plsql code to parse JSON without using third party libraries - json

Question

JSON needs to be parsed using only PL/SQL code like regular expressions to get sentiment and confidence values out of it.
Something similar to this
[
{
"sentiment":"negative",
"confidence":0.6211975044276729
},
{
"sentiment":"neutral",
"confidence":0.3510681601407111
},
{
"sentiment":"positive",
"confidence":0.027734335431616075
}
]
above JSON needs to be parsed to get sentiment and confidence values out of it

Answer 1

The JSON_TABLE function is available starting with Oracle Database 12c Release 1 (12.1.0.2).
SET NUMWIDTH 20 --Use this if SQL*Plus/ SQL developer truncates digits.
--test data
WITH t ( json_col ) AS ( SELECT '[
{
"sentiment":"negative",
"confidence":0.6211975044276729
},
{
"sentiment":"neutral",
"confidence":0.3510681601407111
},
{
"sentiment":"positive",
"confidence":0.027734335431616075
}
]'
FROM dual
) --test data ends
SELECT j.*
FROM t
CROSS JOIN
JSON_TABLE ( json_col,'$[*]'
COLUMNS (
sentiment VARCHAR2(20) PATH '$.sentiment',
confidence NUMBER PATH '$.confidence'
)
)
j;
SENTIMENT CONFIDENCE
-------------------- --------------------
negative .6211975044276729
neutral .3510681601407111
positive .027734335431616075

Answer 2

If you really don't want to use any of the built-in JSON functions and your input does not involve any nested objects then you can use SQL with a recursive sub-query factoring clause:
Oracle Setup:
CREATE TABLE test_data ( json ) AS
SELECT '[
{
"sentiment":"negative",
"confidence":0.6211975044276729
},
{
"confidence":0.3510681601407111,
"sentiment":"neutral"
},
{
"sentiment":"positive",
"confidence":0.027734335431616075
}
]' FROM DUAL
Query:
WITH rsqfc ( json, obj, lvl, cnt ) AS (
SELECT json,
REGEXP_SUBSTR( json, '\{(.*?)\}', 1, 1, 'n' ),
1,
REGEXP_COUNT( json, '\{(.*?)\}', 1, 'n' )
FROM test_data
WHERE REGEXP_COUNT( json, '\{(.*?)\}', 1, 'n' ) > 1
UNION ALL
SELECT json,
REGEXP_SUBSTR( json, '\{(.*?)\}', 1, LVL + 1, 'n' ),
lvl + 1,
cnt
FROM rsqfc
WHERE lvl < cnt
)
SELECT REGEXP_SUBSTR( obj, '"sentiment":\s*"(negative|neutral|positive)"', 1, 1, 'n', 1 ) AS sentiment,
TO_NUMBER( REGEXP_SUBSTR( obj, '"confidence":\s*(\d+(\.\d*)?)', 1, 1, 'n', 1 ) ) AS confidence
FROM rsqfc
Output:
SENTIMENT | CONFIDENCE
:-------- | ------------------:
negative | .6211975044276729
neutral | .3510681601407111
positive | .027734335431616075
PL/SQL:
Or using PL/SQL:
DECLARE
json CLOB := '[
{
"sentiment":"negative",
"confidence":0.6211975044276729
},
{
"confidence":0.3510681601407111,
"sentiment":"neutral"
},
{
"sentiment":"positive",
"confidence":0.027734335431616075
}
]';
cnt PLS_INTEGER;
obj VARCHAR2(4000);
sentiment VARCHAR2(20);
confidence NUMBER;
BEGIN
cnt := REGEXP_COUNT( json, '\{(.*?)\}', 1, 'n' );
FOR i IN 1 .. cnt LOOP
obj := REGEXP_SUBSTR( json, '\{(.*?)\}', 1, i, 'n' );
sentiment := REGEXP_SUBSTR( obj, '"sentiment":\s*"(negative|neutral|positive)"', 1, 1, 'n', 1 );
confidence := TO_NUMBER( REGEXP_SUBSTR( obj, '"confidence":\s*(\d+(\.\d*)?)', 1, 1, 'n', 1 ) );
DBMS_OUTPUT.PUT_LINE( sentiment || ' - ' || confidence );
END LOOP;
END;
/
Output:
dbms_output:
negative - .6211975044276729
neutral - .3510681601407111
positive - .027734335431616075
db<>fiddle here