ant-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Peter Donald <dona...@apache.org>
Subject RE: Parameterized "task-function"
Date Mon, 15 Jan 2001 22:36:26 GMT
At 01:28  15/1/01 -0500, Rosen, Alex wrote:
>[Switching over to ant-dev again...]
>
>Peter Donald wrote:
>> Nope ants build.xml format is a declarative form for
>> describing a build process.
>
>One might even say a build "procedure" :)

right ;)

>As I said before, I'm not wild about the idea of inventing our own
>mini-scripting language. But I think Jason's system would be *by far* the
>easiest to understand and use. Maybe I'm naive, and haven't thought things
>through, but to me this would be a joy to use, whereas using XSLT would be a
>confusing mess. I guess that's somewhat of an exageration - I'm sure once
I got
>my head around it, it wouldn't be too hard to use. But I think it would
require
>me to go "huh?" a lot.

it does that for complex stuff but for easier stuff there should be less
"huh?" I think thou I have been using it for ages so I may have forgotten
what it was like to learn ;)

>When I first tried using Ant, I found it intuitive, and a joy to use. Then I
>ran in to some of the limitations of the current Ant model, mostly related to
>code/data/whatever reuse, and I came to see that Ant was intuitive because it
>was overly simplified. It was great for the easy case, but not the hard case.
>If we can solve the hard case and still keep it intuitive, I think that would
>be fabulous. Maybe adding more scripting to the Ant file is another seductive
>but ultimately unsuccessful idea. But I'd like to try that road first. 

Well thats where I differ I guess. Once we add support for X - no matter
how good or how bad X is we are likely to have to support it. We still have
a whole bunch of bad design choices in Ant1.x (can you say magic properties
;-]) that we have to continue supporting. If we add if/while/case
statements then they can *never* be removed (until Ant3.x). However
experiments with XSLT is easy to remove as it does not change the format of
build.xml file. It is a preprocess stage. 

>I don't
>think that anyone will think that an Ant that requires XSLT to do anything of
>even medium complexity will ever be a joy to use.

I do a lot of things quite easily of medium size - many of the projects at
Apache are of medium size or bigger and they get away with it. Using XSLT
will always be a tradeoff of reuse vs complexity of reuse. 

>Pete, I haven't quite figured out your take on XSLT for build files. Do you
>think it's intuitive, or does it just suck less than the alternatives?

Well I don't much like XSLT but I think it sucks less than the alternatives
;) We can also back out in the future or use CSS instead if we find it was
the wrong choice with minimal hubub. I am loathed to complicate the base
Ant format unless we absolutely have to.

>Does anyone have any experience with IBM's Bean Markup Language and/or Bean
>Scripting Framework? Could they be useful here?

BML is great and we actually use BSF in script task ;) However I am not
sure BML is appropriate here - while some aspects of it has made their way
into my proposal (separation between convertion aspect and
data-container/data-parser aspects) - I am not sure which other aspects
would be useful.

>P.S. Pete, sorry for making you fight "battles" :) on two simultaneous
fronts.

;)

Cheers,

Pete

*-----------------------------------------------------*
| "Faced with the choice between changing one's mind, |
| and proving that there is no need to do so - almost |
| everyone gets busy on the proof."                   |
|              - John Kenneth Galbraith               |
*-----------------------------------------------------*


Mime
View raw message