How to automate creating separate timetables?

Hello all,

I'm trying to create separate timetables based on extracting data from a larger matrix. The larger matrix is being generated using for loops, but I couldn't figure out how to generate new variables (TestX_YZ) in each loop iteration, and I read it's not recommended anyway.

I have done that manually so far, but if there are any ideas to automate this process (especially with the data set getting larger), I'd highly appreciate it. Here's a sample of the code I've written to do the job manually:

Test1_11=array2timetable(TestData(:,1),'SampleRate',SampleRateHz);
Test1_12=array2timetable(TestData(:,2),'SampleRate',SampleRateHz);
Test1_13=array2timetable(TestData(:,3),'SampleRate',SampleRateHz);
Test1_14=array2timetable(TestData(:,4),'SampleRate',SampleRateHz);
Test1_15=array2timetable(TestData(:,5),'SampleRate',SampleRateHz);
...
    

[Steven Lord]

Can you dynamically create variables with numbered names like Test1_11, Test1_12, Test1_13, etc.? Yes.

Should you do this? The general consensus is no. That Discussions post explains why this is generally discouraged and offers several alternative approaches.

What are you hoping/planning to do with these timetables? There may be a way to achieve your ultimate goal without creating these individually-named timetables.

[Ahmed]

My need for separate timetables is related to loading them into signalEditor as separate signals/scenarios, and using that in Simulink. Unfortunately signalEditor doesn't allow multidimensional timetables.

I completely understand that dynamically creating variables is not recommended and inefficient, but it's going to be more efficient than manually copying and pasting 50 of these commands, and manually adjusting the variable names :D

The question is, how can I do so?

[Stephen23]

"How to automate creating separate timetables?"

https://www.mathworks.com/matlabcentral/discussions/tips/849901-tutorial-why-variables-should-not-be-named-dynamically-eval

"I completely understand that dynamically creating variables is not recommended and inefficient, but it's going to be more efficient than manually copying and pasting 50 of these commands, and manually adjusting the variable names... The question is, how can I do so?"

The documentation states that signals can be a "Structure of MATLAB timeseries or timetable objects."

https://www.mathworks.com/help/simulink/slref/signaleditorblock.html#mw_888714a9-54ed-4764-a60e-68c6a0901a5c

Did you try that?

Answer 1

[Star Strider]

Creating separate variables for each one is definitely not recommended.

One option would be to create them as cell arrays, for example:

for k1 = 1:Whatever1
    for k2 = 1:Whatever2
        Test{k1,k2} = array2timetable(TestData(:,k2),'SampleRate',SampleRateHZ);
    end
end

There may be other approaches. This seems to me to be the most obvious, and most easily implemented.

.

[Ahmed]

Thanks for the quick response. That makes sense, but doesn't solve my need for separate timetables of column arrays.

My need for separate timetables is related to loading them into signalEditor as separate signals/scenarios, and using that in Simulink. Unfortunately signalEditor doesn't allow multidimensional timetables.

[Star Strider]

I’m not certain what you intend by ‘multidimensional timetables’. These would be just the same as in your original code. There is nothting ‘mulidimensional’ about them, other than their indexing, and Simulink would never see that.

[Ahmed]

I meant a timetable with multiple variables/columns. Please check the signals that can be used with Signal Editor here: https://www.mathworks.com/help/simulink/slref/signaleditorblock.html

I tried using a timeseries instead (which is also not recommended to be used and a timetable is preferred), but when loading a multi-variable timeseries into signalEditor, all signals are lumped together and the signalEditor output in Simulink doesn't let me select which signal from the different column to simulate.

That's why I ended up creating separate timetables.

[Star Strider]

I don’t understand.

The only difference between your code example and my version of it is the indexing of the original timetable arrays as elements of a cell array. The advantage that my code offers is that it is a straighttforward way of indexing your tables without creating individual new variables.

I have Simullink, however I rarely use it currently and have never used the Signal Editor block. From my reading of that documentation, what that block wants is actually a .mat file for each signal. In order to do that, my code changes to:

for k1 = 1:Whatever1

for k2 = 1:Whatever2

tt = array2timetable(TestData(:,k2),'SampleRate',SampleRateHZ);

filename(k1,k2) = ["Test"+k1+"_"+k2+".mat"]

save(filename(k1,k2),'tt')

end

end

The extra line that defines the .mat file names simply gives you a straightforward way of finding them and working with them. You can use that reference to provide the filenames to the block. See the documentation section on File name for those details.

An alternative approach to the filename would be:

filename(k1,k2) = sprintf("Test%02d_%02d",k1,k2)

That would insert leading zeros for single digit values of ‘k1’ and ‘k2’, for example —

k1 = 1;
k2 = 11;
filename(k1,k2) = sprintf("Test%02d_%02d",k1,k2);

filename(k1,k2)

ans = "Test01_11"

It remains a string array, as in my first example.

.

Answer 2

[Peter Perkins]

As Steve says, "The documentation states that signals can be a "Structure of MATLAB timeseries or timetable objects." Does this help?

X = rand(10,5);
tt = array2timetable(X,SampleRate=1)

tt = 10x5 timetable
    Time        X1          X2          X3         X4          X5   
    _____    ________    _________    _______    _______    ________

    0 sec     0.87877      0.21789    0.53691    0.35512     0.16079
    1 sec    0.045712      0.80786    0.49763    0.67483     0.80661
    2 sec     0.74115      0.16217    0.79375    0.08399    0.010254
    3 sec     0.63341      0.68058    0.64789    0.96002     0.21177
    4 sec     0.34186      0.40133    0.34071    0.67012     0.40378
    5 sec     0.42319      0.35446    0.69305    0.63662     0.99946
    6 sec     0.53082      0.79517    0.31434    0.64664     0.62491
    7 sec     0.53155      0.43744    0.96477    0.90185     0.59736
    8 sec     0.47073      0.67338    0.22334     0.8581     0.39568
    9 sec    0.038269    0.0096788    0.91534    0.11409    0.062523

varNames = tt.Properties.VariableNames

varNames = 1x5 cell array
    {'X1'}    {'X2'}    {'X3'}    {'X4'}    {'X5'}

for i = 1:length(varNames)
    s.(varNames{i}) = tt(:,i);
end
s

s = struct with fields:
    X1: [10x1 timetable]
    X2: [10x1 timetable]
    X3: [10x1 timetable]
    X4: [10x1 timetable]
    X5: [10x1 timetable]