Using JSON_VALUE() when value contains unescaped double quotes - sql

Question

I have a table in the database where in one field (name of the field - JSONDetail) JSON is stored. Recently we encountered a problem where in this field in one of the values there are unescaped double quotes. It's due to migration from another system which allowed double quotes to be stored in the database without backslash before them.
Example (see field "comment"):
{
"noteId": "a34f17c4-f4fd-45ea-b4da-732ef8126a6b",
"memberName": "Test LINKOUS",
"tenantId": "548bead1-bdab-e811-bce7-0003ff21d46b",
"noteType": "General Note",
"memberId": "84cf0adb-850d-e711-80c8-000d3a103f46",
"createdOn": "2020-09-13T17:47:33.2864868Z",
"comment": "test "word" test",
"contacts": [
{
"otherContactType": "",
"communicationType": ""
}
]
}
We need to identify such cases in the database. I tried:
select JSON_VALUE (JSONDetail, '$.comment') as Comment
But instead of test "word" test, it returned
How can I return what is actually stored in key "comment"?

Answer 1

SQL-server does not have "fix_json" function
To find junk records
select *
from table
where ISJSON(json_col) = 0
Fix founded records via back-end language (php, c#, etc)
To prevent such behavior in future add constraint
ALTER TABLE table
ADD CONSTRAINT [record should be formatted as JSON]
CHECK (ISJSON(json_col)=1)

Answer 2

If comment keys are followed by contacts key throught the table within the JSONDetail column, then you can use the following code block which contains SUBSTRING(), PATINDEX(), TRIM() and LEN() functions to extract the whole value of the comment key in order to compare with the value extracted from JSON_VALUE (JSONDetail, '$.comment') :
WITH t(json_extracted,str) AS
(
SELECT JSON_VALUE (JSONDetail, '$.comment'),
SUBSTRING(
JSONDetail,
PATINDEX('%"comment"%', JSONDetail),
PATINDEX('%"contacts"%', JSONDetail)-PATINDEX('%"comment"%', JSONDetail)
)
FROM tab
), t2(json_extracted,str) AS
(
SELECT json_extracted,
TRIM(
SUBSTRING( str, PATINDEX('%:%', str) + 1,
PATINDEX('%,%', str) - PATINDEX('%:%', str) - 1 ) )
FROM t
)
SELECT SUBSTRING(str,2,LEN(str)-2) AS extracted_comment,
CASE WHEN json_extracted = SUBSTRING(str,2,LEN(str)-2)
THEN
'No'
ELSE
'Yes'
END AS "is_it_corrupted"
FROM t2
Demo

Answer 3

[EDIT] It wasn't practical to infer the field location in the JSON string based on length. Based on CHARINDEX search for the field names, this code finds and fixes the 'comments' in the JSON.
Data
drop table if exists #json_to_fix;
go
create table #json_to_fix(
json_col nvarchar(max));
declare #json nvarchar(max)=N'
{
"noteId": "a34f17c4-f4fd-45ea-b4da-732ef8126a6b",
"memberName": "Test LINKOUS",
"tenantId": "548bead1-bdab-e811-bce7-0003ff21d46b",
"noteType": "General Note",
"memberId": "84cf0adb-850d-e711-80c8-000d3a103f46",
"createdOn": "2020-09-13T17:47:33.2864868Z",
"comment": "test "word" test",
"contacts": [
{
"otherContactType": "",
"communicationType": ""
}
]
}';
insert #json_to_fix(json_col) values (#json);
Query
select s.not_escaped, fix.string_to_fix,
replace(fix.string_to_fix, '"', '') fixed
from #json_to_fix j
cross apply
(select charindex('"comment":', j.json_col, 1) strt_ndx) c_start
cross apply
(select charindex('"contacts"', j.json_col, c_start.strt_ndx) end_ndx) c_end
cross apply
(select substring(json_col, c_start.strt_ndx, c_end.end_ndx-c_start.strt_ndx-11) not_escaped) s
cross apply
(select substring(s.not_escaped, 13, len(s.not_escaped)-13) string_to_fix) fix
Output
not_escaped string_to_fix fixed
"comment": "test "word" test" test "word" test test word test