How to plot an array with an hourly time stamp without starting at 00:00:00?

I have a file with basically two colums: time and signal. Time is recorded with the format hh:mm:ss.SSS. First datapoint is at 17:09:47.818 and last one at 08:36:11.706 the next day. I do not have any other info about days, months or anything.

Is there a simple way to plot time vs. signal, and that the x-axis shows the time with the format hh:mm:ss.SSS? If I do simply

figure,plot(T,S)

I get the following graphic:

Which is not ordered in time as I expect. I used the trick of plotting only the signal as:

figure,plot(S)

and defining the xticklabels with the time as I need it, but of course this is a "lie", because as you can see, the data is correct ordered, but the Data Tip does not show the time of the observed signal:

Do you know how can I properly draw a graphic like the second one, but that correctly uses time as in the first figure, and starts at the correct time and not at 00:00?

Thank you all in advance!

[dpb]

If the sample times are uniformly spaced, you can just create a time vector datetime from start to end with the correct number of points.

If the data aren't evenly space, then you need to find the point in the record at which the date change occurs and add 24 hrs to a duration variable that is the difference from the beginning. You can then set the initial time to the origin again by adding that interval magnitude.

Or, convert to a datetime which includes an aribtrary date as well as time -- again you'll have to add the 24 hr point at which time the new day begins. You can set the display format of the datetime axis to not display the date.

Alternatively, use datenum instead, you can add arbitrary date and hours to its representation without worrying about whether a datetime value has a date or not and without the necessity to keep duration and datetime variables separately. That along with datetick will liet you show the plot; the Data Tip, however, will still display the underlying datenum value unless you write a custom callback routine.

[Federico Geser]

Thank you dpb for you help. So, there is no other way but to manually find the date changing point, am I right? I will try to follow your advices, in the mean time, I let you the file with the data attached. I had to delete most of the data because of the 5 MB limit.

Thank you in advance!

Answer 1

[dpb]

You can build complete datetime variables from the input by something like--

dt1=datetime("yesterday")+duration(T(1));       % the first time is before midnight previous day
dt2=datetime(T(end));                           % datetime defaults to "today" for missing date

You can use above logic to apply to the sections of the T vector that belong to which day by logical indexing to find the first location where the time value turns from 23:59.59.SSS to 00:00:00:SSS

Like always, if you would attach the file, folks would have something to play with without having to make up an example dataset, always an incentive to actually do something.

[Federico Geser]

This actually helped a lot. Now I managed to get the correct graphic, basically using logical indexing to find the date changing point, but defining only one datetime variable as follows:

dt(1:ichange-1) = datetime("yesterday",'Format','HH:mm:ss.SSS') + T(1:ichange-1) ;

dt(ichange:length(S)) = datetime( "today",'Format','HH:mm:ss.SSS') + T(ichange:end) ;

Then I simply plotted against the signal S and changed the format of the Data Tip (so that the arbitrary days do not appear):

figure,p = plot(dt,S,'b.');

p.DataTipTemplate.DataTipRows(1).Format = "HH:mm:ss.SSS";

xtickangle(45),

xtickformat('HH:mm:ss.SSS');

And voilà:

Don't mind the days appearing below, but if you know how to remove them, that would be helpful as well.

Thank you dpb!

[dpb]

When plot w/ datetime, that axes becomes a DateTimeRuler object instead of a NumericRuler -- I know this has come up before and I thought I remembered there being a way to turn that off, but I don't seem to find the property to do so...even in the hidden properties...

[Adam Danz]

See this answer by Benjamin Kraus

https://www.mathworks.com/matlabcentral/answers/370074-gca-currentpoint-with-the-datetime-ticks-on-the-x-axis#answer_316069

or this comment by me

https://www.mathworks.com/matlabcentral/answers/438687-how-do-i-set-yaxis-of-axes-from-a-numeric-ruler-to-a-datetime-ruler#comment_1271309

[dpb]

Clever, indeed, for what it does...but it doesn't get rid of the "Dec 15, 2021-Dec 16,2021" displayed date range on the datetime axis leaving only one looking at the time-of-day on the plot without the extraneous clutter/meaningless reference date which is the wart left here with way the DatetimeRuler object is constrained to work in not understanding one can have/want an absolute time of day wiithout really having a real day with which it is associated.

What TMW would have you do here is create a duration variable from the beginning of the first date -- I suppose that isn't any more difficult than creating the datetime, but it certainly is far less intuitive.

ADDENDUM: Unless there's something else I'm not seeing going on here???

[Adam Danz]

I was recently exploring the differences between datetick and xtickformat. The prefered method of formatting datetime tick is by using datetime values along with xtickformat but if you use datetick instead, it omits the trailing date tag.

datetick vs xtickformat demo

dt = datetime('yesterday') + minutes(0:5:200)';
data = nan(size(dt)); 
figure
tiledlayout(2,1)

nexttile
plot(dt, data)
xtickformat('HH:mm'); 
title('xtickformat')
subtitle('contains date stamp')

nexttile
plot(dt, data)
datetick('x', 'HH:MM', 'keepticks') % note change in case
title('datetick')
subtitle('does not contain date stamp')

Answer 2

[Adam Danz]

It sounds like you just need to set the datetime format of the x-tick labels and the x-axis limit.

Use datetime-values for x-coordinates and set datetime format of xticks

xtickformat(axisHandle,'HH:mm:ss.SSS') 
% OR
axisHandle.XAxis.TickLabelFormat='HH:mm:ss.SSS';

Set the XLim

axisHandle.XLim = [min(dtvals), max(dtvals)];  % dtvals are your datetime values

[dpb]

His problem is he doesn't have input enough for a valid datetime variable without building one -- when he converts the input into the datetime, he gets the default date (today) added to the time portion and consequently the two days are merged together, resulting in the first plot where plotted by timeofday and the earlier data that begins at 17:00 hours of day 1 is plotted after the data from day 2 which carries on from midnight of the prior day.

[Adam Danz]

I see. Thanks for the clarification.