how do I convert text to jsonB - postgresql

Question

What is the proper way to convert any text (or varchar) to jsonB type in Postgres (version 9.6) ?
For example, here I am using two methods and I am getting different results:
Method 1:
dev=# select '[{"field":15,"operator":0,"value":"1"},{"field":15,"operator":0,"value":"2"},55]'::jsonb;
jsonb
----------------------------------------------------------------------------------------------
[{"field": 15, "value": "1", "operator": 0}, {"field": 15, "value": "2", "operator": 0}, 55]
(1 row)
Method 2 , which doesn't produce the desired results, btw:
dev=# select to_jsonb('[{"field":15,"operator":0,"value":"1"},{"field":15,"operator":0,"value":"2"},55]'::text);
to_jsonb
----------------------------------------------------------------------------------------------------
"[{\"field\":15,\"operator\":0,\"value\":\"1\"},{\"field\":15,\"operator\":0,\"value\":\"2\"},55]"
(1 row)
dev=#
Here, it was converted to a string, not an array.
Why doesn't the second method creates an array ?

Answer 1

According to Postgres documentation:
to_jsonb(anyelemnt)
Returns the value as json or jsonb. Arrays and composites are
converted (recursively) to arrays and objects; otherwise, if there is
a cast from the type to json, the cast function will be used to
perform the conversion; otherwise, a scalar value is produced. For any
scalar type other than a number, a Boolean, or a null value, the text
representation will be used, in such a fashion that it is a valid json
or jsonb value.
IMHO you are providing a JSON formatted string, then you should use the first method.
to_json('Fred said "Hi."'::text) --> "Fred said \"Hi.\""
If you try to get an array of element using to_json(text) you'll get the next error:
select *
from jsonb_array_elements_text(to_jsonb('[{"field":15,"operator":0,"value":"1"},{"field":15,"operator":0,"value":"2"},55]'::text));
cannot extract elements from a scalar
But if you previously cast it to json:
select *
from jsonb_array_elements_text(to_jsonb('[{"field":15,"operator":0,"value":"1"},{"field":15,"operator":0,"value":"2"},55]'::json));
+--------------------------------------------+
| value |
+--------------------------------------------+
| {"field": 15, "value": "1", "operator": 0} |
+--------------------------------------------+
| {"field": 15, "value": "2", "operator": 0} |
+--------------------------------------------+
| 55 |
+--------------------------------------------+

Answer 2

If your text is just a json format text, you could just explicitly cast it to json/jsonb like this:
select '{"a":"b"}'::jsonb

Answer 3

Atomic type conversion and CSV-to-JSONb
A typical parse problem in open data applications is to parse line by line a CSV (or CSV-like) text into JSONB correct (atomic) datatypes. Datatypes can be defined in SQL jargon ('int', 'text', 'float', etc.) or JSON jargon ('string', 'number'):
CREATE FUNCTION csv_to_jsonb(
p_info text, -- the CSV line
coltypes_sql text[], -- the datatype list
rgx_sep text DEFAULT '\|' -- CSV separator, by regular expression
) RETURNS JSONb AS $f$
SELECT to_jsonb(a) FROM (
SELECT array_agg(CASE
WHEN tp IN ('int','integer','smallint','bigint') THEN to_jsonb(p::bigint)
WHEN tp IN ('number','numeric','float','double') THEN to_jsonb(p::numeric)
WHEN tp='boolean' THEN to_jsonb(p::boolean)
WHEN tp IN ('json','jsonb','object','array') THEN p::jsonb
ELSE to_jsonb(p)
END) a
FROM regexp_split_to_table(p_info,rgx_sep) WITH ORDINALITY t1(p,i)
INNER JOIN unnest(coltypes_sql) WITH ORDINALITY t2(tp,j)
ON i=j
) t
$f$ language SQL immutable;
-- Example:
SELECT csv_to_jsonb(
'123|foo bar|1.2|true|99999|{"x":123,"y":"foo"}',
array['int','text','float','boolean','bigint','object']
);
-- results [123, "foo bar", 1.2, true, 99999, {"x": 123, "y": "foo"}]
-- that is: number, string, number, true, number, object