how to convert .mat file to .m file in matlab R2019a version

how to convert .mat file to .m file in matlab R2019a version

[Walter Roberson]

This is not possible in the general case. It is, however, easy in some cases, especially if the .mat contains a single variable that is a numeric vector or 2D array.

Answer 1

[Stephan]

You can not convert *.mat to *.m files. The *.mat format saves variables from workspace and the *.m file format saves Matlab code. So there is no way to convert it into each other. It are different formats for different use.

Answer 2

[Cloud]

When I directly open a MAT file in MatLab, the software generated this snippet automatically.

function import_mat_file(fileToRead1)
%IMPORTFILE(FILETOREAD1)
%  Imports data from the specified file
%  FILETOREAD1:  file to read

%  Auto-generated by MATLAB on 06-Oct-2022 18:07:48

% Import the file
newData1 = load('-mat', fileToRead1);

% Create new variables in the base workspace from those fields.
vars = fieldnames(newData1);
for i = 1:length(vars)
    assignin('base', vars{i}, newData1.(vars{i}));
end

You may add function to save extract variables to M file.

My runtime environment:

OS: Windows 10

MatLab: R2022a

[Steven Lord]

That code is not a conversion of the MAT-file into a MATLAB program file. That is a MATLAB program file that automates the process of reading the MAT-file. If you were to delete the MAT-file that function would error. Therefore it's not what the original poster asked to do.

[Cloud]

@Steven Lord thanks for indication. My answer is the first process of conversion from MAT to M. The entire process would be: MAT ---> read by the snippet --> saved the exracted variable ---> saved in M. Will it this feasible?

[Walter Roberson]

Does the resulting .m file have to be "readable" ? For example if the file contained

format long g
data = reshape(hex2num(['400921fb54442d18'; 'bff6a09e667f3bcd']), 1, 1, 2)

data = 
data(:,:,1) =

          3.14159265358979


data(:,:,2) =

          -1.4142135623731

would that be acceptable for the 1 x 1 x 2 array πand ?

Does the resulting .m file have to be bit-for-bit accurate for the original data? mat2str() is a very nice convenience function but it is not bit-for-bit accurate

str2num(mat2str(pi))-pi

ans = 
     -3.10862446895044e-15

Is there a need for more than 2 dimensions? Is there a need for non-numeric objects? Is there a need for numeric objects that are not double precision ?

[Cloud]

@Walter Roberson regarding your questions on conversion of MAT file:

Does the resulting .m file have to be bit-for-bit accurate for the original data?

When I open the MAT file in MatLab, it looks like bit for bit accuracy.

Is there a need for more than 2 dimensions?

Do you mean could we save higer than 2D matrix? I think so.

Is there a need for non-numeric objects?

This one I haven't tried but since vars{i}, newData1.(vars{i}) key and value format, so it may be fine.

Is there a need for numeric objects that are not double precision ?

Please, share some example for this case.

[Walter Roberson]

str = 'αφγα'

str = 'αφγα'

bytes = unicode2native(str)

bytes = 1×8

206 177 207 134 206 179 206 177

whos bytes

  Name       Size            Bytes  Class    Attributes

  bytes      1x8                 8  uint8              

Notice that bytes here is numeric but not double precision.

Likewise, technically str is numeric but not double precision (character vectors are uint16 internally)

vars{i}, newData1.(vars{i}) key and value format,

If you need to represent a cell array and a struct, then those would not be numeric.

Question: for your purposes would it be good enough to jsonencode ?

When I open the MAT file in MatLab, it looks like bit for bit accuracy.

Yes, .mat files store exact bits. The question I had was whether the .m file that is generated from the .mat needs to exactly reproduce the data from the .mat file, or whether losing the last bit of double precision numbers would be acceptable ?

[Cloud]

@Walter Roberson Thanks for sharing details including web link:

For your purposes would it be good enough to jsonencode ?

I haven't tried the function.

Whether losing the last bit of double precision numbers would be acceptable ?

I think it depends on application. For me double precision is good enough.