ant-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Jose Alberto Fernandez" <>
Subject Re: [Ant2] Tasks as siblings of <target>
Date Mon, 05 Nov 2001 10:25:58 GMT
From: "Conor MacNeill" <>

> In Ant2, we want to strongly separate parsing from execution. In my 
> view, parsing will contruct an object model of the build file which will 
> be handed to the execution part of Ant. There will no effective 
> communication between these two phases apart from the build model 
> itself. Some things (not necessarily many) need to be processed and 
> validated in the parsing stage. This is why "target" and some other 
> keywords ("projectref", "include", whatever) will be coded into the 
> parser because it allows the parser to contruct the build object model. 

How committed are we to *all* this things? I really do not remember if there
have been votes in all this details. The reason for my question is that I do remember
the huge discussion we had about how AntGUIs were suppose to communicate with
ANT itself (can they manipulate the build-model?) or should all communications
be thru XML (either text or (J)DOM?). 

Independently of what it is decided, I would like that any arquitecture we come up
with will manage this "keyworkds" are its functionality in a pluggable fashion.
That will allow for extensions or customizations in the future on this regard.
I am not saying they are tasks, mayby they are ModelBuilders :)

> Everything that is not specifically recognized will probably be 
> collected into a model component that supports arbitrarily nested XML 
> elements. This will of course be used for tasks/types but may also be 
> used for other constructions which need to be processed in the execution 
> phase. So, there should be sufficient extensibility in the model for 
> other elements without making the parsing phase some sort of stripped 
> down DOM builder or taking on the "everything is a task" view.

One thing that I would like to be able to do is for ANT to have a more reasonable
implementation of <antcall>. Today every <antcall> reads and parses the project

file over and over. Which looks like a waist and makes the file system the only
way for GUIs to communicate with ANT. What we really need here is to be able
to reause the internal representation of the Project at the point where properties
have not been applied to attributes. That would allow for its evaluation as
parameters. Any ideas on how to do this efficiently?

Jose Alberto

To unsubscribe, e-mail:   <>
For additional commands, e-mail: <>

View raw message