ant-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Jon Tirsén <jon.tir...@emm.com>
Subject RE: [PATCH] Dynamic target
Date Fri, 03 Nov 2000 02:20:54 GMT
I'm not sure what you mean by raising the stakes. I do agree that
patternmatching/dynamic-whatever/etc. is advanced features. But just because
there is advanced features doesn't necessarily mean you have to use them. It
just means you have something to grow in. It's like saying Java is an
immensely advanced language just because it contains EJB/Corba/RDBMS etc.
etc. But this is just my 5c.

If the Vision (observe the capital V ;) of ant is that it should be kept
simple, then I guess this definitely goes against that Vision and probably
should need to be dropped. But, still, ant is growing, both in numbers of
users, numbers of features, and lines of code. It is being used in contexts
not thought about when initially developed. Definitely worth thinking about.

Keeping the design of software clean and simple is very sane. Just look at
the sad story of C++.

I'm not sure I would use the term yucky though, it simply just might be out
of scope. Sad, I do consider such a feature to be useful.

-----Original Message-----
From: Peter Donald [mailto:donaldp@apache.org]
Sent: Thursday, November 02, 2000 5:19 PM
To: ant-dev@jakarta.apache.org
Subject: RE: [PATCH] Dynamic target


Hi,

I would be completely and utterly -1 to implement patternmatching or any
dynamic dependency tree building for that matter. If you want make then
there is already jmk, jmak, jam etc. Ant is meant to be simple and
intuitive. Alternatively what you could do is provide a xslt/css/custom
frontend that transforms into ant file. Whatever the case this raises the
stakes too much IMHO and thus is definetly yucky


At 04:54  2/11/00 -0800, you wrote:
>Yes! The ability to specify targets as a pattern is the proper
>generalization/formalization of Jonas' proposal. And since make has had
that
>for a RLTN (Really Long Time Now ;) it's a pretty proven feature. That is
an
>extremely useful tool for build-system-generalization, and probably what I
>miss most from make.
>
>So, was it a target-pattern war or anything? I'm pretty new to this list.
>What was the reason for not having target-patterns? I'm extremely +1 for
>target-patterns.
>
>Also:
><target match="*" property="dynamic.target" ....>
>    ...
></target>
>is not the implementation of what Jonas proposed, basically the
>"dynamic.target" magic property of Jonas is the same as the $< magic
>property of make.
>
>-----Original Message-----
>From: Jose Alberto Fernandez [mailto:JFernandez@viquity.com]
>Sent: Thursday, November 02, 2000 3:02 PM
>To: 'ant-dev@jakarta.apache.org'
>Subject: RE: [PATCH] Dynamic target
>
>
>My concern here is how do you plan to dealt with multiple targets?
>If I do:
>
>  ant unk1 unk2 unk3
>
>what Jonas proposed will not work since the property will only get one
value
>(not mutable).
>
>I do not have a solution for the problem either.
>
>How does this work if I have:
>
>  <target name="known1" depends="unknown2" />
>
>would the execution of known1 cause the execution of the dynamic target
>corresponding to "unknown2"?.
>
>If this does not work, then I think we only have a cluge. Any solid
solution
>should work consistently independently of the situation.
>
>BTW, At some point I proposed having "template" targets. Something like:
>
>  <target match="unk*" property="prop.name"..... />
>
>which matches any target that starts with "unk". If we add to this some
>rules about priorities based on how good the match is, we can get something
>more general and better formulated than the dynamic target concept.
>
>In particular, the dynamic target is nothing else than:
>
>  <target match="*" property="dynamic.target" ....>
>    ...
>  </target>
>
>still there is the issue of what is the scope for "dynamic.target"
>if the rule applies again in a dependence, what should happen?
>Do we need a separate concept for keeping track of the actual target name?
> "${}" could mean the actual name of the current target, which in the
>case of a matching target means the actual value matched. So no "property"
>attribute required.
>
>Opinions?
>
>Jose Alberto
>
>
>
>> -----Original Message-----
>> From: Jon Tirsén [mailto:jon.tirsen@emm.com]
>> Sent: Thursday, November 02, 2000 12:05 PM
>> To: ant-dev@jakarta.apache.org
>> Subject: SV: [PATCH] Dynamic target
>>
>>
>> I'm not sure that the feature is 100% bad. I think that a
>> feature of that
>> kind in combination with the script-task would make much more complex
>> behaviour in a build-system possible. It would for example be
>> a very good
>> tool for generalizing build-files to a higher degree.
>>
>> I do agree upon that the patch is to magic/implicit in it's current
>> manifestation. But I do see the use of such a feature in a
>> very advanced
>> ant-build-implementation. (Such as my own. :-)
>>
>>
>> -----Ursprungligt meddelande-----
>> Från: Conor MacNeill [mailto:conor@m64.com]
>> Skickat: den 2 november 2000 06:29
>> Till: ant-dev@jakarta.apache.org
>> Ämne: RE: [PATCH] Dynamic target
>>
>>
>> Jonas,
>>
>> I agree with Stefan. This is too implicit for my liking.
>> Specially named
>> targets are not a good idea, IMHO. If you did want to have
>> this sort of
>> specific meaning, (and I question whether that is the case, anyway), I
>> think it would require a different element name such as <dynamic>.
>>
>> So for now, I'm -1 on this patch.
>>
>> Conor
>>
>>
>>
>
Cheers,

Pete

*------------------------------------------------------*
| "Nearly all men can stand adversity, but if you want |
| to test a man's character, give him power."          |
|       -Abraham Lincoln                               |
*------------------------------------------------------*


Mime
View raw message