Advanced Assembly-Descriptor Topics

Quick Note on All includes and excludes Patterns

excludes take priority over includes.

Archive file resolution

If two or more elements (e.g., file, fileSet) select different sources for the same file for archiving, only one of the source files will be archived. The assembly plugin decides which source will be archived according to the following behavior (as per 2.2-SNAPSHOT).

  • Case 1: Two fileSets each select a different source for the same archived file.

    For example, one fileSet selects sourceDir1/config.xml and another selects sourceDir2/config.xml to be archived as config/config.xml. In this case, assembly plugin archives the source file selected in the latter fileSet.

    Given these two fileSets, the plugin will archive the config.xml file from source2Dir.

    <fileSets>
     <fileSet>
      <directory>source1Dir</directory>
      <outputDirectory>config</outputDirectory>
      <includes>
       <include>**/config.xml</include>
      </includes>
     </fileSet>
    
     <fileSet>
      <directory>source2Dir</directory>
      <outputDirectory>config</outputDirectory>
      <includes>
        <include>**/config.xml</include>
      </includes>
     </fileSet>
    </fileSets>
  • Case 2: A fileSet and a file element each select a different source file for the same archived file.

    For example, the fileSet selects sourceDir1/config.xml and the file selects sourceDir2/config.xml; both to be archived as config/config.xml.

    In this case, the assembly plugin archives the item selected in the fileSet, not the file element. This rule is true despite the order in which the files and fileSets elements appear.

    For example, given these the following fileSet and file elemetns, the plugin will archive the config.xml file from source1Dir.

    <fileSets>
     <fileSet>
      <directory>source1Dir</directory>
      <outputDirectory>config</outputDirectory>
      <includes>
       <include>**/config.xml</include>
      </includes>
     </fileSet>
    </fileSets>
    
    <files>
     <file>
      <source>source2Dir/config.xml</source>
      <outputDirectory>config</outputDirectory>
     </file>
    </files>

Advanced Artifact-Matching in includes and excludes

When using dependencySet or moduleSet, the <includes/> and <excludes/> sections actually apply to artifacts, not filenames. This can be a good thing, since you don't have to know the artifact's filename in the local repository. However, explicitly specifying the full artifact ID (consisting of groupId, artifactId, version, type, and classifier) for each artifact to be included or excluded can lead to very a verbose descriptor. Starting with version 2.2, the assembly plugin addresses the clumsiness of explicit artifact identification through the use of wildcard patterns.

The following easy rules should be applied when specifying artifact-matching patterns:

  1. Artifacts are matched by a set of identifier strings. In the following strings, type is 'jar' by default, and classifier is omitted if null.
    • groupId:artifactId:type:classifier ( artifact.getDependencyConflictId() )
    • groupId:artifactId ( ArtifactUtils.versionlessKey( artifact ) )
    • groupId:artifactId:type:classifier:version ( artifact.getId() )
  2. Any '*' character in an include/exclude pattern will result in the pattern being split, and the sub-patterns being matched within the three artifact identifiers mentioned above, using String.indexOf(..).
  3. When no '*' is present in an include/exclude pattern, the pattern will only match if the entire pattern equals one of the three artifact identifiers above, using the String.equals(..) method.
  4. In case you missed it above, artifact-identification fields are separated by colons (':') in the matching strings. So, a wildcard pattern that matches any artifact of type 'war' might be specified as *:war.

Example: Include all dependencies of type 'war'

In this example, we'll configure a dependencySet so it only includes those war dependencies.

<assembly xmlns="http://maven.apache.org/plugins/maven-assembly-plugin/assembly/1.1.2"
    xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
    xsi:schemaLocation="http://maven.apache.org/plugins/maven-assembly-plugin/assembly/1.1.2 http://maven.apache.org/xsd/assembly-1.1.2.xsd">
  [...]
  <dependencySets>
    <dependencySet>
      <includes>
        <include>*:war</include>
      </includes>
    </dependencySet>
  </dependencySets>
  [...]
</assembly>
GOTCHA!

In the above example, any war artifacts that happen to have a classifier (not sure why this particular case would happen, but it is possible) will be skipped. If you really want to be careful about catching all of the war artifacts in your project, you might want to use the following pattern:

*:war:*

Example: Exclude all source-jar dependencies.

In this example, we're dealing with the fact that project sources are often distributed using jar files, in addition to normal binaries. We want to filter out any source-jar files (they'll be marked with a sources classifier) from the binary jars.

<assembly xmlns="http://maven.apache.org/plugins/maven-assembly-plugin/assembly/1.1.2"
    xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
    xsi:schemaLocation="http://maven.apache.org/plugins/maven-assembly-plugin/assembly/1.1.2 http://maven.apache.org/xsd/assembly-1.1.2.xsd">
  [...]
  <dependencySets>
    <dependencySet>
      <includes>
        <include>*:jar:*</include>
      </includes>
      <excludes>
        <exclude>*:sources</exclude>
      </excludes>
    </dependencySet>
  </dependencySets>
  [...]
</assembly>

Including Subversion Metadata Directories in a FileSet

For most use cases, it's important to avoid adding metadata files from your source-control system, such as Subversion's .svn directories. Such metadata can increase the size of the resulting assembly vastly. By default, the assembly plugin will exclude metadata files for most common source-control systems from the fileSets specified in the descriptor.

On the other hand, what if you wanted to include Subversion metadata directories? Starting with version 2.2, the assembly plugin offers the useDefaultExcludes option on all fileSet elements, in order to accommodate this use case.

Example: Bundle project sources for a developer-quickstart pack

In this example, let's examine what happens if you have a large project in source control. This project contains a large number of sizable files that haven't changed since the day they were added, in the early stages of the project's lifetime. You want to enable potential developers to get started quickly, without checking out hundreds of 10-megabyte files first.

The compression incorporated with many archiving formats can offer an advantage here. If we create a project assembly, including Subversion metadata directories, developers should be able to download the assembly artifact and expand it, then simply type svn up.

<assembly xmlns="http://maven.apache.org/plugins/maven-assembly-plugin/assembly/1.1.2"
    xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
    xsi:schemaLocation="http://maven.apache.org/plugins/maven-assembly-plugin/assembly/1.1.2 http://maven.apache.org/xsd/assembly-1.1.2.xsd">
  [...]
  <fileSets>
    <fileSet>
      <useDefaultExcludes>false</useDefaultExcludes>
      <excludes>
        <exclude>**/target/**</exclude>
      </excludes>
    </fileSet>
  </fileSets>
  [...]
</assembly>

NOTE: You'll notice that we're excluding all target directories; these are a form of "calculated" and otherwise transient data, and generally shouldn't be included in archives, unless your goal is to create project binaries or similar.

Using Regular Expressions to Exclude Files

Since: 2.2-beta-4

Sometimes, you may find you need to specify an extremely fine-grained inclusion or exclusion pattern for a fileSet. In these cases, you have the option of specifying your pattern in the form of a regular expression by using the %regex[...] syntax.

Note: For completeness, the default pattern type - Ant-style patterns - can also be specified using the new %ant[...] syntax. This will allow room for future expansion of fileSet patterns, including the option to change the default pattern syntax someday.

Example: Including directories named target in the src directory

In this example, we want to produce a buildable source distribution of a Maven project hierarchy. Obviously, each project's target directory is a temporary workspace for the build process, so we want to exclude these directories. However, if one or more of the projects also includes a subdirectory named target in the src directory structure - perhaps as part of a Java package name - we want to make sure the files in this directory are included in the assembly.

<assembly>
  [...]
  <fileSets>
    <fileSet>
      <directory>/home/release15/maven-plugins/maven-assembly-plugin</directory>
      <outputDirectory>/</outputDirectory>
      <excludes>
        <exclude>%regex[(?!.*src/).*target.*]</exclude>
      </excludes>
    </fileSet>
    [...]
  </fileSets>
  [...]
</assembly>

The above fileSet uses a somewhat obscure feature of regular expressions called negative lookahead, which means our exclude pattern will only match paths that contain the word target but don't contain src. Effectively, any target directory within the src directory structure will be preserved in the assembly.

Using Strict-Filtering to Catch Obsolete Patterns or Incorrect Builds

At times, you want to build in a set of sanity checks when creating your assembly, to ensure that what goes into the assembly artifact is what you intended. One way you can do this is by enabling useStrictFiltering on your dependencySets, moduleSets, and fileSets.

useStrictFiltering is a flag that tells the assembly plugin to track each include/exclude pattern to make sure it's used during creation of the assembly. This way, if the assembly-descriptor author intended for a particular file or artifact to be present, he can add an include/exclude pattern to the descriptor to ensure that file/artifact is present, and then set the useStrictFiltering flag. If the pattern isn't used to match at least one file during assembly creation, the build will fail and the user will receive a message notifying him of the unused patterns.

Example: Ensuring the LICENSE.txt file is included in a jar

In this example, we want to make sure that our project jar contains the project's open source license language, in order to be compliant with our software foundation's policies.

<assembly xmlns="http://maven.apache.org/plugins/maven-assembly-plugin/assembly/1.1.2"
    xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
    xsi:schemaLocation="http://maven.apache.org/plugins/maven-assembly-plugin/assembly/1.1.2 http://maven.apache.org/xsd/assembly-1.1.2.xsd">
  [...]
  <fileSets>
    <fileSet>
      <useStrictFiltering>true</useStrictFiltering>
      <outputDirectory>META-INF</outputDirectory>
      <includes>
        <include>LICENSE.txt</include>
      </includes>
    </fileSet>
    [...]
  </fileSets>
  [...]
</assembly>

If a developer inadvertently removes the LICENSE.txt from the project directory, the assembly plugin should refuse to build this assembly.

Using an Alternative Assembly Base Directory

In many cases, assemblies should have all files arranged under one assembly base directory. This way, a user who expands the assembly will have all of the contents collected in a nice, neat directory structure, rather than spread throughout the current working directory. This is achieved using the includeBaseDirectory flag, and this flag is set to true by default, which will result in the project's artifactId-version being used as the assembly base directory.

However, in some special cases you may want to use a different directory name for the root of your assembly. Starting in the 2.2 version of the assembly plugin, this use case is addressed using the baseDirectory element of the assembly descriptor. With this element, you can use POM expressions and static strings to specify the name of the assembly root directory.

Example: Eclipse-style invariable directory name for the Maven assembly

In this example, let's explore what would happen if we wanted Maven to use the Eclipse approach for naming the root directory in its distribution assemblies. This way, instead of expanding the distribution to find a new maven-2.0.4 directory, you'd find a maven directory. Additionally, consider that the distribution assembly is currently built from the maven-core project, which means we shouldn't use the artifactId as part of the assembly root directory.

<assembly xmlns="http://maven.apache.org/plugins/maven-assembly-plugin/assembly/1.1.2"
    xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
    xsi:schemaLocation="http://maven.apache.org/plugins/maven-assembly-plugin/assembly/1.1.2 http://maven.apache.org/xsd/assembly-1.1.2.xsd">
  [...]
  <baseDirectory>maven</baseDirectory>
  [...]
</assembly>

Now, imagine that the distribution assembly were created in the top-level maven project. Now, we can use the artifactId, and probably should, just to minimize the maintenance of these files.

<assembly xmlns="http://maven.apache.org/plugins/maven-assembly-plugin/assembly/1.1.2"
    xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
    xsi:schemaLocation="http://maven.apache.org/plugins/maven-assembly-plugin/assembly/1.1.2 http://maven.apache.org/xsd/assembly-1.1.2.xsd">
  [...]
  <baseDirectory>${artifactId}</baseDirectory>
  [...]
</assembly>

Advanced ModuleSet Topics

One of the most complex sections of the assembly descriptor is the moduleSets section. In fact, so many improvements have been made to this section that we feel it warrants its own "Advanced Topics" page.