Re: jsonb_delete with arrays

On Tue, Jan 17, 2017 at 8:45 PM, Magnus Hagander <magnus(at)hagander(dot)net> wrote:
> On Tue, Jan 17, 2017 at 8:25 AM, Michael Paquier <michael(dot)paquier(at)gmail(dot)com>
> wrote:
>> On Sun, Dec 18, 2016 at 1:27 AM, Dmitry Dolgov <9erthalion6(at)gmail(dot)com>
>> wrote:
>> > * use variadic arguments for `jsonb_delete_array`. For rare cases, when
>> > someone decides to use this function directly instead of corresponding
>> > operator. It will be more consistent with `jsonb_delete` from my point
>> > of
>> > view, because it's transition from `jsonb_delete(data, 'key')` to
>> > `jsonb_delete(data, 'key1', 'key2')` is more smooth, than to
>> > `jsonb_delete(data, '{key1, key2}')`.
>>
>> That's a good idea.
>
> I can see the point of that. In the particular usecase I built it for
> originally though, the list of keys came from the application, in which case
> binding them as an array was a lot more efficient (so as not to require a
> whole lot of different prepared statements, one for each number of
> parameters). But that should be workaround-able using the VARIADIC keyword
> in the caller. Or by just using the operator.

Yes that should be enough:
=# select jsonb_delete('{"a":1 , "b":2, "c":3}', 'a', 'b', 'c');
jsonb_delete
--------------
{}
(1 row)
=# select '{"a":1 , "b":2, "c":3}'::jsonb - '{a,b}'::text[];
?column?
----------
{"c": 3}
(1 row)
That's a nice bonus, perhaps that's not worth documenting as most
users will likely care only about the operator.

>> > I've attached a patch with these modifications. What do you think?
>>
>> Looking at both patches proposed, documentation is still missing in
>> the list of jsonb operators as '-' is missing for arrays. I am marking
>> this patch as waiting on author for now.
>
> Added in updated patch. Do you see that as enough, or do we need it in some
> more places in the docs as well?

I am not seeing other places to update, thanks.

Another victim of 352a24a... Your patch is failing to apply because
now the headers of the functions is generated automatically. And the
OIDs have been taken recently. I have fixed that to test your patch,
the result is attached. The patch is marked as ready for committer.
--
Michael