Preventing "invalid input syntax for json type" in Postgres

Question

I have a text column containing JSON as well as plan text. I want to convert it to JSON and then select a specific property. For instance:

user_data
_________
{"user": {"name": "jim"}}
{"user": {"name": "sally"}}
some random data string

I tried:

select user_data::json#>'{user,name}' from users

I get:

ERROR:  invalid input syntax for type json
DETAIL:  Token "some" is invalid.
CONTEXT:  JSON user_data, line 1: some...

Can this be prevented?

Answer 1

If you want to skip lines with invalid JSON, you must first check if the text is valid JSON. You can do this by creating a function that will try to parse the value and catch an exception for invalid JSON values.

CREATE OR REPLACE FUNCTION is_json(input_text varchar) RETURNS boolean AS $$
  DECLARE
    maybe_json json;
  BEGIN
    BEGIN
      maybe_json := input_text;
    EXCEPTION WHEN others THEN
      RETURN FALSE;
    END;

    RETURN TRUE;
  END;
$$ LANGUAGE plpgsql IMMUTABLE;

, is_json CASE WHERE, .

-- this can eliminate invalid values
SELECT user_data::json #> '{user,name}'
FROM users WHERE is_json(user_data);

-- or this if you want to fill will NULLs
SELECT
  CASE
    WHEN is_json(user_data)
      THEN user_data::json #> '{user,name}'
    ELSE
      NULL
  END
FROM users;

Answer 2

:

create or replace function is_json(text)
returns boolean language plpgsql immutable as $$
begin
    perform $1::json;
    return true;
exception
    when invalid_text_representation then 
        return false;
end $$;

:

with users(user_data) as (
values
    ('{"user": {"name": "jim"}}'),
    ('not json'),
    ('{"user": {"name": "sally"}}'),
    ('also not json')
)

select user_data::json#>'{user,name}' as name
from users
where is_json(user_data);

  name   
---------
 "jim"
 "sally"
(2 rows)