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

IDL compatibility




I had a chance to try an IDL 5.3 installation today.  I know I am
behind the times, sorry :-)

Anyway, I was a bit surprised to find that one of my programs crashed
because of a call to STR_SEP.  Okay, not amazed because one of my
programs crashes, because contrary to popular belief they do that all
the time. :-) Apparently STR_SEP has been declared "obsolete" in IDL
5.3 and above.  A new built-in routine, STRSPLIT, is provided which
gives the same functionality, plus more.

I'm glad RSI keeps "improving" IDL.  The regular expression aspect of
IDL 5.3 looks tantalizing indeed.  However I would urge them to be
careful in the compatibility department.  The revelations about STRTOK
and the date routines make me wonder what's going on there.  Now that
STR_SEP may disappear too I am a bit more worried.  It's been around
since IDL 3 for goodness's sake!  It's entrenched!

In my case I eventually found my own error.  I misconfigured the
!PATH, so lib/obsolete wasn't included.  Thankfully STR_SEP is still
there, for now.

However, here's what I think should happen.  This is me tickling the
RSI secret agents who read this newsgroup. :-) See if people agree:

* when replacing one routine with another (as in the case of STR_SEP
  and STRSPLIT), leave STR_SEP.PRO as a wrapper routine.  Then at
  least people's programs won't break.  So far this is not an issue
  since STR_SEP is still included in the "obsolete" directory.

* when adding undocumented built-in functions, as for STRTOK, the
  function name should really have a unique prefix, like "RSI_".  This
  avoids name clashes with innocent user routines.

And generally, I would advise RSI that they should only break
functional IDL code if:

* it fixes a serious bug (would people classify the Y2K problem in the
  date routines as a serious bug? I'm not sure I do), or;

* in IDL 6.

Let's see if I have any sway with the big boys :-)

Craig


-- 
--------------------------------------------------------------------------
Craig B. Markwardt, Ph.D.         EMAIL:    craigmnet@cow.physics.wisc.edu
Astrophysics, IDL, Finance, Derivatives | Remove "net" for better response
--------------------------------------------------------------------------