I have a usecase where in I had to create a new module in out project. Our main project has multiple modules and each module is a java project. We are using ivy for dependency resolution. Now the probkem is that in the new module , I had to use java 1.7 API (WatchService) which is not there in java 1.6. Now in build.xml I can check for the java version and accordingly build this new module depending on the java -version. The problem comes in ivy.xml of or main web project where I have to mention the jar file of the new module as a dependency to include in the generated war file. If the java version is 1.7 , then in that case problem wont be there as the jar will be build and the its dependency will be resolved. the problem arises when the java version is 1.6. The jar file wont be created and when its time to generate the war file, ivy wont be able to resolve the dependency as the jar file is not there. Maybe the approach that I am trying to apply here is not fine. Please advice me on how to work around this particular use case.
rampal
In ivy you can use configurations to maintain different sets of depenedencies:
<ivy-module version="2.0">
<info organisation="com.myspotontheweb" module="demo"/>
<configurations>
<conf name="compile_jdk7" description="Java JDK7 compile dependencies"/>
<conf name="compile_jdk6" description="Java JDK6 compile dependencies"/>
</configurations>
<dependencies>
<!-- JDK7 dependencies -->
<dependency org="org.myorg" name="module1" rev="latest.integration" conf="compile_jdk7->default"/>
<dependency org="org.myorg" name="module2" rev="latest.integration" conf="compile_jdk7->default"/>
<dependency org="org.myorg" name="module3" rev="latest.integration" conf="compile_jdk7->default"/>
<!-- JDK6 dependencies -->
<dependency org="org.myorg" name="module1" rev="latest.integration" conf="compile_jdk6->default"/>
<dependency org="org.myorg" name="module3" rev="latest.integration" conf="compile_jdk6->default"/>
</dependencies>
</ivy-module>
and in the build file use a condition task to choose which configuration is used at run-time to populate the classpath, using the cachepath task:
<project name="demo" default="compile" xmlns:ivy="antlib:org.apache.ivy.ant">
<condition property="compile.config" value="compile_jdk7">
<equals arg1="${ant.java.version}" arg2="1.7"/>
</condition>
<condition property="compile.config" value="compile_jdk6">
<equals arg1="${ant.java.version}" arg2="1.6"/>
</condition>
<target name="resolve" description="Use ivy to resolve classpaths">
<ivy:cachepath pathid="compile.path" conf="${compile.config}"/>
</target>
<target name="compile" depends="resolve" description="Compile code">
<javac ...... classpathref="compile.path"/>
</target>
</project>
I have two versions of the same jar (3.2 and 2.2.1) I need to use both of them but ivy evicts older revision. How to configure ivy to take two versions?
<dependency org="asm" name="asm-all" rev="3.2">
<artifact name="asm-all" type="jar"/>
</dependency>
<dependency org="asm" name="asm-all" rev="2.2.1">
<artifact name="asm-all" type="jar"/>
</dependency>
You need to use ivy configurations. This is a very flexible mechanism to manage arbitrary groups of dependencies.
The example below places each version of the jar onto a separate configuration. This can be used later to create two classpaths, using the ivy cachepath task.
Example
ivy.xml
<ivy-module version="2.0">
<info organisation="com.myspotontheweb" module="demo"/>
<configurations>
<conf name="compile1" description="Required to compile application1"/>
<conf name="compile2" description="Required to compile application2"/>
</configurations>
<dependencies>
<!-- compile1 dependencies -->
<dependency org="asm" name="asm-all" rev="3.2" conf="compile1->master"/>
<!-- compile2 dependencies -->
<dependency org="asm" name="asm-all" rev="2.2.3" conf="compile2->master"/>
</dependencies>
</ivy-module>
Notes:
Version 2.2.1 does not exist in Maven Central
Note the configuration mapping "??? -> master". In Maven the remote master configuration mapping resolves to the main module artifact without dependencies. (See)
build.xml
<project name="demo" default="init" xmlns:ivy="antlib:org.apache.ivy.ant">
<target name="init" description="Use ivy to resolve classpaths">
<ivy:resolve/>
<ivy:report todir='build/ivy' graph='false' xml='false'/>
<ivy:cachepath pathid="compile1.path" conf="compile1"/>
<ivy:cachepath pathid="compile2.path" conf="compile2"/>
</target>
<target name="clean" description="Clean built artifacts">
<delete dir="build"/>
</target>
<target name="clean-all" depends="clean" description="Additionally purge ivy cache">
<ivy:cleancache/>
</target>
</project>
Notes:
Always a good idea to generate an ivy report. It will tell you which dependencies exist on which ivy configuration.
This example shows ivy managing ANT paths. You can also use ivy configurations with the ivy retrieve task to populate a local "lib" directory when assembling something like a webapp WAR file.
Consider an ivy.xml like the following:
<ivy-module version="2.0">
<info organisation="com.foo" module="FooBar" />
<dependencies>
<dependency org="net.sf.ehcache" name="ehcache-core" rev="2.2.0" />
<!--...-->
</dependencies>
</info>
</ivy-module>
When I run Ivy, it fetches all dependencies for EHCache, even testing dependencies. Specifically, it tries to pull in Hibernate 3.5.1 (which, in the POM file, is listed as a "test" dependency).
How do I prevent Ivy from including test dependencies? I could list it as an excluded dependency, but I don't want to have to do this for every test dependency. I'm new to Ivy and used to the way Maven does things. I was reading about configurations but I don't understand how this aspect of Maven's "scope" maps to "configurations."
You need to define the configuration of the dependency like:
<dependency org="net.sf.ehcache" name="ehcache-core" rev="2.2.0" conf="compile"/>
If you omit conf it is assumed, that you meant conf ="*", which will download all configurations for that dependency.
Here is a simple Example:
<configurations>
<conf name="test" visibility="public" />
<conf name="compile" visibility="public" />
</configurations>
<publications>
<artifact name="${project.name}" type="jar" conf="compile" ext="jar"/>
<artifact name="${project.name}-test" type="jar" conf="test" ext="jar"/>
</publications>
<dependencies>
<!-- COMPILE -->
<dependency org="log4j" name="log4j" rev="1.2.14" conf="compile->*"/>
<dependency org="apache" name="commons-net" rev="2.0" conf="compile->*"/>
<dependency org="itext" name="itext" rev="1.4.6" conf="compile->*"/>
<dependency org="jsch" name="jsch" rev="0.1.29" conf="test->*"/>
<!-- TEST -->
</dependencies>
In this example jsch will be included in the test and the compile configuration.
If you resolve this dependency later with conf ="compile" you will get all dependencies EXCEPT jsch.
If you resolve this dependency with conf ="test" you will get jsch only.
And if test would extend compile, you would get all jars.
<configurations>
<conf name="test" visibility="public" extends="compile" />
<conf name="compile" visibility="public" />
</configurations>
In my web application, there are two separate lib directories:
/lib, and
/web/webroot/WEB-INF/lib.
The idea behind it is that libraries in the latter one are used by front-end code only, and the first one by both the front-end and the business logic code. There is a class loader in place which lets the business logic code not see the jars in /web/webroot/WEB-INF/lib.
How can I tell ivy that certain dependencies should go to the second directory while all others go to first one?
It's not trival since the the web class loader can see jars in both directories and I don't want jars to be in both directories.
Configurations are used to create logical groupings of dependencies:
ivy.xml
<ivy-module version="2.0">
<info organisation="com.myspotontheweb" module="demo"/>
<configurations>
<conf name="frontEnd" description="Jars used by front end"/>
<conf name="businessLogic" description="Jars used for business logic"/>
</configurations>
<dependencies>
<dependency org="commons-lang" name="commons-lang" rev="2.5" conf="businessLogic->default"/>
<dependency org="commons-codec" name="commons-codec" rev="1.4" conf="businessLogic->default"/>
<dependency org="commons-cli" name="commons-cli" rev="1.2" conf="frontEnd->default"/>
<dependency org="commons-logging" name="commons-logging" rev="1.1.1" conf="frontEnd->default"/>
</dependencies>
</ivy-module>
The ivy retrieve ant task can use these configurations to populate your directories:
build.xml
<target name="init" description="--> retrieve dependencies with ivy">
<ivy:retrieve conf="businessLogic" pattern="lib/[artifact].[ext]"/>
<ivy:retrieve conf="frontEnd" pattern="web/webroot/WEB-INF/lib/[artifact].[ext]"/>
</target>
Example
$ find . -type f
./build.xml
./ivy.xml
./lib/commons-lang.jar
./lib/commons-codec.jar
./web/webroot/WEB-INF/lib/commons-cli.jar
./web/webroot/WEB-INF/lib/commons-logging.jar
Are there ant plugins that wrap maven so that I can make use of its dependency management features to download jars for me and place them in my ant build's lib folder?
My specific problem is that I'm using the Crap4j plugin for Hudson, but it doesn't, as of yet, support Maven. Since it's a small project, maven is overkill, but I don't want to go without mvn dependency:copy-dependcies if I don't have to.
Any suggestions? (other than suck it up)
There is a new set of Ant tasks that use Mercury. Mercury is the refactored code that will be the basis of way that Maven 3 interacts with Maven (and OSGi) repositories that is being implemented by Oleg Gusakov. Mercury is well tested, and you can start using it in Ant projects today. Take a look at some of the How-to documents that Oleg has written:
http://people.apache.org/~ogusakov/sites/mercury-ant/mercury-ant-tasks/howto.html
Here's a simple example of using Mercury in an Ant build.xml file. The following build file creates a classpath that depends on verion 3.0 of the asm artifact:
<javac srcdir="src/main/java"
destdir="target/classes">
<classpath>
<deps>
<dependency name="asm:asm:3.0"/>
</deps>
</classpath>
</javac>
There are a lot of advanced features such as support for verifying PGP signatures or MD5 digests. You can also start to define different repositories that Mercury depends on. This XML allows you to define a reference to a repository such as Nexus in addition to using a local directory as a repository:
<repo id="myCentral"
url="http://localhost:8081/nexus/contengs/groups/public"/>
<repository dir="/my/local/repo"/>
<javac srcdir="src/main/java"
destdir="target/classes">
<classpath>
<deps>
<dependency name="asm:asm:3.0"/>
</deps>
</classpath>
</javac>
If you need to reference a repository that requires authentication Mercury has support for storing a username and password:
<repo id="myCentral"
url="http://localhost:8081/nexus/contengs/groups/public">
<auth name="foo" pass="bar"/>
</repo>
<javac srcdir="src/main/java"
destdir="target/classes">
<classpath>
<deps>
<dependency name="asm:asm:3.0"/>
</deps>
</classpath>
</javac>
Most compelling is the ability to publish an artifact to a repository from an Ant build file. If you work in an organization of any scale, you'll want to start thinking about deploying artifacts to a repository manager like Nexus. With Mercury, you can start deploying artifacts to a repository manager without having to adopt Maven. Here's a build file that defines an authenticated repository and writes an artifact:
<repo id="myCentral"
url="http://localhost:8081/nexus/contengs/groups/public">
<auth name="foo" pass="bar"/>
</repo>
<write repoid="myCentral"
name="t:t:1.0"
file="${basedir}/target/t.jar"/>
Mercury is ready to use, and you can expect a lot of developments from Oleg going forward. If you want to start using it, the best place to look is at Oleg's How-to Page. (Note: This information will soon be integrated into the Definitive Guide)
Whilst the mercury tasks work, I haven't used them. I have had good success with their predecessors, the maven-ant-tasks. They're fairly simple to get going, if you already have a POM handy.
<project name="blah" xmlns:artifact="antlib:org.apache.maven.artifact.ant">
<!-- If you drop the maven-ant-tasks in ~/.ant/lib, you don't need these two bits. -->
<taskdef uri="antlib:org.apache.maven.artifact.ant"
resource="org/apache/maven/artifact/ant/antlib.xml"
classpathref="ant.classpath" />
<path id="ant.classpath">
<fileset dir="${ant.tasks.dir}">
<include name="*.jar" />
</fileset>
</path>
<target name="resolve" description="--> retrieve dependencies with maven">
<!-- Resolve dependencies -->
<artifact:dependencies filesetId="dependency.fileset">
<pom file="pom.xml" />
</artifact:dependencies>
<!-- Copy all dependencies to the correct location. -->
<copy todir="${web.dir}/WEB-INF/lib">
<fileset refid="dependency.fileset" />
<!-- This mapper strips off all leading directory information -->
<mapper type="flatten" />
</copy>
</target>
</project>
I like to keep my ant task jars inside the project, so I've added the taskdef and path. But if you want to put maven-ant-tasks-2.0.9.jar in ~/.ant/lib, then you don't need to declare this stuff. I think.
If you think that Maven is overkill in your project, you could/should try Apache Ivy: It's a very powerful dependency management library similar to the Maven one.
If you're hosting a project on the web, take also a look at Ivy Roundup, it's a repository of Ivy definitions for various libraries.
Just use the Maven Ant Tasks. They can be downloaded at the normal maven download page.
Refer this: Why you should use the Maven Ant Tasks instead of Maven or Ivy
I wouldn't really recommend Ivy for reasons given in the link above.
It is very simple to run Maven goal from Ant
<target name="buildProject" description="Builds the individual project">
<exec dir="${source.dir}\${projectName}" executable="cmd">
<arg value="/c"/>
<arg value="${env.MAVEN_HOME}\bin\mvn.bat"/>
<arg line="clean install" />
</exec>
</target>
Using this you are allow to run any kind of Maven goal from Ant...
Enjoy....
In my case i just want an ejb jar to be at the repository so i could use it on another project with maven as dependency so:
<target name="runMaven" depends="deploy" description="LLama al maven.">
<exec executable="cmd">
<arg value="/c"/>
<arg value="mvn.bat install:install-file -DgroupId=com.advance.fisa.prototipo.camel -DartifactId=batch-process -Dversion=1.0 -Dpackaging=jar -Dfile=${jarDirectory}\batch-process.jar"/>
</exec>
</target>
Download Maven Ant Tasks then use this:
<target name="getDependencies">
<path id="maven-ant-tasks.classpath" path="${basedir}${file.separator}maven${file.separator}lib${file.separator}maven-ant-tasks.jar" />
<typedef resource="org/apache/maven/artifact/ant/antlib.xml" uri="antlib:org.apache.maven.artifact.ant" classpathref="maven-ant-tasks.classpath" />
<artifact:dependencies filesetId="dependency.fileset" type="jar">
<pom file="pom.xml" />
</artifact:dependencies>
<!--TODO take care of existing duplicates in the case of changed/upgraded dependencies-->
<copy todir="lib">
<fileset refid="dependency.fileset" />
<mapper type="flatten" from="${dependency.versions}" />
</copy>
</target>
I am working on the same problem right now. I installed all necessary libs in my local Maven repo and from there I put it into our Company Maven Repo.
It is not working quite right yet. Some of the tests fail that work nicely in my Maven test run, but since the outcome of the test is not important for the coverage data, I am quite satisfied.
Here is my Maven snippet. I hope that helps.
<plugin>
<groupId>org.apache.maven.plugins</groupId>
<artifactId>maven-antrun-plugin</artifactId>
<version>1.1</version>
<inherited>false</inherited>
<executions>
<execution>
<phase>site</phase>
<configuration>
<tasks>
<property name="compile_classpath" refid="maven.compile.classpath"/>
<property name="runtime_classpath" refid="maven.runtime.classpath"/>
<property name="test_classpath" refid="maven.test.classpath"/>
<property name="plugin_classpath" refid="maven.plugin.classpath"/>
<property name="CRAP4J_HOME" value="${user.home}/Projects/crap4j"/>
<taskdef name="crap4j" classname="org.crap4j.anttask.Crap4jAntTask">
<classpath>
<fileset dir="${CRAP4J_HOME}/lib">
<include name="**/*.jar"/>
</fileset>
</classpath>
</taskdef>
<crap4j projectdir="${project.basedir}/alm-jar-server"
outputDir="${project.basedir}/crap4jReports"
dontTest="false"
debug="true">
<classes>
<pathElement location="${project.basedir}/target/classes"/>
</classes>
<srces>
<pathElement location="${project.basedir}/src/main/java"/>
</srces>
<testClasses>
<pathElement location="${project.basedir}/target/test-classes"/>
</testClasses>
<libClasspath>
<fileset dir="${user.home}/.m2/repository">
<include name="**/*.jar"/>
</fileset>
</libClasspath>
</crap4j>
</tasks>
</configuration>
<goals>
<goal>run</goal>
</goals>
</execution>
</executions>
<dependencies>
<dependency>
<groupId>org.crap4j</groupId>
<artifactId>crap4j</artifactId>
<version>1.1.6</version>
</dependency>
<dependency>
<groupId>org.kohsuke</groupId>
<artifactId>args4j</artifactId>
<version>2.0.1</version>
</dependency>
<dependency>
<groupId>com.agitar</groupId>
<artifactId>asmlib</artifactId>
<version>1.0</version>
</dependency>
<dependency>
<groupId>commons-httpclient</groupId>
<artifactId>commons-httpclient</artifactId>
<version>3.0</version>
</dependency>
<dependency>
<groupId>com.agitar</groupId>
<artifactId>coverage</artifactId>
<version>1.0</version>
</dependency>
</dependencies>
</plugin>