How do I count the number of rows in a cell array's cells?

Question

Daniel Bridges 2018-1-17

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/377534-how-do-i-count-the-number-of-rows-in-a-cell-array-s-cells

编辑： Daniel Bridges 2018-2-15

My goal is to create a table as a subset of a cell array's contents. I understand that it's better to preallocate memory size and then truncate empty rows rather than have an array grow with each loop repetition. So I wish to count the total number of rows in each cell as the maximum possible size of the resulting table.

This does not work:

>> size(newvoldata)
ans =
       1   200

Nor does this:

>> numel(newvoldata)
ans =
     200

For example, one cell can have a table such that

>> height(newvoldata{160})
ans =
     439

Is there a MATLAB command that more quickly does the following?

counter = 0;
for loop = 1:length(newvoldata)
    counter = counter + size(newvoldata{loop},1);
end
counter

I suspect there is some command similar to numel that does this for cell arrays. Is there one? Or should I use an implicit function ?

1 个评论
显示 -1更早的评论隐藏 -1更早的评论

Greg 2018-1-17

编辑：Greg 2018-1-17

Is your end goal just to know the height of each cell's table? Or are you eventually going to do some form of join on the tables?

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Stephen23 2018-1-17

1
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/377534-how-do-i-count-the-number-of-rows-in-a-cell-array-s-cells#answer_300470

编辑：Stephen23 2018-1-17

在 MATLAB Online 中打开

For a cell array of tables:

sum(cellfun(@height,newvoldata))

For a cell array of numeric/char/cell/struct arrays:

sum(cellfun('size',newvoldata,1))

12 个评论
显示 10更早的评论隐藏 10更早的评论

Stephen23 2018-1-18

编辑：Stephen23 2018-1-18

"The help documentation for cellfun says, "The cell arrays C1,...,Cn all must have the same size." This condition is not met here"

That statement of the documentation actually refers to multiple inputs (cell arrays) of cellfun, which indeed must all have the same size. But it does not apply to your situation, or the answer I gave: you only have one cell array input (the variable newvoldata). Therefore your statement that "This condition is not met here" is incorrect because in fact your (degenerate) case with just one input cell array is always trivially true: it must be the case that one input cell array has the same size... as itself.

This statement certainly does NOT apply to the contents of the cell array/s, which may be where your confusion comes from.

" and perhaps consequently this line yields 199 rather than its correct value, 51200. (Moreover, it yields 199 whether the cellfun argument is 1 or 2.)"

I think that you should upload this variable in a mat file so that we can look at it. Playing this game of tennis is entertaining, but just a waste of your time if you really want a solution.

Daniel Bridges 2018-1-18

编辑：Daniel Bridges 2018-1-18

Erm, right. I misread 'cell array' as 'cells' (thinking of a table within the cell as if it were an array of data). Sorry about that.

The cells are tables of various heights or else empty. The variable newvoldata contains data derived from patients so I am reluctant to upload it: I think one could replicate the situation simply by using the rand function to decide which of 200 cells to leave empty (not empty tables, but empty cells), and then use the rand function again to generate tables of various heights for the remaining cells. There are seven columns, a mixture of string and double.

Now that I think of it, it seems better to remove those empty cells. Instead I'd been using an if ~isempty section in a for loop to skip over them...

That said, my original question was simply whether there was a command to count the number of rows contained in a cell array's cells. Are you confident that there is not, and that writing your own code is necessary to count each cell's data rows? Is the solution I provided in the OP slow relative to what can be done in MATLAB (such as with the cellfun command)?

Stephen23 2018-1-18

编辑：Stephen23 2018-1-18

在 MATLAB Online 中打开

You can easily create a logical mask, something like this (untested):

idx = cellfun('isempty', newvoldata);
sum(cellfun('size',newvoldata(~idx),1)) % or @height

" whether there was a command to count the number of rows contained in a cell array's cells"

There is no single command to do that, in the same way that there is no command "importmydatamakesomeplotsandemailittomysupervisor": a language consists of basic commands which can be built into a practically unlimited number of combinations to achieve different purposes. You put some data into a cell array: there are tools and functions for operating with cell arrays. The data in the cells are tables: there are tools and functions for operating with tables. However having one special command for each permutation of all those different commands would be totally intractable: the total number of possible commands would be astronomical.

" Is the solution I provided in the OP slow relative to what can be done in MATLAB (such as with the cellfun command)?"

That is a hard question to answer. Loops are often faster than cellfun, but the two syntaxes I have shown you (with 'size' and 'isempty') are optimized and quite fast. The evaluation time of loops and cellfun can also change between MATLAB versions, so it is not possible to make a statement that is universally true. Usually time developing and debugging is much more significant than runtime (e.g. we just spent one day discussing this issue, which is unlikely to ever be overtaken by the difference in runtimes of a loop vs. cellfun, therefore you have saved negative time in total). There is nothing wrong with a loop, but sometimes cellfun is just as fast and a bit neater to boot. Testing is the only way to make sure (see timeit).

Stephen23 2018-2-15

编辑：Stephen23 2018-2-15

在 MATLAB Online 中打开

"Just tried both commands and they both failed"

Which "both commands" are you referring to?

Did you use a logical mask like I showed you in a comment nearly one month ago? If you have cells containing a 0x0 numeric then you will need to use a mask, which is why I showed you how to generate and use one.

"The second results in '199', apparently counting the number of non-empty cells"

The second solution I gave in my answer should not count non-empty cells: for a vector cell array it sums the first dimension of all contents of that cell array. To count the non-empty cells you would need something like this:

sum(~cellfun('isempty',newvoldata))

which, depending on the size of the contents of the cell array, may or may not give the same output value. It is possible that the cellfun "backwards compatibility" syntax using 'size' has not been updated to work correctly with tables.

As has been requested before: if you want help then please upload your data in a .mat file.

Daniel Bridges 2018-2-15

编辑：Daniel Bridges 2018-2-15

在 MATLAB Online 中打开

Thank you for this reminder. I did not apply a logical mask as you advised. I tried the two commands you posted in the answer here, sum(cellfun(@height,newvoldata)) and sum(cellfun('size',newvoldata,1)).

You're saying the second one shouldn't be counting non-empty cell [contents?] but it's the only guess I have for why the result is '199', whereas that cell array had 199 tables and 1 0x0 double. The first cell's table did not have height 199 (in fact, I think none of them did).

For the sake of accepting this answer, I just tested to verify that the first command works:

test = table(5,'VariableNames',{'john'});
test2 = table([5;3],'VariableNames',{'billy'});
test3 = table({'jim';'sally';'roger'},[1;5;3],'VariableNames',{'the' 'teh'});
cellarray = cell(1,3);
cellarray{1} = test; cellarray{2} = test2; cellarray{3} = test3;
ans = sum(cellfun(@height,cellarray))