Weird Error

[Diane Stokes]

Steven,

If you have access to the mask that was used to determine land vs ocean
for the sst file, you can create an appropriate descriptor file and use
that along with the maskout function to set the land (NaN's) to missing.

Or, you can try to make a mask based on the data values.  Eg:

  define mask=const(const(maskout(sst,sst-999999),-1),0,-u)

That generates a gridded field named mask that has 0's everywhere the
expression "sst-999999" is negative and -1 elsewhere.  (That expression
would be chosen to ensure you get a negative result from any expected
value from your original data field).  So, mask should have -1's where
you have your NaN's.  Then, you display only the valid ssts using the
maskout function.  Eg:

  d maskout(sst,mask)


Also, I was mistaken about more recent versions giving a more
informative msg.  The msg I was seeing was generated under different
circumstances.

     Diane

Steven C Chan wrote:
> I just checked the data file again.
>
> It is indeed a NaN value not a -9999.99 like number. The original file
> actually is created not for Grads -- it is just some binary file written
> in a way that is consistent with Grads order of x and y with the NaN
> inserted in.
>
> I am not sure if you can set a control file in a way to ignore NaN.
>
> About using set gxout grfill: It did shaded the SST correctly. But Grads
> now is also shading the NaN values (i.e. land) together with the
> correctly shaded SST.
>
> Steven Chan
>
> Mary Jo Nath wrote:
>> Steven,
>>
>> For SST data values over land should by set equal to the missing value,
>> which is not the same thing as "NaN". "NaN" means "not a number", some
>> value that isn't able to be interpreted as a number. These usually are
>> produced as the result of an invalid operation on floating point values
>> (like trying to take the square root of a negative number, for example.)
>>
>> MJ
>>
>> Steven C Chan wrote:
>>> Sorry for the slow response. I thank you everyone who had replied.
>>>
>>> The set ccols command is basically a cut and paste of a repeated color
>>> setting I have used over a lot of plots.
>>>
>>> The Grads built that is installed is 1.94b4. About the issues with NAN
>>> values, it is SST data, so there are a lot of undefined data points
>>> (obviously, over land).
>>>
>>> What would be the best solution for the problem?
>>>
>>> Steven Chan
>>>
>>> Diane Stokes wrote:
>>>> BTW, it would help if you include your build info with this type of
>>>> question.  In this case, I /think/ you are using an older build and a
>>>> more recent version might give you a more informative error msg.
>>>
>