One thing to bear in mind is that bash implemented arrays like ksh, that is as associative arrays where keys are limited to positive integers (contrary to other languages like perl or zsh for instance).
In:
a[123]=foo a[456]=bar a[789]=baz
In bash, you've got an associative array with 3 elements, while in perl, you'd have an array with 790 elements (789 with zsh).
In ksh or bash, ${a[@]:0} returns the first element of the array in the list of elements sorted by indices. So in that case, it returns ${a[123]}, not ${a[0]}.
unset 'a[123]'
(remember to quote it, otherwise it would fail if there was a file called a1 or a2 or a3 in the current directory) makes sense, as it removes a particular key in the array.
unset 'a[@]::2'
makes less sense though. bash only understands unset a, unset 'a[123]' or unset 'a[*/@]', anything after is ignored, so unset 'a[@]::2' and unset 'a[@]please' do the same: unset the whole array.
If you want to unset a range of keys, you'd need to loop through the keys:
To get the list of keys of the array, the syntax is "${!a[@]}". Unfortunately, applying a range to that doesn't work with bash nor ksh, so you'd need a temporary array:
keys=("${!a[@]}")
for i in "${keys[@]::2}"; do unset "a[$i]"; done
Now if you want to consider those arrays like in other languages, you don't want to use unset. Like, if the array is not sparse in the first place and you want to keep it so (that is shift all the elements by 2 instead of unsetting the first two), you can do things like:
a=("${a[@]:2}")
That is reassign the array with the list of elements you want to keep.
For comparison, with zsh.
a=({1..20})
unset 'a[12,16]'
would set an empty value to elements 12 to 16. while unset 'a[16,20]' would shrink the array to 15 elements.
a=({1..20})
a[12,16]=()
(still with zsh) would shift elements 17 to 20 by 5 positions so a[12] would contain 17.
