How to initialize a new matrix of the same type as an existing variable?

I'm writing a function that I want to support either numerical or symbolic inputs. It takes in a matrix, initializes a new matrix, then defines it in a for loop. If the input is a syms variable, I'd like to be able to initialize a syms matrix and populate it. If the input is a double, I want to initialize a matrix of doubles. Is there any function that can do this? I can think of some work-arounds that would suffice for this particular example (mostly just not initializing a new matrix before populating in a loop, or setting the new matrix equal to the incoming one before populating) but I'd like to find a neater way to do it.

Answer 1

[John D'Errico]

Simplest is to use zeroes!!!!!! Ones will also work. For example, suppose that X is an array of some generic numeric class. It might be uint64, it might be sym, it might be a double. As an example, I've created X as a sym. Now, I want my code to create a new array, of some size, here 1x5.

syms X

A = zeros(1,5,class(X))
A =
[ 0, 0, 0, 0, 0]

whos A
  Name      Size            Bytes  Class    Attributes
  A         1x5                 8  sym                

As you can see, zeros creates a new array of the desired size and shape, of the designated class. It is filled with zeros. Ones also works, of course. eye even works.

X = uint64(2);
A = eye(5,class(X))
A =
  5×5 uint64 matrix
   1   0   0   0   0
   0   1   0   0   0
   0   0   1   0   0
   0   0   0   1   0
   0   0   0   0   1

Are there other ways to do this? Well, yes. If X is a scalar, then as long as the array B does not already exist in the workspace, then you can use the dynamic typing ability of MATLAB to solve the problem.

X = single(3)
X =
  single
     3
     
clear B
B(5) = X
B =
  1×5 single row vector
     0     0     0     0     3

As you can see, B now exists, and is a vector of length 5, entirely zero, except for the last element which is the same as X. Or, I could have used repmat.

C = repmat(X,1,3)
C =
  1×3 single row vector
     3     3     3

Or, even just a simple indexing, to replicate X.

D = X(ones(1,4))
D =
  1×4 single row vector
     3     3     3     3

We can overwrite the elements of these arrays as we go, of course. Anything inserted into that array will take on the current class for that variable.

D(2) = 1.25
D =
  1×4 single row vector
            3         1.25            3            3

So D is still a single precision vector.

[J. Alex Lee]

Agree, this would be best for OP since sym is apparently one of the allowed classes for zeros() and ones().

[John Morrell]

Thank you, this is exactly what I was looking for! Very nice well explained answer as well.

Answer 2

[J. Alex Lee]

I am not sure if this works for symbolic type, but you can initialize an array of arbitrary type by specifying the "last" element first, so if you do

clear;
A(10) = 1

A will end up being a 1x10 array of type double, or of type anything that you put in place of 1.

So in a loop, if you know the extents of your desired final array in each dimension, you can step backwards to 1

for j = 10:-1:1
    for i = 20:-1:1
        A(j,i) = ...
    end
end