how to get unique value of datetime and sum all of the value ?

Hello Matlab Community

so i'm really new in Datetime class , so what i want is to sum all the value in the column 5 which is unit sold with it corrosponding dates but the dates are scatter and repeated so i want sum all of the unit with the same date , so how to do it?, here what i wrote and my brian just stop working i guess

date = FinancialSample{:,13};
U_sold = FinancialSample{:,5};
%plot(date,U_sold,'o')
[unqdate,idx,idx1] = unique(date);
for c = FinancialSample.Date == unqdate(:)
    sum1(c)=sum(U_sold(c))
end

here the datetime array from microsoft https://docs.microsoft.com/en-us/power-bi/create-reports/sample-financial-download

edit: is there something not clear in my question plz ask me .

Answer 1

[Jakob B. Nielsen]

Hi Muhannad,

You have made a good start, you have a list of all unique dates in your unqdate. One thing I would recommend is to convert your date strings into serial date numbers, because it makes it much easier (in my opinion, anyway) to use logical indexing later on. You do this by saying

date_nums=datenum(date,'dd/mm/yyyy');
unqdate=unique(date_nums);

After this, you want to have a loop that executes the number of times corresponding to the size of your unqdate variable:

for i=1:size(unqdate,1)

Now you can use logical indexing to solve your problem. Its a little twisty but bear with me: you want to take from U_sold those indexes for which it is true that your full date vector matches the i'th entry in the unique date vector. And then sum the values from those indexes.

    sales(i)=sum(U_sold(date_nums==unqdate(i))); 
    realdate{i}=datestr(unqdate(i),'dd/mm/yyyy') %and convert your unique date number back to a date string again
    
end

Logical indexing is a bit hairy when you first look at it, but think of it this way; you formulate a true/false statement, and only select the indexes for which the statement is true. An illustrative example:

A=[1 2 3 4 5 1 1];
%first, look at 
A==1
%gives this: 1   0   0   0   0   1   1
%Which means if you ask for 
A(A==1)
%you get simply 1 1 1 because it only takes the values out, for which we
%got a logical 1 before. 
%We are essentially doing the same thing in your problem, but for unique dates

[Muhannad AL-hazmi]

Thanke you so much

here is the full code

date = FinancialSample{:,13};
U_sold = FinancialSample{:,5};
%plot(date,U_sold,'o')
[unqdate,idx,idx1] = unique(date);
%% this is ex for one of the date and corrosbonding Value in Unit sold .
c = FinancialSample.Date=="01-Jan-2014 00:00:00";
Z = sum(U_sold(c));
%%
date_nums=datenum(date);
unqdate1=unique(date_nums);
for i=1:size(unqdate1,1)
    sales(i)=sum(U_sold(date_nums==unqdate1(i))); 
    realdate{i}=datestr(unqdate1(i),'dd/mm/yyyy') %and convert your unique date number back to a date string again
end
plot(unqdate,sales,'o');
hold on;
[b,indx] = max(sales);
plot(unqdate(indx),sales(indx),'*','Color','k','MarkerSize',13)
legend('Value of Units',sprintf('Maximum Value: %d',b));    
title("All of the Unit sold world wide")
xlabel("Dates")
ylabel("Unit")