Re: Reflection (was: Re: The best thing/greatest feature summary)

[Tucker Taft <stt@averstar.com>]

Ray Blaak wrote:
> 
> Jeff Carter <jrcarter@acm.org> writes:
> > I've been toying with this:
> [...]
> > generic -- Reflection
> >    -- null;
> > package Reflection is
> >    Unit_Name : constant String;
> >    -- The full name of the unit in which this package is instantiated
> [...]
> > end Reflection;
> [...]
> > This package indicates that this kind of information is available in
> > Ada. We just need an easier way to access it.
> 
> Very cool.

I agree, this is a very clever use of Expanded_Name and T'Tag.
GNAT has a few implementation-defined attributes that provide things
like current line number and file name, I believe, and it would
seem a natural extension to provide things like current scope expanded name,
or even generic actual parameter name.  There are lot's of interesting
possibilities here which don't really require run-time reflection, but rather
generic-instantiation-time reflection, which at least in a macro-expanding
model for generics seems quite doable.

> Now if only tagged types could be declared within routines, and then we would
> be all set.

Tagged types can be declared anywhere.  Type extensions are restricted
to being at the same accessibility level as their ancestor types,
but this isn't a type extension.

> Or does instantiating this package within a routine actually work? (I don't
> have an Ada compiler near me at the moment.)

This should work anywhere.

> 
> --
> Cheers,                                        The Rhythm is around me,
>                                                The Rhythm has control.
> Ray Blaak                                      The Rhythm is inside me,
> blaak@infomatch.com                            The Rhythm has my soul.

-- 
-Tucker Taft   stt@averstar.com   http://www.averstar.com/~stt/
Technical Director, Commercial Division, AverStar (formerly Intermetrics)
(http://www.averstar.com/services/IT_consulting.html)  Burlington, MA  USA