Re: error in Ngl.ftcurv function

Hi,

I change the loop as,

for i in xrange(0, im):
        print i
        x = None
        lat[i,:] = Ngl.ftcurv(ctlpix[:], ctllat[i,:], npix[:])
        lon[i,:] = Ngl.ftcurv(ctlpix[:], ctllon[i,:], npix[:])

but it gives same error. my python version is 2.4.3. may be the problem
could be related with version of the python and its modules (such as,
numeric etc.) and i am using redhat as 3.0.

Ufuk

> Hmmmm. Works for me.
>
> Anyway, try the following. In the loop, before the ftcurv call, insert:
>
> x = None
>
> This may seem stupid, but the problem is that the Python object
> "None" is being garbage collected and when the low level code
> tries to decrease its reference count and that count is already
> zero, you get the error. The above should keep the reference
> count positive.
>
> Again, let me know if this does not work and we will try something
> else.
>
> Fred
>
> On Apr 25, 2007, at 12:09 AM, [BE] Ufuk Utku Turuncoglu wrote:
>
>> Hi,
>>
>> It is not working,
>>
>> 1329
>> 1330
>> Fatal Python error: deallocating None
>> Abort (core dumped)
>>
>> it is terminating in 1330 step. So i try to interpolate the
>> variables in two seperate step. first step from 0 to 1329 and then
>> 1330 to 2040 but the problem still exist. The modified part as
>> follows,
>>
>> for i in xrange(0, 1329):
>> print i
>> lat[i,:] = Ngl.ftcurv(ctlpix[:], ctllat[i,:], npix[:])
>> print "------"
>> for i in xrange(1329, im):
>> print i
>> lat[i,:] = Ngl.ftcurv(ctlpix[:], ctllat[i,:], npix[:])
>>
>> and the output is,
>>
>> 1327
>> 1328
>> ------
>> 1329
>> 1330
>> Fatal Python error: deallocating None
>> Abort (core dumped)
>>
>> so, it is not working in this case.
>>
>> best wishes
>>
>> ufuk
>>
>> Fred Clare wrote:
>>> Hi Ufuk,
>>>
>>> I think I have found the problem. Unfortunately it is in the C
>>> extension
>>> module and will require rebuilding the package. As a temporary
>>> workaround, try splitting up your loop as:
>>>
>>> for i in xrange(0, im):
>>> lat[i,:] = Ngl.ftcurv(ctlpix[:], ctllat[i,:], npix[:])
>>> for i in xrange(0, im):
>>> lon[i,:] = Ngl.ftcurv(ctlpix[:], ctllon[i,:], npix[:])
>>>
>>> Let me know if that does not work for you.
>>>
>>> Fred Clare
>>>
>>> On Apr 24, 2007, at 8:15 AM, [BE] Ufuk Utku Turuncoglu wrote:
>>>
>>>> Hi,
>>>>
>>>> I try to plot the modis sst data and i need to interpolate
>>>> coordinate variables. So i write following script based on an IDL
>>>> script which is found in,
>>>>
>>>> http://seadas.gsfc.nasa.gov/forum/oceancolor/topic_show.pl?tid=758
>>>>
>>>> I am using Ngl.ftcurv function for interpolation but it gives me
>>>> following error,
>>>>
>>>> Fatal Python error: deallocating None
>>>> Abort (core dumped)
>>>>
>>>> so, i does not have any detailed information. I check the memory
>>>> usage of my script but it is around %2-%3 of the total memory.
>>>> The part of my script as follows,
>>>>
>>>> hdf = Nio.open_file("../sst/A2002232102000.L2_LAC_SST.hdf", "r")
>>>>
>>>> sst = hdf.variables['sst']
>>>> ctllat = hdf.variables['latitude']
>>>> ctllon = hdf.variables['longitude']
>>>> ctlpix = hdf.variables['cntl_pt_cols']
>>>>
>>>> im = sst.shape[0]
>>>> jm = sst.shape[1]
>>>> lat = zeros((im, jm), Float)
>>>> lon = zeros((im, jm), Float)
>>>> npix = zeros(jm, Int)
>>>>
>>>> for i in xrange(0, jm):
>>>> npix[i] = i
>>>>
>>>> for i in xrange(0, im):
>>>> print i
>>>> lat[i,:] = Ngl.ftcurv(ctlpix[:], ctllat[i,:], npix[:])
>>>> # lon[i,:] = Ngl.ftcurv(ctlpix[:], ctllon[i,:], npix[:])
>>>>
>>>> so, i have also interpolate the long valuse using last line of
>>>> the script. but in this case it terminates in i=650 in this case
>>>> it terminates the i=1350 the total processed i values is 2040.
>>>> But it is not memory related problem as i mention before. The
>>>> dimensions of the ctlpix, ctllon and npix is,
>>>>
>>>> ctlpix = (170,)
>>>> ctllon = (2040, 170)
>>>> npix = (1354,)
>>>>
>>>> Best regards,
>>>>
>>>> Ufuk Utku Turuncoglu
>>>> Istanbul Technical University
>>>> Informatics Institute
>>>> HPC Lab
>>>>
>>>>
>>>>
>>>> _______________________________________________
>>>> pyngl-talk mailing list
>>>> pyngl-talk_at_ucar.edu
>>>> http://mailman.ucar.edu/mailman/listinfo/pyngl-talk
>>
>

_______________________________________________
pyngl-talk mailing list
pyngl-talk_at_ucar.edu
http://mailman.ucar.edu/mailman/listinfo/pyngl-talk
Received on Wed Apr 25 2007 - 13:07:53 MDT