| Object Programming in IDL [message #30832] |
Tue, 21 May 2002 05:26  |
graham_wilson_1234
Messages: 1
Registered: May 2002
|
Junior Member |
|
|
Just to appease Craig, I have started a new thread so I can avoid putting my
comments after David's 'gosh golly' post ;) I am interested in hearing
others comments...
The first point that we should all be very clear on is that IDL is _NOT_
a particularly good example of an object oriented language. You can
certainly emulate OOP concepts using IDL's objects and a select few
functions/proceedures but if often defeats the purpose of the OOP style.
When someone mentions IDL objects, it is universally assumed that they
really mean 'object graphics' which leads directly to point number 2;
Object oriented programming != object graphics. Unfortunately, it
is very difficult to dispel this myth using IDL because of point number 1.
With regard to writing object oriented code in IDL we are all rather stuck
until RSI implements a more complete feature set. I generally define
polymorphism it as the ability to process objects differently depending on
their data type or class. In this respect, the lack of operator overloading
is an example where IDL fails to offer the full OOP tool set. Yes, you can
overload methods, but operators should be no different. To compensate for
this missing functionality one can write functions and/or procedures but
this better described as an overlay and you must rely on a naming
convention or a path precidence to avoid conflicts. Personally, I'd like
to see true polymorphism (with overloading) and public/private methods
sooner rather than later (is anyone at RSI listening?).
A good technical book describing the merits of using objects in data
analysis is "Programming with Data: A Guide to the S Language"
(ISBN: 0-387-98503-4). The concepts described are specific to S-Plus but
can be adapted to any OOP language. While they may seem abstract at first,
they are very powerful way of manipulating and modelling data. A free
alternative to S-Plus is R (www.r-project.com).
For what it is worth, Matlab has a slightly more complete implementation of
OOP. The one glaring (and annoying) feature missing from Matlab, however,
is the absence of pointers and therefore dynamic structures/sizing. This,
of course, is a grip for a different newsgroup...
I lurk therefore I am.
Graham
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
