faster function than unique for cell arrays

Hi,

I have a slow loop mainly because i use unique(A) where A is a cell array (these are the profiler's analysis). I am wondering if there is a faster unique function for cell arrays.

Answer 1

[Jan]

It depends on the contents of the cell. If you are talking of a cell string, this could be faster for short (< 1000) elements:

function [AA, AI, BI] = CStrUnique(A)
nA = numel(A);
if nA > 1
   [As, SV] = sort(A(:));

     if nargout < 3
        UV(SV)  = [1; strcmp(As(2:nA), As(1:nA - 1)) == 0];
        AI      = find(UV);
     else  % Indices requested:
        UV      = [1; strcmp(As(2:nA), As(1:nA - 1)) == 0];
        UVs(SV) = UV;
        AI      = find(UVs);

        % Complex creation of BI so that AA(BI) == A:
        v      = zeros(1, nA);
        v(AI)  = 1:length(AI);    % Sequence related to AA
        vs     = v(SV);           % Sorted like A
        vf     = vs(find(vs));    %#ok<FNDSB> % Just the filled entries
        BI(SV) = vf(cumsum(UV));  % Inflate multiple elements
     end
  elseif nA  % Comparison of subsequent elements fails for nA == 1
     AI = 1;
     BI = 1;
  else
     AI = [];
     BI = [];
  end

AA = A(AI);

It does not change the sorting order in opposite to UNIQUE. A C-Mex function could be even faster. But before further speculations, it would be helpful if you specify the type and size of the input at first.

A version for numerical input and stable sorting: https://www.mathworks.com/matlabcentral/answers/357818-interpolation-vector-with-a-lot-of-duplicate-values-code-on-r2010a#answer_282652