transform an empty matrix '0x0 double' into a matrix '0x2 double'

Hello! How can I transform an empty matrix '0x0 double' into a matrix '0x2 double' as in the figure?

For example in my case, I have a cell 'test_p'. Rows 3,4,5 (0x0 double) should become like rows 1 and 2 (0x2 double).

test_p = importdata("test_p.mat");
data_0x2_double = test_p{1, 1};
data_0x0_double = test_p{3, 1};

Answer 1

[Bruno Luong]

test_p = importdata("test_p.mat")

test_p = 5×1 cell array
    {0×2 double}
    {0×2 double}
    {0×0 double}
    {0×0 double}
    {0×0 double}

ir = cellfun('size',test_p,1)==0 & cellfun('size',test_p,2)==0

ir = 5×1 logical array
   0
   0
   1
   1
   1

test_p(ir) = {zeros(0,2)}

test_p = 5×1 cell array
    {0×2 double}
    {0×2 double}
    {0×2 double}
    {0×2 double}
    {0×2 double}

Answer 2

[Steven Lord]

Since a 0-by-0 array and a 0-by-2 array have the same number of elements you can reshape one into the other.

A = zeros(0, 0)

A =

     []

B = reshape(A, [0 2])

B =

  0×2 empty double matrix

C = reshape(B, 0, 0)

C =

     []

whos A B C

  Name      Size            Bytes  Class     Attributes

  A         0x0                 0  double              
  B         0x2                 0  double              
  C         0x0                 0  double              

[Bruno Luong]

Is there any difference (internally) between assigning

• a new variable reshaped from an existing empty variable,
• create a new empty variable with desired size?

As they don't have any content and their data pointer is NULL why bother to use reshape?

[Steven Lord]

Your code below made an assumption about the user's data, one that's likely valid (and is valid WRT the original question) but in some cases may not be.

A = zeros(0, 0, 'single');
B = reshape(A, [0 2]);
C = zeros(0, 2);
whos A B C

  Name      Size            Bytes  Class     Attributes

  A         0x0                 0  single              
  B         0x2                 0  single              
  C         0x2                 0  double              

B and C are different classes. You could modify the call that creates C to always create it 'like' A or with the same type as A, but reshape will not change the type and is shorter than either of the calls that create C1 or C2 below.

C1 = zeros(0, 2, 'like', A)

C1 =

  0×2 empty single matrix

C2 = zeros(0, 2, class(A))

C2 =

  0×2 empty single matrix

[Bruno Luong]

OK in short you want to preserve the class of the data with reshape.

Answer 3

[Adam Danz]

To create an empty double with a specified size,

data_0x2_double = zeros(0,2)

data_0x2_double =

  0×2 empty double matrix

Or,

data_0x2_double = double.empty(0,2)

data_0x2_double =

  0×2 empty double matrix

See also Steve Lord's solution using reshape.