Merging two matrices by first column values
显示 更早的评论
I have two matrices (A and B), the first column in each is a date, the second column is a reading (double). The dates in A and B are not contiguous. The result should be a matrix containing both A and B on the dates for which either OR both are relevent.
e.g.
A =
01/01/2014 1.2345
03/01/2014 1.2345
04/01/2014 1.2345
...
B=
02/01/2014 9.8765
03/01/2014 9.8765
04/01/2014 9.8765
...
Merged=
[date] [A] [B]
01/01/2014 1.2345 NaN;
02/01/2014 NaN 9.8765
03/01/2014 1.2345 9.8765
04/01/2014 1.2345 9.8765
I'm not even sure of the proper name of what I'm trying to achieve? Is there one command or a series of commands to do this, as at the moment I am using a script which just loops through each input but this approach will become unsuitable when the input lengths become large.
9 个评论
Rose
2014-4-15
are the dates in ascending order?
Scott
2014-4-15
pietro
2014-4-15
Are A and B cell arrays?
Scott
2014-4-15
Patrik Ek
2014-4-15
Ok now I cannot follow no longer. Do you mean that the date is a string with dates, a string of serial numbers, or a double with serial numbers? You need to give a better explanation.
Scott
2014-4-15
David Sanchez
2014-4-15
The first column of your A and B matrices is in date format or is it the serial date number:
A =
01/01/2014 1.2345
03/01/2014 1.2345
04/01/2014 1.2345
OR:
A =
735600 1.2345
735659 1.2345
735690 1.2345
???
This is key to give a proper answer
Patrik Ek
2014-4-15
Ok but what format do you want? The serial date number or the date string? Both works, except that the date requires a cell
Scott
2014-4-15
采纳的回答
Since the type of the date is only vaguely specified I will select one.
A = {'01/01/2014', 1.2345;'03/01/2014', 1.2345;'04/01/2014', 1.2345};
B = {'02/01/2014', 6.7890;'03/01/2014', 6.7890;'04/01/2014', 6.7890};
q = unique([A(:,1);B(:,1)]); % Unique sorted catenate.
aInd = ismember(cell2mat(q),cell2mat(A(:,1)),'rows'); % All dates in A
bInd = ismember(cell2mat(q),cell2mat(B(:,1)),'rows'); % All dates in b
fullCell = cell(length(q),3);
fullCell(:,1) = q;
fullCell(aInd,2) = A(:,2);
fullCell(not(aInd),2) = {nan};
fullCell(bInd,3) = B(:,2);
fullCell(not(bInd),3) = {nan};
For the case where dates are serial date number you just define matrices instead and aInd and bInd are instead
aInd = ismember(q,A(:,1)); % All dates in A
bInd = ismember(q,B(:,1)); % All dates in b
3 个评论
Scott
2014-4-15
Scott
2014-4-15
Glad I could help. And yes the mex files are always faster. However for the datenum part would be good to solve with a cellfun in the future (if you not does that already, since you most likely uses the RAW format in the excel file).
cellfun(@(x),datenum(x),dateCell);
or if uses the TXT format, then datenum evaluates the date row-wise for a matrix input.
更多回答（0 个）
类别
在 帮助中心 和 File Exchange 中查找有关 Time Series Objects 的更多信息
产品
Translated by 
