matrix - Efficient ways to append new data in Matlab (with example code)

Question

I am looking for methods, built-in functions, good practices... to append new data to a matrix - when the rows and columns are not the same

The data I deal with is structured as follows:

A.values: Ta x Ma matrix of values
A.dates:  Ta x 1 vector of datenum
A.id:     1 x Ma cell array of ids

Now the challenge is how to deal with new (potentially overlapping) data B that I load in and would like to append to a new matrix C:

When new data comes in, it can expand both horizontally and vertically due to:

• new ids
• new dates

It also can have dates that start before min(A.dates) or after max(A.dates) or between min(A.dates) and max(A.dates). The ids can be all unique in B (all new) or some can be overlapping.

Here is an example:

A.values = [2.1 2.4 2.5 2.6; ...
            4.1 4.4 4.5 4.6; ...
            6.1 6.4 6.5 6.6];
A.dates  = [730002; ...
            730004; ...
            730006];
A.id     = {'x1', 'x4', 'x5', 'x6'};

Now new data comes in:

B.values = [1.2 1.9 1.5 1.6 1.7; ...
            3.2 3.9 3.5 3.6 3.7; ...
            7.2 7.9 7.5 7.6 7.7; ...
            8.2 8.9 8.5 8.6 8.7];
B.dates  = [730001; ...
            730003; ...
            730007; ...
            730008];
B.id     = {'x2', 'x9', 'x5', 'x6', 'x7'};

How do we now efficiently and quickly construct the new struct C?

C.values = [NaN 1.2 NaN 1.5 1.6 1.7 1.9; ...
            2.1 NaN 2.4 2.5 2.6 NaN NaN; ...
            NaN 3.2 NaN 3.5 3.6 3.7 3.9; ...
            4.1 NaN 4.4 4.5 4.6 NaN NaN; ...
            6.1 NaN 6.4 6.5 6.6 NaN NaN; ...
            NaN 7.2 NaN 7.5 7.6 7.7 7.9; ...
            NaN 8.2 NaN 8.5 8.6 8.7 8.9];
C.dates  = [730001; ...
            730002; ...
            730003; ...
            730004; ...
            730006; ...
            730007; ...
            730008];
C.id     = {'x1', 'x2', 'x4', 'x5', 'x6', 'x7', 'x9'};

---

Update with timetable

Following the comments, I tried to achieve this with timetable as follows:

function dfmerged = in_mergeCache(dfA, dfB)

dtA = datenum2datetime(dfA.dates); % function datenum2datetime can be found here: https://stackoverflow.com/a/46685634/4262057
dtB = datenum2datetime(dfB.dates);

TTa = array2timetable(dfA.values, 'RowTimes', dtA, 'VariableNames', dfA.id);
TTb = array2timetable(dfB.values, 'RowTimes', dtB, 'VariableNames', dfB.id);

TTs = synchronize(TTa,TTb);

dfmerged.id     = TTs.Properties.VariableNames;
dfmerged.values = table2array(TTs);
dfmerged.dates  = datenum(TTs.Time); %to convert datenum

end 

Problem: However, this gave me a big timetable, where the rows where indeed synchronized, but the columns where just duplicates (9 columns). How can I also synchronize the columns?

C = 

  struct with fields:

        id: {'x1'  'x4'  'x5_TTa'  'x6_TTa'  'x2'  'x9'  'x5_TTb'  'x6_TTb'  'x7'}
    values: [7×9 double]
     dates: [7×1 double]

Answer 1

Here is a solution using unique and indexing:

%combine the data and take unique value of them + their index
[C.id,~,date_i] = unique([A.dates(:);B.dates(:)]);
[C.dates,~,id_i] = unique([A.id B.id]);

C.values = nan(numel(C.dates),numel(C.id));
%use matrix indexing to fill the sub-materices corresponding to elements of A and B
C.values(date_i(1:numel(A.dates)),id_i(1:numel(A.id)))=A.values;
C.values(date_i(numel(A.dates)+1:end),id_i(numel(A.id)+1:end))=B.values;