Ant's buildfiles are written in XML. Each buildfile contains one project
and at least one (default) target. Targets contain task elements.
Each task element of the buildfile can have an id
attribute and
can later be referred to by the value supplied to this. The value has
to be unique. (For additional information, see the
Tasks section below.)
A project has three attributes:
Attribute | Description | Required |
name | the name of the project. | No |
default | the default target to use when no target is supplied. | No; however, since Ant 1.6.0, every project includes an implicit target that contains any and all top-level tasks and/or types. This target will always be executed as part of the project's initialization, even when Ant is run with the -projecthelp option. |
basedir | the base directory from which all path calculations are done. This attribute might be overridden by setting the "basedir" property beforehand. When this is done, it must be omitted in the project tag. If neither the attribute nor the property have been set, the parent directory of the buildfile will be used. | No |
Optionally, a description for the project can be provided as a
top-level <description>
element (see the description type).
Each project defines one or more targets. A target is a set of tasks you want to be executed. When starting Ant, you can select which target(s) you want to have executed. When no target is given, the project's default is used.
A target can depend on other targets. You might have a target for compiling, for example, and a target for creating a distributable. You can only build a distributable when you have compiled first, so the distribute target depends on the compile target. Ant resolves these dependencies.
It should be noted, however, that Ant's depends
attribute
only specifies the order in which targets should be executed - it
does not affect whether the target that specifies the dependency(s) gets
executed if the dependent target(s) did not (need to) run.
Ant tries to execute the targets in the depends
attribute in the order
they appear (from left to right). Keep in mind that it is possible that a target
can get executed earlier when an earlier target depends on it:
<target name="A"/> <target name="B" depends="A"/> <target name="C" depends="B"/> <target name="D" depends="C,B,A"/>
Suppose we want to execute target D. From its
depends
attribute, you
might think that first target C, then B and then A is executed.
Wrong! C depends on B, and B depends on A, so first A is executed, then B, then C, and finally D.
In a chain of dependencies stretching back from a given target such as D above, each target gets executed only once, even when more than one target depends on it. Thus, executing the D target will first result in C being called, which in turn will first call B, which in turn will first call A. After A, then B, then C have executed, execution returns to the dependency list of D, which will not call B and A, since they were already called in process of dependency resolution for C and B respectively as dependencies of D. Had no such dependencies been discovered in processing C and B, B and A would have been executed after C in processing D's dependency list.
A target also has the ability to perform its execution if (or
unless) a property has been set. This allows, for example, better
control on the building process depending on the state of the system
(java version, OS, command-line property defines, etc.). To make a target
sense this property, you should add the if
(or
unless
) attribute with the name of the property that the target
should react to. Note: Ant will only check whether
the property has been set, the value doesn't matter. A property set
to the empty string is still an existing property. For example:
<target name="build-module-A" if="module-A-present"/><target name="build-own-fake-module-A" unless="module-A-present"/>
In the first example, if the module-A-present
property is set (to any value, e.g. false), the target will be run. In the second
example, if the module-A-present
property is set
(again, to any value), the target will not be run.
Only one propertyname can be specified in the if/unless clause. If you want to check multiple conditions, you can use a dependend target for computing the result for the check:
<target name="myTarget" depends="myTarget.check" if="myTarget.run"> <echo>Files foo.txt and bar.txt are present.</echo> </target> <target name="myTarget.check"> <condition property="myTarget.run"> <and> <available file="foo.txt"/> <available file="bar.txt"/> </and> </condition> </target>
If no if
and no unless
attribute is present,
the target will always be executed.
Important: the if
and unless
attributes only
enable or disable the target to which they are attached. They do not control
whether or not targets that a conditional target depends upon get executed.
In fact, they do not even get evaluated until the target is about to be executed,
and all its predecessors have already run.
The optional description
attribute can be used to provide a one-line description of this target, which is printed by the
-projecthelp
-verbose
-debug
It is a good practice to place your tstamp tasks in a so-called
initialization target, on which
all other targets depend. Make sure that target is always the first one in
the depends list of the other targets. In this manual, most initialization targets
have the name "init"
.
If the depends attribute and the if/unless attribute are set, the depends attribute is executed first.
A target has the following attributes:
Attribute | Description | Required |
name | the name of the target. | Yes |
depends | a comma-separated list of names of targets on which this target depends. | No |
if | the name of the property that must be set in order for this target to execute. | No |
unless | the name of the property that must not be set in order for this target to execute. | No |
description | a short description of this target's function. | No |
A target name can be any alphanumeric string valid in the encoding of the XML file. The empty string "" is in this set, as is comma "," and space " ". Please avoid using these, as they will not be supported in future Ant versions because of all the confusion they cause. IDE support of unusual target names, or any target name containing spaces, varies with the IDE.
Targets beginning with a hyphen such as "-restart"
are valid, and can be used
to name targets that should not be called directly from the command line.
A task is a piece of code that can be executed.
A task can have multiple attributes (or arguments, if you prefer). The value of an attribute might contain references to a property. These references will be resolved before the task is executed.
Tasks have a common structure:
<name attribute1="value1" attribute2="value2" ... />
where name is the name of the task, attributeN is the attribute name, and valueN is the value for this attribute.
There is a set of built-in tasks, along with a number of optional tasks, but it is also very easy to write your own.
All tasks share a task name attribute. The value of this attribute will be used in the logging messages generated by Ant.
Tasks can be assigned anid
attribute:
where taskname is the name of the task, and taskID is a unique identifier for this task. You can refer to the corresponding task object in scripts or other tasks via this name. For example, in scripts you could do:<taskname id="taskID" ... />
to set the<script ... > task1.setFoo("bar"); </script>
foo
attribute of this particular task instance.
In another task (written in Java), you can access the instance via
project.getReference("task1")
.
Note1: If "task1" has not been run yet, then it has not been configured (ie., no attributes have been set), and if it is going to be configured later, anything you've done to the instance may be overwritten.
Note2: Future versions of Ant will most likely not be backward-compatible with this behaviour, since there will likely be no task instances at all, only proxies.
A project can have a set of properties. These might be set in the buildfile
by the property task, or might be set outside Ant. A
property has a name and a value; the name is case-sensitive. Properties may be used in the value of
task attributes. This is done by placing the property name between
"${
" and "}
" in the
attribute value. For example,
if there is a "builddir" property with the value
"build", then this could be used in an attribute like this:
${builddir}/classes
.
This is resolved at run-time as build/classes
.
In the event you should need to include this construct literally (i.e. without property substitutions), simply "escape" the '$' character by doubling it. To continue the previous example:
<echo>$${builddir}=${builddir}</echo>will echo this message:
${builddir}=build/classes
In order to maintain backward compatibility with older Ant releases, a single '$' character encountered apart from a property-like construct (including a matched pair of french braces) will be interpreted literally; that is, as '$'. The "correct" way to specify this literal character, however, is by using the escaping mechanism unconditionally, so that "$$" is obtained by specifying "$$$$". Mixing the two approaches yields unpredictable results, as "$$$" results in "$$".
Ant provides access to all system properties as if they had been
defined using a <property>
task.
For example, ${os.name}
expands to the
name of the operating system.
For a list of system properties see the Javadoc of System.getProperties.
In addition, Ant has some built-in properties:
basedir the absolute path of the project's basedir (as set with the basedir attribute of <project>). ant.file the absolute path of the buildfile. ant.version the version of Ant ant.project.name the name of the project that is currently executing; it is set in the name attribute of <project>. ant.java.version the JVM version Ant detected; currently it can hold the values "1.2", "1.3", "1.4" and "1.5".
There is also another property, but this is set by the launcher script and therefore maybe not set inside IDEs:
ant.home home directory of Ant
<project name="MyProject" default="dist" basedir="."> <description> simple example build file </description> <!-- set global properties for this build --> <property name="src" location="src"/> <property name="build" location="build"/> <property name="dist" location="dist"/> <target name="init"> <!-- Create the time stamp --> <tstamp/> <!-- Create the build directory structure used by compile --> <mkdir dir="${build}"/> </target> <target name="compile" depends="init" description="compile the source " > <!-- Compile the java code from ${src} into ${build} --> <javac srcdir="${src}" destdir="${build}"/> </target> <target name="dist" depends="compile" description="generate the distribution" > <!-- Create the distribution directory --> <mkdir dir="${dist}/lib"/> <!-- Put everything in ${build} into the MyProject-${DSTAMP}.jar file --> <jar jarfile="${dist}/lib/MyProject-${DSTAMP}.jar" basedir="${build}"/> </target> <target name="clean" description="clean up" > <!-- Delete the ${build} and ${dist} directory trees --> <delete dir="${build}"/> <delete dir="${dist}"/> </target> </project>
Notice that we are declaring properties outside any target. As of
Ant 1.6 all tasks can be declared outside targets (earlier version
only allowed <property>,<typedef> and
<taskdef>). When you do this they are evaluated before
any targets are executed. Some tasks will generate build failures if
they are used outside of targets as they may cause infinite loops
otherwise (<antcall>
for example).
We have given some targets descriptions; this causes the projecthelp invocation option to list them as public targets with the descriptions; the other target is internal and not listed.
Finally, for this target to work the source in the src subdirectory
should be stored in a directory tree which matches the package names. Check the
<javac> task for details.
Token Filters
A project can have a set of tokens that might be automatically expanded if found when a file is copied, when the filtering-copy behavior is selected in the tasks that support this. These might be set in the buildfile by the filter task.
Since this can potentially be a very harmful behavior,
the tokens in the files must
be of the form @
token@
, where
token is the token name that is set
in the <filter>
task. This token syntax matches the syntax of other build systems
that perform such filtering and remains sufficiently orthogonal to most
programming and scripting languages, as well as with documentation systems.
Note: If a token with the format @
token@
is found in a file, but no
filter is associated with that token, no changes take place;
therefore, no escaping
method is available - but as long as you choose appropriate names for your
tokens, this should not cause problems.
Warning: If you copy binary files with filtering turned on, you can corrupt the files. This feature should be used with text files only.
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.
Wherever path-like values need to be specified, a nested element can be used. This takes the general form of:
<classpath> <pathelement path="${classpath}"/> <pathelement location="lib/helper.jar"/> </classpath>
The location
attribute specifies a single file or
directory relative to the project's base directory (or an absolute
filename), while the path
attribute accepts colon-
or semicolon-separated lists of locations. The path
attribute is intended to be used with predefined paths - in any other
case, multiple elements with location
attributes should be
preferred.
As a shortcut, the <classpath>
tag
supports path
and
location
attributes of its own, so:
<classpath> <pathelement path="${classpath}"/> </classpath>
can be abbreviated to:
<classpath path="${classpath}"/>
In addition, one or more Resource Collections can be specified as nested elements (these must consist of file-type resources only). Additionally, it should be noted that although resource collections are processed in the order encountered, certain resource collection types such as fileset, dirset and files are undefined in terms of order.
<classpath> <pathelement path="${classpath}"/> <fileset dir="lib"> <include name="**/*.jar"/> </fileset> <pathelement location="classes"/> <dirset dir="${build.dir}"> <include name="apps/**/classes"/> <exclude name="apps/**/*Test*"/> </dirset> <filelist refid="third-party_jars"/> </classpath>
This builds a path that holds the value of ${classpath}
,
followed by all jar files in the lib
directory,
the classes
directory, all directories named
classes
under the apps
subdirectory of
${build.dir}
, except those
that have the text Test
in their name, and
the files specified in the referenced FileList.
If you want to use the same path-like structure for several tasks,
you can define them with a <path>
element at the
same level as targets, and reference them via their
id attribute--see References for an
example.
A path-like structure can include a reference to another path-like
structure (a path being itself a resource collection)
via nested <path>
elements:
<path id="base.path"> <pathelement path="${classpath}"/> <fileset dir="lib"> <include name="**/*.jar"/> </fileset> <pathelement location="classes"/> </path> <path id="tests.path"> <path refid="base.path"/> <pathelement location="testclasses"/> </path>The shortcuts previously mentioned for
<classpath>
are also valid for <path>
.For example:
<path id="base.path"> <pathelement path="${classpath}"/> </path>can be written as:
<path id="base.path" path="${classpath}"/>
In Ant 1.6 a shortcut for converting paths to OS specific strings in properties has been added. One can use the expression ${toString:pathreference} to convert a path element reference to a string that can be used for a path argument. For example:
<path id="lib.path.ref"> <fileset dir="lib" includes="*.jar"/> </path> <javac srcdir="src" destdir="classes"> <compilerarg arg="-Xbootstrap/p:${toString:lib.path.ref}"/> </javac>
Several tasks take arguments that will be passed to another
process on the command line. To make it easier to specify arguments
that contain space characters, nested arg
elements can be used.
Attribute | Description | Required |
value | a single command-line argument; can contain space characters. | Exactly one of these. |
file | The name of a file as a single command-line argument; will be replaced with the absolute filename of the file. | |
path | A string that will be treated as a path-like
string as a single command-line argument; you can use ;
or : as
path separators and Ant will convert it to the platform's local
conventions. |
|
pathref | Reference to a path defined elsewhere. Ant will convert it to the platform's local conventions. | |
line | a space-delimited list of command-line arguments. |
It is highly recommended to avoid the line
version
when possible. Ant will try to split the command line in a way
similar to what a (Unix) shell would do, but may create something that
is very different from what you expect under some circumstances.
<arg value="-l -a"/>
is a single command-line argument containing a space character, not separate commands "-l" and "-a".
<arg line="-l -a"/>
This is a command line with two separate arguments, "-l" and "-a".
<arg path="/dir;/dir2:\dir3"/>
is a single command-line argument with the value
\dir;\dir2;\dir3
on DOS-based systems and
/dir:/dir2:/dir3
on Unix-like systems.
Any project element can be assigned an identifier using its
id
attribute. In most cases the element can subsequently
be referenced by specifying the refid
attribute on an
element of the same type. This can be useful if you are going to
replicate the same snippet of XML over and over again--using a
<classpath>
structure more than once, for example.
The following example:
<project ... > <target ... > <rmic ...> <classpath> <pathelement location="lib/"/> <pathelement path="${java.class.path}/"/> <pathelement path="${additional.path}"/> </classpath> </rmic> </target> <target ... > <javac ...> <classpath> <pathelement location="lib/"/> <pathelement path="${java.class.path}/"/> <pathelement path="${additional.path}"/> </classpath> </javac> </target> </project>
could be rewritten as:
<project ... > <path id="project.class.path"> <pathelement location="lib/"/> <pathelement path="${java.class.path}/"/> <pathelement path="${additional.path}"/> </path> <target ... > <rmic ...> <classpath refid="project.class.path"/> </rmic> </target> <target ... > <javac ...> <classpath refid="project.class.path"/> </javac> </target> </project>
All tasks that use nested elements for
PatternSets,
FileSets,
ZipFileSets or
path-like structures accept references to these structures
as shown in the examples. Using refid
on a task will ordinarily
have the same effect (referencing a task already declared), but the user
should be aware that the interpretation of this attribute is dependent on the
implementation of the element upon which it is specified. Some tasks (the
property task is a handy example)
deliberately assign a different meaning to refid
.
Any Ant type which has been declared with a reference can also its string
value extracted by using the ${toString:}
operation,
with the name of the reference listed after the toString:
text.
The toString()
method of the Java class instance that is
referenced is invoked -all built in types strive to produce useful and relevant
output in such an instance.
For example, here is how to get a listing of the files in a fileset,
<fileset id="sourcefiles" dir="src" includes="**/*.java" /> <echo> sourcefiles = ${toString:sourcefiles} </echo>
There is no guarantee that external types provide meaningful information in such a situation
-lib
${user.home}/.ant/lib
${ant.home}/lib
<taskdef name="taskname"
classname="ImplementationClass"/>
<taskdef name="for" classname="net.sf.antcontrib.logic.For" />
<for ... />
<taskdef>
<taskdef resource="net/sf/antcontrib/antcontrib.properties" />
<for ... />
<taskdef>
<taskdef resource="net/sf/antcontrib/antlib.xml" />
<for ... />
antlib:
protocoll handler
<project xmlns:ac="antlib:net.sf.antconrib"/>
<ac:for ... />