choosing unique elements in the order that they appear in a column of a cell vector

 Dear all,

I have the following cell matrix

 Out={    
'MN'    'AER_KL1'           'Bdrerv_1'  
'MN'     'AER_KL1'         'Bdrerv_1'  
'MN'     'AER_KL1'       'Bdrerv_1'  
'MN'    'AER_KL1'       'Bdrerv_1'  
'MN'     'AER_KL1'         'ksitre +2er'     
'MN'      'AER_KL1'       'ksitre +2er'     
'MN'  'AER_KL1'           'ksitre +2er'     
'MN'   'AER_KL1'          'ksitre +2er'        
'MN'  'AER_KL1'           'ksitre +2er'     
'MN'  'AER_KL1'         'ksitre +2er'     
'MN'   'AER_KL1'           'ksitre +2er'     
'MN'     'AER_KL1'          'eorftr>2'     
'MN'   'AER_KL1'           'eorftr>2'     
'MN'    '100_pr'           'se 34'  
'MN'     '100_pr'            'se 34'  
'MN'     '100_pr'            'se 34'  
'MN'    '100_pr'           'se 34'  
'MN'     '100_pr'           'se 34'      
'MN'      '100_pr'          'se 34'  
'MN'  '100_pr'           'se 34'    
'MN'   '100_pr'           'fjgitow f12'   
'MN'  '100_pr'           'fjgitow f12'     
'MN'  '100_pr'           'fjgitow f12'     
'MN'   '100_pr'           'fjgitow f12'      
'MN'     '100_pr'           'fjgitow f12'     
'MN'   '100_pr'           '344 edd'   }

I want to simplify this matrix so as to have the following specific structure

 'MN'    'AER_KL1'           'Bdrerv_1'  
'MN'  'AER_KL1'           'ksitre +2er'     
'MN'     'AER_KL1'          'eorftr>2'     
'MN'    '100_pr'           'se 34'  
'MN'  '100_pr'           'fjgitow f12'    
'MN'   '100_pr'           '344 edd'   }

As you can see I want to select the unique elements from the last column in the order that appear in “out” and construct the above simplified matrix.

IS there any quick/efficient code for doing this?

 The above matrix is an example
My original data set is of dimension 25000 by 4. So a general code for doing this is greatly appreciated

Answer 1

[Jan]

If you have a modern version of Matlab, the unique command can be controlled by teh 'first' flag to reply the first index.

[Sean de Wolski]

@Jan, I believe you meant the 'stable' flag.

[Oleg Komarov]

Or 'stable'.

[idx,idx] = unique(Out(:,end),'stable')
Out(idx,:)

[Jan]

In the linked page of the documentation I find:

 [C,ia,ic] = unique(A,occurrence)
 If occurrence='first', unique returns the index of the first occurrence of each unique value (or row).

Therefore I meant 'first', although 'stable' works also.

[Sean de Wolski]

Well 'first' returns the first index, not necessarily 'stable' order; consider:

>> [uv,idx] = unique([3;2;3;1],'first')

Answer 2

[Thorsten]

You can to it "by feet"

 unique_rows = 1;
 for i = 2:size(Out, 1)
   equals_row = 0;
   for j = 1:numel(unique_rows)
     equal_row = strcmp(Out{i, 3}, Out{unique_rows(j), 3});
     if equal_row, break, end
   end
   if ~equal_row, unique_rows(end+1) = i, end
 end
 Out_simple = Out(unique_rows, :)