Which is to say JSONB is useful, but I wouldn’t throw the relational baby out with the bath water.
However, a bit of duplication is not a terrible trade-off for significantly improved query performance.
Anyone else tried this approach? Anything I should know about it?
So the tables are much simpler to manage, much more portable, so I can serve search off scalable hardware without disturbing the underlying source of truth.
The downside is queries are more complex and slower.
My application couldn't really care less about customer names, for example. But the people who buy my software naturally do care - and what's worse, each of my potential customers has some legacy system which stores their customer names in a different way. So one problem I want to address is, how do I maintain fidelity with the old system, for example during the data migration, while enabling me to move forward quickly?
My solution has been to keep non-functional data such as customer names in JSON, and extract only the two or three fields that are relevant to my application, and put them into a regular SQL database table.
So far this has given me the best of both worlds: a simple and highly customisable JSON API for these user-facing objects with mutable shapes, but a compact SQL backend for the actual work.
Curious what other people think here?
I've now decided this path is a mistake because the performance is so bad. Even with low thousands/tens of thousands of rows it becomes a huge problem, queries that would take <1ms on relational stuff quickly start taking hundreds of ms.
Optimizing these with hand rolled queries is painful (I do not like the syntax it uses for jsonb querying) and for some doesn't really fix anything much, and indexes often don't help.
It seems that jsonb is just many many order of magnitudes slower, but I could be doing something wrong. Take for example storing a dictionary of string and int (number?) in a jsonb column. Adding the ints up in jsonb takes thousands of times longer rather than having these as string and int in a standard table.
Perhaps I am doing something wrong; and I'd love to know it if I am!
When jsonb works, it's incredible. I've had many... suboptimal experiences with mongo, and jsonb is just superior in my experience (although like I said, I haven't used it for performance critical stuff in production). For a long time, it kinda flew under the radar, and still remains an underappreciated feature of Postgres.
An "index on expression" should perform the same regardless of the input column types. All that matters is the output of the expression. Were you just indexing the whole jsonb column or were you indexing a specific expression?
For example, an index on `foo(user_id)` vs `foo(data->'user_id')` should perform the same.