ant-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Jose Alberto Fernandez" <jalbe...@cellectivity.com>
Subject RE: Macrodef and parallel in a recursive situation
Date Mon, 20 Oct 2003 16:44:12 GMT
Peter,

I just made a browse to the changes you posted and I can see
why you are concern with ant/antcall. I see you did quite large
additions to core in order to implement this (using ThreadLocals).

I wonder if we can explore a more simpler implementation that is
less onerous than the one you have. I would think that adding
that making <parallel> pay for the cost of treating the computations
separately will make things simpler and make more posible to 
do the right thing for ant/antcall.

BTW, we are not talking ONLY about ant/antcall. We are talking
about them and about ALL the user implemented versions of <foreach>
<forall>, etc, etc that are around because we never provide one
in core. We cannot just break their them.

Do you see any reasonable way in which <parallel> could create
separate copies of its properties for each execution thread?
I will try to see if I getthe time to propose actual code.
But I have been kind of burden ere at work.

Jose Alberto

> -----Original Message-----
> From: peter reilly [mailto:peter.reilly@corvil.com] 
> Sent: 20 October 2003 16:54
> To: Ant Developers List
> Subject: Re: Macrodef and parallel in a recursive situation
> 
> 
> This is basicly what the local implementation
> does.
> 
> The only problem is for ant/antcall, a new project
> is created and the local properties would need
> to be copied to this, keeping in mind all the
> rules about user/normal/inherited and nested param
> elements.
> 
> Peter
> On Monday 20 October 2003 16:16, Matt Benson wrote:
> > Jose makes a good point about tasks that set
> > properties.  Without having delved into the
> > PropertyHelper* classes, could local properties be
> > handled some other way?  What about a Scope
> > TaskContainer subclass that would push an element onto
> > a stack.  Then any property queries would descend into
> > the stack... then a task that set a property could
> > have that property restricted by nesting it in a set
> > of <scope> tags...
> >
> > -Matt
> >
> >
> > --- Jose Alberto Fernandez <jalberto@cellectivity.com>
> >
> > wrote:
> > > > From: Christopher Lenz [mailto:cmlenz@gmx.de]
> > > >
> > > > Hi Peter,
> > > >
> > > > this looks really cool, but I'd suggest adding an
> > >
> > > attribute like
> > >
> > > > scope="local|global" to the <property> task
> > >
> > > instead of adding
> > >
> > > > a completely
> > > > new task.
> > >
> > > I disagree here. The issue is not <property> by
> > > itself. This local
> > > properties need to work with every <task> that may
> > > set a property:
> > > <condition>, <available>, <uptodate>, <property>,
> > > and a million more
> > > obscure ones. The main use for <local> is when it is
> > > called without
> > > the value attribute, because in that case this other
> > > tasks will
> > > set the local property and not the global one.
> > >
> > > Jose Alberto
> > >
> > > > -chris
> > > >
> > > > peter reilly wrote:
> > > > > I have written the code to support local
> > >
> > > properties.
> > >
> > > > > While I was doing this, I realized that the
> > >
> > > attributes
> > >
> > > > > of a macrodef could/should be local properties
> > >
> > > as well,
> > >
> > > > removing some
> > > >
> > > > > of the issues seen last week (use of attribute
> > >
> > > in a bsf script and
> > >
> > > > > support of parallel/recursive).
> > > > >
> > > > > The following shows it in using a new task
> > >
> > > called local.
> > >
> > > > > <project name="local">
> > > > >   <property name="prop1" value="a global
> > >
> > > value"/>
> > >
> > > > >   <target name="test1">
> > > > >     <local name="prop1" value="a local value"/>
> > > > >     <echo>prop1 is "${prop1}"</echo>
> > > > >   </target>
> > > > >   <target name="test2" depends="test1">
> > > > >     <echo>prop1 is "${prop1}"</echo>
> > > > >   </target>
> > > > > </project>
> > > > >
> > > > > This ant test2 generates the following:
> > > > >
> > > > > test1:
> > > > > prop1 is "a local value"
> > > > >
> > > > > test2:
> > > > > prop1 is "a global value"
> > > > >
> > > > > Each taskcontainer sets up a new local scope:
> > > > >
> > > > >   <target name="sequential">
> > > > >     <local name="prop2" value="in target"/>
> > > > >     <sequential>
> > > > >       <local name="prop2" value="in
> > >
> > > sequential"/>
> > >
> > > > >       <echo>prop2 is "${prop2}"</echo>
> > > > >     </sequential>
> > > > >     <echo>prop2 is "${prop2}"</echo>
> > > > >   </target>
> > > > >
> > > > > will generate the following:
> > > > > sequential:
> > > > > prop2 is "in sequential"
> > > > > prop2 is "in target"
> > > > >
> > > > > The value part of <local> is optional, and  the
> > >
> > > local
> > >
> > > > property may be
> > > >
> > > > > set by a subsequent <property>, <property> will
> > >
> > > only set it if the
> > >
> > > > > value is not set.
> > > > >
> > > > >   <target name="notset">
> > > > >     <local name="prop3"/>
> > > > >     <echo>prop3 is "${prop3}"</echo>
> > > > >     <property name="prop3" value="is set"/>
> > > > >     <property name="prop3" value="is set
> > >
> > > again"/>
> > >
> > > > >     <echo>prop3 is "${prop3}"</echo>
> > > > >   </target>
> > > > >
> > > > > will generate the following:
> > > > > notset:
> > > > > prop3 is "${prop3}"
> > > > > prop3 is "is set"
> > > > >
> > > > > prop3 is still a local variable and will not be
> > >
> > > seen outside the
> > >
> > > > > target.
> > > > >
> > > > > The local properties are thread local so the
> > >
> > > following
> > >
> > > > works as expected:
> > > > >   <target name="parallel">
> > > > >     <local name="prop4"/>
> > > > >     <parallel>
> > > > >       <sequential>
> > > > >         <property name="prop4" value="thread1"/>
> > > > >         <echo>t1: prop4 is "${prop4}"</echo>
> > > > >       </sequential>
> > > > >       <sequential>
> > > > >         <property name="prop4" value="thread2"/>
> > > > >         <echo>t2: prop4 is "${prop4}"</echo>
> > > > >       </sequential>
> > > > >       <sequential>
> > > > >         <property name="prop4" value="thread3"/>
> > > > >         <echo>t3: prop4 is "${prop4}"</echo>
> > > > >       </sequential>
> > > > >     </parallel>
> > > > >   </target>
> > > > >
> > > > > parallel:
> > > > > t2: prop4 is "thread2"
> > > > > t1: prop4 is "thread1"
> > > > > t3: prop4 is "thread3"
> > > > >
> > > > > Use with macrodef.
> > > > > -----------------
> > > > >
> > > > > Attributes may now be implemented as local
> > >
> > > properties, which means
> > >
> > > > > that they will be seen as normal properties by
> > >
> > > ant tasks -
> > >
> > > > including
> > > >
> > > > > script.
> > > > >
> > > > >   <target name="macro">
> > > > >     <macrodef name="callscript">
> > > > >       <attribute name="x"/>
> > > > >       <sequential>
> > > > >         <script language="beanshell">
> > > > >           System.out.println("x is '" +
> > > >
> > > > project.getProperty("x") + "'");
> > > >
> > > > >         </script>
> > > > >       </sequential>
> > > > >     </macrodef>
> > > > >
> > > > >     <callscript x="this is x"/>
> > > > >   </target>
> > > > >
> > > > > will generate:
> > > > > macro:
> > > > > x is 'this is x'
> > > > >
> > > > > Macrodef does not do the attribute substitutions
> > >
> > > so the following
> > >
> > > > >   <target name="macro2">
> > > > >     <macrodef name="callscript">
> > > > >       <attribute name="x"/>
> > > > >       <sequential>
> > > > >         <script language="beanshell">
> > > > >           System.out.println("x is '${x}'");
> > > > >         </script>
> > > > >       </sequential>
> > > > >     </macrodef>
> > > > >
> > > > >     <callscript x="this is x"/>
> > > > >   </target>
> > > > > will generate:
> > > > > macro2:
> > > > > x is '${x}'
> > > > > as <script/> does not do property expansion.
> > > > >
> > > > > A variation of the recurive macrodef last week
> > >
> > > may be done by:
> > > > >   <target name="recur">
> > > > >     <macrodef name="recur">
> > > > >       <attribute name="thread"/>
> > > > >       <attribute name="current"/>
> > > > >       <sequential>
> > > > >         <antcontrib:if>
> > > > >           <equals arg1="0" arg2="${current}"/>
> > > > >           <then>
> > > > >             <echo message="Thread: ${thread}
> > >
> > > done"/>
> > >
> > > > >           </then>
> > > > >           <else>
> > > > >             <antcontrib:math
> > > > >               datatype  = "int"
> > > > >               operand1  = "${current}"
> > > > >               operation = "-"
> > > > >               operand2  = "1"
> > > > >               result    = "current"
> > > > >               />
> > > > >             <echo message = "T: ${thread}, C:
> > >
> > > ${current}" />
> >
> > === message truncated ===
> >
> >
> > __________________________________
> > Do you Yahoo!?
> > The New Yahoo! Shopping - with improved product search 
> > http://shopping.yahoo.com
> >
> > 
> ---------------------------------------------------------------------
> > To unsubscribe, e-mail: dev-unsubscribe@ant.apache.org
> > For additional commands, e-mail: dev-help@ant.apache.org
> 
> 
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: dev-unsubscribe@ant.apache.org
> For additional commands, e-mail: dev-help@ant.apache.org
> 
> 

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@ant.apache.org
For additional commands, e-mail: dev-help@ant.apache.org


Mime
View raw message