Prevent MATLAB from expanding matrix symbolic expression

I wonder if there is any way to prevent MATLAB from expanding the matrix symbolic expression?

For example, let's A = 3x3 matrix and b = constant.

Desired output:

>> A = sym('A', [3,3]);
>> syms b;
>> A^-1 + b*eye(3)
ans =
A^-1 + b*eye(3)

But MATLAB expands the inverse expression of A and gives :

[ b + (A2_2*A3_3 - A2_3*A3_2)/(A1_1*A2_2*A3_3 - A1_1*A2_3*A3_2 - A1_2*A2_1*A3_3 + A1_2*A2_3*A3_1 + A1_3*A2_1*A3_2 - A1_3*A2_2*A3_1),    -(A1_2*A3_3 - A1_3*A3_2)/(A1_1*A2_2*A3_3 - A1_1*A2_3*A3_2 - A1_2*A2_1*A3_3 + A1_2*A2_3*A3_1 + A1_3*A2_1*A3_2 - A1_3*A2_2*A3_1),     (A1_2*A2_3 - A1_3*A2_2)/(A1_1*A2_2*A3_3 - A1_1*A2_3*A3_2 - A1_2*A2_1*A3_3 + A1_2*A2_3*A3_1 + A1_3*A2_1*A3_2 - A1_3*A2_2*A3_1)]
[    -(A2_1*A3_3 - A2_3*A3_1)/(A1_1*A2_2*A3_3 - A1_1*A2_3*A3_2 - A1_2*A2_1*A3_3 + A1_2*A2_3*A3_1 + A1_3*A2_1*A3_2 - A1_3*A2_2*A3_1), b + (A1_1*A3_3 - A1_3*A3_1)/(A1_1*A2_2*A3_3 - A1_1*A2_3*A3_2 - A1_2*A2_1*A3_3 + A1_2*A2_3*A3_1 + A1_3*A2_1*A3_2 - A1_3*A2_2*A3_1),    -(A1_1*A2_3 - A1_3*A2_1)/(A1_1*A2_2*A3_3 - A1_1*A2_3*A3_2 - A1_2*A2_1*A3_3 + A1_2*A2_3*A3_1 + A1_3*A2_1*A3_2 - A1_3*A2_2*A3_1)]
[     (A2_1*A3_2 - A2_2*A3_1)/(A1_1*A2_2*A3_3 - A1_1*A2_3*A3_2 - A1_2*A2_1*A3_3 + A1_2*A2_3*A3_1 + A1_3*A2_1*A3_2 - A1_3*A2_2*A3_1),    -(A1_1*A3_2 - A1_2*A3_1)/(A1_1*A2_2*A3_3 - A1_1*A2_3*A3_2 - A1_2*A2_1*A3_3 + A1_2*A2_3*A3_1 + A1_3*A2_1*A3_2 - A1_3*A2_2*A3_1), b + (A1_1*A2_2 - A1_2*A2_1)/(A1_1*A2_2*A3_3 - A1_1*A2_3*A3_2 - A1_2*A2_1*A3_3 + A1_2*A2_3*A3_1 + A1_3*A2_1*A3_2 - A1_3*A2_2*A3_1)]

Edited:

• The reason is I want to do some complicated matrix symbolic manipulation, and don't want to expand the internal elements of the matrices.

• Fixed the dimension.

[madhan ravi]

str2sym(' ') %?

[Wit]

This doesn't work because it doesn't check the dimension of the matrix. For example:

>> x = sym('x', [2,1])

Obviously, A = 3x3 matrix cannot be multiplied with x. Using

>> str2sym('A*x')
ans =
A*x

doesn't check the dimension requirements.

Answer 1

[Walter Roberson]

No. To do this you would need to code many new MuPAD level routines as it violates the design principles of the existing symbolic linear algebra routines .

It would probably be easier to drop in a bunch of MuPAD level hold() calls and some careful manipulation .