Why is subsref and subscripted reference not equivalent

Assume an struct array with the following content:

A = struct('b', {1, 2});

Following two code blocks do not give the same output:

[A.b]
ans =

     1     2

and

S = struct('type', {'.'}, 'subs', {'b'});
[subsref(A, S)]
ans =

     1

But I would say this two referencing methods are equal. What is the workaround for using subsref?

[Alexandra Harkai]

Not sure if it helps, but it makes a difference whether the underlying data is a cell array or not:

B = struct('b', [1, 2]);

[B.b]
ans =

     1     2

and

S = struct('type', {'.'}, 'subs', {'b'});

[subsref(B, S)]
ans =

     1     2

If it's not a cell array, the two referencing methods seem to be equivalent.

[Jan]

@Alexandra: B = struct('b', [1, 2]) creates a scalar struct, while the detail matters, that A = struct('b', {1, 2}) is a struct array.

[Alexandra Harkai]

Thanks for the clarification!

Answer 1

[Kenneth Eaton]

Not sure if this is a complete answer, but it appears that invoking subsref directly requires a stricter definition of the output arguments. While:

A.b

will automatically dump all possible output arguments in a comma separated list, calling subsref directly requires you to explicitly define how many output arguments you want. The only way I've figured out thus far to get them all is to collect them in a cell array of the appropriate size:

>> [out{1:numel(A)}]=subsref(A, S)

out =

    1×2 cell array

      [1]    [2]

>> [out{:}]

ans =

       1     2

Answer 2

[Jan]

Not an answer: I've played around with

S = struct('type', {'()', '.'}, 'subs', {{':'}, 'b'});
[subsref(A, S)]

because

[A.b]

is actually:

[A(:).b]

But this still replies 1 instead of the expected [1,2]. Then my interest in the question growed and I've voted for it. I expect, that there is a simple solution, by I cannot find it currently.