Apologies, I misinterpreted the results.
I still think it may make sense to change this behavior in a major version
release, although I don't know what the PostgreSQL policy is on such
breakage.
On Fri, Sep 9, 2016 at 7:13 AM, Andrew Gierth <andrew@tao11.riddles.org.uk>
wrote:
> >>>>> "Shay" == Shay Rojansky <roji@roji.org> writes:
>
> >> This is intentional: all empty arrays are alike, with zero
> >> dimensions (not one dimension, but none).
>
> Shay> Also, note that when reading an empty array in binary encoding,
> Shay> the dimension field contains one.
>
> Nope:
>
> postgres=# select array_send('{}'::integer[]);
> array_send
> ----------------------------
> \x000000000000000000000017
>
> (that's ndim=00000000, flags=00000000, element_oid=00000017)
>
> --
> Andrew (irc:RhodiumToad)
>
