How do I modify a single field inside the PostgreSQL JSONB datatype?
Let's say I have a table called animal like this:
id info
------------------------------------------------------------
49493 {"habit1":"fly","habit2":"dive","location":"SONOMA NARITE"}
I'd like to simply change value(say, to upper case or lower case the text) of the location property. so the result, after UPDATE is
id info
------------------------------------------------------------
49493 {"habit1":"fly","habit2":"dive","location":"sonoma narite"}
I tried this below and it does not work
update animal set info=jsonb_set(info, '{location}', LOWER(info->>'location'), true) where id='49493';
----------------------------------
ERROR: function jsonb_set(jsonb, unknown, text, boolean) does not exist
LINE 7: update animal set info=jsonb_set(info, '{con...
^
HINT: No function matches the given name and argument types. You might need to add explicit type casts.
********** Error **********
ERROR: function jsonb_set(jsonb, unknown, text, boolean) does not exist
if I simply know what the updated value would be then I can use just use this:
update animal set info=jsonb_set(info, '{location}', '"sonoma narite"', true) where id='49493';
However, if the text value is unknown and we just want to do some simply operation such as append, prepend, upper/lower case, I can't simply find an answer to it.
I was surprised by the fact that jsonb set function does not offer such a trivial operation that only try to update the case of a text property inside a jsonb.
Can someone help?
The third argument of jsonb_set()
should be of jsonb
type. The problem is in casting a text string to jsonb string, you need a string in double quotes. You can use concat()
or format()
:
update animal
set info =
jsonb_set(info, '{location}', concat('"', lower(info->>'location'), '"')::jsonb, true)
-- jsonb_set(info, '{location}', format('"%s"', lower(info->>'location'))::jsonb, true)
where id='49493'
returning *;
id | info
-------+------------------------------------------------------------------
49493 | {"habit1": "fly", "habit2": "dive", "location": "sonoma narite"}
(1 row)
In Postgres 9.4 you should unnest the json column using jsonb_each_text(), aggregate keys and values modifying the proper value on the fly, and finally build a json object:
update animal a
set info = u.info
from (
select id, json_object(
array_agg(key),
array_agg(
case key when 'location' then lower(value)
else value end))::jsonb as info
from animal,
lateral jsonb_each_text(info)
group by 1
) u
where u.id = a.id
and a.id = 49493;
If you can create functions this solution might be more pleasant:
create or replace function update_info(info jsonb)
returns jsonb language sql as $$
select json_object(
array_agg(key),
array_agg(
case key when 'location' then lower(value)
else value end))::jsonb
from jsonb_each_text(info)
$$
update animal
set info = update_info(info)
where id = 49493;