Guide to Configuring Plug-ins

Introduction

In Maven, there are two kinds of plugins, build and reporting:

  • Build plugins are executed during the build and configured in the <build/> element.
  • Reporting plugins are executed during the site generation and configured in the <reporting/> element.

All plugins should have minimal required information: groupId, artifactId and version.

Important Note: Always define the version of each plugin used to guarantee build reproducibility. A good practice is to specify each build plugin's version in a <build><pluginManagement/></build> element. Often the <pluginManagement/> element is found in the parent POM. For reporting plugins, specify each version in the <reporting><plugins/></reporting> element (and in the <build><pluginManagement/></build> element too).

Generic Configuration

Maven plugins (build and reporting) are configured by specifying a <configuration> element where the child elements of the <configuration> element are mapped to fields, or setters, inside your Mojo. (Remember that a plug-in consists of one or more Mojos where a Mojo maps to a goal.) Say, for example, you have a Mojo that performs a query against a particular URL, with a specified timeout and list of options. The Mojo might look like the following:

@Mojo( name = "query" )
public class MyQueryMojo
    extends AbstractMojo
{
    @Parameter(property = "query.url", required = true)
    private String url;

    @Parameter(property = "timeout", required = false, defaultValue = "50")
    private int timeout;

    @Parameter(property = "options")
    private String[] options;

    public void execute()
        throws MojoExecutionException
    {
        ...
    }
}

To configure the Mojo from your POM with the desired URL, timeout and options you might have something like the following:

<project>
  ...
  <build>
    <plugins>
      <plugin>
        <artifactId>maven-myquery-plugin</artifactId>
        <version>1.0</version>
        <configuration>
          <url>http://www.foobar.com/query</url>
          <timeout>10</timeout>
          <options>
            <option>one</option>
            <option>two</option>
            <option>three</option>
          </options>
        </configuration>
      </plugin>
    </plugins>
  </build>
  ...
</project>

The elements in the configuration match the names of the fields in the Mojo. The mapping is straight forward. The url element maps to the url field, the timeout element maps to the timeout field, and the options element maps to the options field. The mapping mechanism can deal with arrays by inspecting the type of the field and determining if a suitable mapping is possible.

For Mojos that are intended to be executed directly from the CLI, their parameters usually provide a means to be configured via system properties instead of a <configuration> section in the POM. The plugin documentation for those parameters will list an expression that denotes the system properties for the configuration. In the Mojo above, the parameter url is associated with the expression ${query.url}, meaning its value can be specified by the system property query.url as shown below:

mvn myquery:query -Dquery.url=http://maven.apache.org

The name of the system property does not necessarily match the name of the mojo parameter. While this is a rather common practice, you will often notice plugins that employ some prefix for the system properties to avoid name clashes with other system properties. Though rarely, there are also plugin parameters that (e.g. for historical reasons) employ system properties which are completely unrelated to the parameter name. So be sure to have a close look at the plugin documentation.

Help Goal

Most Maven plugins have a help goal that prints a description of the plugin and its parameters and types. For instance, to see help for the javadoc goal, type:

mvn javadoc:help -Ddetail -Dgoal=javadoc

And you will see all parameters for the javadoc:javadoc goal, similar to this page.

Configuring Parameters

Parametrisation of Mojos is relying internally on Plexus Component Configuration API provided by sisu-plexus.

Mapping Value Objects

Mapping value types, like Boolean or Integer, is very simple. The <configuration> element might look like the following:

<project>
...
<configuration>
  <myString>a string</myString>
  <myBoolean>true</myBoolean>
  <myInteger>10</myInteger>
  <myDouble>1.0</myDouble>
  <myFile>c:\temp</myFile>
  <myURL>http://maven.apache.org</myURL>
</configuration>
...
</project>

The detailed type coercion is explained in the table below. For conversion to primitive types their according wrapper classes are used and automatically unboxed.

Parameter Class Conversion from String
Boolean Boolean.valueOf(String)
Byte Byte.decode(String)
Character Character.valueOf(char) of the first character in the given string
Class Class.forName(String)
java.util.Date SimpleDateFormat.parse(String) for the following patterns: yyyy-MM-dd hh:mm:ss.S a, yyyy-MM-dd hh:mm:ssa, yyyy-MM-dd HH:mm:ss.S or yyyy-MM-dd HH:mm:ss
Double Double.valueOf(String)
Enum Enum.valueOf(String)
java.io.File new File(String) with the file separators normalized to File.separatorChar. In case the file is relative, is is made absolute by prefixing it with the project's base directory.
Float Float.valueOf(String)
Integer Integer.decode(String)
Long Long.decode(String)
Short Short.decode(String)
String n/a
StringBuffer new StringBuffer(String)
StringBuilder new StringBuilder(String)
java.net.URI new URI(String)
java.net.URL new URL(String)

Mapping Complex Objects

Mapping complex types is also fairly straight forward. Let's look at a simple example where we are trying to map a configuration for Person object. The <configuration/> element might look like the following:

<project>
...
<configuration>
  <person>
    <firstName>Jason</firstName>
    <lastName>van Zyl</lastName>
  </person>
</configuration>
...
</project>

The rules for mapping complex objects are as follows:

  • There must be a private field that corresponds to name of the element being mapped. So in our case the person element must map to a person field in the mojo.
  • The object instantiated must be in the same package as the Mojo itself. So if your mojo is in com.mycompany.mojo.query then the mapping mechanism will look in that package for an object named Person. The mechanism capitalizes the first letter of the element name and uses that to search for the object to instantiate.
  • If you wish to have the object to be instantiated live in a different package or have a more complicated name, specify this using an implementation attribute like the following:
<project>
...
<configuration>
  <person implementation="com.mycompany.mojo.query.SuperPerson">
    <firstName>Jason</firstName>
    <lastName>van Zyl</lastName>
  </person>
</configuration>
...
</project>

Mapping Collection Types

The configuration mapping mechanism can easily deal with most collections so let's go through a few examples to show you how it's done:

Mapping Collections and Arrays

Mapping to collections works in much the same way as mapping to arrays. Each item is given in the XML as dedicated element. The element name does not matter in that case. So if you have a mojo like the following:

public class MyAnimalMojo
    extends AbstractMojo
{
    @Parameter(property = "animals")
    private List<String> animals;

    public void execute()
        throws MojoExecutionException
    {
        ...
    }
}

where you have a field named animals then your configuration for the plug-in would look like the following:

<project>
  ...
  <build>
    <plugins>
      <plugin>
        <artifactId>maven-myanimal-plugin</artifactId>
        <version>1.0</version>
        <configuration>
          <animals>
            <animal>cat</animal>
            <animal>dog</animal>
            <animal>aardvark</animal>
          </animals>
        </configuration>
      </plugin>
    </plugins>
  </build>
  ...
</project>

Where each of the animals listed would be entries in the animals field. Unlike arrays, collections do not necessarily have a specific component type. In order to derive the type of a collection item, the following strategy is used:

  1. If the first item XML element contains an implementation hint attribute, try to load the class with the given fully qualified class name from the attribute value
  2. If the first item XML element contains a ., try to load the class with the fully qualified class name given in the element name
  3. Try the first item XML element name (with capitalized first letter) as a class in the same package as the mojo/object being configured
  4. Use the parameter type information from either Field.getGenericType() or Method.getGenericParameterTypes()
  5. If the element has no children, assume its type is String. Otherwise, the configuration will fail.

The following collection implementations are being used when there is no implementation hint attribute in the XML element representing the collection:

Collection Class Used for
TreeSet for all types assignable to SortedSet
HashSet for all types assignable to Set
ArrayList for every other Collection type which is not a Map

Since Maven 3.3.9 (MNG-5440), you can list individual items alternatively as comma-separated list in the XML value of animals directly. This approach is also used if configuring collection/array parameters via command line The following example is equivalent to the example above:

<project>
  ...
  <build>
    <plugins>
      <plugin>
        <artifactId>maven-myanimal-plugin</artifactId>
        <version>1.0</version>
        <configuration>
          <animals>cat,dog,aardvark</animals>
        </configuration>
      </plugin>
    </plugins>
  </build>
  ...
</project>

Each item is mapped again according to the rules of this section depending on the type of the collection/array.

Mapping Maps

In the same way, you could define maps like the following:

...
    @Parameter
    private Map<String,String> myMap;
...
<project>
...
  <configuration>
    <myMap>
      <key1>value1</key1>
      <key2>value2</key2>
    </myMap>
  </configuration>
...
</project>

Unlike Collections the value type for Maps is always derived from the parameter type information from either Field.getGenericType() or Method.getGenericParameterTypes(). It falls back to String. The key type must always be String.

In contrast to value objects and collections/arrays there is no string coercion defined for maps, i.e. you cannot give parameters of that type via CLI argument.

The map implementation class is by default TreeMap but can be overridden with an implementation attribute on the XML element representing the map.

Mapping Properties

Properties should be defined like the following:

...
    @Parameter
    private Properties myProperties;
...
<project>
...
  <configuration>
    <myProperties>
      <property>
        <name>propertyName1</name>
        <value>propertyValue1</value>
      </property>
      <property>
        <name>propertyName2</name>
        <value>propertyValue2</value>
      </property>
    </myProperties>
  </configuration>
...
</project>

In contrast to value objects and collections/arrays there is no string coercion defined for properties, i.e. you cannot give parameters of those type via CLI argument.

Configuring Build Plugins

The following is only to configure Build plugins in the <build> element.

Using the <executions> Tag

You can also configure a mojo using the <executions> tag. This is most commonly used for mojos that are intended to participate in some phases of the build lifecycle. Using MyQueryMojo as an example, you may have something that will look like:

<project>
  ...
  <build>
    <plugins>
      <plugin>
        <artifactId>maven-myquery-plugin</artifactId>
        <version>1.0</version>
        <executions>
          <execution>
            <id>execution1</id>
            <phase>test</phase>
            <configuration>
              <url>http://www.foo.com/query</url>
              <timeout>10</timeout>
              <options>
                <option>one</option>
                <option>two</option>
                <option>three</option>
              </options>
            </configuration>
            <goals>
              <goal>query</goal>
            </goals>
          </execution>
          <execution>
            <id>execution2</id>
            <configuration>
              <url>http://www.bar.com/query</url>
              <timeout>15</timeout>
              <options>
                <option>four</option>
                <option>five</option>
                <option>six</option>
              </options>
            </configuration>
            <goals>
              <goal>query</goal>
            </goals>
          </execution>
        </executions>
      </plugin>
    </plugins>
  </build>
  ...
</project>

The first execution with id "execution1" binds this configuration to the test phase. The second execution does not have a <phase> tag, how do you think will this execution behave? Well, goals can have a default phase binding as discussed further below. If the goal has a default phase binding then it will execute in that phase. But if the goal is not bound to any lifecycle phase then it simply won't be executed during the build lifecycle.

Note that while execution id's have to be unique among all executions of a single plugin within a POM, they don't have to be unique across an inheritance hierarchy of POMs. Executions of the same id from different POMs are merged. The same applies to executions that are defined by profiles.

How about if we have a multiple executions with different phases bound to it? How do you think will it behave? Let us use the example POM above again, but this time we shall bind execution2 to a phase.

<project>
  ...
  <build>
    <plugins>
      <plugin>
        ...
        <executions>
          <execution>
            <id>execution1</id>
            <phase>test</phase>
            ...
          </execution>
          <execution>
            <id>execution2</id>
            <phase>install</phase>
            <configuration>
              <url>http://www.bar.com/query</url>
              <timeout>15</timeout>
              <options>
                <option>four</option>
                <option>five</option>
                <option>six</option>
              </options>
            </configuration>
            <goals>
              <goal>query</goal>
            </goals>
          </execution>
        </executions>
      </plugin>
    </plugins>
  </build>
  ...
</project>

If there are multiple executions bound to different phases, then the mojo is executed once for each phase indicated. Meaning, execution1 will be executed applying the configuration setup when the phase of the build is test, and execution2 will be executed applying the configuration setup when the build phase is already in install.

Now, let us have another mojo example which shows a default lifecycle phase binding.

@Mojo( name = "query", defaultPhase = LifecyclePhase.PACKAGE )
public class MyBoundQueryMojo
    extends AbstractMojo
{
    @Parameter(property = "query.url", required = true)
    private String url;

    @Parameter(property = "timeout", required = false, defaultValue = "50")
    private int timeout;

    @Parameter(property = "options")
    private String[] options;

    public void execute()
        throws MojoExecutionException
    {
        ...
    }
}

From the above mojo example, MyBoundQueryMojo is by default bound to the package phase (see the @phase notation). But if we want to execute this mojo during the install phase and not with package we can rebind this mojo into a new lifecycle phase using the <phase> tag under <execution>.

<project>
  ...
  <build>
    <plugins>
      <plugin>
        <artifactId>maven-myquery-plugin</artifactId>
        <version>1.0</version>
        <executions>
          <execution>
            <id>execution1</id>
            <phase>install</phase>
            <configuration>
              <url>http://www.bar.com/query</url>
              <timeout>15</timeout>
              <options>
                <option>four</option>
                <option>five</option>
                <option>six</option>
              </options>
            </configuration>
            <goals>
              <goal>query</goal>
            </goals>
          </execution>
        </executions>
      </plugin>
    </plugins>
  </build>
  ...
</project>

Now, MyBoundQueryMojo default phase which is package has been overridden by install phase.

Note: Configurations inside the <executions> element used to differ from those that are outside <executions> in that they could not be used from a direct command line invocation because they were only applied when the lifecycle phase they were bound to was invoked. So you had to move a configuration section outside of the executions section to apply it globally to all invocations of the plugin. Since Maven 3.3.1 this is not the case anymore as you can specify on the command line the execution id for direct plugin goal invocation. Hence if you want to run the above plugin and it's specific execution1's configuration from the command-line, you can execute:

mvn myquery:query@execution1

Using the <dependencies> Tag

You could configure the dependencies of the Build plugins, commonly to use a more recent dependency version.

For instance, the Maven Antrun Plugin version 1.2 uses Ant version 1.6.5, if you want to use the latest Ant version when running this plugin, you need to add <dependencies> element like the following:

<project>
  ...
  <build>
    <plugins>
      <plugin>
        <groupId>org.apache.maven.plugins</groupId>
        <artifactId>maven-antrun-plugin</artifactId>
        <version>1.2</version>
        ...
        <dependencies>
          <dependency>
            <groupId>org.apache.ant</groupId>
            <artifactId>ant</artifactId>
            <version>1.7.1</version>
          </dependency>
          <dependency>
            <groupId>org.apache.ant</groupId>
            <artifactId>ant-launcher</artifactId>
            <version>1.7.1</version>
          </dependency>
         </dependencies>
      </plugin>
    </plugins>
  </build>
  ...
</project>

Using the <inherited> Tag In Build Plugins

By default, plugin configuration should be propagated to child POMs, so to break the inheritance, you could use the <inherited> tag:

<project>
  ...
  <build>
    <plugins>
      <plugin>
        <groupId>org.apache.maven.plugins</groupId>
        <artifactId>maven-antrun-plugin</artifactId>
        <version>1.2</version>
        <inherited>false</inherited>
        ...
      </plugin>
    </plugins>
  </build>
  ...
</project>

Configuring Reporting Plugins

The following is only to configure Reporting plugins in the <reporting> element.

Using the <reporting> Tag VS <build> Tag

Configuring a reporting plugin in the <reporting> or <build> elements in the pom does not exactly have the same results.

mvn site
Since maven-site-plugin 3.4, it uses the parameters defined in the <configuration> element of each reporting Plugin specified in the <reporting> element, in addition to the parameters defined in the <configuration> element of each plugin specified in <build> (parameters from <build> section were previously ignored).
mvn aplugin:areportgoal
It ignores the parameters defined in the <configuration> element of each reporting Plugin specified in the <reporting> element; only parameters defined in the <configuration> element of each plugin specified in <build> are used.

Using the <reportSets> Tag

You can configure a reporting plugin using the <reportSets> tag. This is most commonly used to generate reports selectively when running mvn site. The following will generate only the project team report.

<project>
  ...
  <reporting>
    <plugins>
      <plugin>
        <groupId>org.apache.maven.plugins</groupId>
        <artifactId>maven-project-info-reports-plugin</artifactId>
        <version>2.1.2</version>
        <reportSets>
          <reportSet>
            <reports>
              <report>project-team</report>
            </reports>
          </reportSet>
        </reportSets>
      </plugin>
    </plugins>
  </reporting>
  ...
</project>

Notes:

  1. To exclude all reports, you need to use:
      <reportSets>
        <reportSet>
          <reports/>
        </reportSet>
      </reportSets>
  2. Refer to each Plugin Documentation (i.e. plugin-info.html) to know the available report goals.

Using the <inherited> Tag In Reporting Plugins

Similar to the build plugins, to break the inheritance, you can use the <inherited> tag:

<project>
  ...
  <reporting>
    <plugins>
      <plugin>
        <groupId>org.apache.maven.plugins</groupId>
        <artifactId>maven-project-info-reports-plugin</artifactId>
        <version>2.1.2</version>
        <inherited>false</inherited>
      </plugin>
    </plugins>
  </reporting>
  ...
</project>