Recursive function not stopping

I'm running the following recursive function. By pausing the execution and checking the values, I find that the execution is happening and values are updated. I have a stopping condition and the workspace is even showing me the final matrix once I pause after sometime. But it's not exiting the function after doing that. It keeps on running. I'm not able to figure out why. I call the function using x_1=no_vel_miss(a);

function r = no_vel_miss( x )      %x is my matrix with all data-time and vel as columns
i=1;
flag=0;
flag=sum(isnan(x(:,2)));
 if flag==0
    r=x;                        %if there is no missing point, return the matrix
 else        
  while i<=length(x(:,1))         
    if isnan(x(i,2))
        for j=i:-1:1
             if x(j,2)==0
                 prezero=j;                %prezero is the nearest zero velocity before missing point
                 break;
             end
        end
        for j=i:length(x(:,1))
            if x(j,2)==0&&x(j+1,2)~=0
                postzero=j;                    %postzero is the nearest zero velocity after which a non-zero velocity occurs
                break;
            end
        end
        x_1=cat(1,x(1:prezero,:),x(postzero:length(x(:,1)),:));   %here I'm removing all that data near the missing point as mentioned above
        i=prezero;
        r=no_vel_miss(x_1); 
    else
        i=i+1;
    end 

    end
   end
  end

[Ameer Hamza]

What is the termination condition i.e. how do you decide that the loop recursion should finish?

[Geoff Hayes]

Nagesh - please provide a sample set of data that you are using. When I run your code I get errors because prezero and postzero are not defined. When I initialize with dummy values, I noticed that the input to no_vel_miss (which is x_1) has one more row than the initial input x to this function...so the recursive call is working on a data set that is larger than the initial input. Which, in my case, leads to a MATLAB error/warning of my maximum recursion limit being met.

[Dennis]

I think it might be your while loop that runs forever. If x(i,2) is NaN for any i, prezero will be equal or smaller than i. Later you set i to prezero so your loop will find the exact same NaN value again and repeat.

[Nagesh A P]

@Ameer Hamza, the loop should terminate when flag=0, i.e., no NaNs in the matrix x. @Geoff Hayes, the real data I'm using is a velocity -time data. The sample data can be a square wave with breaks in it, ie., some data missing. I want to remove the square which has missing data and the zeros in between this square and the next square. @Dennis every time the next iteration starts, the input matrix itself is changing. The nans associated with a particular prezero are deleted.

[Geoff Hayes]

Nagesh - can you provide an example input that exhibits the behaviour that you are observing? have you confirmed that the input to the recursive function is a subset of the data to the outer call?

[Dennis]

i might be missing something...but where?

while i<=length(x(:,1))         
  if isnan(x(i,2))         %assume this is true for i=5
      for j=i:-1:1         %runs from 5 to 1
           if x(j,2)==0    %might be true for j=4
               prezero=j;    %prezero=4         
               break;        %we leave for loop
           end
      end
      for j=i:length(x(:,1))             %runs from 5 to whatever
          if x(j,2)==0&&x(j+1,2)~=0      %needs to be true for same value else you get an error
              postzero=j;                %postzero gets a value
              break;                     %we leave another for loop
          end
      end
      x_1=cat(1,x(1:prezero,:),x(postzero:length(x(:,1)),:)); %we create x_1 - not sure what this is though
      i=prezero;           %i is set to 4 in this example
      r=no_vel_miss(x_1);   %we call a recursive function and get r 
%after this while repeats with i=4
%will enter else ->i=i+1;
%now i is 5 where did you change your matrix?
  else
      i=i+1;
  end

[Nagesh A P]

@Geoff Hayes, If my input is transpose[1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20;0 1 3 4 nan 0 1 2 5 2 0 nan 4 5 nan 0 0 3 6 0] with time and velocity representing the 2 columns respectively, my output should be transpose[1 5 6 7 8 9 10 15 16 17 18 19 20 ;0 0 1 2 5 2 0 0 0 3 6 0] @Dennis, each time the function is called, a different matrix is sent as input, so starting from prezero won't create a problem.

[Dennis]

Yes, a new function is called, but the 'old' function does not simply stop

Answer 1

[OCDER]

Is this what you want to do?

% 
% x = [1  2  3  4    5  6  7  8  9 10  11   12  13  14   15  16  17  18  19  20;
%      0  1  3  4  NaN  0  1  2  5  2   0  NaN   4   5  NaN   0   0   3   6   0]';
% 
% x = no_vel_miss(x)
% 
% x' = 
%      1  6  7  8  9 10 11 16 17 18 19 20
%      0  0  1  2  5  2  0  0  0  3  6  0

function x = no_vel_miss( x )   
NanIdx = find(isnan(x(:, 2)));
if isempty(NanIdx)
    return
end
DelLoc = zeros(size(x, 1), 1, 'logical');
for k = 1:length(NanIdx)
    PreZero = find(x(1:NanIdx(k)-1, 2) == 0, 1, 'last'); 
    if isempty(PreZero)
        PreZero = 1;
    end
    PostZero = find(x(NanIdx(k)+1:end, 2) == 0, 1) + NanIdx(k);
    if isempty(PostZero)
        PostZero = size(x, 1);
    end
    DelLoc(PreZero+1:PostZero-1) = 1;
end
x(DelLoc, :) = [];