Re: idl2matlab translate-o-matic

[Craig Markwardt <craigmnet(at)cow.physics.wisc.edu>]

Pavel Romashkin <pavel@netsrv1.cmdl.noaa.gov> writes:

> > What can you say of a language that is purely array oriented, but
> > cannot comprehend the existence of an empty array?
> 
> Agreeing with D.F., I so far had no use for an empty array. I
> understand it is not flexible, but I usually work on data other than
> nothing.

Forgive him, he knows not what he says.

Empty arrays would be invaluable in both indexing (such as with WHERE)
and array concatenation.  By invaluable, I mean that it would remove a
lot of the special casing.  Consider these examples:

ARRAY INDEXING - indexing with where()
 *With* an empty array:
   wh = where(array GT thresh, /EMPTY)
   array(wh) = 0   ;; indexing with empty array has no effect
 *Without* an empty array
   wh = where(array GT thresh, count)
   if count GT 0 then array(wh) = 0

ARRAY CONCATENATION - growing an array
 *With* an empty array:
   l = empty_array()
   for i = 0, 100 do if expression(values) then l = [l, values]
 *Without* an empty array:
   for i = 0, 100 do $
     if expression then $
        if n_elements(l) EQ 0 then l = [values] else l = [l, values]
   
As you can see, the "with" code is more simple and easy to read.  The
"without" (which represents the status quo) has special cases which
ruin the flow of thought.  For a vectorized language, this is a
painful burden to bear sometimes.  If you don't believe me, try doing
the following (apparently simple) problem:

 * given two arrays, A and B: concatenate all but the last two
   elements of A, with B.  Don't try [A(0:n-3),B], or you will be in a
   world of hurt.
   
Craig

-- 
--------------------------------------------------------------------------
Craig B. Markwardt, Ph.D.         EMAIL:    craigmnet@cow.physics.wisc.edu
Astrophysics, IDL, Finance, Derivatives | Remove "net" for better response
--------------------------------------------------------------------------