| NaN problems [message #43584] |
Thu, 21 April 2005 06:37  |
David Lopez Pons
Messages: 8
Registered: July 2003
|
Junior Member |
|
|
Hi everybody.
I have a problem detecting NaN values. When I write:
err=findgen(10)-5
res=err^2.5
print,res
NaN NaN NaN NaN NaN
0.00000 1.00000 5.65685 15.5885 32.0000
Until here everything is OK, but now i want to look for Nan values in my
result so I write:
print,where(res EQ !values.f_nan)
-1
I can see the NaN values in res. Why they are not equal to !values.f_nan?
|
|
|
|
| Re: NaN problems [message #43703 is a reply to message #43584] |
Sat, 23 April 2005 12:48  |
R.Bauer
Messages: 1424
Registered: November 1998
|
Senior Member |
|
|
David Fanning wrote:
> Y.T. writes:
>
>> I don't have access to IDL on a win-box, but I have a (potentially
>> stupid) question: On that page of yours where you write
>>
>> IDL> print, array ne array
>> 0 0 0 0 0
>>
>> have you tried the converse, i.e.
>>
>> IDL> print, array eq array
>>
>> ?
>>
>> I am asking since, while it is true that a NaN is not equal to a NaN it
>> does not seem to follow (to my wrinkled brain) that a NaN is
>> necessarily UNequal to a NaN.
>>
>> In your version you test for the latter, and I'm kinda curious what
>> would happen if you tried the former...
>
> Well, looks like some things have changed since that
> article was written. Here is what I get in IDL 6.1 on Windows:
>
> IDL> a = [ 1.0, 2.0, !Values.F_NAN, 4.0, !Values.F_NAN ]
> IDL> print, a ne a
> 0 0 1 0 1
> % Program caused arithmetic error: Floating illegal operand
> IDL> print, a eq a
> 1 1 0 1 0
> % Program caused arithmetic error: Floating illegal operand
>
> Consistent, anyway. :-)
>
> Cheers,
>
> David
I got the same result without the warnings on the linux Version.
Probably we both should try it with the actual 6.1.1 version. For windows
it's a 2MB download and I don't know why it is for linux 60MB. ;-)
cheers
Reimar
--
Forschungszentrum Juelich
email: R.Bauer@fz-juelich.de
http://www.fz-juelich.de/icg/icg-i/
============================================================ ======
a IDL library at ForschungsZentrum Juelich
http://www.fz-juelich.de/icg/icg-i/idl_icglib/idl_lib_intro. html
|
|
|
|
| Re: NaN problems [message #43725 is a reply to message #43584] |
Fri, 22 April 2005 20:38 |
David Fanning
Messages: 11724
Registered: August 2001
|
Senior Member |
|
|
Y.T. writes:
> I don't have access to IDL on a win-box, but I have a (potentially
> stupid) question: On that page of yours where you write
>
> IDL> print, array ne array
> 0 0 0 0 0
>
> have you tried the converse, i.e.
>
> IDL> print, array eq array
>
> ?
>
> I am asking since, while it is true that a NaN is not equal to a NaN it
> does not seem to follow (to my wrinkled brain) that a NaN is
> necessarily UNequal to a NaN.
>
> In your version you test for the latter, and I'm kinda curious what
> would happen if you tried the former...
Well, looks like some things have changed since that
article was written. Here is what I get in IDL 6.1 on Windows:
IDL> a = [ 1.0, 2.0, !Values.F_NAN, 4.0, !Values.F_NAN ]
IDL> print, a ne a
0 0 1 0 1
% Program caused arithmetic error: Floating illegal operand
IDL> print, a eq a
1 1 0 1 0
% Program caused arithmetic error: Floating illegal operand
Consistent, anyway. :-)
Cheers,
David
--
David Fanning, Ph.D.
Fanning Software Consulting, Inc.
Coyote's Guide to IDL Programming: http://www.dfanning.com/
|
|
|
|
|
|
|
|
| Re: NAN Problem [message #69392 is a reply to message #43584] |
Thu, 14 January 2010 00:49 |
R.Bauer
Messages: 1424
Registered: November 1998
|
Senior Member |
|
|
David Fanning schrieb:
> Folks,
>
> Consider this sequence of commands on my IDL 7.0.1
> LINUX version.
>
> IDL> a = Findgen(11)
> IDL> a[0] = !Values.F_NAN
> IDL> Print, Min(a), Max(a)
> NaN NaN
> IDL> b = Findgen(11)
> IDL> b[1] = !Values.F_NAN
> IDL> Print, Min(b), Max(b)
> 0.00000 10.00000
>
> What do you make of that? This is giving me a great
> deal of touble today while trying to elminate bad
> values from an image. :-(
>
> I have the same results with my Windows IDL 7.1.2
> version, except I do get a warning about a floating
> illegal operand with variable b, which I don't get
> on LINUX.
>
> Cheers,
>
> David
>
>
>
Hi David
For some reason the nan keyword was introduced
NAN
Set this keyword to cause the routine to check for occurrences of the
IEEE floating-point values NaN or Infinity in the input data. Elements
with the value NaN or Infinity are treated as missing data. (See Special
Floating-Point Values (Application Programming) for more information on
IEEE floating-point values.)
IDL> print, min(a,/nan)
1.00000
IDL> print, min(b,/nan)
0.00000
I try to avoid nan values.
Because not all operators do have the possibility to set nan operations
by a keyword. Frank has added some
http://www.fz-juelich.de/icg/icg-1/idl_icglib/idl_source/idl _work/fh_lib/f_ge.pro
functions because he likes nan.
cheers
Reimar
|
|
|
|
| Re: NAN Problem [message #69394 is a reply to message #43584] |
Wed, 13 January 2010 16:15 |
penteado
Messages: 866
Registered: February 2018
|
Senior Member
Administrator |
|
|
On Jan 13, 10:00 pm, "R.G. Stockwell" <noem...@please.com> wrote:
> i swear i used to get nan as a the result of that.
> I do get a :
> % Program caused arithmetic error: Floating illegal operand
>
> after that min(b) call. But now I am worried about how
> invalid that result is (and how diligent i have been in using
> nan keywords).
Same here. It is probably one of the reasons I once had some
difficulty writing a reader/writer to a file format (ISIS cubes, in
case anyone is interested) which has several special values, that
could be translated to NaN.
|
|
|
|
| Re: NAN Problem [message #69395 is a reply to message #43584] |
Wed, 13 January 2010 16:00 |
R.G.Stockwell
Messages: 163
Registered: October 2004
|
Senior Member |
|
|
"David Fanning" <news@dfanning.com> wrote in message
news:MPG.25b7d9955ac199e398969e@news.giganews.com...
> Folks,
>
> Consider this sequence of commands on my IDL 7.0.1
> LINUX version.
>
> IDL> a = Findgen(11)
> IDL> a[0] = !Values.F_NAN
> IDL> Print, Min(a), Max(a)
> NaN NaN
> IDL> b = Findgen(11)
> IDL> b[1] = !Values.F_NAN
> IDL> Print, Min(b), Max(b)
> 0.00000 10.00000
>
> What do you make of that?
HOLY CRAP!!!
i swear i used to get nan as a the result of that.
I do get a :
% Program caused arithmetic error: Floating illegal operand
after that min(b) call. But now I am worried about how
invalid that result is (and how diligent i have been in using
nan keywords).
cheers,
bob
|
|
|
|
| Re: NAN Problem [message #69398 is a reply to message #43584] |
Wed, 13 January 2010 13:00 |
Foldy Lajos
Messages: 268
Registered: October 2001
|
Senior Member |
|
|
On Wed, 13 Jan 2010, David Fanning wrote:
> Consider this sequence of commands on my IDL 7.0.1
> LINUX version.
>
> IDL> a = Findgen(11)
> IDL> a[0] = !Values.F_NAN
> IDL> Print, Min(a), Max(a)
> NaN NaN
> IDL> b = Findgen(11)
> IDL> b[1] = !Values.F_NAN
> IDL> Print, Min(b), Max(b)
> 0.00000 10.00000
>
> What do you make of that? This is giving me a great
> deal of touble today while trying to elminate bad
> values from an image. :-(
>
The first array element causes the difference. You can reproduce the
results:
min=a[0]
max=a[0]
for j=1,10 do begin
if a[j] lt min then min=a[j]
if a[j] gt max then max=a[j]
endfor
lt and gt are not defined for NaNs, they always return 0:
x=0.0
y=!Values.F_NAN
help, x lt y, x eq y, x gt y
<Expression> BYTE = 0
<Expression> BYTE = 0
<Expression> BYTE = 0
regards,
lajos
|
|
|
|
| Re: NAN Problem [message #69399 is a reply to message #43584] |
Wed, 13 January 2010 12:59 |
penteado
Messages: 866
Registered: February 2018
|
Senior Member
Administrator |
|
|
On Jan 13, 6:23 pm, David Fanning <n...@dfanning.com> wrote:
> Folks,
>
> Consider this sequence of commands on my IDL 7.0.1
> LINUX version.
>
> IDL> a = Findgen(11)
> IDL> a[0] = !Values.F_NAN
> IDL> Print, Min(a), Max(a)
> NaN NaN
> IDL> b = Findgen(11)
> IDL> b[1] = !Values.F_NAN
> IDL> Print, Min(b), Max(b)
> 0.00000 10.00000
>
> What do you make of that? This is giving me a great
> deal of touble today while trying to elminate bad
> values from an image. :-(
>
> I have the same results with my Windows IDL 7.1.2
> version, except I do get a warning about a floating
> illegal operand with variable b, which I don't get
> on LINUX.
Interesting. I found the same here, with 7.1.1 in Linux. I had never
noticed this behavior before.
I do not know how the min and max functions were implemented, but this
would be the result of one algorithms that could be used for it. Since
NaN does not compare larger or smaller than anything, the following
would produce the same result:
function min_test,a
res=a[0]
for i=1,n_elements(a)-1 do if a[i] lt res then res=a[i]
return,res
end
When the NaN is the first element, it will be set as the initial value
for the minimum, and since nothing is smaller, it will be kept as the
result. Conversely, if the first value is not NaN, no NaN value that
comes after will be smaller than it, so an NaN would not replace the
minimum.
|
|
|
|
| Re: NAN Problem [message #69400 is a reply to message #43584] |
Wed, 13 January 2010 12:39 |
David Fanning
Messages: 11724
Registered: August 2001
|
Senior Member |
|
|
jeanh writes:
> heu, hum.... the help for min and max says "Note
> If the MIN function is run on an array containing NaN values and the NAN
> keyword is not set, an invalid result will occur."
I was sort of hoping the "invalid result" would make
itself known to me. :-(
Cheers,
David
--
David Fanning, Ph.D.
Fanning Software Consulting, Inc.
Coyote's Guide to IDL Programming: http://www.dfanning.com/
Sepore ma de ni thue. ("Perhaps thos speakest truth.")
|
|
|
|
| Re: NAN Problem [message #69402 is a reply to message #43584] |
Wed, 13 January 2010 12:28 |
jeanh
Messages: 79
Registered: November 2009
|
Member |
|
|
David Fanning wrote:
> Folks,
>
> Consider this sequence of commands on my IDL 7.0.1
> LINUX version.
>
> IDL> a = Findgen(11)
> IDL> a[0] = !Values.F_NAN
> IDL> Print, Min(a), Max(a)
> NaN NaN
> IDL> b = Findgen(11)
> IDL> b[1] = !Values.F_NAN
> IDL> Print, Min(b), Max(b)
> 0.00000 10.00000
>
> What do you make of that? This is giving me a great
> deal of touble today while trying to elminate bad
> values from an image. :-(
>
> I have the same results with my Windows IDL 7.1.2
> version, except I do get a warning about a floating
> illegal operand with variable b, which I don't get
> on LINUX.
>
> Cheers,
>
> David
hi,
heu, hum.... the help for min and max says "Note
If the MIN function is run on an array containing NaN values and the NAN
keyword is not set, an invalid result will occur."
Jean
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
