Pants for Maven Experts

If you're used to Maven and learning Pants, you're part of a growing crowd. Here are some things that helped other folks come up to speed.

The good news is that Pants and Maven are pretty similar. Both tools use several configuration-snippet files near in source code directories to specify how to build those directories' source code. Both tools use the configuration-snippet files to build up a model of your source code, then execute tasks in a lifecycle over that model. Pants targets tend to be finer-grained than Maven's projects; but if you use subprojects in Maven, Pants targets might feel familiar. (If you're converting Maven pom.xmls to BUILD files, some scripts written by others who have done the same can give you a head start.) Both Maven and Pants expect code to be laid out in directories in a consistent way. If you're used to Maven's commands, many of Pants' goals will feel familiar.

Pants uses Ivy to manage artifact fetching and publishing; Ivy's behavior here is pretty similar to Maven.

Three Pants features that especially confuse Maven experts as they move to pants are:

  • Pants has a first-class mechanism for targets depending on other targets on the local file system
  • Pants targets do not specify version numbers; versions are only determined during release

These points are a significant departure from Maven's handling of inter-project dependencies.

Folks switching a Maven-built codebase to Pants often encounter another source of confusion: they uncover lurking jar-dependency version conflicts. JVM projects can inadvertently end up relying on classpath order for correctness; any two build tools will order their classpaths differently. If your project depends on two versions of the same jar (all too easy to do with transitive dependencies), then your Maven build chose one version, but Pants might end up choosing another: Pants is likely to generate a differently-ordered CLASSPATH than Maven did. You can fix these, making your build configuration more robust along the way; see JVM 3rdparty Pattern for advice.

Pants Equivalents


Run a binary
Maven: exec:java
Pants: run

Run a binary in the debugger
Maven: -Xdebug
Pants: run.jvm --jvm-debug

Run one test in the debugger
Maven: -Dmaven.surefire.debug=true test
Pants: test.junit --jvm-debug

Build a binary package
Maven: package
Pants: binary

Look at dependent projects or artifacts
Maven: dependency:analyze
Pants: depmap
Pants: resolve.ivy --open


Shade with an AppendingTransformer


                    <transformer implementation="org.apache.maven.plugins.shade.resource.AppendingTransformer">


           main = '',
              Duplicate('^reference\.conf', Duplicate.CONCAT_TEXT),

              # We need to add this as it is overridden by adding the reference.conf one above.
              Duplicate('^META-INF/services/', Duplicate.CONCAT_TEXT)

Dependency Scopes

Pants supports dependency scopes. To emulate Maven's provided scope you can specify both the compile and test scopes for a target. For details, see JVM Projects with Pants.




    scala_jar(org='org.apache.spark', name='spark-core', rev='2.1.0'),
  scope='compile test',
Generated by publish_docs from dist/markdown/html/examples/src/java/org/pantsbuild/example/from_maven.html 2022-12-03T01:08:59.499723