How do I combine two DateNumber strings without resulting in all NaN values?

I have two different DateNumber strings DateNumber1 and DateNumber2. I created these from the same dataset as the Date formatting changes from yyyy-MM-dd HH:mm:ss to yyyy-MM-dd HH:mm:ss.SSSSSS resulting in large amounts of NaN values when I import the data into MATLAB. I've imported the two date formats separately and used datenum to create two different Matrixes. I now want to try and combine them, infilling the NaN values with the missing data, however addition will not work and neither will combining.

Any advice would be hugely appreciated!

[Jan]

If you state, that "combining" "will not work", please explain the details of what "combining" is and what "will not work" means. With code and an error message, suggesting an improvement is much easier.

It sounds like a waste of time, if you convert the data twice. You can check in advance, to which format the data match. Posting a small example would be useful also.

[Ben Flett]

Okay thanks Jan. The real problem lies with the Date format, as I said i'm unsure of any way to import the date which has a format of both yyyy-MM-dd HH:mm:ss and yyyy-MM-dd HH:mm:ss.SSSSSS. Hence trying to find a way around this.

Would you suggest anyway of importing the data so that both formats could be read? Sorry for any foolish questions; i'm all very new to this.

[Jan]

Please show us, how the input data are looking. Is this a text file or a MAT file containing a datetime object? Please post the relevant part of your code.

Answer 1

[Jan]

After datenum you have two double vectors. Then all details concerning dates are not relevant for question.

d1  = [1,   NaN, 2,   3,   4,   NaN, 5];
d2  = [NaN, 6,   NaN, NaN, NaN, 7,   NaN];
d12 = d1;
idx = isnan(d1);
d12(idx) = d2(idx);

Answer 2

[Peter Perkins]

You say datenum, but you are using a format for datetimes. And in any case, datenum only accepts three fractional seconds digits, you have six.

It sounds like you have text data with two different formats. If you are using a version of MATLAB since R2014b, I'd suggest using datetime, not dateum.

>> s = {'2017-08-10 09:10:11'; '2017-08-10 09:10:12'; '2017-08-10 09:10:13'; '2017-08-10 09:10:14.123456'; '2017-08-10 09:10:15.234567'}
s =
  5×1 cell array
    {'2017-08-10 09:10:11'       }
    {'2017-08-10 09:10:12'       }
    {'2017-08-10 09:10:13'       }
    {'2017-08-10 09:10:14.123456'}
    {'2017-08-10 09:10:15.234567'}
>> d = datetime(s,'InputFormat','yyyy-MM-dd HH:mm:ss','Format','yyyy-MM-dd HH:mm:ss.SSSSSS')
d = 
  5×1 datetime array
   2017-08-10 09:10:11.000000
   2017-08-10 09:10:12.000000
   2017-08-10 09:10:13.000000
   NaT                       
   NaT                       
>> d(isnat(d)) = datetime(s(isnat(d)),'InputFormat','yyyy-MM-dd HH:mm:ss.SSSSSS')
d = 
  5×1 datetime array
   2017-08-10 09:10:11.000000
   2017-08-10 09:10:12.000000
   2017-08-10 09:10:13.000000
   2017-08-10 09:10:14.123456
   2017-08-10 09:10:15.234567

Answer 3

[Jan]

What about:

s = {'2017-08-10 09:10:11'; ...
     '2017-08-10 09:10:12'; ...
     '2017-08-10 09:10:13'; ...
     '2017-08-10 09:10:14.123456'; ...
     '2017-08-10 09:10:15.234567'};
short    = cellfun('length', s) < 26;
s(short) = strcat(s(short), '.000000');
d = datetime(s, 'InputFormat', 'yyyy-MM-dd HH:mm:ss.SSSSSS', ...
                'Format',      'yyyy-MM-dd HH:mm:ss')