[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Top 10 for old farts

Continuing in the apparently pointless, but extremely 
interesting discussion about "old farts" and IDL...

We're a mid sized (approximately 12 people) research 
group in a physics department.  Quite a range of 
experience with IDL, ranging from novice to over 10 years.

*One* person here is using objects: me.  The only reason
I'm doing that is out of curiosity.  The ideas are intellectually
appealing, and some experience with OOP seemed like a good
idea.  I've put together some classes for various purposes,
have enjoyed the experience, and will probably never use them for 
general purpose work.  They are so completely different from
our existing code, and would require some re-tooling on the part 
of people here, as well as the collaborators that we share
code with.  That's the kind of thing I had plenty of time for
as a student, but won't happen as a researcher.

( Actually, I might argue that the problem with IDL is not
enough object functionality.  If *all* variable types were
first class objects which allowed *all* low level function calls
ie. mag= c->abs(), cstar= c->conj()  then things might
make some more sense.  )

Clearly, there are people out there who find objects very
useful from a programming perspective.  I respect that, but
do worry that the good people at RSI might listen a point
of view that may be the majority in this newsgroup, but does
not at all represent our needs here.

As a side note, David Fanning made some brief comment in a
previous message expressing (possibly tongue in cheek?)
incredulity that not everyone was using widgets yet.  Again,
*none* of our daily work uses widget-based tools.  Don't
get me wrong, I'm a big fan of widgets, and wrote quite a
few during grad school and after.  For certain purposes they
are wonderful.  However, for evolving research they're rarely
useful.  I write functions, try them from the command line,
hook them together with scripts, and look at the results with
direct graphics.  After many iterations I (hopefully) figure
out what the appropriate analysis is, and would be ready to
roll it all up into a nice widget bundle.  Of course by then
it's time to make some figures for the paper, and then move
onto something entirely different.  No time to write the widget,
and no real need.

I'm going to wait until after I've beta-tested 5.4 before
putting together my top 10 wish list, but right now it
simply consists of a general desire for faster more stable
numerical routines and special functions.  The core of what
we do is based on math, and if that's not easy and reliable 
then nothing else matters.  After that, there's I/O to deal
with the different large data sets we use.  That's something
that IDL is already superb at, and doing a good job keeping
up with new file types (ie. PNG, CDF).

				Brian Jackel