How to convert mat files into csvs without changing the names of the files?

I have a folder containing thousands of .mat files (lets call the folder matfolder), I am trying to convert all of the files into csv files and I want their name to be the exact name they had before converting, what would be the right way of doing it? I tried something like

files = dir('/matfolder/*.mat');
for file = files'
    [filepath, name, ext] = fileparts(file);
    csvwrite(name + '.csv', readmatrix(file)); 
end

But this code will not give me any output.

Answer 1

[Ameer Hamza]

This assumes that all the .mat files contain a variable names 'X'.

files = dir('/path_to_matfiles/*.mat');
for i = 1:numel(files)
    mat_filename = fullfile(files(i).folder, files(i).name);
    [~, name] = fileparts(files(i).name);
    csv_filename = fullfile(files(i).folder, [name '.csv']);
    data = load(mat_filename);
    X = data.X;
    writematrix(X, csv_filename)
end

[Stephen23]

Using fileparts is much more robust than strrep, given that we have no idea what the filenames are. Consider that they could include people's names, e.g. "smith.matilda.mat", or a description of the data "test.matrix.mat", both perfectly reasonable and permitted filenames.

[Ameer Hamza]

Yes, thats a good point. I have updated the answer.

[Katie]

Thank for your answer. But my files do not have the variable X? What should I do about it and why is the variable X mentioned here? Can you please update the code without the assumption of having variable X?

[Stephen23]

"But my files do not have the variable X? What should I do about it and why is the variable X mentioned here?"

.mat files can contain any number of variables (arrays), each with its own name. For example, if you did this:

X = [2,3,5];
Y = 'cat';
Z = {true,'hat'};
save('myfile.mat','X','Y','Z')

it would create myfile.mat with three variables named X, Y, and Z, those variables being a numeric array, a character vector, and a cell array of some other arrays. Many variables can be stored in one file.

The recommended way to import data from a .mat file is into an output argument, like this:

data = load(..);

The output data is a scalar structure, where each field of the structure is one variable from the file. Lets try it with the file from the above example:

data = load('myfile.mat');
data.X

ans = 1×3

2 3 5

●

data.Y

ans = 'cat'

data.Z

ans = 1x2 cell array
    {[1]}    {'hat'}

Regardless of what way you import the data from the .mat file you will have to know its name. So far in this thread you have not told us the name/s of the variable/s in the .mat files that you wish to access or even how many there are, so Ameer Hamza quite reasonably used the field X as a placeholder (for you to replace with the actual name of an actual variable saved in the file). You can find out the names by loading it and looking at the structure (e.g. fieldnames), or using whos, or in the File Details panel.

"Can you please update the code without the assumption of having variable X?"

That depends on what is in those files:

• If the data is well-designed and every file contains one (or more) variables with exactly the same names then you can simply select the name of the data that you want: data.nameofdata
• if the data are badly-designed and every file contains exactly one variable with different names then you can use code like this to get the array: S=load(..); C=struct2cell(S); A=C{1};
• if the data are very badly-designed and every file contains multiple variables with different names in each file then it is much more complex (i.e. requires matching fieldnames somehow...).

In order to help you further we would need to know more about the content of those .mat files, in particular:

1. the number of variables in each .mat file.
2. the names (or naming patterns) used for those variables.

[Ameer Hamza]

In addition to the points posted by Stephen, also mention in case of multiple variables, how do you want to export it to a csv file? Do you want to append all the variables together into one matrix (which will only work if their dimensions are compatible).

[Stephen23]

"At the moment, I started off with glootal_flow for example see below ... but I get the error"

We can see that the field glottal_flow contains another structure:

glottal_flow: [1×1 struct]

Apparently the writematrix function will accept numeric matrices, character matrices, categorical matrices, and no doubt a few other types of matrix... but it does not work with structures (I guess mostly because there is no obvious single interpretation of how a structure of any size with any number of fields containing potentially arbitrarily-nested arrays of other classes should be converted into a 2D matrix arrangement like a CSV file or a spreadsheet).

In order to save the data of the glottal_flow structure you will have to investigate it and figure out how its contents can be saved. The contents could be any number of other arrays or any size or class, so there is no simple solution for saving its contents, you have to look inside and decide what would work.

"I am not sure if it is possible to have all of the variables in one csv, but if it is possible, it would be great."

Consider the 'myfile.mat' file from my previous comment: how would you arrange that in a 2D CSV file?

May I ask why this conversion is required? What processing will you perform on them?