Hyukjin Kwon created SPARK-61859:
------------------------------------

             Summary: Make the base data type classes instantiable as 
PySpark's, and give the methods of the types PySpark's parameter names
                 Key: SPARK-61859
                 URL: https://issues.apache.org/jira/browse/SPARK-61859
             Project: Spark
          Issue Type: Bug
          Components: Connect
    Affects Versions: connect-rust-4.4.0
            Reporter: Hyukjin Kwon


The base classes of the data types of PySpark, DataType, AtomicType, 
NumericType, IntegralType, FractionalType, DatetimeType, AnyTimeType, 
SpatialType and AnsiIntervalType, can be instantiated in PySpark 4.3.0: XType() 
has the repr XType(), typeName()/simpleString()/jsonValue() of its stem 
("atomic", ...), needConversion() False, identity toInternal/fromInternal, 
equality by class, and pickles as itself; IntegralType is a singleton. In the 
drop-in they raise "cannot create 'pyspark.sql.types.FractionalType' instances" 
(and DataType() pickles to NullType()), which fails the official 
pandas-on-Spark test data_type_ops/test_parity_base.py::test_data_type_ops.

Also, the parameters of some methods of the types have other names than 
PySpark's: StringType.collationProvider(collationName), 
DateType.toInternal(d)/fromInternal(v), 
TimeType.toInternal(t)/fromInternal(nano), TimestampType and TimestampNTZType 
toInternal(dt)/fromInternal(ts), 
DayTimeIntervalType.toInternal(dt)/fromInternal(micros), 
GeographyType.toInternal(geography), GeometryType.toInternal(geometry).




--
This message was sent by Atlassian Jira
(v10.3.26#10030026)

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

Reply via email to