Inverting and reshaping matrix which includes NaNs

I have a 3056 x 29 matrix where the rows represent time and columns represent current speed at various heights above the seabed - giving a current speed profile (where column 1 is closest the seabed and 29 furthest away). As there is a tide, which affects the height of water above the seabed over time, I have replaced all values above a certain height (the sea surface) with NaN (as shown below). I would now like to effectively invert the current speed profile for each row (i.e. time step) so that the columns represent water current speed at various depths below the sea surface. I have tried a variety of inverting and reshaping functions but, as the matrix contains NaNs, when these are removed the matrix is irregularly sized - which Matlab cannot work with.

Any suggestions would be appreciated! Many thanks

Jamie

Answer 1

[Thorsten]

You can rearrange the matrix by reversing the order of the non-NaN elements in each row such that the first column represents the velocity closest to the surface:

A = [1 2 3 4 NaN NaN; 1 2 3 NaN NaN NaN; 1 2 NaN NaN NaN NaN]; % small sample matrix to test
idx = sum(isnan(A),2);
B = A;
for i=1:size(A,1)
   B(i,1:end-idx(i)) = A(i,end-idx(i):-1:1);
end

[Jamie Dollman]

Thanks Thorsten. That worked perfectly! All I've got to do now is understand how it works...!

[Thorsten]

It is hopefully not too hard to understand. The idx = sum(isnan(A),2) line determines the number of NaNs in a row. Because they all appear in the end, the non-NaNs are at positions 1:end-idx(i). You reverse the order of these numbers if you count down from end-idx(i) to 1, i.e., end-idx(i):-1:1. So the code flips all numbers in a row from left to right.