Unique rows of 2 cells with 2 columns each.

Hi

I have got 2 cells with 2 columns each:

IdxMembers(1636x2)

IdxMembers2(1630x2)

I am looking of the unique values of each cells in form of (nr. of unique values in first row x 2 columns)

I tried the following 2 codes:

unique(vertcat(IdxMembers,IdxMembers2),'rows')
and
unique(vertcat(IdxMembers,IdxMembers2))

I get the unique values in dim (nr. of unique values x 1 column). However, I would like get the output with the second column.

Do you have any suggestions?

[madhan ravi]

Mind sharing the data as .mat file??

[BdS]

Hi Madhan

Enclosed you will find my data. I changed the variable names:

IdxMembers=old

IdxMembers2=new

With this formula unique(vertcat(old,new)) I am still not getting a cell with dimensions (nr. of unique values in first row x 2 columns) even if variables old and new have 2 columns.

Do you have any suggestions?

[Guillaume]

Is the question resolved or not? You've accepted an answer but seem to still be asking something. if you're still asking something, what do you want as an output? A 1886x2 cell array whose first column is either 1 or 2 (n. of repeat) and 2nd column is the matching value from column2 of either input?

[BdS]

Hi Guilaume,

this question is not solved. I am looking for the following output:

a cell array with the first column containing the unique values and 2nd column with their corresponding values. cell(nr. unique values x 2)

Thank you

[Guillaume]

I understand "containing the unique values" as containing for example:

column1 = {'1 HK Equity';
           '10 HK Equity';
           '101 HK Equity';
           '1038 HK Equity';
           '11 HK Equity';
           ...
           };

I understand "nr. unique value" as the count of each unique value, so:

column1 = [2;
           2;
           1;
           2;
           ...
           ];

so I'm confused as to what you want.

[BdS]

I am sorry for the confusion.

I mean as your wrote in your first example. But with the corresponding value in column 1

Thanks

[BdS]

The unique values should be based on the information based on column 1

Answer 1

[Matt J]

I tried the following 2 codes: unique(vertcat(IdxMembers,IdxMembers2),'rows')... I get the unique values in dim (nr. of unique values x 1 column).

No, you really should have gotten 2 column output, as in the example below. Something about your situation must be different from what you are describing.

>> [IdxMembers,IdxMembers2] =deal( randi(3,4,2)  , randi(3,3,2))

IdxMembers =

     2     1
     2     2
     3     2
     3     2


IdxMembers2 =

     3     3
     3     2
     1     1

>> unique(vertcat(IdxMembers,IdxMembers2),'rows')

ans =

     1     1
     2     1
     2     2
     3     2
     3     3

Answer 2

[Guillaume]

merged = [new; old];  %simpler way to write vertcat(new, old)
[~, row] = unique(merged(:, 1));  %get row index of unique values in the 1st column
result = merged(row, :)  %extract these rows

[BdS]

Thank you. It worked. But why doesn't it work with just the code given by Matt J? I have other examples in which the coded worked well.

[Guillaume]

The 'rows' option of unique does not work with cell arrays. It gives you a warning if you try to do it (in my opinion, it should throw an error):

>> result = unique([new; old], 'rows');
Warning: The 'rows' input is not supported for cell array inputs. 
> In cell/unique>celluniqueR2012a (line 236)
  In cell/unique (line 148) 

However, the 'rows' option does work with string arrays, so another option would be to convert your cell arrays of char vectors into string arrays (requires R2016b or later):

result = unique(string([new; old]), 'rows');

You can then keep working with the string array or if you really want to convert back to cell array:

resultcell = cellstr(result);

[BdS]

thank you for your explanation.

Option 1 seems to more efficient: 0.007129 seconds in comparison with 0.028908

[Guillaume]

Option 1 seems to more efficient: 0.007129 seconds

As long as you consider efficiency to be just the execution speed and that you care about a few milliseconds difference.

Indeed string arrays may have a slight speed impact but if you consider that if you'd used string arrays to start with your original code would have worked straight away but instead it took you 4 days to resolve the problem, in term of development time string arrays would have won hands down.