Unable to detect datetime or duration data using readtimetable with a seconds index

Here is a toy example of my problem:

% Create a timetable with a seconds index
data = timetable(seconds(1:3)', randn(3,1), randn(3,1))
writetimetable(data, "test.csv")
data_in = readtimetable("test.csv")

Output:

data =

3×2 timetable

Time Var1 Var2

_____ ________ _________

1 sec -0.10224 0.31286

2 sec -0.24145 -0.86488

3 sec 0.31921 -0.030051

Error using readtimetable (line 190)

Unable to detect datetime or duration data in file '/Users/username/test.csv' for row times.

The content of the CSV is as follows:

Time,Var1,Var2
1 sec,-0.225584402271252,0.0325574641649735
2 sec,1.11735613881447,0.552527021112224
3 sec,-1.08906429505224,1.10061021788087

For comparison, this works fine:

% Create a timetable with a regular datetime index
dt = datetime({'2015-12-18 08:00:00';'2015-12-18 08:01:00';'2015-12-18 08:02:00'});
data = timetable(dt, randn(3,1), randn(3,1));
writetimetable(data, "test.csv")
data_in = readtimetable("test.csv")

Answer 1

[Cris LaPierre]

The issue appears to be that seconds get written in a format that MATLAB cannot read back in. The only supported input formats for durations are

• 'dd:hh:mm:ss'
• 'hh:mm:ss'
• 'mm:ss'
• 'hh:mm'
• Any of the first three formats, with up to nine S characters to indicate fractional second digits, such as 'hh:mm:ss.SSSS'

The values are written with the units: 1 sec

For now, I would recommend creating and saving the timetable using one of the recognized formats.

% Create a timetable with a seconds index
data = timetable(duration(zeros(3,1),zeros(3,1),(1:3)'), randn(3,1), randn(3,1))

data = 3×2 timetable
      Time        Var1        Var2  
    ________    ________    ________

    00:00:01    -0.35574     0.35245
    00:00:02    -0.60954    -0.22619
    00:00:03      0.6993      1.9414

writetimetable(data, "test.csv");
data_in = readtimetable("test.csv")

data_in = 3×2 timetable
      Time        Var1        Var2  
    ________    ________    ________

    00:00:01    -0.35574     0.35245
    00:00:02    -0.60954    -0.22619
    00:00:03      0.6993      1.9414

[Bill Tubbs]

Thanks, this seems the best solution for now. Although this format (00:00:00) is not as nice-looking for plots of less than a few 100 seconds duration.

Also, I noticed that this simplification works:

duration(0,0,(1:3)')

[Cris LaPierre]

Just modify your plot command to still plot seconds.

data = timetable(duration(0,0,[1 100 500]'), randn(3,1), randn(3,1))

data = 3×2 timetable
      Time        Var1        Var2   
    ________    ________    _________

    00:00:01     0.75423       1.4009
    00:01:40    -0.51516     -0.83689
    00:08:20     0.25414    -0.079847

plot(seconds(data.Time),data.Var1)

[Bill Tubbs]

The problem I've got now with this solution is that durations of a fraction of a second are getting rounded in the csv file to the nearest second.

data = timetable(duration(0,0,(1.1:0.1:1.3)'), randn(3,1), randn(3,1));
writetimetable(data, "test.csv");
data_in = readtimetable("test.csv");
data_in.Time(1) == data_in.Time(3)

ans =

logical

1

How do I make writetimetable use a format like 'hh:mm:ss.SSS'? (Surprised it doesn't do this by default).

[Cris LaPierre]

I'll admit I'm a little surprised as well, but confirmed the behavior with datetimes. It will write the data to file using the specified format. Any date/time information not visible in the displayed format will be lost.

Therefore, try setting your Time format before writing your table to csv.

data = timetable(duration(0,0,(1.1:0.1:1.3)'), randn(3,1), randn(3,1));
data.Time.Format = 'hh:mm:ss.SSS'

data = 3×2 timetable
        Time          Var1        Var2  
    ____________    ________    ________

    00:00:01.100     0.96684      1.3273
    00:00:01.200       1.342     -1.0984
    00:00:01.300    0.029173    -0.36432

writetimetable(data, "test.csv");
data_in = readtimetable("test.csv")

data_in = 3×2 timetable
        Time          Var1        Var2  
    ____________    ________    ________

    00:00:01.100     0.96684      1.3273
    00:00:01.200       1.342     -1.0984
    00:00:01.300    0.029173    -0.36432

[Bill Tubbs]

Great, this solves the problem, thanks.

[Jiri Hajek]

I just met with this same issue in R2022, update 2. Hopefully the incompatibility between writetimetable and readtimetable will be reslved in a future MATLAB update...

[Simon Sigl]

Issue is still present in R2023a Update 3