Hyukjin Kwon created SPARK-61116:
------------------------------------
Summary: Port the Read, Range and SQL plans, Column.to_plan, and
the type and conversion helpers of the connect import path, in the Python
client as PySpark does
Key: SPARK-61116
URL: https://issues.apache.org/jira/browse/SPARK-61116
Project: Spark
Issue Type: Bug
Components: Connect
Affects Versions: connect-rust-4.3.0
Reporter: Hyukjin Kwon
With the generated messages of {{pyspark.sql.connect.proto}} and
{{DataFrame(plan, session)}} (SPARK-61111), the rest of the connect import path
that builds or reads messages can be PySpark's:
* In {{pyspark.sql.connect.plan}}, {{Read}}, {{Range}} and {{SQL}} are
placeholders. Their constructors take any arguments and build no relation, so
{{DataFrame(Range(0, 3, 1), spark)}} fails. PySpark's build the {{read}},
{{range}} and {{sql}} relations, which
{{pyspark.testing.connectutils.PlanOnlyTestFixture}} uses.
* {{Column.to_plan(session)}}, the {{spark.connect.Expression}} of a column, is
missing.
* {{pyspark.sql.connect.types}} lacks {{parse_attr_name}}, {{verify_col_name}}
and {{verify_numeric_col_name}}.
* {{pyspark.sql.connect.conversion}} ({{storage_level_to_proto}},
{{proto_to_storage_level}} and {{proto_to_remote_cached_dataframe}}) is missing.
The fix is to port them from Apache Spark v4.3.0, over the messages of the core.
--
This message was sent by Atlassian Jira
(v10.3.26#10030026)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
