How do I combine rows of a 2D cell array with 1D cell arrays of varying lengths?

I have a cell array of mixed datatypes, basically a table where I wanted the features of cell arrays

data(1,:) = {'sid001' 'status1' datetime(2020,1,1) datetime(2021,1,1)};
data(2,:) = {'sid002' 'status1' datetime(2020,1,1) datetime(2021,1,1)};
data(3,:) = {'sid003' 'status2' datetime(2020,1,1) datetime(2021,1,1)};

Now, I wish to append varying lengths of cell arrays to each row:

tempdata = {[1] [0] [1]}
data(1,:) = [data(1,:) tempdata]
% ...
tempdata = {[0] [0] [1] [1] [1]}
data(2,:) = [data(2,:) tempdata]
% ...
tempdata = {[1] [0] [0] [0]}
data(3,:) = [data(3,:) tempdata]

But this fails, since the size of the left side is smaller than that of the right side. What does work is this:

[n,m] = size(data);
ctemp = cell(n,5); % 5 is maximum number of columns added
data = [data ctemp];

tempdata = {[1] [0] [1]};
mtemp = length(tempdata);
data(1,m+1:m+mtemp) = tempdata;

tempdata = {[0] [0] [1] [1] [1]};
mtemp = length(tempdata);
data(2,m+1:m+mtemp) = tempdata;

tempdata = {[1] [0] [0] [0]};
mtemp = length(tempdata);
data(3,m+1:m+mtemp) = tempdata;

However, this seems like too much of a workaround, albeit I admit the idea is dirty in the first place.

Answer 1

[Peter Perkins]

Mixed data are what tables (or timetables) are for:

>> SID = ["sid001";"sid002";"sid003";"sid004";"sid005"];
>> Status = categorical(["Status1";"Status1";"Status2";"Status1";"Status2"]);
>> Date1 = datetime(2020,1,1:5)';
>> Date2 = datetime(2020,1,6:10)';
>> t = table(SID,Status,Date1,Date2)
t =
  5×4 table
      SID       Status        Date1          Date2   
    ________    _______    ___________    ___________
    "sid001"    Status1    01-Jan-2020    06-Jan-2020
    "sid002"    Status1    02-Jan-2020    07-Jan-2020
    "sid003"    Status2    03-Jan-2020    08-Jan-2020
    "sid004"    Status1    04-Jan-2020    09-Jan-2020
    "sid005"    Status2    05-Jan-2020    10-Jan-2020

A table is perfectly happy to contain a cell array, which is good because if your numeric vectors have different lengths, that's what you need:

>> t.Vectors = {[1 0 1]; [0 0 1 1 1]; [1 0 0 0]; [0]; []}
t =
  5×5 table
      SID       Status        Date1          Date2          Vectors    
    ________    _______    ___________    ___________    _____________
    "sid001"    Status1    01-Jan-2020    06-Jan-2020    {[    1 0 1]}
    "sid002"    Status1    02-Jan-2020    07-Jan-2020    {[0 0 1 1 1]}
    "sid003"    Status2    03-Jan-2020    08-Jan-2020    {[  1 0 0 0]}
    "sid004"    Status1    04-Jan-2020    09-Jan-2020    {[        0]}
    "sid005"    Status2    05-Jan-2020    10-Jan-2020    {0×0 double }

Note that you do NOT want t.Vectors to be a cell array of cell arrays.

Now you can do this:

>> t.Vectors{2}
ans =
     0     0     1     1     1
>> t.Vectors{2}(3)
ans =
     1

[Ulf Molich]

I will attempt to switch to tables, but I am unsure if I will have the same functionality available as I had with only cell arrays, like if I wish to lookup each value in Vectors in a specific row of table 3 based on a value in the current row of table 4? I guess I just have to add "t1." in front instead, then I will get my cell array functionality?

data3(strcmp(data3(:,1),data4(i,1)), find(dateMat(:,1)==data4{i,5}):-1:7)

Also, I wish to be able to sum the columns of the Vectors, ignoring rows without a value.

[Peter Perkins]

Yes. As far as the Vectors variable goes, there's not a lot of difference in using t.Vectors and in using your data{:,end} (or whatever). But for the other variables in the table, your life will be easier, for example, finding all the rows for Status2.

It'sa hard to say exactly what that kine would be, because it looks like you've changed things from your OP, and that line seems to have nothing to do with your vectors.

data3(:,1),data4(i,1) probably becomes something like t3.SiD==t4.SID(i)

find(dateMat(:,1)==data4{i,5}) probably becomes something like find(t3.Date1==t4.Date1(i))

[Ulf Molich]

Alright, thank you very much. You have given me the courage to use tables. They seem pretty neat!