How to properly convert variable names into a chain of strings in Matlab?

I'm using the save command to store useful data of certain variables a, b_123, cc42,.... I want to have the variable names converted to something like a chain form of 'a','b_123','c42',...: chain(a,b_123,cc42,...)

so that instead of manually writing:

save(file,'a','b_123','cc42',...);

I can simply write:

save(file,chain(a,b_123,cc42,...));

Now that way if for example I want to change the name of cc42 to say cc43 (and using shift+enter to change it in all locations) I won't need to manually change it in the save command from 'cc42' to 'cc43'.

[Stephen23]

From a comment: "...the fact that the names chosen for the variables may change, and it's also easier to work with when you have lots of variables"

Both of those are actually very good reaons to avoid messing around with variable names. Your data design could be significantly improved by using one structure or table, which then makes this entire problem go away.

[Walter Roberson]

Not really. This is a difficulty in refactoring in MATLAB that arises because MATLAB mostly needs variables in "plain" form, but in some situations needs variable names in quoted form (especially for "save"). This problem occurs no matter what data structure you have; the best you can do is reduce down to using a single variable to reduce the number of strings you have to look for. (But that is not actually possible, as for loop control variables must be entire variables, not indexed in any way.)

Answer 1

[Bruno Luong]

If you use number in variable name then you'll run into trouble soon or later. Just don't do it.

Not sure the advantage of using variable it-self rather typing the name directly, beside that the name might change, but then how do you know the right variable after reading back the file ?

a = 1;
b = 2;
s = chain(a,b);
save('myfile.mat', s{:});

function s = chain(varargin)
s = cell(1,nargin);
for k=1:nargin
    s{k} = inputname(k);
end

[tensorisation]

I just gave an example of variables, the names themselves are meaningless. The main reason for this is due to the fact that the names chosen for the variables may change, and it's also easier to work with when you have lots of variables.

Note sure what exactly do you mean by "how do you know the right variable after reading back the file ?"

The method you suggested here appears to be working and is probally what I'm looking for, I just need to make sure it is working properly.

Answer 2

[Walter Roberson]

This is not possible in any released version of MATLAB.

However you can define

function savevars(varargin)
  N = nargin();
  Vn = cell(N,1);
  Vn{1} = varargin{1};  %content! 
  for k = 2 : N
    Vn{k} = inputname(k);
  end
  cmd = ['save ', strjoin(Vn, ''', ''')] ;
  evalin('caller', cmd)

This is untested

[Walter Roberson]

By the way, the fundamental difference between this and the syntax you suggested is that your function chain(), when called the way you do, can only return a single output, but in the context of save() you need multiple arguments at that point.

It would be possible to write something like

function C = chain(varargin)

and have it return a cell array of character vectors, but you would need to use something like

C = chain(a,b_123,cc42) ;
save(filename, C{:})

Unfortunately it is not possible to to {:} the result of a function call without storing into a variable

[tensorisation]

It doesn't appear to be working for me.

I may be missing something, but in regards to what you said: "Unfortunately it is not possible to to {:} the result of a function call without storing into a variable", the method suggested by Bruno appears to be working:

a = 1;
b = 2;
s = chain(a,b);
save('myfile.mat', s{:});
function s = chain(varargin)
s = cell(1,nargin);
for k=1:nargin
    s{k} = inputname(k);
end

Is this method not valid? It appears to be working but I want to make sure that it won't create problems.

[Walter Roberson]

s = chain(a,b);

That stores the result of the function call in a variable.

save('myfile.mat', s{:});

and that does {:} on the stored variable. Two operations.

What would be nice is if you could do something like

save('myfile.mat', chain(a,b){:});

so that no temporary variable was needed, but that is not permitted in MATLAB.

[tensorisation]

Agreed, it would be nicer if I could do this without the need of creating a temporary cell in-between, but other than that it's quite simple and it works. It's good enough for now I guess. Thanks for the input.

Answer 3

[Bruno Luong]

a = 1;
b = 2;
savevars('myfile.mat', a,b);

function savevars(file, varargin)
v = varargin;
f = cell(size(v));
for k=1:length(f)
    f{k} = inputname(k+1);
end
sarg = [f; v]; 
s = struct(sarg{:});
save(file,'-struct','s');
end

[tensorisation]

In this method you have s as a huge structure which is probally not good if some of your variables are big (then s will be too big). How is this better than what you suggested earlier?

a = 1;
b = 2;
s = chain(a,b);
save('myfile.mat', s{:});
function s = chain(varargin)
s = cell(1,nargin);
for k=1:nargin
    s{k} = inputname(k);
end

[Bruno Luong]

no way s is big it contains only names of the variable not the variable it self.

I'll leave my posts 15 more minutes for you or anyone else to read. I'll delete all of them after.

[tensorisation]

But how is this better than what you suggested earlier?

a = 1;
b = 2;
s = chain(a,b);
save('myfile.mat', s{:});
function s = chain(varargin)
s = cell(1,nargin);
for k=1:nargin
    s{k} = inputname(k);
end

Why would you delete your posts?

[Walter Roberson]

Even if the contents of the variable were being used intead of the names of the variables, MATLAB does "copy on write", so if you have

a = something_big;
b = something_else_big;
s = {a, b};

then s would not contain complete copies of something_big an something_else_big : a would contain a pointer to where something_big's data is stored, and b would contain a pointer to where something_else_big's data is stored, and s would contain copies of those two pointers, not the data itself.

[tensorisation]

Thanks for the clarification Walter.