lines from the corner

[Adam Sobel]

Matt,

thanks.  I managed to fix the problem myself by finding an easy way to
recreate
the file so that it had no NaNs.  Your patch might be useful at some
point ... but
since my problem is solved I think there is no need for you to go to the
trouble,
if it becomes important later I'll remember and bug you then.

thanks
Adam

Matthias Munnich wrote:

>Hi Adam:
>
>A while ago I  patched GrADS to replace NaN's, Inf and -Inf internally
>by  missing value. I had forwarded a patch to COLA for inclusion in the
>main distribution.
>
>I thought it was put in. But now I remember that Brian was hesitant as
>it uses some bit-pattern trickery around IEEE binary numbers to make it
>work for big- and little-endian numbers on both 32 and 64 bit machines.
>
>I can put together a patch and send it to you if you like.  Getting rid
>of NaN's can be awkward as  IEEE standards require tests  like "x==NaN"
>to return NaN no matter what.
>
>Matt
>
>
>Adam Sobel wrote:
>
>
>
>>Diane,
>>
>>thanks a lot.  You are right.  The file has NaNs in it.  Looking over
>>old posts on this
>>problem I was hoping to find a way to treat NaN as a missing value
>>without having to
>>actually go and rewrite the file.  So far it looks like this is not
>>doable, but if I'm wrong
>>maybe someone could let me know how to do that.
>>
>>thanks
>>Adam
>>
>>Diane Stokes wrote:
>>
>>
>>
>>>Adam,
>>>
>>>Looking closely at your plot, I wonder if you have bad values at some
>>>coastal points.
>>>
>>>I suggest you zoom in on one of those areas and use gxout grid or
>>>print to check that you have good values or the correct "missing"
>>>value at all points.
>>>
>>>If that's not the cause, what version of grads are you running?  A
>>>user here encountered similar problems (with presumably good data) but
>>>his plot was fine when he switched to version 1.9b4.
>>>
>>>   Diane
>>>
>>>
>>>Adam Sobel wrote:
>>>
>>>
>>>
>>>>Hello,
>>>>
>>>>I have a data file (netcdf, opens fine with sdfopen) which when I try
>>>>to make a contour plot, gives me straight lines from the corner which
>>>>cover up the plot, which otherwise would look fine. Example attached.
>>>>I suspect it has something to do with the fact that this data is all
>>>>missing over the land (it is an ocean-only data set) but other than
>>>>that can't see what's wrong. Changing the contour interval doesn't
>>>>help. I found another post on this behavior in the archive but no one
>>>>seems to have answered that one (a few years ago). Any insight
>>>>appreciated.
>>>>
>>>>thanks
>>>>Adam
>>>>
>>>>--
>>>>Adam H. Sobel
>>>>Department of Applied Physics and Applied Mathematics
>>>>Department of Earth and Environmental Sciences
>>>>Columbia University
>>>>500 W. 120th Street, Room 217
>>>>New York, NY 10027
>>>>tel: 212-854-6587
>>>>fax: 212-854-8257
>>>>e-mail: ahs129 at columbia.edu
>>>>web: http://www.columbia.edu/~ahs129/home.html
>>>>
>>>>
>>>>------------------------------------------------------------------------
>>>>
>>>>
>>>>
>>>>
>>--
>>Adam H. Sobel
>>Department of Applied Physics and Applied Mathematics
>>Department of Earth and Environmental Sciences
>>Columbia University
>>500 W. 120th Street, Room 217
>>New York, NY 10027
>>tel: 212-854-6587
>>fax: 212-854-8257
>>e-mail: ahs129 at columbia.edu
>>web: http://www.columbia.edu/~ahs129/home.html
>>
>>

--
Adam H. Sobel
Department of Applied Physics and Applied Mathematics
Department of Earth and Environmental Sciences
Columbia University
500 W. 120th Street, Room 217
New York, NY 10027
tel: 212-854-6587
fax: 212-854-8257
e-mail: ahs129 at columbia.edu
web: http://www.columbia.edu/~ahs129/home.html