This is a theoretical question to understand what is going behind the scenes.
If I run this:
q=( 11 22 33 )
q="${q-()}"
declare -p q
it outputs the expected:
declare -a p=([0]="11" [1]="22" [2]="33")
But if I run:
q=( 11 22 33 )
q="${q[@]-()}"
declare -p q
then I get this:
declare -a p=([0]="11 22 33" [1]="22" [2]="33")
I understand which way is the correct one, but I cannot explain why the latter produces the result it does. Can someone explain this?
Assigning to q, when it already has its array attribute set, is equivalent to assigning to q[0]. That is, you would get the same result with
q[0]="${q[@]-()}"
In your first code, ${q-()} also expands to ${q[0]-()} for the same reason, and that value is assigned back to q[0], resulting in an apparent no-op. (In every case, ${...-()} is the same as ${...}, since you aren't dealing with any unset variables.)
q then it gets set to the new array, eg q=(aa bb cc) or q=("{other_arr[@]}").