ant-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Dominique Devienne" <DDevie...@lgc.com>
Subject RE: local properties
Date Wed, 20 Oct 2004 16:47:32 GMT
> From: Jose Alberto Fernandez [mailto:jalberto@cellectivity.com]
> > From: Dominique Devienne [mailto:DDevienne@lgc.com]
> > > From: Stefan Bodewig [mailto:bodewig@apache.org]
> > 1) I don't like the <let> name. Perhaps it shows how ignorant I am
> >    about other languages not in the C family, but it doesn't speak
> >    to me, and the name to not convey the purpose. Thus I'm -1 to
> >    the <let> name. <scope> or <local> or else are not perfect,
but
> >    at least convey more meaning to my ignorant self.
> 
> :-(. <let> comes from mathematics.

Maybe in your part of the world...

> That is the reason for using <let>, it is also used in some functonal
> languages.

I though so. I'm still -1, or at best -0
if I hear more convincing arguments.

> The scope, is the scope of the attribute notation (i.e., the
macrodef).
> Now you can use this name to create a new property, using <property/>
> or to watever else you please, there are no expectation whatsoever.

I'm not following. I think I understand what a scope is.
I don't confuse scope for the notation to define explicitly what
should 'go out of scope' when the explicit scope ends.

I consider running a <macrodef> as starting/entering a new scope for
names,
doing something, then ending/leaving the scope, restoring shadowed
properties and removing local properties, as defined by the propertyset.

> And I have mentioned several times that one could use propertysets
> to stop things leaking through <antcalls> and such. But you are still
> thinking only on properties. There are other things that we create
> dynamically in ANT. Like references, scripts, etc.

Not at all. I'm thinking properties and references. What else is there?
Scripts are scripts, id'd or not. That's still references.

> The main hurdle is that ANT uses a flat namespace for things and
> that anypart can see any property/reference defined by any other part
at
> any time on the life of the project.

So? As Peter points out, a flat namespace doesn't mean it cannot
be implemented using a stack of maps per scope, similar the nesting
of Properties when you provide a default. It's still flat, but uses
nested Properties to provide correct compartimentation. This is how
<antcall> should have been implemented instead of the copy going on
right now.

> I know they are very old concepts here, but just as old are the data
> structures we used in ANT. And on top, the fact that something does
not
> exist is meaningful (i.e., unless) I do not see how you can reconcile
> all this things. Maybe this should be done as part of ANT 2.0  (joke).
> And forget about BC,

Sorry, but I still don't see why it can't be made BC... The Project API
has to be unchanged of course, but the actual impl should move to a
lighter weight Context object or something that would be stacked
correctly, and to which Project would delegate to.

Dealing with <parallel> would be tricky, but I think we *should* break
BC by not allowing the different 'threads' of a <parallel> to share
properties, at least unless explicitly requested.

But that's going into too much detail for now. --DD

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@ant.apache.org
For additional commands, e-mail: dev-help@ant.apache.org


Mime
View raw message