How to import multiple .mat files into the same workspace

Hello,

I have multiple .mat files that contain different arrays.

These arrays have the same name in each .mat file; I need to open two or more .mat files to elaborate the data.

The problem is that the content of the most recent .mat file overwrites the previous one since the name of the arrays is always the same.

For example, let's say I have test1.mat that contains:

accel

effort

gyro

and test2.mat that contains:

accel

effort

gyro

Is there a way to be able to load into the workspace the arrays coming from test1 and test2 at the same time?

[Mark]

Thank you a lot, guys, for your support!

Unfortunately, these mat files are generated by an external software system that run on a robot, so I cannot change the variables name.

I have something like this:

current_folder/folder1/test1.mat
current_folder/folder1/test2.mat
current_folder/folder1/test3.mat
current_folder/folder2/test1.mat
current_folder/folder2/test2.mat
current_folder/folder2/test3.mat
current_folder/folder3/test1.mat
current_folder/folder3/test2.mat
current_folder/folder3/test3.mat
..

and I run my matlab script from

current_folder

.do you think it is possible to automatize the loading of each mat file even in this case where there are different directories nested in the current directory?

I'm sorry if this is a stupid question, I'm not very used to Matlab, unfortunately.

[Stephen23]

"Unfortunately, these mat files are generated by an external software system that run on a robot, so I cannot change the variables name."

Why do you write "unfortunately" ? Neither of us asked you to change the variable names.You seem imagine that to be a problem, but it isn't. In fact, keeping the variable names unchanged is by far the best data design.

"do you think it is possible to automatize the loading of each mat file even in this case where there are different directories nested in the current directory?"

Of course, see the comment under my answer.

Answer 1

[Stephen23]

"The problem is that the content of the most recent .mat file overwrites the previous one since the name of the arrays is always the same."

Having exactly the same variable names in each MAT file is the opposite of a problem, it is the best and reccommended way of designing data in multiple MAT files, because it allows you to write robust and efficient code to process it. Whatever you do, do not try to LOAD all of those variables directly into the workspace... there be dragons!

The most important change is to always LOAD into an output variable (which is a scalar structure):

S = load(..);

What to do then depends rather on the nature of that data and what you are going to do with it. But here is a general approach:

P = '.';  % absolute or relative path to where the MAT files are saved.
S = dir(fullfile(P,'test*.mat'));
S = natsortfiles(S); % optional, must be downloaded, see link below.
for k = 1:numel(S)
    F = fullfile(P,S(k).name);
    S(k).data = load(F);
end
D = [S.data]; % optional

After that all of the imported data will be in the non-scalar structure S. You can easily access it using indexing, e.g: the 2nd file:

S(2).name % filename
S(2).data % complete imported data
D(2).accel
D(2).effort
D(2).gyro

NATSORTFILES can be used to sort the files into alphanumeric order, it can be downloaded here:

https://www.mathworks.com/matlabcentral/fileexchange/47434-natural-order-filename-sort

See also:

https://www.mathworks.com/matlabcentral/discussions/tips/847976-tutorial-comma-separated-lists-and-how-to-use-them

https://www.mathworks.com/help/matlab/matlab_prog/access-multiple-elements-of-a-nonscalar-struct-array.html

[Stephen23]

Given the folder/file structure shown in your comment here, you could easily do something like this:

P = '.';  % absolute or relative path to the parent folder.
S = dir(fullfile(P,'*','*.mat'));
S = natsortfiles(S); % optional, must be downloaded, see link in answer.
for k = 1:numel(S)
    F = fullfile(S(k).folder,S(k).name);
    S(k).data = load(F);
end
D = [S.data]; % optional

Accessing the data is exactly as I showed in my answer. You can easily access it using indexing, e.g: the 2nd file:

S(2).name % filename
S(2).folder % folder where that file is saved
S(2).data % complete imported data
D(2).accel
D(2).effort
D(2).gyro

[Mark]

Thank you a lot! It works perfectly!

Answer 2

[Walter Roberson]

projectdir = '.';  %path to .mat files
dinfo = dir( fullfile(projectdir, '*.mat') );
data = struct();
for K = 1 : length(dinfo)
    fullname = fullfile( dinfo(K).folder, dinfo(K).name );
    [~, basename, ~] = fileparts(fullname);
    data.(basename) = load(fullname);
end

The result will be a structure that has one field for each file loaded, with the field named after the file. Each field will be a struct that contains one field for each loaded variable.

So data.test1.accel, data.test1.effort, data.test2.accel, data.test2.effort

It is common that a different organization of data is desireable, such as

projectdir = '.';  %path to .mat files
dinfo = dir( fullfile(projectdir, '*.mat') );
filenames = fullfile( {dinfo.folder}, {dinfo.name});
numfiles = length(filenames);
accels = cell(numfiles,1);
efforts = cell(numfiles,1);
gyros = cell(numfiles,1);
for K = 1 : numfiles
    data = load(filenames{K});
    accels{K} = data.accel;
    efforts{K} = data.accel;
    gyros{K} = data.gyro;
end

This gives cell arrays accels efforts gyros containing the entries; the filenames can be pulled out of filenames

Thirdly, it is not uncommon for the best organization data to be multidimensional arrays.

projectdir = '.';  %path to .mat files
dinfo = dir( fullfile(projectdir, '*.mat') );
filenames = fullfile( {dinfo.folder}, {dinfo.name});
numfiles = length(filenames);
for K = numfiles : -1 : 1     %count backwards
    data = load(filenames{K});
    accels(:,:,K) = data.accel;
    efforts(:,:,K) = data.effort;
    gyros(:,:,K) = data.gyro;
end

This assumes the accels and efforts and gyros are vectors or 2d arrays of consistent size, and splices the information of the various files together into 3D arrays.