What does it mean to extend a path with another path in ANT?
if I have the folloing:
<path>
<pathelement path="build/classes;lib/junit.jar" />
</path>
Does it mean that Ant breaks build/classes;lib/junit.jar into build/classes and lib/junit.jar
in other words, does ant think that we are defining two path elements?
Thank you so much in advance. Any help will be appreciated.
Your example:
<path>
<pathelement path="build/classes;lib/junit.jar" />
</path>
is equivalent to:
<path>
<pathelement path="build/classes" />
<pathelement path="lib/junit.jar" />
</path>
in both cases the path that created includes the contents of the build/classes directory plus the contents of the jarfile lib/junit.jar.
This mechanism allows you to dynamically create the required paths within your build file, rather than having to construct the combined string build/classes;lib/junit.jar.
For example, if you had the classes directory and the path to the junit jar specified in properties, you could than combine them into a single path:
<property name="classes.dir" location="build/classes" />
<property name="junit.jar" location="lib/junit.jar" />
...
<path>
<pathelement path="${classes.dir}" />
<pathelement path="${junit.jar}" />
</path>
This path would be the same as the examples above.
You could also refer to each of these items individually in other parts of your build file, and would only have to make changes in one place if either value needed to be changed.
The operation of the pathelement tag is documented here
You can specify PATH- and
CLASSPATH-type references using both
":" and ";" as separator characters.
Ant will convert the separator to the
correct character of the current
operating system.
Related
I have been googling for the "Differences between fileset and path" article for some time, but have found nothing useful.
For example, what is the difference between the following (say, there is a someDir directory, which contains .jar files and has no subdirectories):
<path id="somePathId">
<pathelement path="someDir"/>
</path>
<path id="someId">
<path refid="somePathId" />
</path>
and
<path id="someId">
<fileset dir="someDir">
<include name="*.*">
</fileset>
</path>
?
They are used in different situations.
fileset is used to specify a group of files. You can use selectors and patternsets to get only the files you want.
classpath is used to specify classpath references. classpath can be specified with a single jar (location="..."), a ; or : separated list of jars (path="...") or with nested resource collections (like fileset).
Also if you want to debug them, it is different:
<echo message="Build-path: ${toString:build-path}" />
vs
<property name="debug.classpath" refid="classpath"/>
<echo message="Classpath = ${debug.classpath}"/>
As for your scripts,
<path id="somePathId">
<pathelement location="someDir"/>
</path>
I did not test it but according to the documentation path= expects a ; or : separated list of jars. This is not the same as your second example.
The major difference between a <path> and a <fileset> is that in <fileset> you can specify if you want to include or exclude certain type of files (Basically, its a group of files within a path... not necessary all the files), for eg:
<path id="someId">
<fileset dir="someDir">
<include name="*.java">
<include name="*.properties">
</fileset>
</path>
How does ant behave if I define a path with a pathelement which points to a non-existent directory?
<path id="foo.bar">
<pathelement location="this/might/not/exist/">
</path>
The scenario is that the ant file is used for several projects - some have this additional folder, and some do not.
Does ant just ignore it, or does it fail?
It depends on the context.
When used as a classpath for the javac task, the missing directories are simply ignored:
This task will drop all entries that point to non-existent
files/directories from the classpath it passes to the compiler.
But if you use a path containing a non-existent directory, say as the source for a copy, you'll get an error.
For example, here directories 'one' and 'three' exist, but 'two' does not:
<path id="mypath">
<pathelement path="one" />
<pathelement path="two" />
<pathelement path="three" />
</path>
<copy todir="dest">
<path refid="mypath" />
</copy>
BUILD FAILED
/.../build.xml:14: Warning: Could not find resource file ".../two" to copy.
You could use a dirset to filter out the missing items perhaps:
<pathconvert property="dirs.list" pathsep="," refid="mypath">
<map from="${basedir}/" to="" />
</pathconvert>
<dirset id="exists.dirs" dir="." includes="${dirs.list}" />
<copy todir="dest">
<dirset refid="exists.dirs" />
</copy>
[copy] Copied 2 empty directories to 2 empty directories under /.../dest
In a project we have several source paths, so we defined a reference path for them:
<path id="de.his.path.srcpath">
<pathelement path="${de.his.dir.src.qis.java}"/>
<pathelement path="${de.his.dir.src.h1.java}"/>
...
</path>
Using the reference works fine in the <javac> tag:
<src refid="de.his.path.srcpath" />
In the next step, we have to copy non-java files to the classpath folder:
<copy todir="${de.his.dir.bin.classes}" overwrite="true">
<fileset refid="de.his.path.srcpath">
<exclude name="**/*.java" />
</fileset>
</copy>
Unfortunately, this does not work because "refid" and nested elements may not be mixed.
Is there a way I can get a set of all non-java files in my source path without copying the list of source paths into individual filesets?
Here's an option. First, use the pathconvert task to make a pattern suitable for generating a fileset:
<pathconvert pathsep="/**/*,"
refid="de.his.path.srcpath"
property="my_fileset_pattern">
<filtermapper>
<replacestring from="${basedir}/" to="" />
</filtermapper>
</pathconvert>
Next make the fileset from all the files in the paths, except the java sources. Note the trailing wildcard /**/* needed as pathconvert only does the wildcards within the list, not the one needed at the end:
<fileset dir="." id="my_fileset" includes="${my_fileset_pattern}/**/*" >
<exclude name="**/*.java" />
</fileset>
Then your copy task would be:
<copy todir="${de.his.dir.bin.classes}" overwrite="true" >
<fileset refid="my_fileset" />
</copy>
For portability, instead of hard-coding the unix wildcard /**/* you might consider using something like:
<property name="wildcard" value="${file.separator}**${file.separator}*" />
I'm writing an Ant script to package a project into a WAR file. The software consists of several projects with their own source directories, libraries, etc.
The WAR task has a nested element lib which I'm currently working on. I currently have a reference of the required libs as a Path (containing several FileSets, which I use in a classpath reference. The lib, however, wants the input to be a FileSet, and it refuses a Path.
I tried converting my Path into a FileSet, but then I didn't get it to work as a classpath elsewhere.
Is there a way to convert a Path into a FileSet? I would hate to copy-paste the directories.
<path id="compile.libs">
<fileset dir="${common.path}/lib" includes="*.jar"/>
<fileset dir="${data.path}/lib" includes="*.jar"/>
<fileset dir="${gui.path}/lib" includes="*.jar"/>
<fileset dir="${gui.path}/WebContent/WEB-INF/lib" includes="*.jar"/>
</path>
...when used with <war ..><../> <lib refid="compile.libs"/> </war> leads to:
BUILD FAILED
build.xml:173: compile.libs doesn't denote a zipfileset or a fileset
Assuming the paths are absolute, you can first convert the Path to a comma-delimited list using <pathconvert>, and then convert the list back into a Fileset:
<!-- create path -->
<path id="foo.path">
<pathelement location="/foo/bar/baz.txt"/>
<pathelement location="/qux/quux/quuux.txt"/>
</path>
<!-- convert foo.path to foo.list -->
<pathconvert
refid="foo.path"
property="foo.list"
pathsep=","
dirsep="/"
>
<!--
<fileset> will want relative paths, so we need to strip
the leading /. result: "foo/bar/baz.txt,qux/quux/quuux.txt"
-->
<map from="/" to=""/>
</pathconvert>
<!-- convert foo.list to fileset -->
<fileset id="foo.fileset" dir="/" includes="${foo.list}"/>
(Note the above assumes Unix; you may need to fiddle a bit with separators and whatnot if you're on Windows or you want to make it platform-independent.)
You may have several choices.
You may provide more than one
<lib> nested element to <war>
task. Maybe this would be enough.
You may preassemble all of your
lib files in one temporary
directory and then just reference that
directory as a fileset.
There is an ant-contrib
PathToFileSet task, but it
requires a central root directory,
and this may not be a case with your
compile.libs layout.
Since Ant 1.8.0 you can use a mappedresources. Source: Ant script: Prevent duplication of JAR in javac-classpath war-lib
I think I would try option 1.
I solved this by staging the libs like this :
<copy todir="stage/libs" flatten="true">
<path refid="classpath" />
</copy>
and then using a in the WAR task.simple.
The jars in the classpath used to compile are not the same that needs to be packaged inside the war. For example: I'm sure you need servlet-api.jar to compile your project but you don't need it inside the war because the container provides it. And some jars aren't needed at compile time but at runtime.
I know I'm not answering your question, just want you to think what you are doing.
When declaring external ant tasks using taskdef, for instance ant-contrib, the proposed setup is to use the followin taskdef:
<taskdef resource="net/sf/antcontrib/antcontrib.properties">
<classpath>
<pathelement location="lib/ant-contrib/ant-contrib-1.0b3.jar"/>
</classpath>
</taskdef>
This works when antcontrib.properties is located in net/sf/antcontrib relative to the build.xml file.
But when I put it in lib/net/sf/antcontrib and changes the taskdef into
<taskdef resource="lib/net/sf/antcontrib/antcontrib.properties">
<classpath>
<pathelement location="lib/ant-contrib/ant-contrib-1.0b3.jar"/>
</classpath>
</taskdef>
Ant is not able to find the properties file, it gives the error
[taskdef] Could not load definitions from resource
lib/net/sf/antcontrib/antcontrib.properties. It could not be found.
It seems like ant treats the lib directory separately and fails to load a taskdef resource from there.
As Alex said, you shouldn't need to unzip the jar. The <taskdef> can load antcontrib.properties directly out of the jar.
The error you got is because you changed the resource path, but the path to the file inside the compressed jar/zip is still the same. The taskdef isn't paying attention to the properties file you moved because the <classpath> you provided to <taskdef> tells it to only look in the jar.
Use antlib.xml resource:
Here is the taskdef definition that I use:
<property name="ant-contrib.jar" location="..."/>
<taskdef
resource="net/sf/antcontrib/antlib.xml"
uri="http://ant-contrib.sourceforge.net"
>
<classpath>
<pathelement location="${ant-contrib.jar}"/>
</classpath>
</taskdef>
You do not need to extract anything from the jar file. Also, uri attribute is optional if you do not want to use namespaces with antcontrib tasks.
To handle classpath for tasks definitions, I use a classpath ref in Ant, it's way easier. You can link either a directory containing classes, either a directory containing many .jar, either (of course) a single .jar.
For example :
<!-- Properties -->
<property name="lib" value="lib/" />
<property name="classes" value="bin/" />
<!-- Classpath definition -->
<path id="runtime-classpath" >
<pathelement location="${bin}" />
<fileset dir="${lib}">
<include name="*.jar"/>
</fileset>
</path>
<!-- Taskdefs definitions -->
<taskdef name="myTask" classname="org.stackoverflow.tasks.MyTask" classpathref="runtime-classpath" />
<!-- Tasks -->
<target name="test" description="Test Action">
<myTask parameter1="value" />
</target>