Re: IDL interpreter questions - can someone (D.Fanning) explain - TIA

[JD Smith <jdsmith(at)astro.cornell.edu>]

dadada wrote:
> 
> Topics in question:
> 
> How are variables referenced by default?

I'm not sure what you mean here.  Pointer references?  They are explicit
only... i.e. you can't create a reference of an existing variable.

> Does IDL have first class procedures?

Only in-as-much as run-time evaluation is allowed.  So the answer is no,
I guess.  You can call methods and routines by string name... e.g.

call_procedure,'mypro',arg1,arg2

The overhead is only slightly higher than real anonymous functions, but
the syntax obviously isn't transparent (i.e. you always *know* you're
calling a procedure by reference to it's name, vs directly).

> Continuations?

No continuations.  Non-local exits from recursions are possible using
standard argument passing, or common blocks.

> Suspensions (ie. thunks)?

No thunks or closures.  

> How is IDL interpreted in relation to Scheme or LISP?

IDL is a very straightforward procedural language similar to FORTRAN,
and lately with more and more C-like paradigms, but with vector operator
overloading and built-in, convenient support for arrays of up to 8
dimensions (once you learn the peculiarities of slinging them around). 
It is a data-based, not function-based language, and sometimes it really
shows.  A simple object-oriented framework was added in the last few
years that introduces very little complexity, and trades features for
ease of development.

In part because of this simple design, IDL is actually quite fast, and
the interpreter can assemble large chunks of precompiled code after the
first pass. 

If you are comfortable in C and/or FORTRAN or other imperative
languages, you will be utterly amazed at the speed with which you can
accomplish things in IDL.  

If you are a LISP/SCHEME or other functional language guru, you will
wonder how anyone can get anything done in a language so devoid of
higher-level constructs (a lambast you'd also levy against C, FORTRAN,
and the rest).  If you are a hardcore OO programmer (C++, Smalltalk,
Objective-C, Java), you might feel IDL's OOP is somewhat toylike.  If
you are a guy with a bunch of data, probably without a CS degree, and a
need to process, visualize, and combine your data, you'll feel right at
home in IDL.

JD