On 6/5/15 3:51 PM, Alvaro Herrera wrote:
> Jim Nasby wrote:
>> On 6/5/15 2:08 PM, Petr Jelinek wrote:
>>> That's a good point, and it won't get any better if/when we add the json
>>> point support in 9.6 since the syntax would be something like select
>>> jsonb '{"a":"1", "b":"2", "c": {"a": "2"}}' - '/c/a'; and we will again
>>> silently do nothing. That's going to cause bugs in applications using this.
>>
>> Yeah, this is a miniature version of the pain I've felt with variant: trying
>> to get sane casting for a data type that encompasses other types in current
>> Postgres is essentially impossible.
>
> I'm not sure this is the same problem. But anyway I think requiring
> explicit casts in this stuff is a good thing -- relying on implicit
> cast to text, when most useful behavior uses other types, seems bad.
I'm not sure I'm following, at least for jsonb. If there's only jsonb -
json_pointer operator, why shouldn't we be able to resolve it? I suspect
part of the answer to that problem is that we need to make the
resolution of unknown smarter, or perhaps somehow configurable.
--
Jim Nasby, Data Architect, Blue Treble Consulting, Austin TX
Data in Trouble? Get it in Treble! http://BlueTreble.com