ant-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "James Cook" <jimc...@iname.com>
Subject RE: Did somebody say Shut up and Write? :)
Date Wed, 20 Dec 2000 05:27:57 GMT
My mail is kind of screwed up, so I don't get all of the posts in order. It
seems that Sim responded to a post to Peter regarding frANTic, and he made a
lot of very good points. Now Peter is responding to your post, but I haven't
seen your original post.

I know at one time that you posted a nice list of "features" that you
wanted, Diane. I hope to add to my current design document and address each
of your features and Peter Vogel's nice list of needs. These are the two
"feature" lists that stand out for me right now. I am interested in Ant
integration into Java IDEs, so I have a bunch of desires that I would want
Ant 2.0 to achieve.

> -----Original Message-----
> At 01:23  19/12/00 -0800, Diane Holt wrote:
> >Would you be able to say if your "frANTic" proposal will help
> simplify the
> >way I currently have to use Ant to deal with the build issues I have, and
> >if so, how?

> From: Peter Donald [mailto:donaldp@apache.org]
> Nope - no difference in anyway at all as I see it ;)

I'm not sure where Peter is coming from, but I can give you a summary of
what is in my design document, and you can see if it answers your question.
Unfortunately, I don't have your original email yet, but I think I can glean
from Peter's response it seems that you have an interest in templating (I
like that better than rules also).

I think frANTic is well suited to the templating task because of the way the
execution engine operates. Unlike Ant 1.x, frANTic doesn't try to protect
the script builder from themselves by trying to resolve circular references.

In fact, it allows a Task to be referenced by name (with a full hierarchical
namespace) and invoked as many times as necessary. You can also change the
value of "properties", so a Task can be executed multiple times with varied
behaviors depending on the new values of the properties.

I tried to model the execution engine on a compiler model. There is full
support for a call stack (task stack), the Tasks are analagous to methods,
and "parameters" are akin to programming variable. The parameters even have
the same scope definitions that you would expect in a language.

This model also makes scripting almost trivial to implement. I have been
looking at IBM's BSF framework for integrating a variety of scripting
languages from JPython, javascript, and TCL (among others). Check it out at
http://oss.software.ibm.com/developerworks/opensource/bsf/languages.html
[Our friend Sam Ruby has had a hand in this.]

jim


Mime
View raw message