idl 5.3, problem with plots and mapping environments

[Vapuser <vapuser(at)catspaw.jpl.nasa.gov>]

<rant>
!version={ mipseb IRIX unix 5.3 Nov 11 1999}

  I've just run into a strange problem with PLOTS and mapping
environments that seems to be new with IDL 5.3. I say this because I'm
not doing anything radically different from before and now I'm getting
this strange ... well... effect. You'll see in a moment why I hesitate
to be more specific in naming it.

  Do the following

  IDL> map_set,0,180,lim=[-10,100,60,290] ,/grid,/lab,/cont    
  IDL> plots,[100,200],[80,91]

   This will produce the message 

  % PLOTS: Value of Latitude is out of allowed range.

  Fine. I don't really have a problem with this. Yet.


  But, if you put this into a routine, with 'catch,error' code to try
  to catch the 'error,' you'll never will. At least, you won't on my
  SGI. I did

;------------------------------
pro junk2

  catch, error
  IF error NE 0 THEN BEGIN 
    Message,!error_state.msg,/cont
    stop
  ENDIF 
  map_set,0,180,lim=[-10,100,60,290] ,/grid,/lab,/cont    
  plots,[100,200],[80,91]

END

;------------------------------

  The error handling code is never called, running this puts you right
  back at the prompt at %MAIN.

  *However*

IDL> help,!error_state,/st
 
** Structure !ERROR_STATE, 7 tags, length=52:
   NAME            STRING    'IDL_M_BADARGVAL'
   BLOCK           STRING    'IDL_MBLK_CORE'
   CODE            LONG               -82
   SYS_CODE        LONG      Array[2]
   MSG             STRING    'PLOTS: Value of Latitude is out of allowed ra'...
   SYS_MSG         STRING    ''
   MSG_PREFIX      STRING    '% '

Apparently all errors are equal, but some errors are *more* equal.

  I also tried 'on_error' and 'check_math().' No dice.

  So we have a routine sending a message which is impossible to turn
  off because it never generates an error that can be caught with any
  error catching code! I find this *very* frustrating. You try to
  write code that only generates exceptions in, well, exceptional
  circumstances and here comes IDL creating a message which suggests
  that somethings wrong, but it won't let you find which vector of the
  hundreds or thousands is the problem!


  Now, why am I worried about plotting over the poles, you may ask.

  I make animations where I move vectors around in response to wind
  fields, static or dynamic. Occasionally I make them close to the
  poles and the vector which I want to draw may go over the
  pole. Here's the problem. I wrote the vector plotter to be
  independent of the environment, it doesn't know from mapping
  environments and, if I could possibly help it, I'd rather that it
  continued this way. When I call the plotter I don't know whether the
  vector will go over the pole. Up until now, IDL has been either
  smart enough to deal with the problem or blessedly unaware/silent
  about the existance of one. But now I have to find a way to
  determine that a mapping environemt is in place inside the plotter
  -- and a cursory, but fairly thorough examination of the !map
  environment variable revealed no obvious way to make this
  determination, or send in a keyword telling the plotter that it's in
  a mapping environment and please check for boundary cases like
  plotting over a pole. Which means I have to chase through all my
  code changing the calls.

  More annoying is the fact that IDL can wrap longitudes without any problem. For instance, you can do this:

  IDL> map_set,90,0,/stereo,/grid,/lab,/cont & plots,[90,-90],[30,30] 

  and this:

  IDL> map_set,90,0,/stereo,/grid,/lab,/cont & plots,[90,270],[30,30] 

  of event this

  IDL> map_set,90,0,/stereo,/grid,/lab,/cont & plots,[90,-90]+360,[30,30]  

  It's all the same to IDL.

  All that would be necessary to fix this problem is if the core of
  IDL was smart enough to do what I'll have to do anyway, take any
  reference to a latitude greater than 90, subtract 90 and change the
  longitude to the antipodal longitude. (mutatis mutandis for negative
  latitude). 
  

  Finally, the most annoying thing is that the vectors in question
  weren't  even *in* the scene, so IDL was complaining about vectors
  which would've been impossible to plot, even if it was smart enough
  to wrap the latitudes like it does the longitudes.


  *But at the very least* if they're going to send a message, there
  should be a way to turn it off by catching an error! What good is
  the message if you can't catch the error!

  And this is not a small thing for me. Each one of these vectors
  generates 1 to 8 messages, because I draw an arrow and any one of
  the endpoints can generate one of these errors. I draw thousands of
  arrows for each frame of an animation and I make animations that
  have between 60 and 800 frames. In one 60 frame animation, IDL
  produced 624 of these unstoppable error messages!

  It's this kind of *passive aggressive* help that I could very well
  do without.

  I understand that mapping is complicated, so I very well may be
  missing something that makes the solution to this sort of problem
  difficult if not impossible. If that's the case, though, I really
  wish someone would tell me what would make this fix hard. I've
  already implemented this code for the 'locations' of the vectors
  (defined as the midpoint), now I'm going to have to implement it for
  all endpoints as well. If I can do it, RSI can too.

</rant>

William
-- 
William Daffer: 818-354-0161: vapuser@catspaw.jpl.nasa.gov