Shawn's json_group_array(json(o)) works indeed, but it's also 30% slower in
my case than using
'[' || ifnull(group_concat(o, ','), '') || ']'
which is however more case specific and less obvious.
Would be nice to see the subtype passing be improved, as otherwise query
planner improvements
could
On Tue, Jan 8, 2019 at 11:04 AM Dominique Devienne
wrote:
>
> On Tue, Jan 8, 2019 at 10:50 AM Eric Grange wrote:
>>
>> Can someone confirm whether this is a bug ?
>
>
> My guess is that it works as "designed", even if this is surprising...
>
> I believe that JSON1 leverages value "sub-types" [1],
Try using json_group_array(json(o))
On Tue, Jan 8, 2019, 1:50 AM Eric Grange Thanks.
>
> I think I may have encountered a "real" bug while ordering in a subquery.
> I have simplified it in the following exemples:
>
> select json_group_array(o) from (
>select json_object(
> 'id', sb.id
>
On Tue, Jan 8, 2019 at 10:50 AM Eric Grange wrote:
> Thanks.
>
> I think I may have encountered a "real" bug while ordering in a subquery.
> I have simplified it in the following exemples:
>
> select json_group_array(o) from (
>select json_object(
> 'id', sb.id
>) o
>from (
>
Thanks.
I think I may have encountered a "real" bug while ordering in a subquery.
I have simplified it in the following exemples:
select json_group_array(o) from (
select json_object(
'id', sb.id
) o
from (
select 1 id, 2 field
) sb
)
the json_group_array returns an a
I don't recall that any (aggregate) function is concerned at all about the
order in which rows are visited. The effect is only visible in non-commutative
aggregates (e.g. concatenation).
If you want the arguments presented to an aggregate function in a specific
order, then you need to enforce t
6 matches
Mail list logo
