Filling missing times in timetable with NaN using retime

I am trying to create rows in a timetable for days where data is missing. I have the attached table as an example.

Can anyone explain why the following does not work? Instead of getting one row filled with NaN for the only day missing data (15th Nov), all rows after the missing date are replaced with NaN.

clc;clear;clf
load('data.mat')
data.ChunkEnd(4:end)=data.ChunkEnd(4:end)+seconds(1); %think this is necessary because the timing is slightly off?
data_retime=retime(data,data.ChunkEnd(1):days(1):max(data.ChunkEnd),'fillwithmissing');

[Louise Wilson]

I can get it to work if I do

data.ChunkEnd=data.ChunkEnd-timeofday(data.ChunkEnd) 
data_retime=retime(data,data.ChunkEnd(1):days(1):max(data.ChunkEnd),'fillwithmissing');

since the time isn't actually necessary in this context, but I'm wondering why it doesn't work...

Answer 1

[dpb]

load(websave('data.mat','https://www.mathworks.com/matlabcentral/answers/uploaded_files/1148080/data.mat'))
ttD=data_site_num_tt1;      % just get a shorter variable name to work with
clear data_site_num_tt1;
rule=-1*double(second(ttD.ChunkEnd)==0);                    % don't round up 00 seconds rule
ttD.ChunkEnd=dateshift(ttD.ChunkEnd,'end','minute',rule);   % round up any those with seconds
newt=datetime(ttD.ChunkEnd(1):days(1):ttD.ChunkEnd(end));   % fill in the time vector to match
ttD=retime(ttD,newt)                                        % and retime -- will insert missing automagically

ttD = 8×5 timetable
          ChunkEnd          podN    DPM       Nall       MinsOn    FalsePositiveDPM
    ____________________    ____    ___    __________    ______    ________________

    12-Nov-2017 23:59:00    2714      7          1756     1440             0       
    13-Nov-2017 23:59:00    2714     16         11348     1440             0       
    14-Nov-2017 23:59:00    2714      0          3310     1440             0       
    15-Nov-2017 23:59:00     NaN    NaN           NaN      NaN           NaN       
    16-Nov-2017 23:59:00    1533      0         32098     1440             0       
    17-Nov-2017 23:59:00    1533      0         48293     1440             3       
    18-Nov-2017 23:59:00    1533      0    3.4973e+05     1440             0       
    19-Nov-2017 23:59:00    1533      0         18702     1440             0       

The above is a little esoteric with the use of the secondary rule argument and the logic test to set it so would not turn 23:59 into 24:00 and then 00:00 of the next day, granted, but figured may as well demonstrate a corner feature of retime while at it and had the opportunity.

One might just use regular timing and round to the next day since all are so near midnight; that's a decision the data owner has to decide about which date to report on.