Indexed deletion with an empty index transforms a matrix into a row vector since R2016a

When using indexed deletion on a matrix, this matrix is ALWAYS transformed into a row vector, independent if the empty index is an "1x0" empty double column OR row vector.

This behavior originated in R2016a and can be reproduced by the following commands:

>> A = rand(2)
A =
    0.4218    0.7922
    0.9157    0.9595

Create an empty index:

>> ind = 1:0
ind =
  1×0 empty double row vector

Use logical indexing:

>> A(ind)
ans =
  1×0 empty double row vector

Performing  indexed deletion using the the empty index creates a row-vector, even if the empty index is a column vector

>> A(ind) = []
A =
    0.4218    0.9157    0.7922    0.9595

>> A(transpose(ind)) = []
A =
    0.4218    0.9157    0.7922    0.9595

As a comparison, the output of the same commands in MATLAB R2015b looks like:

>> A = rand(2)
A =
    0.8147    0.1270
    0.9058    0.9134

>> A(1:0) = []
A =
    0.8147    0.1270
    0.9058    0.9134

Answer 1

[MathWorks Support Team]

This was a deliberate change for the MATLAB release R2016a, in order to get consistent behavior for deletion operations. When you do indexed deletion with a single, non-empty subscript, the matrix is always reshaped to a row vector.

>> B = rand(2)
B =
    0.2785    0.9575
    0.5469    0.9649

>> B(1) = []
B =
    0.5469    0.9575    0.9649

In R2016a, this behavior was enforced for single empty subscripts as well. The shape of the empty index makes no difference for indexed deletion.