How to join matching rows of cells in a particular column order

I have a cell array containtin strings and numbers. I'd like to combine rows that have matching key values, but I need the colums to stay in the same place. Most rows will have a single matching row- more than one match is not possible. Some cells in every row will have empty (0x0 double) values- these essentially will be overwritten by the data in the matching row column.

a={'201901' 'timeA' 123 []; ...
   '201901' 'timeB' [] 22; ...
   '201901' 'timeA' [] 456; ...
   '201902' 'timeA' 999 []; ...
   '201901' 'timeB' 11 []}

the output should be:

b={'201901' 'timeA' 123 456;...
   '201901' 'timeB' 11 22; ...
   '201902' 'timeA' 999 []}

Not sure if there is something like join for cells. Seem ls like unique(a(:,1:2),'rows') will not work on cells either.

Answer 1

[Matt J]

Not sure why you're working with cells, rather than tables, but it's easy enough to switch back and forth.

A=cell2table(a);

B=varfun(@(z){max([z{:}])}, A,'GroupingVariables',[1,2]);  %EDITED
B.GroupCount=[];

b=table2cell(B)

[Marcus Glover]

Thanks Matt. One odd thing is that it is inserting 1 in the last column where the empty cell was in the 'unmatched' row. I'd need that to remain blank when there is no value.

[Matt J]

Here's what I get,

>> A=cell2table(a);
B=varfun(@(z){max([z{:}])}, A,'GroupingVariables',[1,2]);
B.GroupCount=[];
b=table2cell(B)

b =

  3×4 cell array

    {'201901'}    {'timeA'}    {[123]}    {[     456]}
    {'201901'}    {'timeB'}    {[ 11]}    {[      22]}
    {'201902'}    {'timeA'}    {[999]}    {0×0 double}

Maybe you were using a previous edit ...

[Marcus Glover]

Thanks, works now.