How to reverse logical indexing?

Hi, let's say I have a data set A with NaN values that is a 3-D array.

B = ~isnan(A); AA = A(B);

Unfortunately now AA is a one-dimensional vector. How do I regain 3-D array from AA? There is a reshape function AAA = reshape(AA, X, Y, Z); but I cannot use this function because many elements got lost from logical indexing.

[Stephen23]

@June Yang: can you please tell what should happen in this example:

A = [1, 2
     3, 999];
Z = A(A<100)

What shape should Z be? If you think it should be square, what value should replace the 999?

Z = [1, 2
     3, ???];

Answer 1

[John D'Errico]

You cannot, based on only what remains after the operation. You threw away information. It is now gone. Therefore the result is returned as a column vector.

A = [1 NaN; 2 3];
B = ~isnan(A)
B =
  2×2 logical array
   1   0
   1   1

AA = A(B)
AA =
     1
     2
     3

Arrays in MATLAB MUST always be rectangular. One row cannot have more elements than another row.

The are somethings you could do, if perhaps you wanted to just replace the NaNs with zeros, or perhaps inf, that would be easy. But there are limits, and irregular arrays fall beyond those limits.

Answer 2

[Image Analyst]

You can't regain it from AA, but no problem - you still have A, so what's the problem?