.app module exports a greenpages package containing Directory and Listing interfaces.

The greenpages.jpa module imports the greenpages package and uses the javax.sql.DataSource service to access the external database and publishes its contents as a greenpages.Directory service.

The greenpages.web module imports the greenpages package and uses the greenpages.Directory service to respond to web requests.

3.2 Obtaining GreenPages

This document provides instructions for building the complete GreenPages application and running it in Web Server as well as running tests and executing it from the SpringSource Tool Suite. In addition, the application is examined in development stages from SpringSource Tool Suite, stepping through the development process and comparing this to the solution.

To get the completed GreenPages application, including tests and explanatory skeleton parts:

  1. download the latest ZIP file from

    http://eclipse.org/virgo/download/
  2. extract all the files from the ZIP file to a convenient directory (preserving the directory structure).

To extract the files on Windows:

prompt> mkdir c:\springsource\samples
prompt> cd c:\springsource\samples
prompt> jar xf c:\path\to\greenpages-2.3.0.RELEASE.zip
prompt> set GREENPAGES_HOME=c:\springsource\samples\greenpages-2.3.0.RELEASE

To extract the files on Unix systems:

prompt$ mkdir -p /opt/springsource/samples 
prompt$ cd /opt/springsource/samples 
prompt$ unzip /path/to/greenpages-2.3.0.RELEASE.zip 
prompt$ export GREENPAGES_HOME=/opt/springsource/samples/greenpages-2.3.0.RELEASE

The environment variable GREENPAGES_HOME set here is not used by the projects, but is used as a shorthand in the instructions that follow.

The GreenPages zip file contains two main directories called solution and start. The solution directory contains the completed application which can be built and tested (as described in the next section). The start directory contains an initial skeleton of the GreenPages application which can be used to generate the full application. The examination steps refer to this skeleton by way of illustration.

To follow the examination steps read Chapter 4, The Web Module after installing the full application in the following section.

3.3 Building and installing GreenPages

Building with Apache Maven

GreenPages uses Apache Maven as its primary build system. Each module of the application can be built separately and the entire application can built and assembled into a PAR file from a single location. To build the application and assemble it into a PAR file:

  1. Make $GREENPAGES_HOME/solution the current directory.

  2. Run the command mvn package. The first time this is run several files will be downloaded from Maven repositories and SpringSource repositories. Subsequent runs will not need to do this.

  3. Verify that the greenpages-2.3.0.RELEASE.par file exists in $GREENPAGES_HOME/solution/greenpages/target.

Installing dependencies into Web Server

Unlike traditional Java EE applications, GreenPages does not package all of its dependencies inside its deployment unit. Instead, it relies on the mechanisms of OSGi to locate its dependencies at runtime. When running an OSGi application on Web Server, these dependencies can be loaded into memory as needed, but first they must be made available to Web Server.

The Maven build included with GreenPages uses the dependency:copy-dependencies plugin to gather all the artifacts that GreenPages depends on that are not supplied by the Web Server runtime. These dependencies can then be installed into the Web Server repository. Dependencies are gathered automatically during the package phase. These dependencies can be found in $GREENPAGES_HOME/solution/greenpages/target/par-provided. To install dependencies simply copy all the *.jar files from this directory into $VWS_HOME/repository/usr.

Installing dependencies on Windows:

prompt> cd %GREENPAGES_HOME%\solution\greenpages 
prompt> copy target\par-provided\* %VWS_HOME%\repository\usr

Installing Dependencies on UNIX:

prompt$ cd $GREENPAGES_HOME/solution/greenpages 
prompt$ cp target/par-provided/* $VWS_HOME/repository/usr

Notice that Web Server will not necessarily see these dependencies unless its repository indexes are rebuilt. Different repositories behave differently in this respect; some are passive (their indexes are built only once upon startup) and some are active (they can detect new files or files being removed dynamically). The usr repository is active so there is no need to restart Web Server when copying these files. The next time Web Server is started the -clean option will cause Web Server to re-scan the repository directories in any case. It is always safe to start Web Server with the -clean option.

Starting and configuring the database

GreenPages uses the H2 database to store all its data. Before starting the application, start the database server and populate the database with data.

  1. Change to the $GREENPAGES_HOME/db current directory. On Unix:

    prompt$ cd $GREENPAGES_HOME/db

    On Windows:

    prompt> cd %GREENPAGES_HOME%\db
  2. Run the database startup script appropriate to the operating system. For Unix, this is run.sh, run in the background:

    prompt$ sh run.sh &

    Press Return to continue.

    On Windows, run the run.bat command:

    prompt> run

    For both platforms, the command might invoke a browser window offering a connection to the database; close this window.

  3. Run the data population script appropriate to the operating system. For Unix, this is data.sh:

    prompt$ sh data.sh

    On Windows, run the data.bat command:

    prompt> data

Run these commands once to start a database server for H2; the server will continue to run in the background.

Installing and starting GreenPages PAR

To install the GreenPages PAR into Web Server and start it:

  1. Copy the GreenPages PAR to the $VWS_HOME/pickup directory. On Unix:

    prompt$ cd $VWS_HOME
    prompt$ cp $GREENPAGES_HOME/solution/greenpages/target/greenpages-solution-2.3.0.RELEASE.par pickup/

    On Windows:

    prompt> cd %VWS_HOME%
    prompt> copy %GREENPAGES_HOME%\solution\greenpages\target\greenpages-solution-2.3.0.RELEASE.par pickup\
  2. Start Web Server with the -clean option. On Unix:

    prompt$ $VWS_HOME/bin/startup.sh -clean

    On Windows:

    prompt> "%VWS_HOME%"\bin\startup.bat -clean
  3. Verify that GreenPages starts correctly by checking in the Web Server output for the log message:

    <DE0005I> Started par 'greenpages' version '2.3.0.RELEASE'. 

3.4 Browsing the GreenPages application

Once installed and started, the GreenPages application can be accessed with a web browser using the address http://localhost:8080/greenpages.

From the home page, a search query can be entered into the search box:

After entering a query into the search box, the results page shows all the matches from the directory:

Clicking on view next to an entry in the search listing displays the full details for that listing entry:

3.5 Running GreenPages from Eclipse

Using Eclipse and the Web Server tools, it is possible to run applications directly from the IDE. As changes are made to the application in the IDE, they can be automatically applied to the running application allowing for rapid feedback of changes in function.

Importing the GreenPages projects into Eclipse

Before starting the GreenPages application from Eclipse, import the projects:

  1. Open the Import Wizard using FileImport.

  2. From the Import Wizard select GeneralExisting Projects into Workspace and click Next:

  3. Click Browse… and select $GREENPAGES_HOME/solution as the root directory.

  4. In the Import Projects window, select all the projects and click Finish:

  5. Validate that the imported projects appear in Package Explorer:

    There may be compilation errors at this stage.

Configuring Web Server target runtime

Projects for Web Server are associated with a Virgo Web Server runtime environment in Eclipse. This is to allow launching and testing from within Eclipse, and also to allow classpath construction in Eclipse to mirror the dynamic classpath in the Web Server runtime.

Compilation errors in the previous step will be resolved here.

To configure a Web Server runtime environment:

  1. Open WindowShow ViewOther….

  2. In the Show View dialog choose ServerServers to make the servers view visible:

  3. Right-click in the Servers (which may not be empty) view and select NewServer.

  4. In the New Server dialog, choose EclipseRTVirgo Web Server and click Next.

  5. Click Browse and select the $VWS_HOME directory. Ensure that a JRE is selected supporting Java 1.6 or above. Click Finish to complete creation of the server:

  6. Select all projects (except Servers) in Package Explorer. Right-click on the projects and choose Close Project and then Open Project.

It is possible that there remain spurious build errors from Eclipse (see the Problems view), in which case a project clean build may clear the problems. Select ProjectClean… from the main menu, and choose to Clean all projects. It may be necessary to repeat this on a few projects. (This process is sometimes known as the “Eclipse dance”.)

Despite the dance steps outlined, there will remain some Warnings like this:

It is safe to ignore these.

When STS starts the Web Server it uses a ‘warm start’ by default. It is useful to set the -clean option so that every server start is a clean one. This is done by an option on the Web Server Overview window, which is obtained by opening the Web Server entry in the Servers window in STS. (Double-click, or right-click and choose Open.) The check box is labelled ‘Start server with -clean option’. Close the window before proceeding.

Running GreenPages from within Eclipse

Now that GreenPages is successfully imported into Eclipse, run the project directly from within the IDE.

If the GreenPages PAR file was copied to the pickup directory, be sure it is now removed so that it does not conflict with the deployment of the Eclipse project. On Unix:

prompt$ cd $VWS_HOME/pickup
prompt$ rm greenpages-solution-2.3.0.RELEASE.par

On Windows:

prompt> cd %VWS_HOME%\pickup
prompt> del greenpages-solution-2.3.0.RELEASE.par

Also, to prevent conflicts with the server configured in Eclipse, stop a currently-running Web Server by typing Control-C in the console window.

To run GreenPages from within Eclipse:

  1. Right click on the Web Server instance in the Servers view and select the Add and Remove… menu item.

  2. Add greenpages-solution (which is the containing project or PAR) to the server and finish.

  3. To start Web Server from within Eclipse right-click on the Web Server node in the Servers window and choose Start. The Servers view should now show the server and the added project:

  4. Verify that GreenPages is started correctly by checking for <DE0005I> Started par 'greenpages' version '2.3.0.RELEASE'. in the Console window.

(If errors are shown implying that GreenPages failed to be installed, this may be because some dependencies were not copied to Web Server, as described in section the section called “Installing dependencies into Web Server”. Check this.)

Once installed and started GreenPages is again available from a web browser at the address http://localhost:8080/greenpages.

4. The Web Module

4.1 Introduction

In common with most Enterprise Java applications GreenPages uses a web-based interface for user interactions. The following steps show how the controller for the application is constructed, using a service from the OSGi Service Registry.

It is assumed that the instructions in Chapter 2, Installation have been followed already and that the GreenPages zip file has been downloaded and unzipped as described in Chapter 3, Installing and exploring GreenPages.

4.2 GreenPages set up

Before beginning, configure the development environment to work with the application. In the case of GreenPages this means Maven and Eclipse.

Setting up Eclipse (STS)

The following sections are most easily followed in the Java (or Spring) perspective (not, for example, the Java EE perspective). If not already in an appropriate perspective, switch in SpringSource Tool Suite using the Open Perspective menu:

In this step create a reference to the Web Server instance that the GreenPages application integrates with (the instance may already be created if the GreenPages solution has been run from STS).

In STS open PreferencesServerRuntime Environments. Select Add… to create a new Web Server runtime environment. In the dialog that opens, select the Virgo Web Server (runtime) v2.1 and check the box to Create a new local server. When complete, press Next.

In the next dialog, set the Virgo Web Server installation directory field to the value of $VWS_HOME and check that the JRE: option is set to Java 1.6 or later. This may not be the workbench default. When complete, press Finish.

After returning to the Preferences window, press OK to return to Eclipse. The Servers view has opened and now shows an instance of Virgo Web Server in it.

There is also a Servers project, in which the server is listed.

4.3 The controller

The Spring’s MVC style of web application development is used in which the central type is the Controller class.

Import the web project

The GreenPages application is divided into OSGi bundles that are represented as Eclipse projects. In this step import the greenpages.web project.

Starting with no projects, import the web project by right-clicking in the Package Explorer view and selecting the Import… menu item. In the dialog that opens, choose GeneralExisting Projects into Workspace and select Next. In the following dialog set the root directory to the value of $GREENPAGES_HOME/start/greenpages.web and press Finish.

(Initially this project may have compiler errors; this is to be expected particularly if the Maven repository hasn’t yet been created.) When this project is imported go to the next step.

The controller class

In the src/main/java source folder of the greenpages.web project the package greenpages.web should contain the controller class named GreenPagesController. Create this by right-clicking on the greenpages.web package in the src/main/java source folder and selecting NewClass. (If Class is not offered on the New menu the Java perspective may not be being used, in which case look for the Class option under Other… in the Java section.)

Name the new class GreenPagesController and press Finish.

The code should be edited to look like this:

@Controller
public class GreenPagesController {
	…
    @RequestMapping("/home.htm")
    public void home() {
    }
    …

The annotations Controller and RequestMapping are from Spring Framework and are imported by adding the lines:

import org.springframework.stereotype.Controller;
import org.springframework.web.bind.annotation.RequestMapping;

STS will offer (as a Quick Fix) to insert imports for these Spring Framework annotations the first time they are used. (Java 1.6 supports annotations, and the Spring Framework libraries are accessible by linking to the correct Web Server runtime environment or generating the correct dependencies for the Maven plug-in.)

Enabling component scanning

Spring will detect the @Controller annotation and create a bean of controller type, provided that it scans the classpath for these. Spring’s component scanning is enabled by inserting a context tag in one of the Spring bean definition files.

Open the WEB-INF/greenpages-servlet.xml file in the src/main/webapp folder and ensure the following lines are present:

	<!-- enable classpath scanning -->
	<context:component-scan base-package="greenpages.web" />

Experiment by adding and removing this line, saving the file after each change. (Easily done by commenting it—use the Toggle Comment shortcut in STS.) Look in the Spring Explorer view for a bean named greenPagesController dynamically created by the component-scan tag.

4.4 Deploying a bundle

During development time, it can be helpful to run an application inside of the deployment container. In the case of GreenPages, this means deploying the greenpages.web bundle to the Virgo Web Server.

Deploying the greenpages.web bundle and starting the Web Server

The Web Server can be used while working in Eclipse. In this step the greenpages.web bundle is deployed and the Web Server instance is started.

Drag the greenpages.web project from the Package Explorer and drop it on the Web Server instance in the Servers view. Because greenpages.web is a web bundle the server will start automatically, and a browser window may open. Expand the Web Server instance and the bundle greenpages.web will be listed as a child.

If deployment is successful the console will contain the message:

<DE0005I> Started bundle 'greenpages.web' version '2.3.0'

Leave the server instance running and go to the next step.

Creating web module metadata

The Web Server has special knowledge of web application bundles. In this step web bundle metadata is added to the bundle and a web browser is used to navigate to it.

Open a web browser and navigate to http://localhost:8080/greenpages. If the link is not currently served by any bundle in the Web Server there may be an error displayed:

or else there is a blank page displayed. No pages are served.

To fix this issue the greenpages.web bundle must be declared to be a web bundle and a context path defined.

Open the template.mf file (at the top level under the greenpages.web project) and add (and save) the following entry (using the template.mf pane of the editor):

Web-ContextPath: greenpages

Be careful not to insert any blank lines or trailing spaces in this file.

Once added, right-click on the greenpages.web project and select Spring ToolsRun generation of MANIFEST.MF file. This will use a tool called Bundlor (included in STS) to update the OSGi metadata in the MANIFEST.MF file. Once Bundlor has finished running, open the META-INF/MANIFEST.MF file in the src/main/webapp folder.

It should look something like the following:

Manifest-Version: 1.0
Bundle-Name: GreenPages Web
Import-Library: org.springframework.spring;version="[3.0, 3.1)"
Import-Bundle: com.springsource.org.apache.taglibs.standard;version="[
 1.1.2,1.3)"
Web-ContextPath: greenpages
Import-Package: org.eclipse.virgo.web.dm;version="[2.0.0, 3.0.0)
 ",freemarker.cache;version="[2.3.15,2.3.15]",javax.servlet.jsp.jstl.c
 ore;version="[1.1.2,1.2.0)",javax.sql,org.apache.commons.dbcp,org.spr
 ingframework.core.io,org.springframework.stereotype,org.springframewo
 rk.web.bind.annotation,org.springframework.web.context,org.springfram
 ework.web.servlet
Bundle-ManifestVersion: 2
Bundle-SymbolicName: greenpages.web
Tool: Bundlor 1.0.0.RELEASE
Bundle-Version: 2.3.0

although the order of the entries may be different.

The server (if it is still running) will track these changes and automatically refresh (or restart) the greenpages.web bundle as required. Observe the context path for the web bundle being announced (it should now be '/greenpages' whereas previously it would have been a default context path derived from the bundle name: '/greenpages.web').

By default, Bundlor generates Import-Package entries with no version range specified. In the absence of a version range, the OSGi default (which denotes any version) is used. While this is very flexible it is generally a good idea to restrict an import by specifying a narrower range. This can be achieved by providing Bundlor with some additional information in the manifest template, as in the next step.

Add (and save) the following entry to the template.mf file:

Import-Template: 
 org.springframework.*;version="[3.0.0, 3.1.0)"

(Again, be careful not to leave trailing spaces on lines or insert blank lines in this file, except that there must be one trailing space after the colon to indicate that the header continues on the next line.)

Re-run the MANIFEST.MF generation as described earlier. In the MANIFEST.MF file the Import-Package entry should now have version ranges on each of the springframework packages:

Import-Package: org.eclipse.virgo.web.dm;version="[2.0.0, 3.0.0)
 ",freemarker.cache;version="[2.3.15,2.3.15]",javax.servlet.jsp.jstl.c
 ore;version="[1.1.2,1.2.0)",javax.sql,org.apache.commons.dbcp,org.spr
 ingframework.core.io;version="[3.0.0, 3.1.0)",org.springframework.ste
 reotype;version="[3.0.0, 3.1.0)",org.springframework.web.bind.annotat
 ion;version="[3.0.0, 3.1.0)",org.springframework.web.context;version=
 "[3.0.0, 3.1.0)",org.springframework.web.servlet;version="[3.0.0, 3.1
 .0)"

Behind the scenes the Web Server Tools have refreshed the deployed bundle as changes were made. Once again navigate to http://localhost:8080/greenpages. This page now displays an entry field.

Put any characters into the entry field and press Submit. This should display a “404” error page with the description:

description   The requested resource () is not available.

This is because there is no search page (search.htm) to process this request yet. The next section will address this.

4.5 Creating a PAR

At the end of the previous step, the Web Server instance was started and the greenpages.web bundle deployed. This bundle shows a static home page but a search value causes an error. The error appears because the URL for that search is not serviced by the controller. The application logic behind the search request is not in the greenpages.web project but in another project called greenpages.app. This section creates the greenpages.app project and then combines the two projects into a PAR so as to deploy them together as a single unit.

While executing these instructions it is not necessary to remove bundles from the Web Server instance, nor to stop the instance. As changes are made the bundle will be refreshed (or redeployed) and the server instance may report errors if the changes are incomplete. These may safely be ignored. Alternatively, the greenpages.web bundle can be removed from the Web Server instance, or the server can be stopped while these changes are made.

Importing the parent project

All of the bundles in the GreenPages sample are Maven projects which refer to a ‘parent’ project. To allow STS to accommodate references to the parent, import this project first.

In the same way that the starting greenpages.web project was imported (see Section 4.3, “The controller”) import the $GREENPAGES_HOME/start/parent project.

Importing the greenpages.app project

In this step, the greenpages.app project is imported which contains the business interfaces (and stub implementations of these interfaces).

In the same way that the starting greenpages.web project was imported (see Section 4.3, “The controller”) import the $GREENPAGES_HOME/start/greenpages.app project.

When Eclipse finishes importing the project, go to the next step.

Adding the controller implementation

The controller implementation will depend on the Directory and Listing interfaces found in the greenpages.app project. In this step, the implementation is added.

Open the GreenPagesController class. Add the following field and methods to the class:

@Autowired
private Directory directory;

@RequestMapping("/search.htm")
public List<Listing> search(@RequestParam("query") String query) {
  return this.directory.search(query);
}

@RequestMapping("/entry.htm")
public Listing entry(@RequestParam("id") int id) {
  return this.directory.findListing(id);
}

Add the (Quick Fix) suggested imports for the annotations Autowired and RequestParam, and choose the import for List< > from java.util.List.

Eclipse will not be able to suggest import statements for the Listing and Directory types. This is because the greenpages.web and greenpages.app projects are not linked together and therefore cannot see each other’s types.

Proceed to the next step.

Creating a PAR project

In Web Server, applications consisting of multiple bundles can be packaged as part of a PAR. In this step a PAR project containing the greenpages.web and greenpages.app bundles is created and deployed to the server.

Right-click in the Package Explorer and select NewProject…. In the dialog that opens select EclipseRTPAR Project and press Next:

In the New PAR Project dialog, ensure the Use default location option is unchecked, name the project greenpages, set the location to $GREENPAGES_HOME/start/greenpages and press Next.

In the next dialog, some of the PAR properties are pre-populated. Change the Application Name to Greenpages PAR and the Version to 2.3.0, then ensure that the Target Runtime is set to Virgo Web Server (Runtime) v2.1 and press Next.

In the next dialog, select the greenpages.app and greenpages.web bundles so that they are contained in the PAR and press Finish.

The project greenpages.web still shows errors; these are soon to be fixed.

The package explorer view will now show the following:

PAR project creation is complete, go to the next section.

4.6 Referencing an OSGi Service

In an OSGi-based application, the business logic behind a controller is typically accessed through the OSGi Service Registry.

Exporting Packages

By default, Bundlor detects and exports all packages in a bundle. In this step Bundlor is told what to export from the greenpages.app bundle and which types from those packages to use in the greenpages.web bundle.

Add and save the following entry to the template.mf file in the greenpages.app project and then run the MANIFEST.MF generation on the project as explained in the section called “Creating web module metadata”.

Excluded-Exports: 
 greenpages.internal

(As before, be careful not to leave trailing spaces on the ends of lines, except for the one space after the colon, and not to add any blank lines to the file. The second line of this entry has a leading space—do not omit it.)

Check that the package is no longer exported in the greenpages.app MANIFEST.MF file which should look something like this:

Manifest-Version: 1.0
Bundle-Name: GreenPages Service
Bundle-ManifestVersion: 2
Bundle-SymbolicName: greenpages
Tool: Bundlor 1.0.0.RELEASE
Export-Package: greenpages;version="2.3.0"
Bundle-Version: 2.3.0

Go to the next step.

Referencing Projects and Packages

Now that the greenpages.app bundle exports the package that the Directory and Listing interfaces reside in, the greenpages.web bundle must import it. In this step the Maven pom.xml file is updated to depend on the greenpages.app bundle and import the package.

Open the pom.xml file in the greenpages.web project. (Edit the source directly by using the pom.xml tab in the editor.) In this file add the following entry (between the <dependencies> tags):

<dependency>
  <groupId>com.springsource.dmserver</groupId>
  <artifactId>greenpages.app</artifactId>
  <version>${project.version}</version>
</dependency>

Open the GreenPagesController class and import the Listing and Directory types. (Eclipse should now offer these as a Quick Fix. It it does not, set greenpages.app as a project dependency of greenpages.web in the Build Path of the web project.) The class should now compile cleanly.

The following imports should now have been added to the GreenPagesController class:

import greenpages.Directory;
import greenpages.Listing;

import java.util.List;

import org.springframework.beans.factory.annotation.Autowired;
import org.springframework.stereotype.Controller;
import org.springframework.web.bind.annotation.RequestMapping;
import org.springframework.web.bind.annotation.RequestParam;

Add the following package clause to the Import-Template entry in the template.mf file in the greenpages.web project. When added run the MANIFEST.MF generation on the project as described in the section called “Creating web module metadata”.

greenpages.*;version="[2.3, 2.4)"

Be careful to include the “.*” in the package pattern.

Once Bundlor has finished, go to the next step.

Deploying a PAR

Currently the Web Server instance has a single web module bundle deployed. In this step, the greenpages.web bundle is undeployed and greenpages PAR is deployed.

Right-click on the Web Server in the Servers view, and select Add and Remove…. In the dialog that opens, remove the greenpages.web bundle and add the greenpages PAR to the server. When the configuration is complete, press Finish.

Eclipse automatically undeploys the greenpages.web bundle and deploys the greenpages PAR. When this happens, the deployment may fail with an error. If it does not, open the browser again at http://localhost:8080/greenpages and observe the failure which should have a root cause similar to:

org.springframework.beans.factory.NoSuchBeanDefinitionException: 
	No matching bean of type [greenpages.Directory] found for dependency: 
	expected at least 1 bean which qualifies as autowire candidate for this dependency. 
	Dependency annotations: {@org.springframework.beans.factory.annotation.Autowired(required=true)}

This error is caused by there being no instance of Directory to inject into the controller. The next section will supply one.

Referencing an OSGi Service

There is no instance of Directory to be injected into the controller. In the GreenPages application, it is intended that this implementation is used through an interface in the OSGi Service Registry. Using a service in the Service Registry enables another bundle to provide an implementation without revealing the implementation or the provider to all clients of the service. Web Server supports the use of the Spring DM namespace for referencing elements in the OSGi Service Registry. This step adds an OSGi Service Reference to an implementation of the Directory interface.

In the webapp/WEB-INF/applicationContext.xml file in the greenpages.web projects add a reference to a greenpages.Directory instance in the OSGi service registry using the <osgi:reference/> tag as follows:

<osgi:reference id="directory" interface="greenpages.Directory"/>

The tools will automatically redeploy the greenpages.web bundle when the change to the bean definition has been saved. The web bundle will not completely start.

This is because there is no provider of a greenpages.Directory in the Service Registry. The next step will address this.

The error is re-issued as the Web Server instance waits for the service to be supplied. After about five minutes the server will “time-out” and the deploy will be abandoned. This same error (and time-out) will occur each time the PAR is redeployed as each change is made.

Stop the server instance by right-clicking on the server in the Servers view and selecting Stop. This will avoid unnecessary delays as changes are made.

4.7 Publishing an OSGi Service

At the end of the previous step, a dependency was created on an OSGi Service Registry exposed instance of greenpages.Directory. The application would not start because no other bundle was contributing an instance of this service to the Service Registry.

Stop the server instance before proceeding.

Add Implementation

In this step Spring’s context scanning is added which will create an instance of the DirectoryImpl class.

Open the greenpages.internal.DirectoryImpl class in the greenpages.app project. Add the @Component annotation to the class:

@Component("directory")
public class DirectoryImpl implements Directory {
…

generating imports with Eclipse’s help if necessary.

Open the META-INF/spring/module-context.xml in the greenpages.app project. Add component scanning to this file:

<context:component-scan base-package="greenpages.internal"/>

When complete, go to the next step.

Publish OSGi Service

In this step the DirectoryImpl instance is published to the OSGi Service Registry.

Open the META-INF/spring/osgi-context.xml file. Add the <osgi:service/> tag to publish the directory bean with an interface of greenpages.Directory.

<osgi:service ref="directory" interface="greenpages.Directory"/>

A Working Web Application

Start (or restart) the Web Server instance from the Servers view. If the GreenPages PAR was not removed before, it will be automatically deployed, otherwise deploy it as before. There should be no errors reported. When GreenPages is deployed successfully, open a web browser and navigate to http://localhost:8080/greenpages. On the home page type wilkinson into the search field and press Submit. Unlike the previous attempt, this should return a list (of size 1) of search results. From here, select view to get the “detailed” listing.

This uses a stub implementation of the Directory interface which only knows about “Andy Wilkinson”.

The web interface is complete enough. Go to the next chapter to see the middle tier implementation.

5. The Middle Tier

5.1 Introduction

GreenPages’ middle-tier provides implementations of the Directory and Listing interfaces that can be used by the Web bundle. The implementation will use EclipseLink JPA to access a database via a DataSource published in the OSGi service registry.

The database

The GreenPages application uses a very simple database that contains a single table. The table, named LISTING, consists of four columns:

LISTING_NUMBERFIRST_NAMELAST_NAMEEMAIL_ADDRESS

Scripts are provided with the sample source code (in $GREENPAGES_HOME/db) to start, create, and populate the database. These will be used during the creation of the middle tier.

Using JPA

The middle tier will provide JPA-based implementations of the Directory and Listing interfaces with the four attributes of a Listing (first name, last name, email address, and id) being mapped to the corresponding columns in the LISTING. JPA will be used to implement the queries that search the database and return Listings.

Structure

The middle tier consists of two bundles, greenpages.jpa that publishes a Directory implementation for consumption by the Web bundle, and greenpages.db to configure and publish the DataSource used to access the database.

5.2 Creating the DataSource project

This section describes how to use the bundle project creation wizard to create a new Bundle Project. The project’s Spring bean definition files will also be created using the Spring bean configuration file creation wizard.

Creating a new Bundle Project

Create a new project by right-clicking in the Package Explorer view and selecting NewProject…. In the resulting dialog select EclipseRTBundle Project and press Next:

In the New Bundle Project dialog, name the project greenpages.db. Choose the create the project from an existing source location and specify a location that will place the new greenpages.db alongside the project skeletons that were imported into the workspace earlier. If the start directory of the GreenPages sample is being used this will be $GREENPAGES_HOME/start/greenpages.db (and, mutatis mutandis on Windows). Click Next.

In this page of the wizard, many of the Bundle Properties are already populated. The Bundle-SymbolicName is the name of the project. The Bundle-Name is derived from the Bundle-SymbolicName. The Bundle-Version is set, and there is no Bundle-Description.

Change the Bundle-Name to “GreenPages DataSource” to more accurately describe the bundle’s purpose. An option to ‘Enable Bundle Classpath Container’ is already selected. It should be de-selected, as a Maven Classpath container will be configured later. Ensure that the Target Runtime is set to Virgo Web Server (Runtime) v2.1. Click Finish.

The greenpages.db project appears in the Package Explorer view.

Configuring the project’s classpath container

Before a Maven Classpath Container can be added to the project, a pom.xml file must be created. Create a new file in the root of the greenpages.db project named pom.xml and add the following contents to it:

<?xml version="1.0" encoding="UTF-8"?>
<project
	xmlns="http://maven.apache.org/POM/4.0.0"
	xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
	xsi:schemaLocation="http://maven.apache.org/POM/4.0.0 http://maven.apache.org/maven-v4_0_0.xsd">

	<parent>
		<groupId>com.springsource.dmserver</groupId>
		<artifactId>greenpages.parent</artifactId>
		<version>2.3.0.RELEASE</version>
		<relativePath>../parent</relativePath>
	</parent>

	<modelVersion>4.0.0</modelVersion>
	<groupId>com.springsource.dmserver</groupId>
	<artifactId>greenpages.db</artifactId>
	<name>greenpages.db</name>
	<packaging>jar</packaging>

	<dependencies>
	</dependencies>

</project>

Save the file.

A Maven Classpath Container can now be added to the project. Right-click the greenpages.db project in the Package Explorer and select Maven 2Enable dependency management. Eclipse will perform some workspace building, and the greenpages.db project will now be marked as a Maven project. (If the error Cannot find artifact for parent POM occurs check that the version is correct. It may differ from the one given here.)

Configuring the source folders

The last part of the setup of the project is to configure its source folders. Return to the Properties dialog of the greenpages.db project (from the Package Explorer view). Select Java Build Path on the left-hand side and the Source tab on the right-hand side. Remove any pre-configured source folders by selecting them and clicking Remove.

Now click Add folder and then Create new folder…. Specify src/main/resources as the folder name and click Finish, then OK and OK again.

The final change to be made is to drag the META-INF folder from src to src/main/resources. Once these changes have been made the project will appear similar to the following in the Package Explorer view:

(It is useful to check that the Maven project just configured has the correct Java System library associated with it and that the MANIFEST.MF file that we have just moved is correctly generated when necessary. To ensure this, disable the dependency management (enabled in the previous section) and then re-enable it, saving the changes in between. This step may also change the Java system libraries associated with the build (right-click on the greenpages.db project, under Properties, Java build path and Libraries). If the Java system libraries look incorrect, simply remove them in the Libraries window and reset them.)

Configuring the DataSource

The DataSource bundle’s main rôle is to configure and create a DataSource object and to publish this to the OSGi service registry. This will be done by creating a handful of Spring beans.

By default, Spring DM looks for application context files in a bundle’s META-INF/spring directory. Create a new folder named spring in the greenpages.db project’s META-INF folder. Having created the new folder, right-click it in the Package Explorer and select New Spring Bean Configuration File. This will open the wizard for creating Spring bean configuration files.

In the wizard enter a File name of module-context.xml and click Next:

Add the p - http://www.springframework.org/schema/p namespace declaration to the pre-selected beans declaration and then click Finish.

Update the newly-created file (which is opened by Eclipse) to declare a bean that defines the DataSource object that will be used to access the GreenPages database. Do this by adding the following bean declaration:

	<bean id="dataSource" class="org.apache.commons.dbcp.BasicDataSource" 
	p:driverClassName="org.h2.Driver" p:url="jdbc:h2:~/greenpages-db/greenpages"
	p:username="greenpages" p:password="pass"
	init-method="createDataSource" destroy-method="close" />

The new bean has introduced a dependency on Commons DBCP, which will cause an error to be reported by Eclipse.

This dependency must be recorded in the project’s pom file. Open the pom file for greenpages.db and add the following dependency between the <dependencies> tags:

        <dependency>
            <groupId>org.apache.commons</groupId>
            <artifactId>com.springsource.org.apache.commons.dbcp</artifactId>
        </dependency>

Save the updated pom and then switch back to the editor for module-context.xml. Save the updated file and observe that the previously reported problem is now resolved as Commons DBCP is available on the classpath.

Now that the DataSource bean is available, it can be published into the OSGi service registry.

Right-click the spring folder and select New Spring Bean Configuration File again. This time specify a name of osgi-context.xml, click Next, and add the osgi namespace declaration (ensure that the resultant schema location string contains the URL http://www.springframework.org/schema/osgi/spring-osgi-1.2.xsd as http://www.springframework.org/schema/osgi/spring-osgi-2.0-m1.xsd will result in an obscure deployment failure due to an invalid 'cacheTarget' property). Click Finish and then add the following to the new file to publish the DataSource as a service:

    <!--
        export the dataSource bean to the OSGi service registry under the
        DataSource interface
    -->
    <osgi:service ref="dataSource" interface="javax.sql.DataSource" />

Configuring Bundlor’s manifest template

Bundlor uses a manifest template to control the contents of the generated manifest. Create a new file named template.mf in the root of the greenpages.db project. Open the existing MANIFEST.MF and switch to the MANIFEST.MF tab to view its source. Copy the contents. Switch to the editor for template.mf, switch to the template.mf tab and paste the contents from MANIFEST.MF. These entries will tell Bundlor what the resulting manifest’s bundle symbolic name, bundle version, etc. should be. Save the updated template.

Still in the template.mf editor switch to the Overview tab and click Update MANIFEST.MF which is under the “Bundle Actions” section.

At this point Bundlor will scan the project to determine its dependencies. It will scan both module-context.xml and osgi-context.xml looking for references to classes. For each class to which it finds a reference, an import for the class’s package will be added to the resulting manifest.

In this case, Bundlor will generate imports for both javax.sql and org.apache.commons.dbcp. These imports may not be resolved. The greenpages.db project needs to be associated with a Web Server instance which has the Commons DBCP bundle in its repository to resolve them. In any event the next step adds the greenpages.db project to the GreenPages PAR and will result in it inheriting the PAR project’s targetted runtime configuration.

Double-click the MANIFEST.MF file in the greenpages project in the Package Explorer view. Switch to the Dependencies tab and click Add…. Select greenpages.db and click OK. Save the updated file. A problem concerning the org.apache.commons.dbcp dependency should now be resolved (along with any other resolution errors) and (if the server is running) the GreenPages application will be redeployed due to the addition of the greenpages.db module. Start the server if it is not already running and observe that this deployment fails.

The deployment will fail because the org.h2.Driver class that is referenced in the DataSource bean’s definition in module-context.xml is not available to the bundle. (Check for the exception org.springframework.beans.factory.BeanCreationException with text something like:

Error creating bean with name 'dataSource'
	defined in URL [bundleentry://68.fwk504117357/META-INF/spring/ module-context.xml]: 
	Invocation of init method failed; 
	nested exception is org.apache.commons.dbcp.SQLNestedException: Cannot load JDBC driver class 'org.h2.Driver'

though the numbers might be different.)

There are a few cases where Bundlor will not identify a dependency on a class and, at the moment, this is one of them, although this is an area of Bundlor that is being improved all the time. Thankfully, it is easy to add the required import by making a simple update to the template.

Open the editor for the template.mf file in the greenpages.db project and add the following Import-Package header and save the updated manifest:

Import-Package: org.h2;version="[1.0.71,1.0.71]"

Saving the manifest will trigger a redeployment (or click on Update MANIFEST.MF as before) which will fail if the H2 database is not available. (Refer to the section the section called “Starting and configuring the database” in Chapter 3, Installing and exploring GreenPages to run and configure the database.)

If the database is running the GreenPages application should correctly deploy. Although the application web front-end will run, the database contents is not visible, of course, because we are still running with the stub version of the search method on the controller. The implementation of the Directory service needs to be changed to exploit the database.

(One possible cause of failure here is the username and password on the dataSource bean defined in module-context.xml. Check that these are exactly correct.)

5.3 Building the JPA module

In this section the JPA module in GreenPages is created, building upon an existing skeleton. JPA and its metadata are configured, and a JPA-based Directory service implementation is published which is then consumed by the application’s Web bundle.

Completing the JPA-based Directory implementation

The greenpages.jpa starter project provides the beginnings of a JPA-based implementation of Directory named JpaDirectory. Import the greenpages.jpa project from the $GREENPAGES_HOME/start directory.

Open the JpaDirectory.java source file in the greenpages.jpa package of greenpages.jpa project (under src/main/java).

The source file contains a Java Persistence Query Language (JPQL) search query that will be used to retrieve listings from the database, and empty implementations of the search and findListing methods.

First add an EntityManager to it. Before the new field can be added, EntityManager must be available on the classpath. Open the pom for greenpages.jpa and add the following dependency:

        <dependency>
            <groupId>javax.persistence</groupId>
            <artifactId>com.springsource.javax.persistence</artifactId>
        </dependency>

Now return to JpaDirectory and add the following field to the class along with an import for javax.persistence.EntityManager (which should be suggested by Eclipse):

    private EntityManager em;

This EntityManager can now be used to implement the search and findListing methods. Update the implementations of these two methods to match the following implementations and then save the updated class:

   public Listing findListing(int id) {
        return em.find(JpaListing.class, id);
    }

    @SuppressWarnings("unchecked")
    public List<Listing> search(String term) {
        return em.createQuery(SEARCH_QUERY).setParameter("term",
                "%" + term.toUpperCase() + "%").getResultList();
    }

(Warnings from Eclipse should now be absent.)

The application context now needs to be updated to create JpaDirectory and to create an EntityManager that can be injected into JpaDirectory.

Open module-context.xml in the META-INF/spring folder of the greenpages.jpa. Add the following beans that will create JpaDirectory and an EntityManager, enable load-time weaving that is required by JPA, and enable annotation-based configuration that will allow the EntityManager to be injected into JpaDirectory:

    <!--
        Activates a load-time weaver for the context. Any bean within the
        context that implements LoadTimeWeaverAware (such as
        LocalContainerEntityManagerFactoryBean) will receive a reference to
        the autodetected load-time weaver.
    -->
    <context:load-time-weaver aspectj-weaving="on" />

    <!-- JPA EntityManagerFactory -->
    <bean id="entityManagerFactory"
        class="org.springframework.orm.jpa.LocalContainerEntityManagerFactoryBean"
        p:dataSource-ref="dataSource">
        <property name="jpaVendorAdapter">
            <bean id="jpaVendorAdapter"
                class="org.springframework.orm.jpa.vendor.EclipseLinkJpaVendorAdapter"
                p:databasePlatform="org.eclipse.persistence.platform.database.HSQLPlatform"
                p:showSql="true" />
        </property>
    </bean>

    <!--
        Activates various annotations to be detected in bean classes: Spring's
        @Required and @Autowired, as well as JSR 250's @PostConstruct,
        @PreDestroy and @Resource (if available) and JPA's @PersistenceContext
        and @PersistenceUnit (if available).
    -->
    <context:annotation-config />

    <bean id="directory" class="greenpages.jpa.JpaDirectory" />

The addition of the new beans to the context has introduced a new dependency upon Spring’s ORM support and upon EclipseLink and its JPA implementation. Add the following dependencies to the pom file for greenpages.jpa and save it:

        <dependency>
            <groupId>org.springframework</groupId>
            <artifactId>org.springframework.spring-library</artifactId>
            <type>libd</type>
        </dependency>
        <dependency>
            <groupId>org.eclipse.persistence</groupId>
            <artifactId>com.springsource.org.eclipse.persistence</artifactId>
        </dependency>
        <dependency>
            <groupId>org.eclipse.persistence</groupId>
            <artifactId>com.springsource.org.eclipse.persistence.jpa</artifactId>
        </dependency>

Now switch back to module-context.xml for greenpages.jpa and observe that the errors relating to Spring’s ORM types have now been resolved. Save module-context.xml.

The application context now contains a factory that will create an EntityManager and is configured for annotation-based configuration. The last step in completing JpaDirectory is to annotate the EntityManager field so that Spring will inject the EntityManager created by the factory into the field.

Open JpaDirectory.java again and add an annotation @PersistenceContext to the EntityManager field.

@PersistenceContext
private EntityManager em;

Eclipse will suggest an import for javax.persistence.PersistenceContext; accept this and save the file.

Providing the JPA metadata

JPA uses a file named META-INF/persistence.xml to describe persistence units. persistence.xml refers to a second file, typically named META-INF/orm.xml, to define entity mappings. In the case of GreenPages the persistence.xml file specifies a single persistence unit that points to the greenpages.JpaListing class. The specified mapping file (META-INF/orm.xml) tells the JPA implementation how to map JpaListing to the LISTING database table described above. (For more information on JPA consult the Documentation section in the appendix.)

Create a new file named persistence.xml in the META-INF folder of the greenpages.jpa project. Add the following contents to the new file and then save it:

<?xml version="1.0" encoding="UTF-8" ?>
<persistence xmlns="http://java.sun.com/xml/ns/persistence"
    xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
    xsi:schemaLocation="http://java.sun.com/xml/ns/persistence 
                        http://java.sun.com/xml/ns/persistence/persistence_1_0.xsd"
    version="1.0">

    <persistence-unit name="GreenPages" transaction-type="RESOURCE_LOCAL">
        <class>greenpages.jpa.JpaListing</class>
    </persistence-unit>

</persistence>

Now create a new file named orm.xml also in the META-INF folder alongside persistence.xml. Add the following contents to the new file and then save it:

<?xml version="1.0" encoding="UTF-8" ?>
<entity-mappings xmlns="http://java.sun.com/xml/ns/persistence/orm"
    xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
    xsi:schemaLocation="http://java.sun.com/xml/ns/persistence/orm 
                        http://java.sun.com/xml/ns/persistence/orm_1_0.xsd"
    version="1.0">
    <package>greenpages.jpa</package>
    <entity class="greenpages.jpa.JpaListing" name="Listing">
        <table name="LISTING" />
        <attributes>
            <id name="listingNumber">
                <column name="LISTING_NUMBER" />
                <generated-value strategy="TABLE" />
            </id>
            <basic name="firstName">
                <column name="FIRST_NAME" />
            </basic>
            <basic name="lastName">
                <column name="LAST_NAME" />
            </basic>
            <basic name="emailAddress">
                <column name="EMAIL_ADDRESS" />
            </basic>
        </attributes>
    </entity>
</entity-mappings>

Consuming the DataSource from the service registry

The entityManagerFactory bean that was added earlier depends upon a bean named dataSource which it will use to connect the EntityManager to the GreenPages database. The greenpages.db module already publishes a DataSource to the service registry. greenpages.jpa must now be updated to consume this.

Open osgi-context.xml in the META-INF/spring folder of the greenpages.jpa project and add the following:

    <!-- import the DataSource from OSGi -->
    <osgi:reference id="dataSource" interface="javax.sql.DataSource" />

This will result in a bean being created in the application context that is named dataSource. The bean will be of type javax.sql.DataSource and will be backed by a service found in the OSGi service registry that implements the javax.sql.DataSource interface. (Some warnings concerning the dataSource bean will now disappear.)

Publishing the Directory implementation to the service registry

To make the JPA-based Directory implementation available to GreenPages’ Web module it must be “published” to the OSGi service registry.

Open osgi-context.xml in the META-INF/spring folder of the greenpages.jpa project, add the following and then save the updated file:

    <!-- export the directory bean to OSGi under the Directory interface -->
    <osgi:service ref="directory" interface="greenpages.Directory" />

Generating greenpages.jpa’s manifest using Bundlor

Open the template.mf file in the root of the greenpages.jpa project and switch to the template.mf tab. Add the following entries to the template and save it.

Import-Bundle: com.springsource.org.eclipse.persistence;version="[1.0.0,1.0.0]",
 com.springsource.org.eclipse.persistence.jpa;version="[1.0.0,1.0.0]"
Import-Package: org.springframework.context.weaving;version="[3.0,3.1)",
 org.springframework.transaction.aspectj;version="[3.0,3.1)"
Excluded-Exports: greenpages.jpa

The Excluded-Exports header tells Bundlor that the greenpages.jpa should not be exported from the greenpages.jpa bundle.

The Import-Package entries for org.springframework.context.weaving and org.springframework.transaction.aspectj are needed as Bundlor cannot, yet, detect that these packages are required.

Lastly, the Import-Bundle entries for EclipseLink and its JPA implementation are needed as Bundlor cannot, yet, detect that EclipseLink is the JPA implementation that is being used by GreenPages.

Switch to the Overview tab and click Update MANIFEST.MF. As with greenpages.db before, this update may result in some errors being reported in the manifest as the project is not associated with a targetted runtime. Double-click the MANIFEST.MF file in the greenpages project in the Package Explorer. Switch to the Dependencies tab and click Add…. Select greenpages.jpa and click OK. Save the updated file. The problems in the manifest should now be resolved and the GreenPages application should be redeployed due to the addition of the greenpages.jpa module. This redeployment should succeed and it’s now time to try the application again.

(A possible action if this fails is to Update (Maven) Dependencies on the project right-click menu in the Maven sub-menu.)

5.4 Trying out the JPA middle tier

Open a Web browser and navigate to http://localhost:8080/greenpages. Click the Submit button. Unfortunately the search will not return any results as the Web bundle is still using the stub Directory implementation provided by the greenpages.app module, rather than the JPA-based implementation that is provided by greenpages.jpa. This can be confirmed by using the Equinox console or the web-based admin console to examine the services being used by greenpages.web.

The service which is being used by the Web bundle can be changed at runtime without having to restart the application or the Web Server. This can be achieved by changing greenpages.app so that it no longer publishes its Directory implementation. As a result of this Directory service no longer being available, the Web bundle will automatically switch to using the JPA-based implementation.

Open the osgi-context.xml file in the META-INF/spring folder of the greenpages.app project and comment out the publication of the directory service:

<!-- <osgi:service interface="greenpages.Directory" ref="directory"/> -->

Now save the updated file which will cause the application to be updated and refreshed on the server. Switch back to the Web browser and click Submit again.

This time eight results should be returned. Clicking on any of the View links will display the listing’s details. The application is now working. All that remains is to apply some best practices to the middle tier.

5.5 Applying best practices to the middle tier

While the application middle tier now works as required, it does not observe a few Spring-related best practices.

Using transactions

At the moment, the middle tier does not make any use of transactions. This isn’t a problem while the database access methods are only running single queries, but could lead to problems in the future if the application is made more complex. Thankfully, adding the use of transactions to the middle tier is simple.

Open module-context.xml in the META-INF/spring folder of greenpages.jpa. Add the following bean definition to create a transaction manager and associate it with the context’s EntityManager:

    <!--
        Transaction manager for a single JPA EntityManagerFactory (alternative to JTA)
    -->
    <bean id="transactionManager" class="org.springframework.orm.jpa.JpaTransactionManager"
        p:entityManagerFactory-ref="entityManagerFactory" />

(Save it, and the greenpages.jpa module will be refreshed.)

Next, Spring must be told to enable transaction management. In keeping with the use of annotation-based configuration for the EntityManager, annotation-based transaction configuration will also be used. Add the following to enable AspectJ-powered transaction demarcation for appropriately annotated beans:

    <!--
        Instruct Spring to perform declarative transaction management
        automatically on annotated classes.
    -->
    <tx:annotation-driven mode="aspectj" />

Save the updated file which will trigger (another) successful refresh of greenpages.jpa.

Lastly, JpaDirectory needs to be annotated so that it is identified as requiring Spring-based transaction management. Open JpaDirectory.java in greenpages.jpa. Annotate the class with @Transactional and add an import for org.springframework.transaction.annotation.Transactional, which Eclipse should suggest:

import org.springframework.transaction.annotation.Transactional;

@Transactional
final class JpaDirectory implements Directory {
…

Save the updated file triggering another successful refresh: JpaDirectory is now transactional.

Enabling exception translation

When using JPA, the standard exceptions are somewhat out of keeping with Spring’s exception model. Spring provides support for automatically translating these exceptions into Spring’s DataAccessException hierarchy.

Open module-context.xml for greenpages.jpa again and add the following bean definition to add the exception translator to the application context:

    <!--
        Post-processor to perform exception translation on @Repository classes
        (from native exceptions such as JPA PersistenceExceptions to
        Spring&rsquo;s DataAccessException hierarchy).
    -->
    <bean class="org.springframework.dao.annotation.PersistenceExceptionTranslationPostProcessor" />

Save the updated file. The translation will only occur on classes that are annotated with Spring’s @Repository stereotype annotation. JpaDirectory needs to have this annotation added to it complete the enabling of the exception translation.

Open JpaDirectory.java again, annotate the class with @Repository and add an import for org.springframework.stereotype.Repository:

import org.springframework.stereotype.Repository;

@Transactional
@Repository
final class JpaDirectory implements Directory {

Save the updated file.

At this point the redeploy of the GreenPages application may fail with an error similar to this:

<SPDE0100E> The class with name 'org.springframework.dao.annotation.PersistenceExceptionTranslationPostProcessor', 
referenced by bean 'org.springframework.dao.annotation.PersistenceExceptionTranslationPostProcessor#0', 
could not be loaded by class loader 'ServerBundleClassLoader: [bundle=greenpages-1-greenpages.jpa_2.3.0]':
…

which indicates that there is some package (org.springframework.dao.annotation) which is not available to the “BundleClassLoader” for bundle greenpages-1-greenpages.jpa_2.3.0. We should look in the MANIFEST.MF file for this bundle, and see that this package is not imported (in the Import-Package header). Since Bundlor generated this file (controlled by the template file template.mf) we should check that the manifest was re-generated on our last change.

Open template.mf in greenpages.jpa and, in the Overview pane, click on Update MANIFEST.MF in the Bundle Actions section. The MANIFEST.MF file is updated, and the application is redeployed, this time successfully. It might be worthwhile checking the option Automatically update MANIFEST.MF in the background on the template.mf Overview pane so that the MANIFEST.MF is kept up to date as the project is changed.

Versioning imports

By default, Bundlor generates Import-Package entries with no version range specified. In the absence of a version range, the OSGi default of “any version” is used. Whilst this is very flexible it’s generally a good idea to restrict an import by specifying a narrower range. This can be achieved by providing Bundlor with some additional information in the manifest template.

Open template.mf for greenpages.jpa and add the following Import-Template header:

Import-Template: org.springframework.*;version="[3.0,3.1)",
 greenpages;version="[2.0,2.1)",
 javax.persistence;version="[1.0.0,1.0.0]"

If there is already an Import-Template header in the template, extend it to include the above package version range specifications.

This header tells Bundlor that all imports of org.springframework packages should be in the range 3.0 inclusive to 3.1 exclusive, that an import of the greenpages package should be in the range 2.0 inclusive to 2.1 exclusive, and that an import of javax.persistence should be at exactly version 1.0.0.

Bundlor has also generated an import for the javax.sql package due to the greenpages.jpa module’s use of javax.sql.DataSource. This class is provided by the JRE and as such is generally considered to be unversioned, that is it has the default OSGi version of zero. If version zero is precisely what is required then add the following to the Import-Template header:

,javax.sql;version="[0,0]"

but if “any” version is acceptable add the following instead:

,javax.sql;version="0"

Either of these will successfully allow GreenPages to deploy and work correctly. The difference is in the level of flexibility allowed with the external dependency, something which is probably irrelevant in this case, but with other package sources might be important.

Congratulations!

The GreenPages middle tier is now complete and observes some “best practice” development with Spring and OSGi.

6. Testing GreenPages

6.1 Introduction

Testing is one of the most important aspects of software development. Without testing it would be difficult to determine if a piece of code worked properly, changes would have undetected consequences, and the quality of products would generally be lower.

There are two major categories of testing generally recognised today: unit testing and integration testing. In the context of the GreenPages application, unit testing means testing a single class in isolation from other application code. This type of testing does not change at all when developing for Web Server.

In our application integration testing means testing an application or portion of an application with other code. This kind of testing does look a bit different when developing for Web Server. In most cases Web Server applications are made up of small bundles that consume services through the OSGi registry. In the following steps a single bundle and the entire GreenPages application will be integration tested outside the container.

6.2 Single bundle integration testing

One of the most common forms of integration testing is ensuring that the object relational mapping in an application is working properly. This kind of testing typically uses a data access object to retrieve data from a live database. In this step a test case for the JpaDirectory class is created.

Before proceeding, stop any Web Server instance that was previously running.

Open the greenpages.jpa.JpaDirectorySpringContextTests class in the src/test/java source folder of the greenpages.jpa project. This class contains a method that uses JUnit to test that a search completes correctly. Rather than instantiate this class directly in the test, the Spring Test Framework is used to instantiate and inject a JpaDirectory based on the META-INF/spring/module-context.xml file.

Add Spring Test Framework declarations to the test class. These declarations run the test with the SpringJunit4ClassRunner and configure the test with the classpath:/META-INF/spring/module-context.xml file:

@RunWith(SpringJUnit4ClassRunner.class)
@ContextConfiguration(locations = "classpath:/META-INF/spring/module-context.xml")
@TestExecutionListeners(value = DependencyInjectionTestExecutionListener.class)
public class JpaDirectorySpringContextTests {
…

Use Eclipse to suggest the necessary imports until there are no errors.

When this configuration is complete, click on the Run drop-down menu and select Run Configurations…. In the the dialog that opens select JUnitJpaDirectorySpringContextTests and press Run.

This test run will fail because there is no DataSource bean to be injected; it is typically sourced from the OSGi service registry at runtime:

Caused by: org.springframework.beans.factory.NoSuchBeanDefinitionException: 
    No bean named 'dataSource' is defined

The next step will correct this error.

6.3 Contributing OSGi sourced dependencies

In the previous step the JpaDirectorySpringContextTests test failed because it did not have a DataSource to be injected. In this step, an “in-process” database is instantiated and populated with data for testing.

Open the test-context.xml file in the src/test/resources META-INF/spring folder. In this file, define two beans; a DataSource and a TestDataPopulator. These two beans will provide a test DataSource complete with test data.

    <bean id="dataSource" class="org.apache.commons.dbcp.BasicDataSource"
        p:driverClassName="org.h2.Driver" p:url="jdbc:h2:.~/greenpages-db/greenpages"
        p:username="greenpages" p:password="pass" init-method="createDataSource"
        destroy-method="close" />

    <bean class="greenpages.jpa.TestDataPopulator" init-method="populate">
        <constructor-arg ref="dataSource" />
        <constructor-arg value="file:../../db/db.sql" />
    </bean>

Open the JpaDirectorySpringContextTests class and update the ContextConfiguration annotation to point at both the module-context.xml file and the test-context.xml file:

@ContextConfiguration(locations = {
        "classpath:/META-INF/spring/module-context.xml",
        "classpath:/META-INF/spring/test-context.xml" })

Once again use the JpaDirectorySpringContextTests JUnit profile to run the test class. Now that there is a DataSource being contributed, the test will pass. If the test fails, check that the H2 database server is still running in the background.

Proceed to the next step.

6.4 Multi bundle integration testing

Earlier a single bundle was integration tested by providing a test implementation of its DataSource dependency. When integration testing it is often a good idea to test the entire application outside of the container. In this step a test case for the entire GreenPages application is created, starting with the GreenPagesController class and descending all the way to a database. It would be sensible to create this in a separate test bundle but as one of the bundles involved here is a web bundle the tests will have to go in there.

Since this project will be testing the GreenPages application as a whole, it needs to depend on the bundles that make up the application. Open the pom.xml file for the greenpages.web project and add a dependency declaration for the greenpages.jpa bundle:

        <dependency>
            <groupId>com.springsource.dmserver</groupId>
            <artifactId>greenpages.jpa</artifactId>
            <version>${project.version}</version>
            <scope>test</scope>
        </dependency>

noting that the scope is test.

Open the GreenPagesSpringContextTests class and add the Spring Test Framework declarations. These declarations should run the test with the SpringJunit4ClassRunner and configure the test with the classpath*:/META-INF/spring/module-context.xml, file:src/main/webapp/WEB-INF/greenpages-servlet.xml and classpath:/META-INF/spring/test-context.xml files. Note the use of classpath*: with respect to the module-context.xml path. This will cause Spring to look for files that match that path in all of the bundles on the classpath meaning that all the application beans will be instantiated. Also, as we do not want the WEB-INF folder on the classpath we must reference the servlet context for GreenPages with a full file path:

@RunWith(SpringJUnit4ClassRunner.class)
@ContextConfiguration(locations = {
        "classpath*:/META-INF/spring/module-context.xml",
        "file:src/main/webapp/WEB-INF/greenpages-servlet.xml",
        "classpath:/META-INF/spring/test-context.xml" })
@TestExecutionListeners(value = DependencyInjectionTestExecutionListener.class)
public class GreenPagesSpringContextTests {
…

It may be necessary to click on Update MANIFEST.MF on the template overview pane and Update Dependencies from the Maven menu, before Eclipse will suggest appropriate imports here.

When this configuration is complete, click on the Run drop-down and select Run Configurations…. In the the dialog that opens select JUnitGreenPagesSpringContextTests and press Run;

When this test is run, Spring creates an ApplicationContext that is built from the module-context.xml configuration files from all of the bundles. Because of this all of the internal dependencies are satisfied by the beans created directly by the bundles.

The test should pass. If it doesn't, try the usual Eclipse dance steps of opening and closing the project and doing a clean rebuild to clear the problem.

There are warnings output by this test concerning log4j:

log4j:WARN No appenders could be found for logger
                                (org.springframework.test.context.junit4.SpringJUnit4ClassRunner).
log4j:WARN Please initialize the log4j system properly.

These warnings are benign, and do not influence the tests in any way.

The next chapter constructs an automated build system that might be used to build GreenPages (and run its tests) outside of an interactive development environment.

7. Automated Build

7.1 Introduction

One of the most important components in application development is the automated build. This permits application artifacts to be created outside of the developer’s IDE. The application can be created and tested in a variety of environments including continuous integration.

7.2 Setting up for Automated Build

Before building and deploying from the command line, it is important to clean up the artifacts that Eclipse has deployed. In this section the GreenPages application will be undeployed within Eclipse and all of the GreenPages bundles built from the command line.

Right-click on the greenpages application in the Servers view and select Remove. Once this is complete close Eclipse: it is no longer needed.

Run the following command from a command prompt with the $GREENPAGES_HOME/start as the current directory. This will build the individual bundles that make up the GreenPages application:

mvn clean install

The first time this is run will cause Maven to download quite a few packages. It is likely also that this does not build successfully on the first try, due to warnings from Bundlor. These warnings are due to the lack of information regarding some of the packages required by greenpages.db and greenpages.web. For example warnings like the following may be issued:

[WARNING] Bundlor Warnings:
[WARNING]     <SB0001W>: The import of package javax.sql does not specify a version.
[WARNING]     <SB0001W>: The import of package org.apache.commons.dbcp does not specify a version.
[INFO] ------------------------------------------------------------------------
[ERROR] BUILD ERROR
[INFO] ------------------------------------------------------------------------
[INFO] Bundle transformer returned warnings.
       Please fix manifest template at '/opt/greenpages-2.3.0.RELEASE/start/greenpages.db/template.mf'
       and try again.

which indicate that there is no information in the template.mf file in the greenpages.db project to inform Bundlor what version of these packages to generate in the MANIFEST.MF for that bundle.

To correct these problems add the following lines to the template.mf file for the greenpages.db bundle:

Import-Template: javax.sql;version="0",
 org.apache.commons.dbcp;version="[1.2.2.osgi, 1.2.2.osgi]"

and, if further warnings are issued, in the template.mf file of other bundles (for example, greenpages.jpa).

When the mvn command returns successfully, go to the next step.

7.3 Create POM

All of the projects except the PAR project have Maven POM files for building. In this step an initial POM file for the PAR is created.

Using a text editor create a file called $GREENPAGES_HOME/start/greenpages/pom.xml. Open this file and add the following skeleton to it:

<?xml version="1.0" encoding="UTF-8"?>
<project
    xmlns="http://maven.apache.org/POM/4.0.0"
    xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
    xsi:schemaLocation="http://maven.apache.org/POM/4.0.0 http://maven.apache.org/maven-v4_0_0.xsd">

  <parent>
    <groupId>com.springsource.dmserver</groupId>
    <artifactId>greenpages.parent</artifactId>
    <version>2.3.0.RELEASE</version>
    <relativePath>../parent</relativePath>
  </parent>

  <modelVersion>4.0.0</modelVersion>
  <groupId>com.springsource.dmserver</groupId>
  <artifactId>greenpages</artifactId>
  <name>GreenPages PAR</name>
  <packaging>par</packaging>

  <dependencies>
  </dependencies>

  <build>
    <plugins>
    </plugins>
  </build>

</project>

ensuring that the version numbers are consistent (for example, 2.3.0.RELEASE might be 2.3.0 depending on which version of greenpages being developed).

This skeleton defines a basic configuration with a parent POM. Notice that the packaging type is par. After this file is created execute the following command from the $GREENPAGES_HOME/start/greenpages directory.

mvn clean package

This command returns an error indicating that Maven does not know how to build a PAR:

[INFO] ------------------------------------------------------------------------
[ERROR] BUILD ERROR
[INFO] ------------------------------------------------------------------------
[INFO] The plugin 'org.apache.maven.plugins:maven-par-plugin' does not exist
[INFO] or no valid version could be found
[INFO] ------------------------------------------------------------------------

The next step will correct this.

7.4 Adding the par plugin

Thorsten Maus contributed a Maven plugin to SpringSource (see Section A.2, “Documentation”) that builds a PAR file from a list of dependencies. In this step the Maven par plugin is added to properly build a PAR artifact type.

In the <build><plugins>…</plugins></build> section, add a plugin declaration for the par plugin.

<plugin>
	<groupId>org.apache.maven.plugins</groupId>
	<artifactId>maven-par-plugin</artifactId>
	<version>1.0.0.RELEASE</version>
	<configuration>
		<applicationSymbolicName>greenpages</applicationSymbolicName>
		<applicationDescription>GreenPages</applicationDescription>
	</configuration>
</plugin>

Declare the list of bundles to be packaged in the PAR as dependencies of the PAR project.

<dependency>
  <groupId>com.springsource.dmserver</groupId>
  <artifactId>greenpages.app</artifactId>
  <version>${project.version}</version>
</dependency>
<dependency>
  <groupId>com.springsource.dmserver</groupId>
  <artifactId>greenpages.jpa</artifactId>
  <version>${project.version}</version>
</dependency>
<dependency>
  <groupId>com.springsource.dmserver</groupId>
  <artifactId>greenpages.db</artifactId>
  <version>${project.version}</version>
</dependency>
<dependency>
  <groupId>com.springsource.dmserver</groupId>
  <artifactId>greenpages.web</artifactId>
  <version>${project.version}</version>
  <type>war</type>
</dependency>

Now, run the following command.

mvn clean package

This command will now complete successfully and build a PAR into target/:

[INFO] Scanning for projects...
			[INFO] ------------------------------------------------------------------------
			[INFO] Building GreenPages PAR
			[INFO]    task-segment: [clean, package]
			[INFO] ------------------------------------------------------------------------
			[INFO] [clean:clean {execution: default-clean}]
			[INFO] [resources:resources {execution: default-resources}]
			[INFO] [par:par {execution: default-par}]
			[INFO] Assembling Artifacts for PAR '…/start/greenpages/target/greenpages-2.3.0.RELEASE.par'
			[INFO]   Added 'greenpages.app.jar'
			[INFO]   Added 'greenpages.jpa.jar'
			[INFO]   Added 'greenpages.db.jar'
			[INFO]   Added 'greenpages.web.war'
			[INFO] [com.springsource.bundlor.:transform {execution: bundlor}]
			[INFO] Ignored project with non-bundle packaging: [par]
			[INFO] ------------------------------------------------------------------------
			[INFO] BUILD SUCCESSFUL
			[INFO] ------------------------------------------------------------------------

Proceed to the next step.

7.5 Adding the dependency plugin

Maven now successfully builds the PAR for the application, however the dependencies of the PAR are not apparent. In this step the Maven dependency plugin is added to collect the transitive dependency graph for the PAR.

In the <build><plugins>…</plugins></build> section (after the par plugin declaration), add a plugin declaration for the dependency plugin:

<plugin>
  <groupId>org.apache.maven.plugins</groupId>
  <artifactId>maven-dependency-plugin</artifactId>
  <executions>
    <execution>
      <id>copy-dependencies</id>
      <phase>package</phase>
      <goals>
        <goal>copy-dependencies</goal>
      </goals>
      <configuration>
        <outputDirectory>${project.build.directory}/par-provided</outputDirectory>
        <overWriteIfNewer>true</overWriteIfNewer>
		<excludeGroupIds>com.springsource.dmserver,org.apache.log4j</excludeGroupIds>
      </configuration>
    </execution>
  </executions>
</plugin>

A dependency on Freemarker needs to be added to the other dependencies. This is required to ensure the Web bundle has the correct set of dependencies as well as the other bundles. Normally they would simply be resolved transitively from the bundle projects but the ‘war’ project does not pass on its dependencies; it expects them to be contained in its ‘lib’ directory. For this reason its dependencies must be given explicitly.

<!-- Required for the web bundle as dependencies are not propagated up from war build types -->
<dependency>
	<groupId>org.freemarker</groupId>
	<artifactId>com.springsource.freemarker</artifactId>
	<scope>provided</scope>
</dependency>

The next step is to stop the Web bundle including its dependencies in a lib directory as they will be provided by the runtime enviroment. Add the following build section to the greenpages.web POM file.

<build>
	<plugins>
		<plugin>
			<artifactId>maven-war-plugin</artifactId>
			<version>2.1-beta-1</version>
			<configuration>
				<packagingExcludes>WEB-INF/lib/**</packagingExcludes>
			</configuration>
		</plugin>
	</plugins>
</build>

Run the following command.

mvn clean package

When the command has completed, it will have copied all of the PAR’s dependencies into the target/par-provided directory. The output from Maven should include lines like these

[INFO] [par:par]
[INFO] Assembling Artifacts for PAR '/Users/chrisfrost/Repos/GIT/greenpages/solution/
       greenpages/target/greenpages-solution-2.3.0.RELEASE.par'
[INFO]   Added 'greenpages.app-solution.jar'
[INFO]   Added 'greenpages.jpa-solution.jar'
[INFO]   Added 'greenpages.db-solution.jar'
[INFO]   Added 'greenpages.web-solution.war'

If the dependencies are produced, proceed to the next step.

7.6 Automatically running the tests

Although the application is built, and dependencies produced for separate deployment, the tests are not run as part of that build.

Add (or replace) the following plug-in entry in the pom.xml file in the parent directory under start:

<plugin>
	<groupId>org.apache.maven.plugins</groupId>
	<artifactId>maven-surefire-plugin</artifactId>
	<configuration>
		<includes>
			<include>**/*Tests.java</include>
		</includes>
		<excludes>
			<exclude>**/Abstract*.java</exclude>
		</excludes>
		<junitArtifactName>org.junit:com.springsource.org.junit</junitArtifactName>
		<argLine>-javaagent:${user.home}/.m2/repository/org/springframework/org.springframework.instrument/3.0.0.M3/org.springframework.instrument-3.0.0.M3.jar</argLine>
	</configuration>
</plugin>

where the location of the user Maven repository is hard-coded.

Now run mvn clean install from the start directory. Observe that the tests we constructed before are now run.

7.7 Deploying the application

Maven can now build both the PAR application and the collection of dependencies required for the application. In this step the PAR and dependencies are copied to the Web Server and the PAR is started.

Change directory to start/greenpages.

Copy the JARs in the target/par-provided directory into the $VWS_HOME/repository/usr/ directory.

Copy the PAR (greenpages-2.3.0.RELEASE.par) in the target/ directory into the $VWS_HOME/pickup directory.

Start the Web Server and look for a message similar to:

<DE0005I> Started par 'greenpages' version '2.3.0.RELEASE'.

in the console output.

Once deployment of the GreenPages application has completed, navigate to http://localhost:8080/greenpages.

The GreenPages application has been built from the command line, with a complete dependency set generated for independent deployment.

The automated build and test procedure is to run mvn clean install from the base directory, generating the component bundles, and then to run mvn clean package from the greenpages directory to generate the PAR and produce all its dependencies.

Appendix A. Further Resources

A.1 Projects

  1. Virgo (http://www.eclipse.org/virgo) — homepage for Virgo.

  2. SpringSource.org (http://www.springsource.org) — homepage for Spring Framework.

  3. OSGi (http://www.osgi.org) — homepage for OSGi.

  4. H2 Database (http://www.h2database.com — homepage for the H2 database.

  5. FreeMarker (http://freemarker.sourceforge.net) — homepage for FreeMarker templating engine.

  6. Commons DBCP (http://commons.apache.org/dbcp) — homepage for Commons DBCP.

  7. Eclipse IDE (http://www.eclipse.org/eclipse) — homepage for Eclipse IDE.

  8. EclipseLink (http://www.eclipse.org/eclipselink) — homepage for EclipseLink JPA.

A.2 Documentation

  1. Virgo Web Server Documentation (http://www.eclipse.org/virgo/documentation) – cover page for all Virgo documentation.

  2. Spring DM Reference Guide (http://static.springsource.org/osgi/docs/1.2.0/reference/html/).

  3. Spring Framework 3.0 documentation (http://static.springsource.org/spring/docs/3.0.x/spring-framework-reference/html/).

  4. FreeMarker documentation (http://freemarker.sourceforge.net/docs).

  5. Eclipse IDE documentation (http://www.eclipse.org/documentation).

  6. EclipseLink documentation wiki (http://wiki.eclipse.org/EclipseLink/UserManual).

  7. Maven PAR plugin (http://blog.springsource.com/2009/06/24/maven-par-plugin-100m1/).