Details
-
Improvement
-
Status: Reopened
-
Trivial
-
Resolution: Unresolved
-
Impala 2.8.0
-
None
Description
If you attempt to create a Kudu table with a bogus value for the ENCODING attribute, the error message gives a list of attribute values that are supposed to be recognized:
create table xyz (x int primary key encoding foo) stored as kudu; ERROR: AnalysisException: Unsupported encoding value 'FOO'. Supported encoding values are: UNKNOWN, AUTO_ENCODING, PLAIN_ENCODING, PREFIX_ENCODING, GROUP_VARINT, RLE, DICT_ENCODING, BIT_SHUFFLE
However, several of these suggested values are not recognized or otherwise don't work:
create table xyz (x int primary key encoding unknown) partition by hash (x) partitions 2 stored as kudu; ERROR: IllegalStateException: Error parsing encoding/compression values for Kudu column 'x': Unsupported encoding: UNKNOWN
create table xyz (x int primary key encoding group_varint) partition by hash (x) partitions 2 stored as kudu; ERROR: ImpalaRuntimeException: Error creating Kudu table 'impala::jrussell.xyz' CAUSED BY: NonRecoverableException: invalid encoding for column 'x': encoding GROUP_VARINT not supported for type INT32
The spec of GROUP_VARINT in Kudu sounds like it should apply to all sizes of integers. I tried TINYINT, SMALLINT, INT, and BIGINT but Impala didn't recognize GROUP_VARINT encoding for any of those types.
Let's make the error message only refer to choices that are actually available.