On 5/25/21 4:31 PM, Tom Lane wrote:
> Andrew Dunstan <andrew@dunslane.net> writes:
>> On 5/25/21 4:10 PM, Tom Lane wrote:
>>> Also, even if ZSON was "100% compatible with JSONB" back in 2016,
>>> a whole lot of features have been added since then. Having to
>>> duplicate all that code again for a different data type is not
>>> something I want to see us doing. So that's an independent reason
>>> for wanting to hide this under the existing type not make a new one.
>> I take your point. However, there isn't really any duplication. It's
>> handled by [ creating a pair of casts ]
> If that were an adequate solution then nobody would be unhappy about
> json vs jsonb. I don't think it really is satisfactory:
>
> * does nothing for user confusion (except maybe make it worse)
>
> * not terribly efficient
>
> * doesn't cover all cases, notably indexes.
>
>
Quite so. To some extent it's a toy. But at least one of our customers
has found it useful, and judging by Aleksander's email they aren't
alone. Your ideas downthread are probably a useful pointer of how we
might fruitfully proceed.
cheers
andrew
--
Andrew Dunstan
EDB: https://www.enterprisedb.com