带有Servlet3.0的嵌入式Tomcat --如何在扫描时跳过某些jars？

Question

下面是一个用于扫描Servlet3.0初始化器的嵌入式Tomcat服务器的简单方法：

public static void main(String[] args) throws ServletException, LifecycleException {
    Tomcat tomcat = new Tomcat();
    tomcat.setPort(8080);

    StandardContext ctx = (StandardContext) tomcat.addWebapp("/", new File("web").getAbsolutePath());

    //declare an alternate location for your "WEB-INF/classes" dir:     
    File additionWebInfClasses = new File("target/classes");
    VirtualDirContext resources = new VirtualDirContext();
    resources.setExtraResourcePaths("/WEB-INF/classes=" + additionWebInfClasses);
    ctx.setResources(resources);

    tomcat.start();
    System.out.println("Started");
    tomcat.getServer().await();
}

我知道属性tomcat.util.scan.DefaultJarScanner.jarsToSkip在catalina.properties中允许您限制扫描哪些jars以加快部署时间。我将如何将同样的想法整合到这里嵌入的Tomcat代码中？

奖励:是否有一种方法可以指定要包含哪个jars而不是跳过哪个jars？

Answer 1

虽然我没有加载单个catalina.properties属性的特定方法，但我发现在Tomcat容器上下文中重载JarScanner本身已经足够方便了，如下所示：

答案在Groovy中(对不起，我很懒)

public static void main(String[] args) throws ServletException, LifecycleException {
    Tomcat tomcat = new TomcatWithFastJarScanner()
    tomcat.setPort(8080)

    StandardContext ctx = (StandardContext) tomcat.addWebapp("/", new File("web").getAbsolutePath())

    //declare an alternate location for your "WEB-INF/classes" dir:     
    VirtualDirContext resources = new VirtualDirContext()
    resources.setExtraResourcePaths("/WEB-INF/classes=" + new File("target/classes"))
    ctx.setResources(resources)

    tomcat.start()
    println "Started server on port 8080"
    tomcat.getServer().await()
}

private static class TomcatWithFastJarScanner extends Tomcat {
    @Override
    public void start() throws LifecycleException {
        getServer().findServices().each { service ->
            service.getContainer().findChildren().each { container ->
                container.findChildren().each { c ->
                    ((Context) c).setJarScanner(new FastJarScanner())
                }
            }
        }
        super.start()
    }
}


private static class FastJarScanner extends StandardJarScanner {
    def jarsToInclude = [ 'spring-web.*' ]

    @Override
    public void scan(ServletContext context, ClassLoader classloader,
            JarScannerCallback callback, Set<String> jarsToSkip) {
        jarsToSkip = new HashSet<String>();

        ((URLClassLoader) classloader.getParent()).getURLs().each {
            def jar = it.path.find(/[^\/]+\.jar$/)
            if(!jar) return
            for(String inclusionPattern : jarsToInclude) {
                if(jar.find(inclusionPattern))  
                    println "including jar: " + jar
                else jarsToSkip.add(jar)
            }
        }

        super.scan(context, classloader, callback, jarsToSkip);
    }
}

基本思想是，我们正在查看类加载器可以看到的所有jars，而不包括我们不想包含的所有jars。

Tomcat嵌入式启动非常快，就像这样！

Answer 2

以下是推测，只读规格。我们可以使用一个web片段，它有自己的metadata/web.xml和元数据-完全=真正的信号，没有注释扫描。

Answer 3

您可以将Tomcat设置scanClassPath扩展为false，类似于-这是user jkschneider here所指出的)：

new Tomcat() {
    @Override
    public void start() throws LifecycleException {
        for (final org.apache.catalina.Service service : getServer().findServices()) {
            for (final Container container : service.getContainer().findChildren()) {
                for (final Container subContainer : container.findChildren()) {
                    final StandardJarScanner jarScanner = (StandardJarScanner) ((Context) subContainer).getJarScanner();
                    jarScanner.setScanClassPath(false);
                }
            }
        }
        super.start();
    }
};

顺便说一句，更好的办法可能是：

final Tomcat tomcat = new Tomcat();
final StandardHost tomcatHost = (StandardHost) tomcat.getHost();
        tomcatHost.setContextClass("com.fullclassname....FastStandardContext");

FastStandardContext在哪里

public class FastStandardContext extends StandardContext {
    public FastStandardContext() {
        ((StandardJarScanner) getJarScanner()).setScanClassPath(false);
    }
}

这将涵盖在tomcat运行后添加新的new应用程序的情况。