从 JSON 数组字段中查找和/或删除某些元素的 PostgreSQL 查询

Question

We have a table with a jsonb[] column that has an array of JSON objects with various keys. The DB is PostgreSQL (v9.6)

----------------------------------------------------
| id | data (jsonb[])                              |
----------------------------------------------------
|  1 | [{a:1, B:2, c:3}, {a:4, c:5}, {a:6, B:7}]   |
|  2 | [{a:8}, {B:9, c:10}]                        |
|  3 | [{a:11}]                                    |
----------------------------------------------------

We'd like to remove or set to empty all the values of B:

so, we'd like to get

---------------------------------------------------------------
| id | data (jsonb[])                                         |
---------------------------------------------------------------
|  1 | [{a:1, B:null/"", c:3}, {a:4, c:5}, {a:6, B:null/""}]  |
|  2 | [{a:8}, {B:null/"", c:10}]                             |
|  3 | [{a:11}]                                               |
---------------------------------------------------------------

or

------------------------------------------
| id | data (jsonb[])                    |
------------------------------------------
|  1 | [{a:1, c:3}, {a:4, c:5}, {a:6}]   |
|  2 | [{a:8}, {c:10}]                   |
|  3 | [{a:11}]                          |
------------------------------------------

If there is no simple way to update the B field, we'd be happy to find a way to get all records that have B: in data field, in that case, we can load all the record and process them outside DB.

So far we can't find a way to do it, not even to find all records that have the B: field.

Any ideas will be great, thank you!

Answer 1

You can achieve it by UNNEST Array --> Subtract Element --> Aggregate approach

If you want to remove the Element then try this:

select 
id,
jsonb_agg(y-'B') 
from test cross join lateral unnest(data) x(y)
group by id

DEMO

If you want to set the value of element NULL then try this:

select 
id, 
jsonb_agg(jsonb_set(y,'{B}','null',false)) 
from test cross join lateral unnest(data) x(y)
group by id

DEMO

EDIT as per Comment:

with cte as (
select 
id,
array_agg(y-'B') "data"
from test cross join lateral unnest(data) x(y)
group by id
)

update test t1 set data = t2.data from cte t2 where t1.id=t2.id;

FIDDLE