ant-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Steve Loughran <>
Subject Re: Python implementation for Ant
Date Fri, 09 Feb 2007 10:14:38 GMT
Matt Benson wrote:
> Here's a link to Terence Parr's "Humans should not
> have to grok XML" article:
> In it he explains the distinction between XML as a
> data representation format, the use for which it was
> AFAIK intended, vs. as an explicit interface for human
> consumption.  Regardless of your own position, I found
> it an interesting read.  Terence is an example of a
> noted personality in the OSS community who develops in
> Java but has never touched Ant precisely _because_ of
> its XML interface.  So such developers definitely
> exist, thought it's quite possible he is the only one
> whose aversion to human-targeted XML is a strong
> enough deterrent in and of itself.  ;)
> -Matt

I think somewhere in Ant-in-Anger I'm critical of XML too. If XML is a 
language for humans and machines, then humans have already lost the war 
they talk about in terminator.

Peter said:
 > XML is actually easier that HTML as it is less forgiving with regard 
to mistakes.

Is that easier or worse? Its harder to write well formed XML than a 
simple HTML page. Easier to parse XML, yes, but write?

Now, regarding Ant languages. I'm going to be even more heretical. I 
find it ironical that we are using a declarative language -Ant XML- to 
describe how to build, test and deliver procedural code.

It seems to me that long term the goal should not be 'improve the syntax 
of Ant', but 'radically improve how we code'. Does the syntax of setting 
up the classpath for javac, junit, etc, matter more or less than the 
fact that to run junit I first have to compile my code, then run the 
tests? Whereas when I was doing some prolog work in december, I just 
declared a set of prolog unit tests [1] at the bottom of my prolog file, 
and I could reload the file and rerun the tests in one go. And because 
prolog is a logic langauge, pretty much any prolog clause with real 
values in constitutes a test case.

Now, I am not going to evangelise Prolog on everyone, but I found some 
things really nice about going back to it
  -its very graph centric. If you can walk the graph (especially 
depth-first), then you get graph code for free
  -lists are intrinsic to the language. Java only has arrays and 
strings; no lists. So working with lists drops you down to objects and 
unwiedly operations.
  -incredibly fast dev cycle. That is despite the fact it doesnt come 
with IDEs as good as IDEA.

Admittedly, this is very post-java, but that time will happen. And then 
the language used for the java build tools will be a mere detail.



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

View raw message