O'Reilly logo

Maven: The Definitive Guide by Sonatype Company

Stay ahead with the world's most comprehensive technology and business learning platform.

With Safari, you learn the way you learn best. Get unlimited access to videos, live online training, learning paths, books, tutorials, and more.

Start Free Trial

No credit card required

Chapter 1. Introducing Apache Maven

Convention over Configuration

Convention over configuration is a simple concept. Systems, libraries, and frameworks should assume reasonable defaults without requiring that unnecessary configuration systems should “just work.” Popular frameworks such as Ruby on Rails and EJB3 have started to adhere to these principles in reaction to the configuration complexity of frameworks such as the initial Enterprise JavaBeans™ (EJB) specifications. An illustration of convention over configuration is something like EJB3 persistence. All you need to do to make a particular bean persistent is to annotate that class with @Entity. The framework will then assume table names and column names from the name of the class and the names of the properties. Hooks are provided for you to override these names if the need arises, but, in most cases, you will find that using the framework-supplied defaults results in a faster project execution.

Maven incorporates the concept by providing sensible default behaviors for projects. Without customization, source code is assumed to be in ${basedir}/src/main/java and resources are assumed to be in ${basedir}/src/main/resources. Tests are assumed to be in ${basedir}/src/test, and a project is assumed to produce a JAR (Java ARchive) file. Maven assumes that you want to compile byte code to ${basedir}/target/classes and then create a distributable JAR file in ${basedir}/target. Although this might seem trivial, consider the fact that most Ant-based builds have to define the locations of these directories in every subproject. Maven’s adoption of convention over configuration goes further than just simple directory locations; Maven’s core plugins apply a common set of conventions for compiling source code, packaging distributions, generating web sites, and many other processes. Maven’s strength comes from the fact that it is “opinionated.” It has a defined lifecycle and a set of common plugins that know how to build libraries and web applications. If you follow the convention, Maven will require almost zero effort—just put your source in the correct directory, and Maven will take care of the rest.

One side effect of using systems that follow “convention over configuration” is that end users might feel that they are forced to use a particular setup. While it is certainly true that Maven has some central opinions that shouldn’t be challenged, most of the defaults can be customized. For example, the location of a project’s source code and resources can be customized, names of JAR files can be customized, and through the development of custom plugins, almost any behavior can be tailored to your specific environment’s requirements. If you don’t follow convention, Maven will allow you to customize defaults in order to adapt to your requirements.

A Common Interface

Before Maven provided a common interface for building software, every single project had someone dedicated to managing a completely custom build system, and developers had to take time away from developing software to learn about the idiosyncrasies of each new project they wanted to contribute to. In 2001, you’d take a completely different approach to building a project such as Apache Turbine than you would to building a project such as Tomcat. If a new source analysis tool came out that would perform static analysis on source code, or if someone developed a new unit testing framework, everyone would have to drop what they were doing and figure out how to fit it into each project’s custom build environment. How would you run unit tests? There were a thousand different answers. This environment was characterized by endless arguments about tools and build procedures. The age before Maven was an age of inefficiency—the age of the “Build Engineer.”

Today, most open source developers have used or are currently using Maven to manage new software projects. This transition is less about developers moving from one build tool to another and more about developers starting to adopt a common interface for project builds. As software systems have become more modular, build systems have become more complex, and the number of projects has skyrocketed. Before Maven, when you wanted to check out a project such as Apache ActiveMQ or Apache ServiceMix from Subversion and build it from source, you really had to set aside about an hour to figure out the build system for each particular project. What does the project need to build? What libraries do I need to download? Where do I put them? What goals can I execute in the build? In the best case, it took a few minutes to figure out a new project’s build, and in the worst cases (like the old Servlet API implementation in the Jakarta Project), a project’s build was so difficult it would take many hours just to get to the point where a new contributor could edit source and compile the project. These days, with Maven, you check it out from source, and you run mvn install.

Although Maven provides an array of benefits, including dependency management and reuse of common build logic through plugins, the core reason it has succeeded is that it has defined a common interface for building software. When you see that a project such as Apache Wicket uses Maven, you can assume that you’ll be able to check it out from source and build it with mvn install without much hassle. You know where the ignition key goes, and you know that the gas pedal is on the right and the brake is on the left.

Universal Reuse Through Maven Plugins

The core of Maven is pretty dumb; it doesn’t know how to do much beyond parsing a few XML documents and keeping track of a lifecycle and a few plugins. Maven has been designed to delegate most responsibility to a set of Maven plugins that can affect the Maven lifecycle and offer access to goals. Most of the action in Maven happens in plugin goals that take care of things like compiling source, packaging bytecode, publishing sites, and any other task that needs to happen in a build. The Maven you download from Apache doesn’t know much about packaging a WAR file or running JUnit tests; most of Maven’s intelligence is implemented in the plugins, and the plugins are retrieved from the Maven repository. In fact, the first time you run something like mvn install with a brand new Maven installation, it retrieves most of the core Maven plugins from the central Maven repository. This is more than just a trick to minimize the download size of the Maven distribution; this is behavior that allows you to upgrade a plugin to add capability to your project’s build. The fact that Maven retrieves both dependencies and plugins from the remote repository allows for universal reuse of build logic.

The Maven Surefire plugin is responsible for running unit tests. At some point between version 1.0 and the version that is in wide use today, someone decided to add support for the TestNG unit testing framework in addition to the support for JUnit. This happened in a way that didn’t break backward compatibility—if you were using the Surefire plugin to compile and execute JUnit 3 unit tests, and you upgraded to the most recent version of the Surefire plugin, your tests continued to execute without fail. You also gained new functionality, so if you wanted to execute unit tests in TestNG, you now had that ability, thanks to the efforts of the maintainers of the Surefire plugin. You also gained the ability to run annotated JUnit 4 unit tests. You gained all of these capabilities without having to upgrade your Maven installation or install new software. Most importantly, nothing about your project had to change aside from a version number for a plugin in a POM.

It is this mechanism that affects much more than the Surefire plugin: projects are compiled with a Compiler plugin, projects are turned into JAR files with a Jar plugin, and there are plugins for running reports, plugins for executing JRuby and Groovy code, as well as plugins to publish sites to remote servers. Maven has abstracted common build tasks into plugins that are maintained centrally and shared universally. If the state of the art changes in any area of the build, if some new unit testing framework is released or if some new tool is made available, you don’t have to be the one to hack your project’s custom build system to support it. You benefit from the fact that plugins are downloaded from a remote repository and maintained centrally. This is what is meant by universal reuse through Maven plugins.

Conceptual Model of a “Project”

Maven maintains a model of a project: you are not just compiling source code into bytecode, you are developing a description of a software project and assigning a unique set of coordinates to a project. You are describing the attributes of the project. What is the project’s license? Who develops and contributes to the project? What other projects does this project depend on? Maven is more than just a “build tool”; it is more than just an improvement on tools such as make and Ant; it is a platform that encompasses a new semantics related to software projects and software development. This definition of a model for every project enables such features as:

Dependency management

A project is defined as unique coordinates that consists of a group identifier, artifact identifier, and version. Projects can now use these coordinates to declare dependencies.

Remote repositories

Related to dependency management, we can use the coordinates defined in the Maven Project Object Model (POM) to create repositories of Maven artifacts.

Universal reuse of build logic

Plugins are coded to work with the POM; they are not designed to operate on specific files in known locations. Everything is abstracted into the model—plugin configuration and customization happens in the model.

Tool portability and integration

Tools such as Eclipse, NetBeans, and IntelliJ now have a common place to find information about a project. Before the advent of Maven, every integrated development environment (IDE) had a different way to store what was essentially a custom POM. Maven has standardized this description, and although each IDE continues to maintain custom project files, they can be easily generated from the model.

Easy searching and filtering of project artifacts

Tools such as Nexus allow you to index and search the contents of a repository using the information stored in the POM.

Maven has provided a foundation for the beginnings of a consistent semantic description of a software project.

Is Maven an Alternative to XYZ?

So, sure, Maven is an alternative to Ant, but Apache Ant continues to be a great, widely used tool. It has been the reigning champion of Java builds for years, and you can integrate Ant build scripts with your project’s Maven build very easily. This is a common usage pattern for a Maven project. On the other hand, as more and more open source projects move to Maven as a project management platform, working developers are starting to realize that Maven not only simplifies the task of build management, it is helping to encourage a common interface between developers and software projects. Maven is more of a platform than a tool. Although you can consider Maven an alternative to Ant, you are comparing apples to oranges. “Maven” includes more than just a build tool.

This is the central point that makes all of the Maven versus Ant, Maven versus Buildr, Maven versus Gradle arguments irrelevant. Maven isn’t totally defined by the mechanics of your build system. It isn’t about scripting the various tasks in your build as much as it is about encouraging a set of standards, a common interface, a lifecycle, a standard repository format, a standard directory layout, etc. It certainly isn’t about what format the POM happens to be in, i.e., XML versus YAML versus Ruby. Maven is much larger than that, and Maven refers to much more than the tool itself. When this book talks about Maven, it is referring to the constellation of software, systems, and standards that support it. Buildr, Ivy, Gradle—all of these tools interact with the repository format that Maven helped create, and you could just as easily use a tool such as Nexus to support a build written entirely in Buildr. Nexus is introduced in Chapter 16.

Although Maven is an alternative to many of these tools, the community needs to evolve beyond seeing technology as a zero-sum game between unfriendly competitors in a contest for users and developers. This might be how large corporations relate to one another, but it has very little relevance to the way that open source communities work. The headline “Who’s winning? Ant or Maven?” isn’t very constructive. If you force us to answer this question, we’re definitely going to say that Maven is a superior alternative to Ant as a foundational technology for a build; at the same time, Maven’s boundaries are constantly shifting and the Maven community is constantly trying to seek out new ways to become more ecumenical, interoperable, and cooperative. The core tenets of Maven are declarative builds, dependency management, repository managers, and universal reuse through plugins, but the specific incarnation of these ideas at any given moment is less important than the sense that the open source community is collaborating to reduce the inefficiency of “enterprise-scale builds.”

Comparing Maven and Ant

Although the previous section should convince you that the authors of this book have no interest in creating a feud between Apache Ant and Apache Maven, we are cognizant of the fact that most organizations have to make a decision between Ant and Maven. In this section, we compare and contrast the tools.

Ant excels at build process; it is a build system modeled after make with targets and dependencies. Each target consists of a set of instructions that are coded in XML. There is a copy task and a javac task as well as a jar task. When you use Ant, you supply it with specific instructions for compiling and packaging your output. Look at the simple build.xml file shown in Example 1-1.

Example 1-1. A simple Ant build.xml file
<project name="my-project" default="dist" basedir=".">
    <description>
        simple example build file
    </description>
  <!-- set global properties for this build -->
  <property name="src" location="src/main/java"/>
  <property name="build" location="target/classes"/>
  <property name="dist"  location="target"/>

  <target name="init">
    <!-- Create the time stamp -->
    <tstamp/>
    <!-- Create the build directory structure used by compile -->
    <mkdir dir="${build}"/>
  </target>

  <target name="compile" depends="init"
        description="compile the source " >
    <!-- Compile the java code from ${src} into ${build} -->
    <javac srcdir="${src}" destdir="${build}"/>
  </target>

  <target name="dist" depends="compile"
        description="generate the distribution" >
    <!-- Create the distribution directory -->
    <mkdir dir="${dist}/lib"/>

    <!-- Put everything in ${build} into the MyProject-${DSTAMP}.jar file -->
    <jar jarfile="${dist}/lib/MyProject-${DSTAMP}.jar" basedir="${build}"/>
  </target>

  <target name="clean"
        description="clean up" >
    <!-- Delete the ${build} and ${dist} directory trees -->
    <delete dir="${build}"/>
    <delete dir="${dist}"/>
  </target>
</project>

In this simple Ant example, you can see how you have to tell Ant exactly what to do. There is a compile goal that includes the javac task, which compiles the source in the src/main/java directory to the target/classes directory. You have to tell Ant exactly where your source is, where you want the resulting bytecode to be stored, and how to package this all into a JAR file. Although some recent developments help make Ant less procedural, a developer’s experience with Ant is in coding a procedural language written in XML.

Contrast the previous Ant example with a Maven example. In Maven, to create a JAR file from some Java source, all you need to do is create a simple pom.xml, place your source code in ${basedir}/src/main/java, and then run mvn install from the command line. The example Maven pom.xml that achieves the same results as the simple Ant file listed in Example 1-1 is shown in Example 1-2.

Example 1-2. A simple Maven pom.xml
<project>
  <modelVersion>4.0.0</modelVersion>
  <groupId>org.sonatype.mavenbook</groupId>
  <artifactId>my-project</artifactId>
  <version>1.0</version>
</project>

That’s all you need in your pom.xml. Running mvn install from the command line will process resources, compile source, execute unit tests, create a JAR, and install the JAR in a local repository for reuse in other projects. Without modification, you can run mvn site and then find an index.html file in target/site that contains links to Javadoc and a few reports about your source code.

Admittedly, this is the simplest possible example project: a project that contains only source code and produces a JAR; a project that follows Maven conventions and doesn’t require any dependencies or customization. If we want to start customizing the behavior, our pom.xml is going to grow in size, and in the largest of projects, you can see collections of very complex Maven POMs that contain a great deal of plugin customization and dependency declarations. But even when your project’s POM files become more substantial, they hold an entirely different kind of information from the build file of a similarly sized project using Ant. Maven POMs contain declarations: “This is a JAR project,” and “The source code is in src/main/java.” Ant build files contain explicit instructions: “This is project,” “The source is in src/main/java,” “Run javac against this directory,” “Put the results in target/classses,” “Create a JAR from the ....”, etc. Where Ant has to be explicit about the process, there is something “built-in” to Maven that just knows where the source code is and how it should be processed.

The differences between Ant and Maven in this example are:

Apache Ant
  • Ant doesn’t have formal conventions such as a common project directory structure; you have to tell Ant exactly where to find the source and where to put the output. Informal conventions have emerged over time, but they haven’t been codified into the product.

  • Ant is procedural; you have to tell Ant exactly what to do and when to do it. You have to tell it to compile, then copy, then compress.

  • Ant doesn’t have a lifecycle; you have to define goals and goal dependencies. You have to attach a sequence of tasks to each goal manually.

Apache Maven
  • Maven has conventions: in the example, it already knew where your source code was because you followed the convention. It put the bytecode in target/classes, and it produced a JAR file in target.

  • Maven is declarative; all you had to do was create a pom.xml file and put your source in the default directory. Maven took care of the rest.

  • Maven has a lifecycle, which you invoked when you executed mvn install. This command told Maven to execute a series of sequence steps until it reached the lifecycle. As a side effect of this journey through the lifecycle, Maven executed a number of default plugin goals that did things such as compile and create a JAR.

Maven has built-in intelligence about common project tasks in the form of Maven plugins. If you want to write and execute unit tests, all you need to do is write the tests, place them in ${basedir}/src/test/java, add a test-scoped dependency on either TestNG or JUnit, and run mvn test. If you want to deploy a web application and not a JAR, all you need to do is change you project type to WAR and put your docroot in ${basedir}/src/main/webapp. Sure, you could do all of this with Ant, but you would be writing the instructions from scratch. In Ant, you would first have to figure out where the JUnit JAR file should be, and then you would have to create a classpath that includes the JUnit JAR file, and then you would tell Ant where it should look for test source code, write a goal that compiles the test source to bytecode, and execute the unit tests with JUnit.

Without supporting technologies such as antlibs and Ivy (and even with these supporting technologies), Ant has the feeling of a custom procedural build. An efficient set of Maven POMs in a project that adheres to Maven’s assumed conventions has surprisingly little XML compared to the Ant alternative. Another benefit of Maven is the reliance on widely shared Maven plugins. Everyone uses the Maven Surefire plugin for unit testing, and if someone adds support for a new unit testing framework, you can gain new capabilities in your own build just by incrementing the version of a particular Maven plugin in your project’s POM.

The decision to use Maven or Ant isn’t a binary one, and Ant still has a place in a complex build. If your current build contains some highly customized process, or if you’ve written some Ant scripts to complete a specific process in a specific way that cannot be adapted to the Maven standards, you can still use these scripts with Maven. Ant is made available as a core Maven plugin. Custom Maven plugins can be implemented in Ant, and Maven projects can be configured to execute Ant scripts within the Maven project lifecycle.

Summary

This introduction has been kept purposefully short. We have covered a basic outline of what Maven is and how it stacks up to and improves on other build tools throughout time. The next chapter will explain how to install and run Maven, and Chapter 3 will dive into a simple project and show how Maven can perform phenomenal tasks with the smallest amount of configuration.

With Safari, you learn the way you learn best. Get unlimited access to videos, live online training, learning paths, books, interactive tutorials, and more.

Start Free Trial

No credit card required