comp.lang.idl-pvwave archive: archive » Re: Setting values to NaN

Home » Public Forums » archive » Re: Setting values to NaN

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

Re: Setting values to NaN [message #39926 is a reply to message #39925]

Fri, 25 June 2004 09:30

Michael Wallace
Messages: 409
Registered: December 2003

Senior Member

>> OK I spotted it. F_NAN and D_NAN apply to floats.
>>
>> So is there an alternative I can use with integers?
>
>
> How would you define a bit pattern for an invalid integer that's not a
> valid integer?, as opposed to some user-set value (like -99 or or -32767
> or -2147483647L etc.) ?
>
> You need to define some user-set value (like -99 or or -32767 or
> -2147483647L etc.) based on what you knwo about your data.

Just to expand on Paul's answer a little, computers work with floating
point numbers very differently than they work with integers. For
floating point numbers certain bit patterns are reserved for NaN,
positive infinity and negative infinity. If you want to see the gory
details, look up the IEEE 754 spec for floating point numbers.

The integer specification does not set aside special bit patterns for
NaN or the infinities. Therefore, if you want create a NaN for
integers, you have to create an arbitrary definition based on your data.
Typically, a good rule of thumb is to use the smallest or largest
possible integers as these will most likely be outside your data set.
The problem is that since your definition will be arbitrary you need to
make sure that if others use your data you indicate that there is a
number in the data which needs to be interpreted as NaN.

For example, let's say I have a data set of all positive numbers and I
use -1 to indicated NaN. If someone else comes along and doesn't know
that -1 represents NaN, they will do their processing using the -1
values and potentially get results that are very far off. I have run
into this very thing before. Someone used -999 to represent that no
data was present, but didn't document it. It took me the major part of
an afternoon to figure out why my code was failing so badly.

-Mike

Report message to a moderator

[Message index]

		Re: Setting values to NaN By: Craig Markwardt on Fri, 25 June 2004 13:17
		Re: Setting values to NaN By: b_gom on Fri, 25 June 2004 11:38
		Re: Setting values to NaN By: Matt Feinstein on Fri, 25 June 2004 09:34
		Re: Setting values to NaN By: Michael Wallace on Fri, 25 June 2004 09:30
		Re: Setting values to NaN By: Craig Markwardt on Fri, 25 June 2004 09:14
		Re: Setting values to NaN By: Paul Van Delst[1] on Fri, 25 June 2004 08:50
		Re: Setting values to NaN By: matt_westmore on Fri, 25 June 2004 08:27
		Re: Setting values to NaN By: David Fanning on Fri, 25 June 2004 06:43
		Re: Setting values to NaN By: Klaus Scipal on Fri, 25 June 2004 04:41
		Re: Setting values to NaN By: matt_westmore on Wed, 30 June 2004 04:37

Previous Topic:	inverse function of binary.pro
Next Topic:	Local solar time

-=] Back to Top [=-

[ Syndicate this forum (XML) ] [

] [

]

Current Time: Sat Oct 11 00:12:24 PDT 2025

Total time taken to generate the page: 0.16018 seconds