Re: missing_value attribute

From: David Brown <dbrown_at_nyahnyahspammersnyahnyah>
Date: Tue Oct 11 2011 - 12:20:10 MDT

Hi Dave,
We certainly have no plans to remove the existing recognition and special handling for the missing_value attribute.
 -dave

On Oct 11, 2011, at 12:13 PM, Dave Allured wrote:

> Dennis,
>
> As of CF conventions 1.5, October 2010, the missing_value attribute
> is recognized as a valid alternate for _FillValue. _FillValue is
> preferred, but missing_value is no longer deprecated.
>
> http://cf-pcmdi.llnl.gov/documents/cf-conventions/1.5/ch02s05.html
>
> So it is important for CF as well as COARDS, that NCL continues to
> handle the missing_value attribute when reading Netcdf files, as you
> described in [5] below. Converting to _FillValue for NCL internal
> use is a reasonable way to do this.
>
> --Dave
>
> On 10/11/2011 6:23 AM, Dennis Shea wrote:
>> [1] _FillValue is what the CF convention states should be used
>> to indicate missing values. Further, Unidata says the same thing.
>> NCL recognizes _FillValue only.
>>
>> [2] missing_value is a COARDS convention attribute. It has no
>> special status under CF or within NCL.
>>
>> [3] _MissingValue, etc....., mean nothing to any known software
>>
>> [4] By rule, the type of the _FillValue should be the same as
>> the variable with which it is associated. However, there is
>> nothing in the netCDF software to prevent this from happening.
>>
>> [5] NCL does try to be 'helpful' in the following way. When a variable
>> is read and NCL sees it has a missing_value attribute but
>> no _FillValue, NCL will assign an virtual _FillValue' attribute.
>> NCL checks the type of the missing_value attribute and the type
>> of the associated variable. If they are the same, NCL creates
>> the _FillValue attribute; if the are different and NCL can
>> coerce the missing_value to the type of the variable, NCL will
>> do it. If not, you get the warning message indicated earlier.
>> This is *not* an NCL problem. The file creators 'screwed up'.
>> NCL can be used to address the issue but it must be done manually.
>>
>> [6] NCL can check for NaNs and replace the values.
>> http://www.ncl.ucar.edu/Document/Functions/Built-in/isnan_ieee.shtml
>>
>> http://www.ncl.ucar.edu/Document/Functions/Built-in/replace_ieeenan.shtml
> _______________________________________________
> ncl-talk mailing list
> List instructions, subscriber options, unsubscribe:
> http://mailman.ucar.edu/mailman/listinfo/ncl-talk

_______________________________________________
ncl-talk mailing list
List instructions, subscriber options, unsubscribe:
http://mailman.ucar.edu/mailman/listinfo/ncl-talk
Received on Tue Oct 11 12:20:17 2011

This archive was generated by hypermail 2.1.8 : Wed Oct 12 2011 - 15:59:45 MDT