| Re: A big BUG in PLOT, SURFACE, ... [message #4475] |
Wed, 21 June 1995 00:00 |
boswell
Messages: 7
Registered: October 1994
|
Junior Member |
|
|
In OpenVMS V6.1 and IDL V4.0, there are no error messages, but the negative
number is plotted positive. Making d(10)=1e15 instead results in the off-scale
value being plotted as though it were negative.
Jonathan Boswell
FDA/CDRH
|
|
|
|
| Re: A big BUG in PLOT, SURFACE, ... [message #4482 is a reply to message #4475] |
Fri, 16 June 1995 00:00  |
offenbrg
Messages: 31
Registered: August 1993
|
Member |
|
|
Philippe LATREMOLIERE <phlatrem@cnrs-orleans.fr> writes:
> Try this :
> d = FINDGEN(100)
> d(10) = -1e15
> PLOT, d, YRANGE = [0, 100]
> Don't you fing something strange ?. On my screen, the negative value is
> transformed to a positive one
> It is the same with SURFACE...
> Is it a new bug from the great RSI ?
Hm. Interesting.
The error does not appear when I run this on SUN SPARCstations running
Solaris 2.4 and SunOS 4.1.3. However, it does appear when I try it on a
VAXstation running OpenVMS.
Anyone see it on any other platforms?
Joel
--
"...And I am unanimous in this." - Mrs. Slocumbe
Joel D. Offenberg | offenbrg@fondue.gsfc.nasa.gov
Hughes STX, NASA/GSFC/LASP | (301)-286-5801
|
|
|
|
| Re: A big BUG in PLOT, SURFACE, ... [message #4483 is a reply to message #4482] |
Fri, 16 June 1995 00:00 |
theil
Messages: 2
Registered: June 1995
|
Junior Member |
|
|
Running on sun4 with SunOs 4.1.3 I get the arithmetic error but the
plot looks fine. when the yrange is large enough, the error goes away,
so the error probably occurs when IDL tries to calculate where on the
plot the -1e15 point should lie. I do not know what the plotting bug
looks like in the other OS's but from what I read, I guessed
that some number in the graphics interface with the OS wraps around
positive in the OS end when it goes too far negative in the IDL end.
I Tried the following experiment;
I defined d=dindgen(100).
I made a series of plots with d(10)=-1*double(10)^X, where X was a
number I steadily increased. The X value where the plots start
displaying improperly was 39, as in the first exponent which is beyond
the range of a 32 bit float! So I made a series of plots where
d(10)=-1*Y*double(10)^38 and incremented Y. I saw some plots with
d(10) very negative and some with d(10) very positive; but never less
than -3.40282E38. This is clearly some kind of wrap around problem
when you reach the limit of a float.
I decided that it is a probably a problem in IDL rather than the OS,
because when I did not specify the yrange, IDL would supply a yrange
that fit the DISPLAY and not the DATA. with out knowing details of
how the plotting routines work I can't be sure; perhaps IDL feeds info
to the graphics of what to plot; the graphics driver wraps around and
then tells IDL what is about to draw and then IDL comes up with
tick labels based on this. Perhaps there is some other sort of mischief
afoot. Regardless, the problem seems to occur at the limit of a float
on my system and probably the limit of a long 2147483647 on the pc's
etc.
MORAL: Don't stake your career on any plot if IDL complained while making
it! If IDL does complain, find some other meaningful way to scale/display
the data.
--
David Theil
campus box 389 Boulder, Colorado 80309 (303)492-0895
No guts, no glory.
|
|
|
|
|
|
| Re: A big BUG in PLOT, SURFACE, ... [message #4486 is a reply to message #4482] |
Fri, 16 June 1995 00:00 |
thompson
Messages: 584
Registered: August 1991
|
Senior Member |
|
|
"Rolf P. W\"urtz" <rolf> writes:
> Philippe LATREMOLIERE <phlatrem@cnrs-orleans.fr> wrote:
>> Try this :
>>
>> d = FINDGEN(100)
>> d(10) = -1e15
>> PLOT, d, YRANGE = [0, 100]
>>
>> Is it a new bug from the great RSI ?
>> --
> Well, it's not new. IDL 3.6.1 does the same. At least I get an error
> message, namely two instances of "Program caused arithmetic error:
> Floating overflow". Can't imagine what they're calculating there, though.
> BTW, does anybody collect bugs like this -- would be good to scan that list
> from time to time?
In OSF/1 v2.0 I don't have any problems with IDL v3.6.1. In IDL 4.0 I do get
the error messages, but the plot still comes out correctly. There certainly
seems to be an OS dependence.
Bill Thompson
|
|
|
|
| Re: A big BUG in PLOT, SURFACE, ... [message #4487 is a reply to message #4482] |
Fri, 16 June 1995 00:00 |
Rolf P. W\"urtz[1]
Messages: 1
Registered: June 1995
|
Junior Member |
|
|
Philippe LATREMOLIERE <phlatrem@cnrs-orleans.fr> wrote:
> Try this :
>
> d = FINDGEN(100)
> d(10) = -1e15
> PLOT, d, YRANGE = [0, 100]
>
> Is it a new bug from the great RSI ?
> --
Well, it's not new. IDL 3.6.1 does the same. At least I get an error
message, namely two instances of "Program caused arithmetic error:
Floating overflow". Can't imagine what they're calculating there, though.
BTW, does anybody collect bugs like this -- would be good to scan that list
from time to time?
Rolf
+---------------------------------+------------------------- ------------------+
| Rolf P. W"urtz | URL: http://www.cs.rug.nl/~rolf/rolf.html |
| Department of Computing Science | Email: rolf@cs.rug.nl |
| University of Groningen | |
| P.O. Box 800 | Phone: +31 50 63-6496 or |
| 9700 AV Groningen | -3939 (dept. secr.) |
| The Netherlands | Fax: +31 50 63-3800 |
+---------------------------------+------------------------- ------------------+
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
