[深入Maven源代码]maven绑定命令行参数到具体插件

maven的插件

我们知道Maven具体构建动作都是由插件执行的，maven本身只是提供一个框架，这样就提供了高度可定制化的功能，我们用maven命令执行比如mvn clean package这样的命令时maven会将package这个阶段（phase）绑定到相应的生命周期（lifecycle），再寻找项目（project）里配置的plugin，执行具体的plugin完成持续构建

maven绑定插件（plugin）

maven在读取命令行之后会根据命令行参数是系统默认的phase还是其他的自定义插件(goal)来解析成task参数，继而根据这些task参数来生成插件执行列表。

for ( Object task : tasks ) { if ( task instanceof GoalTask ) { String pluginGoal = ( (GoalTask) task ).pluginGoal;

                MojoDescriptor mojoDescriptor = mojoDescriptorCreator.getMojoDescriptor( pluginGoal, session, project );

                MojoExecution mojoExecution =
                    new MojoExecution( mojoDescriptor, "default-cli", MojoExecution.Source.CLI );

                mojoExecutions.add( mojoExecution );
            }
            else if ( task instanceof LifecycleTask )
            {
                String lifecyclePhase = ( (LifecycleTask) task ).getLifecyclePhase();

                Map<String, List<MojoExecution>> phaseToMojoMapping =
                    calculateLifecycleMappings( session, project, lifecyclePhase );

                for ( List<MojoExecution> mojoExecutionsFromLifecycle : phaseToMojoMapping.values() )
                {
                    mojoExecutions.addAll( mojoExecutionsFromLifecycle );
                }
            }
            else
            {
                throw new IllegalStateException( "unexpected task " + task );
            }
        }

　　

代码里先判断task的类型，如果是GoalTask就说明参数是clean:clean这一种带goal的格式，这种情况可能是maven内置插件也可能是开发者自己开发的插件，这种情况下maven就会先根据插件的goal去找到具体的插件，找的方法是先从项目定义的插件里找，找不到的话再去仓库里找，这里把核心部分代码贴出来:

   PluginDescriptor pluginDescriptor =
                    pluginManager.loadPlugin( plugin, request.getRepositories(), request.getRepositorySession() );

          if ( request.getPrefix().equals( pluginDescriptor.getGoalPrefix() ) )
                {
                    return new DefaultPluginPrefixResult( plugin );
                }

这个这部分比较简单，就是遍历项目里的插件一个个加载，然后比较这个插件的goal前缀是否和命令行的请求相同，如果相同的话直接封装下该插件返回。其他用groupId:artifactId:version:goal格式的命令行解析也是差不多的方法，根据这些信息加载响应的plugin插件。接下来要重点说的是比较复杂的情况，即clean package这种比较内置的phase，下面代码一一解读： 先根据lifecyclephase来决定是哪个lifecycle，比如package这个phase就是在default生命周期里，maven内置定义了clean、default、site三个生命周期，maven采用plexus作为IOC容器，这个defaultLifeCycles的依赖是在maven-core的component.xml中定义的，与此同时定义了各个生命周期里的phase，这个读者感兴趣可以去看相应的代码，此处略去不表。回到这里的代码，maven根据phase去默认找生命周期，这里通过package找到了default生命周期。

/*
         * Determine the lifecycle that corresponds to the given phase.
         */

        Lifecycle lifecycle = defaultLifeCycles.get( lifecyclePhase );

        if ( lifecycle == null )
        {
            throw new LifecyclePhaseNotFoundException(
                "Unknown lifecycle phase "" + lifecyclePhase + "". You must specify a valid lifecycle phase" +
                    " or a goal in the format <plugin-prefix>:<goal> or" +
                    " <plugin-group-id>:<plugin-artifact-id>[:<plugin-version>]:<goal>. Available lifecycle phases are: " +
                    defaultLifeCycles.getLifecyclePhaseList() + ".", lifecyclePhase );
        }

接下来遍历default lifecycle下所有的phase直到package这个phase就退出遍历循环，这里体现了maven的一个特性，就是如果你指定了某个生命周期中某个phase，那这个phase之前的phase都会被执行，这里主要是初始化mappings包含哪些phase，每个phase的插件列表之所以是TreeMap是因为后面要根据优先级，也就是Key来排序遍历确定插件执行顺序，每个phase具体要执行的插件到下一段代码再写入。

    /*
         * Initialize mapping from lifecycle phase to bound mojos. The key set of this map denotes the phases the caller
         * is interested in, i.e. all phases up to and including the specified phase.
         */

        Map<String, Map<Integer, List<MojoExecution>>> mappings =
            new LinkedHashMap<String, Map<Integer, List<MojoExecution>>>();

        for ( String phase : lifecycle.getPhases() )
        {
            Map<Integer, List<MojoExecution>> phaseBindings = new TreeMap<Integer, List<MojoExecution>>();

            mappings.put( phase, phaseBindings );

            if ( phase.equals( lifecyclePhase ) )
            {
                break;
            }
        }

接下来遍历本项目中所有插件，每个插件在遍历所有执行配置，如果execution配置里已经指定了phase，则将这个execution下所有goal对应的Mojo加到对应phase的执行map里，如果execution配置里没有指定phase的话，那就要去遍历这个execution下所有goal，依次获取该goal的Mojo描述信息，根据每个Mojo绑定的phase来将该Mojo加到对应phase的执行map里。

        /*
         * Grab plugin executions that are bound to the selected lifecycle phases from project. The effective model of
         * the project already contains the plugin executions induced by the project's packaging type. Remember, all
         * phases of interest and only those are in the lifecyle mapping, if a phase has no value in the map, we are not
         * interested in any of the executions bound to it.
         */

        for ( Plugin plugin : project.getBuild().getPlugins() )
        {
            for ( PluginExecution execution : plugin.getExecutions() )
            {
                // if the phase is specified then I don't have to go fetch the plugin yet and pull it down
                // to examine the phase it is associated to.
                if ( execution.getPhase() != null )
                {
                    Map<Integer, List<MojoExecution>> phaseBindings = mappings.get( execution.getPhase() );
                    if ( phaseBindings != null )
                    {
                        for ( String goal : execution.getGoals() )
                        {
                            MojoExecution mojoExecution = new MojoExecution( plugin, goal, execution.getId() );
                            mojoExecution.setLifecyclePhase( execution.getPhase() );
                            addMojoExecution( phaseBindings, mojoExecution, execution.getPriority() );
                        }
                    }
                }
                // if not then i need to grab the mojo descriptor and look at the phase that is specified
                else
                {
                    for ( String goal : execution.getGoals() )
                    {
                        MojoDescriptor mojoDescriptor =
                            pluginManager.getMojoDescriptor( plugin, goal, project.getRemotePluginRepositories(),
                                                             session.getRepositorySession() );

                        Map<Integer, List<MojoExecution>> phaseBindings = mappings.get( mojoDescriptor.getPhase() );
                        if ( phaseBindings != null )
                        {
                            MojoExecution mojoExecution = new MojoExecution( mojoDescriptor, execution.getId() );
                            mojoExecution.setLifecyclePhase( mojoDescriptor.getPhase() );
                            addMojoExecution( phaseBindings, mojoExecution, execution.getPriority() );
                        }
                    }
                }
            }
        }

　　

经过前面几个步骤之后，已经拿到了所有phase对应的Mojo执行列表，接下来需要将所有phase的Mojo串起来到一个总的列表里，这里注意mappings是一个LinkedHashMap，所以遍历的时候是有顺序的，而每个phase的execution map是TreeMap，根据优先级排序，这样最后总体的顺序是先按照总体的phase顺序，再按照phase内的优先级进行排序。

 Map<String, List<MojoExecution>> lifecycleMappings = new LinkedHashMap<String, List<MojoExecution>>();

        for ( Map.Entry<String, Map<Integer, List<MojoExecution>>> entry : mappings.entrySet() )
        {
            List<MojoExecution> mojoExecutions = new ArrayList<MojoExecution>();

            for ( List<MojoExecution> executions : entry.getValue().values() )
            {
                mojoExecutions.addAll( executions );
            }

            lifecycleMappings.put( entry.getKey(), mojoExecutions );
        }