Replace Nan in array with the previous value/text/number

function A = fill_nans(A)
% Replaces the NaN in each column with 
% previous non-NaN values.

for i = 1:size(A,1)
   I = A(1,i)
   for j = 2:size(A,2)
    if cellfun(@isnan,A(i,j))
        A(i,j) = I
    else
        I  = A(i,j)
    end
   end
end

I use this function 


data={'Robert',NaN,5,6,7,8,NaN,NaN,NaN};
fill_nans(data)  %working ok

data2={'Robert','pai',NaN,5,6,7,8,NaN,NaN,NaN};
fill_nans(data2) %Not working error is

data3={'Robert',NaN,NaN,5,6,7,8,'pai',NaN,NaN,NaN};
fill_nans(data2) %Not working error is

Error using cellfun

Non-scalar in Uniform output, at index 1, output 1.

Set 'UniformOutput' to false.

Error in fill_nans (line 8)

if cellfun(@isnan,A(i,j))

Answer 1

[Stephen23]

Explanation: you forgot to consider what isnan returns with a non-scalar input (such as when you provide it with those character vectors) (hint: it returns a non-scalar logical array of the same size as the input array)

Solution: simply define an anonymous funciton that only return TRUE for numeric scalars (as well as being NaN) (i.e. the output of this function is always scalar):

fun = @(a) isnumeric(a)&&isscalar(a)&&isnan(a);
...
if cellfun(fun,A(i,j))
    ...
end

[babyelephant]

Thank you for this help.

[Rik]

Just for the record my suggestion:

function A = fill_nans(A)
% Replaces the NaN in each column with previous non-NaN values.
% If there is no non-NaN value, no change will happen.
fun=@(x) isnumeric(a)&&isscalar(a)&&isnan(a);
for col = 2:size(A,2)
    nan_ind=cellfun(fun,A(:,col));
    A(nan_ind,col)=A(nan_ind,col-1);
end
end