Find the number of non-empty cells in each row of a cell array

I have a cell array of tables where each row contains data for one specimen and each column of that row contains results of repeated tests for that specimen. There are different numbers of tests in each row.

I would like to process each row based on the number of tests. How can I determine the number of cells in each row that contain tables? For example, the first row contains 12 cells, the second contains 14, etc.

Answer 1

[Adam Danz]

Here's a demo that

1. creates a cell array of tables and empty cells
2. determines which cells contain a table
3. counts the number of tables per row

c = cell(3,4); 
c([1,2,3,4,6,10,12]) = {array2table(rand(5))}

c = 3×4 cell array
    {5×5 table}    {5×5 table }    {0×0 double}    {5×5 table }
    {5×5 table}    {0×0 double}    {0×0 double}    {0×0 double}
    {5×5 table}    {5×5 table }    {0×0 double}    {5×5 table }

y = cellfun(@istable, c)

y = 3×4 logical array
   1   1   0   1
   1   0   0   0
   1   1   0   1

nTablesPerRow = sum(y,2)

nTablesPerRow = 3×1

3 1 3

If you'd rather count any cell that is not empty,

y2 = ~cellfun(@isempty, c)

y2 = 3×4 logical array
   1   1   0   1
   1   0   0   0
   1   1   0   1

nNonEmptiesPerRow = sum(y2,2)

nNonEmptiesPerRow = 3×1

3 1 3

[Jim McIntyre]

Perfect. Since my data is always left-justified, I can use:

cellfun(@istable,c)

and

1:sum(c(x,:)) as the limits in a for loop to operate on the data.

In your example, with 0s in the middle, I'd use 1:size(y,2) and an if statement

[Adam Danz]

Sounds right, Jim. Alternatively, you could use logical indexing - the "y" variable in my example, but your approach is perfectly fine if all of the empties are guaranteed to be at the end of the rows.

[Jim McIntyre]

Okay, so please educate me. The cell array contains tables of time-series data. For each of these tables I want to filter two of the columns to create new columns. Here is the code I wrote. It works.

y = cellfun(@istable, data) %Determine the non-empty cells in the data array

%Filter the data to new channels

for j=1:size(y,1)

for k=1:sum(y(j,:))

data{j,k}.FiltAccel=filter(Hd,data{j,k}.Accel);

data{j,k}.FiltSlip=filter(Hd,data{j,k}.Slip);

end

end

Is there a more-efficient way to do this? I'm not sure how I would use logical indexing to add the filtered columns.

[Adam Danz]

> Is there a more-efficient way to do this?

Your approach is readable and reasonable so I don't see a need to change it. I can think of other approaches and if we were to do a proper investigation of their efficiencies, they might be a tad faster or a tad slower by but the difference wouldn't be big enough to sacrifice readability, IMO.

For example, here are two suggestions below. The first is merely preferential and easier to read. The second is an improvement on flexibility but is likely a tad slower.

for j=1:height(y) % <---- changed to height(), just my preference

for k=find(y(j,:)) % <--- no longer assumes empties are all at the end, but slower

...

end

end

Another idea is to reduce two for-loops to one for-loop which, IMO, is more readable.

% *not tested....

y = cellfun(@istable, data);

ind = find(y);

for i = ind(:)' % must be row vector

data{i}.FiltAccel=filter(Hd,data{i}.Accel);

data{i}.FiltSlip=filter(Hd,data{i}.Slip);

end

[Jim McIntyre]

Thank you. I really like the second idea, but it gave the following error:

-----------

Intermediate brace '{}' indexing produced a comma-separated list with 113 values, but it must produce a single value when followed by subsequent indexing operations.

-----------

[Adam Danz]

Could you show me what data{i} is when the error occurs, assuming the error happens on one of the two lines within the for-loop?

[Jim McIntyre]

The error is occuring the first time the data{i}.FiltAccel = line is called.

In the Command window:

ind=find(y)

then

data{i}

shows all the data, but

data{i}.Accel

gives the error.

[Adam Danz]

What I'd like to see is the contents of data{i} at the line that is breaking.

For example, copy-paste all of data{i} from the command window:

% *not tested....
y = cellfun(@istable, data);
ind = find(y); 
for i = ind
    
    % PRINT data{i} % % % % %
    % TO COMMAND WINDOW % % %
    data{i} % % % % % % % % %
    % % % % % % % % % % % % %
    
   data{i}.FiltAccel=filter(Hd,data{i}.Accel);  % ERROR HERE
   data{i}.FiltSlip=filter(Hd,data{i}.Slip);
end

[Jim McIntyre]

There are 113 tables in data{}. The first part looks like this:

>> ind=find(y);

>> for i = ind

data{i}

end

ans =

280×9 table

Time Speed Distance Acceleration RRSpeed PctSlipRR SW_Angle Throttle Gear

____ _____ ________ ____________ _______ _________ ________ ________ ____

0 0.32 0 0.15 0 0 1.5 21 1

0.01 0.35 0.01 0.13 0 0 1.5 21 1

0.02 0.37 0.01 0.13 0 0 1.5 21 1

0.03 0.39 0.02 0.12 0 0 1.5 21 1

0.04 0.39 0.02 0.12 0 0 1.5 21 1

: : : : : : : : :

2.75 4.8 9.95 0.17 5.6323 17.926 3 29.5 1

2.76 4.79 10.02 0.18 5.6581 17.388 3 29.5 1

2.77 4.8 10.09 0.18 5.6581 16.48 3 29.5 1

2.78 4.86 10.16 0.2 5.7739 18.408 3 29.5 1

2.79 4.93 10.23 0.2 5.7739 18.865 3 29.5 1

Display all 280 rows.

ans =

317×9 table

Time Speed Distance Acceleration RRSpeed PctSlipRR SW_Angle Throttle Gear

____ _____ ________ ____________ _______ _________ ________ ________ ____

0 0.33 0 0.13 0 0 -1.5 22.5 1

0.01 0.33 0.01 0.13 0 0 -1.5 22.5 1

0.02 0.33 0.01 0.13 0 0 -1.5 22.5 1

0.03 0.35 0.02 0.13 0 0 -1.5 23 1

0.04 0.36 0.02 0.13 0 0 -1.5 23 1

: : : : : : : : :

3.12 4.91 13.38 0.14 10.484 116.94 -6 34 2

3.13 4.95 13.45 0.12 10.079 106.69 -6 34 2

3.14 4.98 13.52 0.14 9.8829 101.64 -6 34 2

3.15 4.99 13.59 0.12 9.4256 91.573 -6 34 2

3.16 4.99 13.67 0.08 9.2232 87.22 -6 34 2

Display all 317 rows.

This causes the error:

>> for i = ind

data{i}.Speed

end

Intermediate brace '{}' indexing produced a comma-separated list with 113 values, but it must produce a single value when followed by subsequent indexing operations.

[Adam Danz]

@Jim McIntyre the reason you are getting that error is because ind must be a row vector and my bet is that yours is a column vector, although I don't know why it's not a row vector.

If I'm right, forcing ind to be a row vector should fix it

>> ind=find(y);
>> for i = ind(:)'

[Jim McIntyre]

The ' transform definitely made the difference.

Y is a 7x18 logical, yet ind is a 113x1 double, and data{i} wants a 1x113.

From the help for find:

If X is a multidimensional array, then find returns a column vector of the linear indices of the result.

Thank you again for all your help on this!

[Adam Danz]

Ah, of course. Thanks @Jim McIntyre. That's what I get for running code in my head rather than actually testing it 😊

I'll update my previous comment to avoid confusions for any future visitors.