delete rows depend on repeated value from 1 col

Hi I have a 20000x60 cell matrix I want to delete rows depend on repeated values from the first columns.

 s = {
  1  '2013-10-01'  '05:35:00'  '01:05:00'
  2  '2013-10-01'  '10:20:00'  '00:00:00'
  2  '2013-10-01'  '10:20:00'  '00:00:00'
  3  '2013-10-01'  '11:00:00'  '00:40:00'
  3  '2013-10-01'  '11:00:00'  '00:40:00'
  3  '2013-10-01'  '11:00:00'  '00:40:00'
  3  '2013-10-01'  '11:00:00'  '00:40:00'
  5  '2013-10-01'  '14:50:00'  '00:50:00'
  5  '2013-10-01'  '14:50:00'  '00:50:00'
  6  '2013-10-01'  '15:15:00'  '01:15:00'
  7  '2013-10-01'  '15:55:00'  '00:05:00'
  7  '2013-10-01'  '15:55:00'  '00:05:00'
  7  '2013-10-01'  '15:55:00'  '00:05:00'
  7  '2013-10-01'  '15:55:00'  '00:05:00'}

the result should like:

s1 = 

1  '2013-10-01'  '05:35:00'  '01:05:00'
2  '2013-10-01'  '10:20:00'  '00:00:00'
3  '2013-10-01'  '11:00:00'  '00:40:00'
5  '2013-10-01'  '14:50:00'  '00:50:00'
6  '2013-10-01'  '15:15:00'  '01:15:00'
7  '2013-10-01'  '15:55:00'  '00:05:00'
8  '2013-10-01'  '19:00:00'  '00:05:00'
9  '2013-10-01'  '19:10:00'  '00:10:00'

I use unique for it but because the different type of data there is Error using cell/unique (line 85) Input A must be a cell array of strings."

Answer 1

[Star Strider]

If the number in the first column is the same for all repeated values in the rest of the row, just use it.

Using your example data:

s = { 1  '2013-10-01'  '05:35:00'  '01:05:00'
      2  '2013-10-01'  '10:20:00'  '00:00:00'
      2  '2013-10-01'  '10:20:00'  '00:00:00'
      3  '2013-10-01'  '11:00:00'  '00:40:00'
      3  '2013-10-01'  '11:00:00'  '00:40:00'
      3  '2013-10-01'  '11:00:00'  '00:40:00'
      3  '2013-10-01'  '11:00:00'  '00:40:00'
      5  '2013-10-01'  '14:50:00'  '00:50:00'
      5  '2013-10-01'  '14:50:00'  '00:50:00'
      6  '2013-10-01'  '15:15:00'  '01:15:00'
      7  '2013-10-01'  '15:55:00'  '00:05:00'
      7  '2013-10-01'  '15:55:00'  '00:05:00'
      7  '2013-10-01'  '15:55:00'  '00:05:00'
      7  '2013-10-01'  '15:55:00'  '00:05:00'};

sc1 = cellfun(@(x)num2str(x, '%.0f'), s(:,1));
[s1u, ia] = unique(sc1);
s1 = s(ia,:)

s1 = 
    [1.0000e+000]    '2013-10-01'    '05:35:00'    '01:05:00'
    [2.0000e+000]    '2013-10-01'    '10:20:00'    '00:00:00'
    [3.0000e+000]    '2013-10-01'    '11:00:00'    '00:40:00'
    [5.0000e+000]    '2013-10-01'    '14:50:00'    '00:50:00'
    [6.0000e+000]    '2013-10-01'    '15:15:00'    '01:15:00'
    [7.0000e+000]    '2013-10-01'    '15:55:00'    '00:05:00'

You may have to modify this slightly if your actual cell array is different, but it works with the array you posted, and as I interpreted it.

[bero]

As I see it works for the sample data you test,but in my sample data from the attach file there is a error,could you please check the error

[Star Strider]

The only change necessary is to set ( 'UniformOutput',false ) in the ‘sc1’ assignment and it will work with your data:

sc1 = cellfun(@(x)num2str(x, '%.0f'), s(:,1), 'Uni',0);

Here, ( 'Uni',0 ) does the same thing.

[bero]

Ok, I need to sort the results based on first col 1-2-3... can i use : sortrows(s1,1) is it sort all results based on the first columns?? thx

[Star Strider]

With that change, the full code is now:

sc1 = cellfun(@(x)num2str(x, '%.0f'), s(:,1), 'UniformOutput',false);
[s1u, ia] = unique(sc1, 'stable');
s1 = s(ia,:);

That should do what you want.

[bero]

Great...Done could you please explain (%.0f)?? short comment about the code thanks

[Stephen23]

@bero: MATLAB's great documentation tells us exactly what '%.0f' does:

http://www.mathworks.com/help/matlab/ref/num2str.html#inputarg_formatSpec

[Star Strider]

The '%.0f' is one of a number of format descriptors (see the link Stephen provided for a full discussion of all of them), this one telling MATLAB to produce a string representing a floating-point number with no digits to the right of the decimal. So using that format descriptor, pi=3.1415... would print as 3 with no trailing decimal point. When you read the documentation, I leave it for you to determine the reason I chose this format rather than, for example, '%d'.

[Walter Roberson]

bero commented "good"

Answer 2

[MHN]

Consider this example (it is not the fastest and easiest way, but it solves your problem)

 s = {3 '2013-10-01' '11:00:00' '00:40:00'; 3 '2013-10-01' '11:00:00' '00:40:00'; 3 '2013-10-01' '11:00:00' '00:40:00'; 5 '2013-10-01' '14:50:00' '00:50:00'; 5 '2013-10-01' '14:50:00' '00:50:00'; 6 '2013-10-01' '15:15:00' '01:15:00'; 7 '2013-10-01' '15:55:00' '00:05:00'};
  col1 = cell2mat(s(:,1));
  uniqcol1 = union(col1,col1);
  % change all the repeated rows to '0' (or some number which is unused in the first column)  except the first one  and then remove them.
  for i=1:length(uniqcol1)
      for j = 1:size(s,1)
          if uniqcol1(i)==s{j,1}
              for k = j+1: size(s,1)
                  if uniqcol1(i)==s{k,1}
                      s{k,1} = 0;
                  end
              end
          end
      end
  end         
  col1 = cell2mat(s(:,1));
  s(col1==0,:)=[];

[bero]

It is so slowly for my my big matrix, I need more efficient and faster process...