Hyukjin Kwon created SPARK-59101:
------------------------------------

             Summary: Release the Rust Spark Connect client from a chosen 
apache upstream branch on manual dispatch
                 Key: SPARK-59101
                 URL: https://issues.apache.org/jira/browse/SPARK-59101
             Project: Spark
          Issue Type: Improvement
          Components: Project Infra
    Affects Versions: 5.0.0
            Reporter: Hyukjin Kwon


The Rust-backed Spark Connect client (apache/spark-connect-rust) drop-in 
under-exposes
several public pyspark v4.2.0 APIs. A rigorous introspection-based audit of the 
v4.2.0
pyspark.sql.connect.* surface vs the current drop-in found the following gaps, 
to be
closed while mirroring the official pyspark implementation exactly and keeping 
test coverage:

Layer A (Rust core already implements it; only Python exposure missing):
- functions higher-order: transform, filter, exists, forall, aggregate, reduce, 
zip_with,
  transform_keys, transform_values, map_filter, map_zip_with
- functions misc: cume_dist, broadcast, call_function, call_udf, column
- DataFrame.repartitionById; RuntimeConfig.getAll
- DataType/StructType/StructField: json, jsonValue, simpleString, typeName, 
fieldNames, fromDDL

Layer B (missing in both core and Python):
- Catalog: analyzeTable, createDatabase, dropDatabase, dropTable, dropView,
  getCreateTableString, getTableProperties, listPartitions, listViews, 
truncateTable
- types model: toDDL, treeString, toNullable, fromJson, fromInternal, 
toInternal, needConversion
- DataFrame.zipWithIndex; Column.transform; DataFrameStatFunctions.sampleBy
- DataFrameReader.changes; DataStreamReader.changes; DataStreamReader.xml; 
StreamingQueryManager.close
- functions: udtf, arrow_udtf, arrow_udf
- UDFRegistration.registerJavaFunction, registerJavaUDAF
- ml.connect base classes / param mixins / *Model / ArrayAssembler / 
TorchDistributor
- Row.count, Row.index

Module structure: expose pyspark.sql.avro.functions, 
pyspark.sql.protobuf.functions,
pyspark.sql.functions.partitioning submodules (functions currently exist only 
flat).




--
This message was sent by Atlassian Jira
(v8.20.10#820010)

---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to