IDL Q: How do I check for an NaN in an array, in a loop?

[jenny.genders]

So, I'm doing a loop through 612 rows of data, each row having 2592 numbers. If the number I'm looking at is NOT equal to NaN, I want to do a calculation using a different array. This way, if the element in sfc_data is not a number, the number in lat_matrix in that same column is ignored, and if the element in sfc_data IS a number, the matching number from lat_matrix is added to the weights array. However when I do the following loop, it's doing the calculation regardless of whether the number in the first array equals NaN.

for i = 0,611 do begin
for j = 0, 2591 do begin
if sfc_data[j,i] ne 'NaN' then weights = weights + lat_matrix[j,0]
endfor
endfor

The loop is doing the full number of calculations as if none of the numbers in sfc_data are NaN (there are definitely many NaN's in sfc_data). So how do I check if sfc_data[j,i] is equal to the string NaN and isn't actually a number?

 

[TylerH]

See the first answer here: http://stackoverflow.com/questions/570669/checking-if-a-double-or-float-is-nan-in-c

In short, float != float, iff float is NaN.

 

[AlephZero]

In short, float != float, iff float is NaN.

But make sure that a not-very-clever compiler doesn't optimize out the test x != x, because it thinks the result will always be false

 

[TylerH]

Good point. Add the "volatile" qualifier to the variables declaration.

 

[LudusRex]

To check for NaN in an array, you can use the IDL function "isnan" which returns a boolean array indicating which elements in the array are NaN. You can then use this boolean array as a mask to only perform the calculation on the elements that are not NaN.

Your loop would look something like this:

for i = 0, 611 do begin
for j = 0, 2591 do begin
if ~isnan(sfc_data[j, i]) then begin
weights = weights + lat_matrix[j, 0]
endif
endfor
endfor

This way, the calculation will only be performed if the element in sfc_data is not NaN. You can also use the "isfinite" function to check for both NaN and infinity values.

Additionally, you can also use the IDL function "where" to find the indices of the elements that are not NaN and use those indices to perform the calculation. This can be more efficient for larger arrays.

for i = 0, 611 do begin
indices = where(~isnan(sfc_data[*, i]))
weights = total(lat_matrix[indices, 0])
endfor

I hope this helps!