Joining data from different cell arrays in Matlab - matlab

Question

I have data in Matlab that is in cell array format with columns representing different items. The cell arrays have different columns, as in the following example:
a = {'A', 'B', 'C' ; 1, 1, 1; 2, 2, 2 }
a =
'A' 'B' 'C'
[1] [1] [1]
[2] [2] [2]
b = {'C', 'D'; 3, 3; 4, 4}
b =
'C' 'D'
[3] [3]
[4] [4]
I would like to be able to join the different cell arrays in the following manner:
c =
'A' 'B' 'C' 'D'
[1] [1] [1] [NaN]
[2] [2] [2] [NaN]
[NaN] [NaN] [3] [3]
[NaN] [NaN] [4] [4]
In the real example I have hundreds of columns and several rows, so creating a new cell array manually is not an option for me.

Answer 1

If you were willing to store your data in dataset arrays (or convert them to dataset arrays for this purpose), you could do the following:
>> d1
d1 =
A B C
1 1 1
2 2 2
>> d2
d2 =
C D
3 3
4 4
>> join(d1,d2,'Keys','C','type','outer','mergekeys',true)
ans =
A B C D
1 1 1 NaN
2 2 2 NaN
NaN NaN 3 3
NaN NaN 4 4

Answer 2

I'm assuming you want to join the two arrays based on their first row only.
% get the list of all keys
keys = unique([a(1,:) b(1,:)]);
lena = size(a,1)-1; lenb = size(b,1)-1;
% allocate space for the joined array
joined = cell(lena+lenb+1, length(keys));
joined(1,:) = keys;
% add a
tf = ismember(keys, a(1,:));
joined(2:(2+lena-1),tf) = a(2:end,:);
% add b
tf = ismember(keys, b(1,:));
joined((lena+2):(lena+lenb+1),tf) = b(2:end,:);
This will give you the joined array except that it has empty cells instead NaNs. I hope this is OK.

Answer 3

Here is my solution adapted from an old another to a similar question (simply transpose rows/columns):
%# input cell arrays
a = {'A', 'B', 'C' ; 1, 1, 1; 2, 2, 2 };
b = {'C', 'D'; 3, 3; 4, 4};
%# transpose rows/columns
a = a'; b = b';
%# get all key values, and convert them to indices starting at 1
[allKeys,~,ind] = unique( [a(:,1);b(:,1)] );
indA = ind(1:size(a,1));
indB = ind(size(a,1)+1:end);
%# merge the two datasets (key,value1,value2)
c = cell(numel(allKeys), size(a,2)+size(b,2)-1);
c(:) = {NaN}; %# fill with NaNs
c(:,1) = allKeys; %# available keys from both
c(indA,2:size(a,2)) = a(:,2:end); %# insert 1st dataset values
c(indB,size(a,2)+1:end) = b(:,2:end); %# insert 2nd dataset values
Here is the result (transposed to match original orientation):
>> c'
ans =
'A' 'B' 'C' 'D'
[ 1] [ 1] [1] [NaN]
[ 2] [ 2] [2] [NaN]
[NaN] [NaN] [3] [ 3]
[NaN] [NaN] [4] [ 4]
Also here is the solution using the DATASET class from the Statistics Toolbox:
aa = dataset([cell2mat(a(2:end,:)) a(1,:)])
bb = dataset([cell2mat(b(2:end,:)) b(1,:)])
cc = join(aa,bb, 'Keys',{'C'}, 'type','fullouter', 'MergeKeys',true)
with
cc =
A B C D
1 1 1 NaN
2 2 2 NaN
NaN NaN 3 3
NaN NaN 4 4