Re: json indexing and data types

On Wed, Dec 2, 2015 at 7:11 PM, Jim Nasby <Jim.Nasby@bluetreble.com> wrote:
> On 12/2/15 7:06 PM, Merlin Moncure wrote:
>>
>>  > The basics is, that I have a column with what is essentially json data;
>> a
>>  > number of data structures of different depths. Perhaps 10 - 30 top
>> levels,
>>  > and probably no more than 3, max 4 levels deep. In total there are some
>>  > hundred thousands of rows in each table. It would probably be best
>> stored as
>>  > jsonb. Right now it's text, because it's only used by the application
>>  > itself.
>
>
> After re-reading this part...
>
> Are you in control of the JSON itself, and are the number of permutations
> known in advance? It might be that something like table inheritance is a
> better solution...

Yeah, or other data storage strategies that are btree friendly but not
'schema rigid', like EAV (perhaps a modified variant to support
storing the document structure).  There are definitely tradeoffs
involved but you have to consider all the options.

merlin