Phil Weighill-Smith wrote:
My opinion regarding the disadvantages of this approach:
* Antcall has to create a whole new Project in memory in order to work
and is therefore an inefficient task
Yes. If the project is large this could be a large overhead. It seems
the semantics of antcall is not like a sub target but more like a target
in a sub project (even though the project happens to be the same
project). Is there a more lightweight solution planned in this area?
* If something invoked via Antcall depends on a target that is also
depended on by something depending on the target invoking Antcall then this
dependency target will be executed more than once because dependencies are not
handled across Antcall invocations
Yes.
* The dependency tree is "interrupted" and graphing tools that can show
ant build script structures will not (generally) work correctly and show the whole
dependency tree
I did not think about the graphing tools, but that is a good point also.
Given the fact that you did not list any advantages it seems this is not
a good idea.
It might be better to add "if" and "unless" to the standard ant Task to allow for conditional
execution, or even add a nested "condition" to the standard ant Task to allow for conditional execution. To
provide BC with the standard "execute" method, the condition/if/unless processing would need to happen
outside this method.
This seems like this is the real answer. However I read somewhere that
it is an architectural decision to not support "if" and "unless" etc. at
the task level. Can anyone point me to a discussion/document on that?
What about using scripting? Is that not recommended either?
Google search revealed that many people are looking for solutions for
similar problems.
Regards,
Sandip
Phil :n.
-----Original Message-----
From: Sandip Chitale [mailto:[EMAIL PROTECTED]
Sent: Sat 28/05/2005 18:56
To: dev@ant.apache.org
Cc:
Subject: A possible solution for conditional execution of tasks?
To conditionally execute a step in Ant one has to resort to setting up a
target structure like this:
:
<target name="predicate">
<condition property="condition-satisfied">
<available .../>
:
</condition>
</target>
<target name="conditional-step" if="condition-satisfied">
<!-- conditional tasks here -->
:
:
</target>
<target name="conditional" depends="predicate, conditional-step"/>
<target name="main" depends="conditional">
:
:
</target>
:
This is because of several reasons:
* The ant tasks do not have something like *if* attribute.
* One cannot get away with only two targets instead of three because
the dependencies are executed before the dependent. Using the
above example it is not possible to do what target predicate does
in the main target and avoid using the predicate target.
* Ensure order of execution
However, I tried a solution making use of antcall task and it worked. It
works as follows:
:
<target name="conditional-step" if="condition-satisfied">
<!-- conditional tasks here -->
:
:
</target>
<target name="main" depends="conditional-step">
:
<condition property="condition-satisfied">
<available .../>
:
</condition>
<antcall target="condition-satisfied"/>
:
</target>
The advantage of this approach is to quickly have some tasks execute
conditionally by putting them in a target and calling that target using
antcall after setting some property.
And it seemed to work. My question is - is there a problem using this
approach? Why or why isn't this a preferred approach?
Thanks in advance,
Sandip
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
