Re: IDL's handling of LOGICAL quantities (WHERE)

[Mark Fardal <fardal(at)weka.astro.umass.edu>]

James Tappin <sjt@star.sr.bham.ac.uk> writes:

> \begin{rant}
> I've finally decided to have a public moan about one of the weaknesses of=
>  IDL's
> handling of logical operations: to boot -- that the WHERE function  follo=
> ws
> a C-like interpretation while most other things are  Fortran-like.
> 
...
> IDL> print, where(a eq 0)
>            0           2
> IDL> print, where(not (a ne 0))
>            0           1           2           3
> 
> I guess the proper answer isto have  aproper  logical or boolean type and
> functions like FINITE and logical operations should return it, and of cou=
> rse
> WHERE should accept it.

Actually, it would be fine if IDL were more C-like in this case.  C
doesn't have a boolean type either.  But it does maintain a consistent
definition of "true", which helps a lot.  In C, 2.0 and 2 are both
true.  In IDL, 2.0 is true but 2 is false.  Even worse, IDL conflates
the boolean with the bitwise operators.  AND, OR, XOR, and NOT are
boolean operators for floating types and bitwise operators for the
integral types.  In C you don't have this problem, because there are
different symbols for the boolean and bitwise operators.  (Did I get
that explanation right?  It only took me eight years with IDL to
figure out how it works...)

This illogical logic has been around for so long that I think we're
stuck with it.  But if anyone out there is writing an IDL replacement
(hey Arno, you can handle this right?), this is my nomination for the
second thing to fix.

Mark Fardal
UMass