ant-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Michael Ludwig <>
Subject Re: Doing Ant builds
Date Fri, 12 Jun 2009 19:31:32 GMT
Steve Loughran schrieb am 12.06.2009 um 11:49:48 (+0100):

> Ant is a language primarily for Java projects. Basing it on Java is
> not just an ideological purity game, but the only way to get at those
> internal bits of the JDK in the same process. all the original JDK
> library tasks: javac, javadoc, rmic, etc do this: we go in and use the
> underlying code.

I see.

> >But why the clunky XML syntax? When Ant was conceived, XML was a hot
> >new thing, the final solution to almost everything, so doing XML was
> >en vogue.
> I know its easy to dismiss it now, but it does have strengths
> -any XML editor can work with it
> -easy to use with XSLT operations

I use XSLT a lot, and I really like XML. It lets you operate at a high
level, and with great flexibility. Sorry for conveying the expression
that it was the XML nature of Ant that I found clunky. It's rather that
by comparison of the Ant and XSLT syntaxes, I found Ant much more
difficult to memorize: path, location, name, file, dir - these are all
too similar to me, and I pick the wrong one. Just lack of habit, I gues.
Well, XSLT certainly has been given much more time for language design.

> The XML editing meant that before IDEs had explicit support for the
> semantics of Ant (properties, target dependencies), they could let you
> edit it in a structured manner.

I see; makes sense.

> we do strive to be more declarative than fully procedural languages,
> we don't have loops and so lack full turing-equivalence. There are
> also limits to what you can do in java

> Notice how Ant deliberately leaves out all fault handing too.

It's good guidance to know what not to attempt in Ant.

> >Somehow, Ant was adopted. How did it happen, and why? Does anyone
> >know?
> ant was written by James Duncan Davidson while sun was opening up
> Tomcat, moving it from a sun project which used make to something for
> anyone to use. Ant was a solution to a tooling problem.
> 1. There was no open source IDE at that time, open source projects
> couldnt mandate a single IDE the way in-house java teams could.
> 2. there was no way open source projects could mandate a single
> platform for the same reason. Today, I'd say "linux 1st, other unixes
> second, ignore windows" -this is effectively what Hadoop does.
> 3. It turned out to offer a profound advantage over IDEs. The lack of
> an IDE meant no debugger. All we had left was testing, and as JUnit
> came out at the same time, the <junit> task got written, and the world
> became a better place. Before then testing wasn't that mainstream, you
> sat in front of the IDE and debugged. Now you add tests and wait for
> email from the CI tool.

Thanks for these historical insights!

> >(Still no Perl.)
> There is a bit of perl in Ant, still works.

C:\jlib\ant-1.7.1 :: dir /s /b *.pl


Michael Ludwig

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

View raw message