Hi again,

> > Some time ago I propose XML-like based communication - maybe its time
comes?
> > How do you see it now when there was enough time to consider it?
>
> I don't like the we-do-anything-in-xml approach. XML is a
> nice thing, but it certainly adds some overhead to
> communications. I would prefer an invisible property system
> based on keys. This would allow to store informations in one
> task and to read it in another one, without dealing with
> properties set by the user.

I'm in no way "xml guy". XML is nice thing for some purpose but definitely
not usable for everything. What I propose is to introduce _some_ system how
to pass information from one task to other. Since speed is not main interest
in automated builds (sorry Matt - I still appreciate your optimization
work!) XML should be straightforward solution. But I never think we will
serialize XML on every task. XML hierarchy is pretty efficiently stored in
memory like object tree and that should be enough (what is exactly what you
propose, isn't it?)

---

>Perhaps the problem is just that I can't think of an example for which
>verbose inter-task communication would be a good fit.  Do you have an
>example that is more extensive than the exec example that you can share?

More complex example? Ok here is one from mine real needs:

I have <solution> without .sln file but I include fileset of <projects>. But
compilation itself is only small part of every complex build. I need to
check,sign,ftp,copy,... builded assemblies. If one of those assemblies
failed to compile, whole <solution> failed and now I'm forced to scan
directory what is builded to know what I have to process further. And I'm
totally short to make some meaningful error log in such case.

If <solution> have ability to construct some result block I could parse it
and construct result-fileset as well as error-log and whatever information I
need. E.g. (in xml syntax) it could return:
<solution-result>
<projects>
<errorproperty value="true"/>
<project name="prj1" status="successful" output="prj1.dll"/>
<project name="prj2" status="successful" output="prj2.exe" warnings="2" />
<project name="prj3" status="error" warnings="5" errors="1">
  <error class="Bar" file="foo1.cs" line="15" text="class Xyz not found"/>
</project>
<project name="prj4" status="skipped" text="skipped because of dependancy on
failed project prj3"/>
</projects>
</solution-result>

Having that one could easily condition other tasks with <if>+<xmlpeek>.


---

>dependencies... but in practice, if a particular task is so dependent on
>the output of another task, then why doesn't it extend (or compose) the
>other task.  This allows the task to communicate with its predecessor in
>the trivial way, through object properties and method calls, without
>having to modify how NAnt works.

True. But why contruct some task for everything? This is case of <solution>
or even <csc> (which calls <resgen>) and it is ok. I do not want to call
every task myself - solution is great! On the other hand I do not want to
write task for every one time stuff.

Regards,
Martin



-------------------------------------------------------
This SF.Net email sponsored by Black Hat Briefings & Training.
Attend Black Hat Briefings & Training, Las Vegas July 24-29 - 
digital self defense, top technical experts, no vendor pitches, 
unmatched networking opportunities. Visit www.blackhat.com
_______________________________________________
nant-developers mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/nant-developers

Reply via email to