On Mon, Dec 5, 2011 at 3:12 PM, Bruce Momjian <bruce@momjian.us> wrote:
> Where are we with adding JSON for Postgres 9.2? We got bogged down in
> the data representation last time we discussed this.
We're waiting for you to send a patch that resolves all
previously-raised issues. :-)
In all seriousness, I think the right long-term answer here is to have
two data types - one that simply validates JSON and stores it as text,
and the other of which uses some binary encoding. The first would be
similar to our existing xml datatype and would be suitable for cases
when all or nearly all of your storage and retrieval operations will
be full-column operations, and the json types is basically just
providing validation. The second would be optimized for pulling out
(or, perhaps, replacing) pieces of arrays or hashes, but would have
additional serialization/deserialization overhead when working with
the entire value. As far as I can see, these could be implemented
independently of each other and in either order, but no one seems to
have yet found the round tuits.
--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company