Understanding the structfun() or cellfun() commands

Dear All,

I want to avoid the use of for loops by using the structfun() and cellfun() commands.

I have a folder with a bunch of "nnn_M.csv" files, where the "nnn" prefix corresponds to numbering of the files, and the "_M" suffix being constant for all files. My goal is to create a double array of the "nnn" values.

The code I currently use with a for loop: (Works)

Files = dir('*.csv');           % Create the structure of file descriptions (name,datenum,...)
N = length(Files);              % Determind # of files for the for loop
for ii = 1:N
    FileNames{ii} = Files(ii).name;                 % Create cell array of the file names
    nnn(ii) = sscanf(FileNames{ii},'%d_M');         % Create double array of file prefixes
end

This is the code without a for loop: (Does not work)

Files = dir('*.csv');                                           % Create the structure of file descriptions (name,datenum,...)
FileNames = structfun(@(x) Files(x).name, Files);               % Create cell array of the file names
nnn = cellfun(@(x) sscanf(FileNames{x},'%d_M'),FileNames);      % Create double array of file prefixes 

The 2nd and 3rd lines give me errors, repectively:

Error using structfun
Inputs to STRUCTFUN must be scalar structures.

% and

Index exceeds the number of array elements (14).
Error in @(x)sscanf(FileNames{x},'%dK')
% When using the correct 'FileNames' the for loop gives
% There are 14 *.csv files in the folder

I welcome all suggestions you may have. Thank you for helping me understand these functions!

Cheers,

-Jackson

Answer 1

[dpb]

As the error message and doc says, structfun applies a function to each field in a scalar struct; you have a struct array -- not what you want.

And, for the specific desire you don't need either function nor a loop construct, either -- use MATLAB vectorized notation--

 nums=str2double(extractBefore(string({Files.name}),'_'));

[Jackson Kock]

Thank you dpb for the answer, which does work as needed.

A couple follow up comments/ questions if I may:

1) I learned that the {} around Files.name creates the cell array of the file names. Such as,

A = {Files.name};

The reason I need to {} is to create the array? Otherwise the command,

A = Files.name;

only knows to grab the first file's name?

2) I do not understand my initial error with the cellfun() command. If you couple please enlighten me on this regard. It seems like the function was doing something but the indexing was incorrect?

3) Could I still use a cellfun() to get the numbers in front of the file name? For example:

A = {Files.name};
B = cellfun(??,A);        % I do not know what to put here

I want to understand how to use an arbitrary function inside the cellfun() command to be able to use them in the future. I understand the case of simpler functions, such as:

C = {[1,2],[3,4]};
D = cellfun(@mean,C);

Most appreciated.

[dpb]

1) Try at the command line and see what Files.name returns. (Hint: search for "comma-separated list" in doc)

2) Since the first failed, I don't know what the content of the cell array was when you tried it so it's not possible to say just what, but probably you had a smaller array than thought and passed inconsistent ones

3a) Sure, but why? You want to avoid loops; cellfun is a loop in sheep's clothing; underneath it is the loop with more overhead than just the direct for...end construct. It has its place, certainly, but isn't always the better solution. But, for illustration

>> cellfun(@(s)sscanf(s,'%d_M'),{Files.name})
ans =
             0          1.00
>> 

3b) As the above illustrates, you write an anonymous function in place of the function handle--it can be any one-line expression. Or, if you can't manage it in one line, use a handle to the m-file function you write. NB: one feature of anonymous functions is that the embed any workspace variables not in their argument list in the function body itself -- the same result as above could be obtained by

>> arrayfun(@(i)sscanf(Files(i).name,'%d_M'),1:numel(Files))
ans =
             0          1.00
>> 

where the Files struct is embedded in the anonymous function. While not optimal here, this can be extremely useful when other parameters are needed to evaluate the function.

Answer 2

[Stephen23]

Avoiding CELLFUN or STRUCTFUN is simpler and much more efficient:

S = dir('*_M.csv');
V = sscanf([S.name],'%f_M.csv')

V = 3×1

1.2000 3.4000 5.6000