Re: idl2matlab translate-o-matic

["J.D. Smith" <jdsmith(at)astro.cornell.edu>]

Craig Markwardt wrote:
> 
> "J.D. Smith" <jdsmith@astro.cornell.edu> writes:
> > > Since I can't pass a testing function to that routine (IDL doesn't have
> > > higher order functions), I will accept a routine, for illustrative purposes,
> > > that removes all even values from the array.
> > >
> > > Now suppose some joker passes an array containing only even values to that
> > > routine...
> > >
> > > - DM
> > >
> >
> > wh=where(array mod 2, cnt)
> > if cnt gt 0 then return,array[wh] else return, -1
> >
> > I use scalars (often -1) as cheap and easy to use empty arrays.  Anything with:
> >
> >  size(x,/N_DIMEN) eq 0
> >
> > is patently *not* an array.
> >
> >
> > And as far as the lack of "higher order testing functions":
> >
> > function evens, arr
> >    return, arr mod 2 eq 0
> > end
> >
> > function odds, arr
> >    return, arr mod 2
> > end
> >
> > function exclude,arr, exc_func
> >    wh=where(call_function(exc_func,arr) eq 0,cnt)
> >    if cnt gt 0 then return,arr[wh] else return,-1
> > end
> >
> > and to get rid of the odds, e.g.:
> >
> > IDL> a=exclude(b,"odds")
> 
> Okay, but let's say now you wanted to merge two lists like that
> together.  Wouldn't this be nice:
> 
> IDL> c = [exclude(a,'odds'), exclude(b,'evens')]
> 
> The way I say it makes it sound like it's just an inconvenience, which
> it is.  But for gosh sakes, its a *completeness* issue too.  We don't
> have a general purpose number system without zero!  It would be silly.
> Why should we have lists without the empty list?  Instead we have to
> drag around this extra notion of the COUNT or play tricks by returning
> scalars.
> 

I totally agree with you about the convenience of such an entity.  I'm not
trying to deny that.  What I'm trying to show is that what we might think of as
an "empty array" or "empty list" is just an abstract notion, actually
implemented in code in some way analogous to what I've done.  In stark contrast
to the issue you raise of a complete number system, in which the internal
representation for "0" is equivalent to that for any other number, an empty
array is achieved only by special case programming, which just happens to be
hidden from our sight.  Now, IDL is not C, and lots of special case programming
is hidden from our sight, so I'm certainly not arguing that hidden conveniences,
if well implemented, are to be avoided.  I just want everyone to understand that
this would be an addition purely motivated by convenience, and that there really
is no fundamental "incompleteness".  

Having said that, I see no reason that it couldn't be done pretty easily. 
Variables can already be marked "undefined", so why not extend that somewhat and
allow "undefined" arrays and lists to exist.  Dimensionality is important of
course, so the concept of a 2x2 empty array need be addressed, etc., but I
wouldn't think it's prohibitive.

JD


 
-- 
 J.D. Smith                             |*|      WORK: (607) 255-5842    
 Cornell University Dept. of Astronomy  |*|            (607) 255-6263
 304 Space Sciences Bldg.               |*|       FAX: (607) 255-5875 
 Ithaca, NY 14853                       |*|