13.12. Release 0.231
There is a bug in this release that will cause queries with the predicate on bucketed columns to produce incorrect results.
- Fix failures caused by invalid plans for queries with repeated lambda expressions in the order by clause.
- Fix accounting of raw input bytes in ScanFilterAndProjectOperator.
- Fix an issue where server fails to start when two function namespace managers of the same type are specified.
- Fix SQL function compilation error when input parameters contain lambda.
- Fix
IS DISTINCT FROM
for long decimals. - Improve performance of
array_join()
. - Improve efficiency of . The new implementation produces fewer buffer points on rounded corners, which will produce very similar but different results. It also better handles buffering with small (<1e-9) distances.
- Improve OOM killer log output to put memory heavy nodes and queries first.
- Improve efficiency by using JTS instead of Esri for many geometrical operations and fix incorrect calculation of extreme points in certain cases (#14031).
- Add forms of accepting an accuracy parameter.
- Add a new session property to configure partition merging strategy when adding exchange around aggregation node (#11262).
- Add
IGNORE NULLS
clause to various .
- Fix schema mismatch with Parquet
INT64
&Timestamp
. - Improve string column stats by including the sum of all strings’ lengths when the individual min/max of the string is too long in ORC writer.
- Allow reading data from HDFS while caching the fetched data on local storage. Turn on the feature by specifying the cache directory config
cache.base-directory
. - Add support for the AWS Default Credentials Provider for S3.
- Add the ability to read and write with delta deletes.
- Fix an issue where checksum query text and ID are not recorded if the checksum query fails.
- Improve verifier to skip verification when checksum query fails to compile.
- Add checks for the cardinalities sum when validating an array column.
- Add new columns and
test_session_properties
toverifier_queries
, and remove columnsession_properties_json
. The value of the removed column can be copied to the two new columns for the schema change. - Add details of determinism analysis runs to the output.
- Add configuration property
max-determinism-analysis-runs
to control maximum number of determinism analysis runs in case of column mismatch. - Add configuration property to allow disabling teardown for determinism analysis runs.
- Add API to check if the query is unexpectedly modified using the credentials passed in the identity.