checking for nan's in 2d numpy arrays

Question

I am working on a small piece of code that starts with an interpolated surface I made previsouly. The interpolation filled in gaps in the surface with nan's. Part of my processing involves looking at a local window around a particular point, and calculating some measures using the local surface. I would ideally like this code to only be able to do any calculations if the entire local surface does not contain nan values. The code iterates through the original large surface and checks to see if the local window about a point has a nan.

I know this is not the most efficent way to go about doing it, time-efficiency is not something I have to worry about.

Here is what I have so far:

for in in range(startz,endx):
    imin = i - half_tile
    imax = i + half_tile +1   

    for j in range(starty,endy):
        jmin = i - half_tile
        jmax = i + half_tile +1 

        #Test the local surface for nan's
        z = surface[imin:imax,jmin:jmax]
        Test = np.isnan(sum(z))

        #conditional statement
        if Test:
            print 'We have a nan'
            #set measures I want to calculate to zero

        else:
            print 'We  have a complete window'
            #do a set of calculations

the variable surface is the interpolated surface I created originally. The half_tile variables are just defining the size of the local window I want to use. The startx,endx,starty,endy are defining the size of the original surface to iterate through.

Where I am running into issues is that my conditional statement doesn't seem to be working. It will tell me that the local window that I am evaluating doesn't have any nan's in it, but then the rest of my code (which I didn't show here) will not work because it says there are nan's in the array.

An example of this might be:

 [[ 7.07494104  7.04592032  7.01689961  6.98787889  6.95885817  6.92983745
  6.90081674  6.87179602  6.8427753   6.81375458  6.78473387  6.75571315
  6.72669243]
 [ 7.10077447  7.07175376  7.04273304  7.01371232  6.98469161  6.95567089
 6.92665017  6.89762945  6.86860874  6.83958802  6.8105673   6.78154658
 6.75252587]
 [ 7.12660791  7.09758719  7.06856647  7.03954576  7.01052504  6.98150432
 6.9524836   6.92346289  6.89444217  6.86542145  6.83640073  6.80738002
  6.7783593 ]
  [ 7.15244134  7.12342063  7.09439991  7.06537919  7.03635847  7.00733776
 6.97831704  6.94929632  6.9202148   6.89105825  6.86190169  6.83274514
 6.80358859]
 [ 7.17804068  7.14888413  7.11972758  7.09057103  7.06141448  7.03225793
  7.00310137  6.97394482  6.94478827  6.91563172  6.88647517  6.85731862
      nan]]

Here is an example of the local window that my code is evaluating. In my code this would be z. The entire array has good values except for the last value which is a nan.

The "checking" function in my code is not picking up that there is a nan in the array. The conditional statement is returning a false when it should be a true to indicate that there is a nan present. I am missing anything fundamental in the way I am checking the array? or are my methods just totally wrong?

Answer 1

isnan() returns an array with true or false for each element in array. you need np.any() in addition to isnan(). see below example

import numpy as np
a = np.array([[1,2,3,4],[1,2,3,np.NaN]])
print np.isnan(a)
print np.any(np.isnan(a))

results in

[[False False False False]
 [False False False  True]]
True