| Object-oriented IDL? [message #5884] |
Mon, 26 February 1996 00:00 |
Ray Osborn
Messages: 1
Registered: February 1996
|
Junior Member |
|
|
I have only a limited acquaintance with IDL and am not a regular
follower of this group, so forgive me if this has been discussed before.
I've checked that it's not discussed in the FAQ. While I'm very
impressed with many of the features of IDL (and presumably PV-Wave), and
shall probably be using it a good deal in the near future, I have been
put off by what I believe is a fundamental limitation in its abilities
to handle composite data structures. I would be interested to hear from
more experienced IDL users, and perhaps the company itself, if they
think that this is a genuine limitation and whether it's something that
can/will be addressed in the future.
IDL is an array-based language, but when I perform data analyses, I
don't deal with arrays but rather with spectra. These spectra comprise
multi-dimensional arrays grouped with other arrays defining dimension
scales, data errors, spectrum titles, axis labels, sample temperature
etc. I gather that in IDL, I can group all these arrays and other
variables into a record structure, but if I want to add two such
structures, I would have to do all the book-keeping by hand i.e. check
that the x-arrays are compatible, add the y-arrays, propagate the errors
in quadrature and copy all the labels and other variables over. Doing
this once or twice would be no problem but for more complex analyses, I
am liable not to bother, and drop, for example, error-propagation to
keep it manageable.
If IDL were an object-based language, rather than an array-based one, I
could overload the "+" operator to perform all these operations in a
consistent fashion. As I understand it (and I'm not an OOP expert),
operator overloading just involves defining a standard set of procedures
that are invoked when performing mathematical operations on these
spectrum "objects". Once written, I can perform these operations much
more transparently with far fewer lines of IDL code knowing that, for
example, the errors are properly handled with each operation.
Incidentally, data plotting could also be simplified since the
independent axes and their labels could be carried around with the
spectrum, and would not have to be specified in the plot command.
I am not suggesting that IDL should have the complete flexibility of C++
or other object-oriented languages, but just defining a limited set of
classes could, I think, enormously enhance the power of IDL, at least
for the sorts of data analysis that I perform. I would be interested if
other IDL users agree with this, or whether they know of other ways of
getting what I want within the current version. If not, would it be
possible for IDL to be developed along these lines without being rebuilt
from scratch?
--
R. 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
