Vinod KC created SPARK-58793:
--------------------------------
Summary: Support ANSI SQL IS [NOT] JSON predicate
Key: SPARK-58793
URL: https://issues.apache.org/jira/browse/SPARK-58793
Project: Spark
Issue Type: Sub-task
Components: SQL
Affects Versions: 4.4.0
Reporter: Vinod KC
ANSI SQL defines the IS [NOT] JSON predicate for checking whether a string is
well-formed JSON, with an optional shape qualifier constraining the top-level
type:
{code:java}
expr IS [NOT] JSON [VALUE | ARRAY | OBJECT | SCALAR]{code}
{*}Shape qualifiers{*}:
* (none) / VALUE – any well-formed JSON value; VALUE is equivalent to the
unqualified form
* OBJECT – well-formed JSON object ( \{...})
* ARRAY – well-formed JSON array ([...])
* SCALAR – JSON primitive only: string, number, boolean, or null (not
object/array)
IS NOT JSON is the negation of IS JSON and accepts the same optional qualifier.
*Behavior:*
{code:sql}
SELECT '{"a":1}' IS JSON -- true
SELECT '[1,2]' IS JSON ARRAY -- true
SELECT '{"a":1}' IS JSON VALUE -- true (VALUE accepts objects, arrays,
and scalars)
SELECT '{"a":1}' IS JSON SCALAR -- false (SCALAR rejects objects and arrays)
SELECT '42' IS JSON SCALAR -- true
SELECT 'null' IS JSON SCALAR -- true (JSON null literal is a scalar)
SELECT '{"a":1}' IS JSON ARRAY -- false
SELECT 'bad' IS JSON -- false
SELECT 'bad' IS NOT JSON -- true
SELECT NULL IS JSON -- NULL (three-valued logic: NULL
propagates)
{code}
Note: IS JSON vs JSON_EXISTS
These are different predicates. IS JSON checks whole-value well-formedness with
no path navigation. JSON_EXISTS navigates inside a JSON document using a path
expression and checks for the existence of a matching node.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
