Ganesha S created SPARK-58729:
---------------------------------
Summary: Support the ANSI SQL/JSON JSON_OBJECT constructor function
Key: SPARK-58729
URL: https://issues.apache.org/jira/browse/SPARK-58729
Project: Spark
Issue Type: Sub-task
Components: SQL
Affects Versions: 4.3.0
Reporter: Ganesha S
Add the SQL:2016 JSON_OBJECT constructor (feature T811). It builds a JSON
object from key/value pairs written as key VALUE value or key : value, controls
null-valued keys via (NULL | ABSENT) ON NULL (default NULL ON NULL, i.e. null
values are emitted as "k": null), and supports an optional RETURNING type
(STRING default). Keys must be non-null strings; a null/non-string key raises
an error. Nested JSON_OBJECT / JSON_ARRAY compose.
Today the only workaround is to_json(named_struct(...)), which cannot express
ABSENT ON NULL and does not match the standard name used by Oracle, SQL Server
(2022+), MySQL, Trino, BigQuery, and DuckDB, forcing manual rewrites when
migrating.
Scope: add reserved tokens and a dedicated SqlBaseParser production (the
argument list carries non-expression syntax), an AstBuilder visitor, a new
Catalyst constructor expression reusing the existing Jackson JSON-generation
path (to_json / StructsToJsonEvaluator), a nullBehavior enum, error conditions
for null-key handling, docs, and golden-file + unit test coverage. VARIANT
RETURNING is deferred (STRING-only v1).
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
