comp.lang.idl-pvwave archive: archive » Re: VALUE

Home » Public Forums » archive » Re: VALUE_LOCATE and NaNs

Show: Today's Messages :: Show Polls :: Message Navigator
E-mail to friend

Re: VALUE_LOCATE and NaNs [message #81817 is a reply to message #81816]

Thu, 25 October 2012 07:59

lecacheux.alain
Messages: 325
Registered: January 2008

Senior Member

Le jeudi 25 octobre 2012 16:30:22 UTC+2, Craig Markwardt a écrit :
> On Thursday, October 25, 2012 5:14:47 AM UTC-4, Fab wrote:
>
>> Dear IDLers,
>
>>
>
>> I am quite an adept of using NaNs as missing values for my data, because
>
>> many IDL routines go along with them. However, some of the routines are
>
>> not really documented about how they handle NaNs.
>
>>
>
>> See the following example:
>
>> IDL> data = FINDGEN(10) & data[0] = !VALUES.F_NAN
>
>> IDL> p = VALUE_LOCATE(FINDGEN(10), data) & print, p[0]
>
>> -1
>
>> IDL> p = VALUE_LOCATE(INDGEN(10), data) & print, p[0]
>
>> 0
>
>> % Program caused arithmetic error: Floating illegal operand
>
>>
>
>>
>
>>
>
>> Which is quite dangerous! If I didn't debug my code to find the origin
>
>> of the Floating illegal operand warning, value_locate's output would be
>
>> wrong.
>
>>
>
>>
>
>>
>
>> Someone else than me thinking of this as a bug?
>
>
>
> NAN is neither greater than or less than any other finite number. For the purposes of VALUE_LOCATE(), there is no way to indicate "complete failure" other than the message you saw. I think the results of VALUE_LOCATE() are undefined, but this should be documented more explicitly in the reference documentation.
>
>
>
> The lesson is: you can't use NAN's with VALUE_LOCATE().
>
>
>
> Craig

In addition, there is one more inconsistency: 'value_locate' is considering NAN as the smallest value, while 'sort' is doing the opposite. Both being wrong as Craig said.
In IDL 8.2.1 and Win7, no warning/error messages are provided.
IDL> data=findgen(10) & data[1] = !Values.F_NAN
IDL> print,value_locate(findgen(10),data)
0 -1 2 3 4 5 6 7 8 9
IDL> print,sort(data)
0 2 3 4 5 6 7 8 9 1
alx.

Report message to a moderator

[Message index]

		Re: VALUE_LOCATE and NaNs By: Fabzi on Thu, 25 October 2012 08:07
		Re: VALUE_LOCATE and NaNs By: Jeremy Bailin on Thu, 25 October 2012 08:05
		Re: VALUE_LOCATE and NaNs By: wlandsman on Thu, 25 October 2012 08:01
		Re: VALUE_LOCATE and NaNs By: lecacheux.alain on Thu, 25 October 2012 07:59
		Re: VALUE_LOCATE and NaNs By: Craig Markwardt on Thu, 25 October 2012 07:30
		Re: VALUE_LOCATE and NaNs By: Fabzi on Thu, 25 October 2012 06:48
		Re: VALUE_LOCATE and NaNs By: Kai Muehlbauer on Thu, 25 October 2012 06:41
		Re: VALUE_LOCATE and NaNs By: lecacheux.alain on Thu, 25 October 2012 04:10
		Re: VALUE_LOCATE and NaNs By: Fabzi on Thu, 25 October 2012 02:57
		Re: VALUE_LOCATE and NaNs By: lecacheux.alain on Thu, 25 October 2012 02:34

Previous Topic:	random numbers with gamma distribution
Next Topic:	new graphics legend function: problems with symbol fill color

-=] Back to Top [=-

[ Syndicate this forum (XML) ] [

] [

]

Current Time: Sun Oct 12 03:53:16 PDT 2025

Total time taken to generate the page: 2.55916 seconds