How do I make assignments to my workspace using variable names and values in a spreadsheet?

I have an excel file that has names of structures in the top row, names of fields of those structures in the 2nd row and in the remaining rows, values to be assigned to those fields in a columnwise manner. How can I read this into arrays of values for each field? Assume the values are floats.

Example: 4x4 Excel file might look like

   P P V V 
   X Y X Y 
   0 1 1 2 
   3 2 1 2

Result should be that I have in my workspace

P.X = [0 3]; 
P.Y = [1 2]; 
V.X = [1 1]; 
V.Y = [2 2];

Can I do this without using eval()?

Answer 1

[per isakson]

AFAIK: one cannot do it without "eval".

Yes, there are good reasons to avoid eval. However, in your situation it might be justified to use "eval".

The code below is slightly more readable than a basic eval solution. It uses fewer blips at least. Try

cac = {
   'P' 'P' 'V' 'V' 
   'X' 'Y' 'X' 'Y' 
   0 1 1 2 
   3 2 1 2 };

for jj = 1 : size(cac,2)
    assign2struct( cac{1,jj}, cac{2,jj}, [cac{3:4,jj}] )
end

>> P,V
P = 
    Y: [1 2]
    X: [0 3]
V = 
    X: [1 1]
    Y: [2 2]

where

function    assign2struct( name, field, value )
    sas = evalin( 'caller', ['whos( ''', name, ''' );' ] );
    if isempty( sas )
        assignin( 'caller', name, struct(field,value) )
    elseif strcmp( sas.class, 'struct' ) 
        S = evalin( caller, name );
        S.(field) = value;
        assignin( 'caller', name, S )
    else
        error( '%s is not a struct', name ) 
    end
end

Answer 2

[Walter Roberson]

Technically you can do it without using eval(), but it would require writing a routine that used assignin('caller') to assign the value into the workspace you are already in, with much the same effect.

Doing what you want to do is not a good idea. See http://matlab.wikia.com/wiki/FAQ#How_can_I_create_variables_A1.2C_A2.2C....2CA10_in_a_loop.3F

I recommend that you push everything down one level into a master structure, so instead of creating P.X and so on you would create (say) vars.P.X . This can be done easily using dynamic field names without the risks over eval() or of a structure name in the file accidentally being a MATLAB reserved word or accidentally being the name of a variable or function you are using in the reading routine itself. For example suppose you read your lines using fscanf() and the user has requested a structure name of 'fscanf', then as soon as you created that as a variable you would no longer be able to call the fscanf routine.

[John Petersen]

The variable names in the excel file are already defined varialbes in a simulink model. The file defines different scenarios for regression testing. I just need to read them in and apply them to the model. So my question is how to do it?

[Walter Roberson]

If you are applying them to a model then you would be using set_param(), which does not care what the MATLAB-level variable name is, only that you give the right string for the Simulink variable name and that you give a valid value for the parameter.

Or are these variables for a From Workspace block?

[John Petersen]

Some are variable names used in block parameters and others are for constant blocks. Some of the constants propagate as bus signals, and so have predefined structures.

[Walter Roberson]

This is getting beyond what I have read about Simulink. Perhaps if you used a Model Workspace layer, and specified a .mat file that has the variables stored? A MATLAB file is also an option there but it is not obvious what form that file would have to take.