How to save a mixed data (strings and integers) from workspace into a matlab data file?

Hello,

I want to save information processed from a large group of data into a matlab data file. For example:

a = 1253, b = 5869, c = 3569
char1 = "gyiuh"
column1  column2  
value1     value1  
value2     value2
value3     value3
.
.
.
N                N

The savefile has both individual variables and tabulated data too. How can such a combination be saved in a single file so that they can be later retrieved for further processing? Thank you.

Answer 1

[Mathieu NOE]

hello

To save all variables in the current workspace to a .mat file:

save('filename.mat')

You can load the saved workspace later using the load command:

load('filename.mat')

[dpb]

NOTA BENE: that the above syntax while very easy, will overwrite any existing variables in the workspace of the same name. This may be intentional in use, but also has the potential to override a later calculation with unintended results.

This can be avoided if use the return variable form of save as

NewSet=load('filename.mat');

which will return the struct variable NewSet and the original variable names will be the fieldnames. This style can lead to automated processing rather than duplicating code repeating the same thing over and over with sequentially named variables, etc. as application progresses.

SECOND COMMENT:

Something like

column1  column2  
value1     value1  
value2     value2
value3     value3
.
.
.
N                N

would be an ideal candidate for a single table variable instead ...

x=rand(3,1); y=rand(size(x)); % create the two arrays
tData=table(x,y)

tData = 3×2 table
       x           y   
    ________    _______

     0.74549    0.56811
     0.81703    0.56595
    0.047118    0.79113

after which could

clear x y

and then save.

[Vijayananda]

Is it possible to append additional data in non-tabular form to this table variable ?

[dpb]

No. While very powerful and extremely useful to hold variables of different types in a single object, the table similar as other variables must be regular in height of each variable and all elements of each variable must be defined (even if "defined" is the missing value).

The only ways to put a mish-mash of stuff together in a single variable is either as fields of a stuct or as elements in a cell array or array of cells. But, even a cell array itself, while each cell can be different in type or size, must be a regular array; MATLAB simply does not have the concept of "jagged" arrays.

Read up on <Data Types> to become more familiar with what MATLAB can and cannot do...

[dpb]

In the following

char1 = "gyiuh"
column1  column2  
value1     value1  
value2     value2
value3     value3

the existence what appears may be sequentially-numbered variables is indicative of not using the power of MATLAB as designed. If you do have such variables, particularly if they are numeric, it would be much better to use arrays where the above could be at 2D array column or as shown before a table of two columns. Eitheer way allows writing generic code to address the individual data whereas otherwise one has to write explicit code referrng to each specifically by name or revert to complex, hard-to-debug obfuscated coding practices that are to be avoided.

Answer 2

[Walter Roberson]

Consider creating a struct of data to be represented, and then using jsonencode to create a text representation, and then writelines to write the text to a file.

Answer 3

[dpb]

As example, I have a set of scripts the most basic two of which are

BACKUP_WORKSPACE.m

% Backup current workspace to "WORKING ddMMMyyyy hhmm.mat"
fn="WORKING-"+string(datetime(now,'ConvertFrom','datenum','Format','yyyy_MM_dd(MMM)_HHmm'));
if exist('comment','var')
  fn=fn+"-"+string(comment);
end
clear d comment                                % don't save the dir() struct or comment
save(fn)

LOADLAST.m

% load last WORKING backup file w/o needing date...
dd=dir('WORKING*.mat');
[~,ix]=max([dd.datenum]);
load(dd(ix).name)
disp(['Loaded: ' dd(ix).name])
clear dd

Which SAVEs a .mat file with the present date/time in the filename as well as an optional comment string for ID purposes and the second that will retrieve the most recent of these sets for picking up after the previous save state. This does, in fact, deliberately overwrite local variables so it will reproduce those in the workspace as they were at the time. (Note it also does NOT clear existing workspace first as is such a penchant one sees in instructor-led code so new data/variables aren't destroyed in order to continue on without having to recreate new work done since.)

I have a couple of others that will find a given date or comment and one that lets select from a dropdown list as well...the possibiliities are limited only by your imagination.