| Re: Object based/oriented IDL ? Ever likely ? [message #5985 is a reply to message #5975] |
Tue, 19 March 1996 00:00   |
ROsborn
Messages: 2
Registered: March 1996
|
Junior Member |
|
|
In article <julien.827283540@marsh>, julien@cs.curtin.edu.au (Julien
Flack) wrote:
> I am very impressed with a number of features of IDL. However, I think
> that its lacking support for structure (primarily data structures) due to
> its historical affiliation with Fortran (no flames please). This weakness
> becomes noticable when you reach a 3,000+ line application (IMHO).
>
> I think that a version of IDL using object based/oriented technology would
> be immensely powerful and would reach a far wider audience. Is there a
> desire for OO technology in the scientific community, or is Fortran still
> predominant ? Have RSI made any moves in this direction ?
>
I'm amused that you've posted this question, because I posted an identical
question only a couple of weeks ago. I got a couple of responses, mostly
stating that IDL was flexible enough to cope with any programming needs,
and expressing some scepticism of the need for OO concepts. Otherwise,
the thread sank like a trace.
Just in case it does spark some interest this time, I'll briefly repeat my
reason for raising the issue. The concern I have with IDL is that I deal
with composite data objects, or spectra, comprising several arrays
(usually x, y and error arrays with axis labels, various attributes
etc.). When I combine different spectra e.g. in subtracting a background
run, I have a considerable amount of book-keeping to do, such as checking
that the x-arrays are compatible, passing the labels across, propagating
the errors etc. If IDL incorporated some OO features, it would be
possible to define a spectrum class so that I could overload the
arithmetic operators and hide this book-keeping from my interactive
session, making my online analyses much more productive IMHO.
You have raised another equally valid issue, that of simplifying large IDL
programs. One of the responses to my posting drew my attention to a suite
of IDL procedures developed at the Institut Laue Langevin, Grenoble,
called LAMP, which attempts to treat multi-dimensional spectra in the way
I wanted. Strangely enough, it was this suite which made me concerned
about the limitations of IDL in the first place. It is a very impressive
achievement, and contains some very nice features, but it has required
well over 10000 lines of code, many of which I am sure would be
unnecessary if IDL contained some OO concepts. I think that it would also
be a lot easier to maintain.
I also would be interested if RSI has any intention of moving in this direction.
--
Ray Osborn Tel: +1 (708) 252-9011
Materials Science Division Fax: +1 (708) 252-7777
Argonne National Laboratory E-mail: ROsborn@anl.gov
Argonne, IL 60439-4845
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
