12.47. Release 0.199
General Changes
- Allow users to create views for their own use when they do not have permission
to grant others access to the underlying tables or views. To enable this,
creation permission is now only checked at query time, not at creation time,
and the query time check is skipped if the user is the owner of the view.
- Add support for spatial left join.
- Add
hmac_md5(), hmac_sha1(), hmac_sha256(), and hmac_sha512() functions.
- Add
array_sort() function that takes a lambda as a comparator.
- Add
line_locate_point() geospatial function.
- Add support for
ORDER BY clause in aggregations for queries that use grouping sets.
- Add support for yielding when unspilling an aggregation.
- Expand grouped execution support to
GROUP BY and UNION ALL, making it possible
to execute aggregations with less peak memory usage.
- Change the signature of
round(x, d) and truncate(x, d) functions so that
d is of type INTEGER. Previously, d could be of type BIGINT.
This behavior can be restored with the deprecated.legacy-round-n-bigint config option
or the legacy_round_n_bigint session property.
- Accessing anonymous row fields via
.field0, .field1, etc., is no longer allowed.
This behavior can be restored with the deprecated.legacy-row-field-ordinal-access
config option or the legacy_row_field_ordinal_access session property.
- Optimize the
ST_Intersection() function for rectangles aligned with coordinate axes
(e.g., polygons produced by the ST_Envelope() and bing_tile_polygon() functions).
- Finish joins early when possible if one side has no rows. This happens for
either side of an inner join, for the left side of a left join, and for the
right side of a right join.
- Improve predicate evaluation performance during predicate pushdown in planning.
- Improve the performance of queries that use
LIKE predicates on the columns of information_schema tables.
- Improve the performance of map-to-map cast.
- Improve the performance of
ST_Touches(), ST_Within(), ST_Overlaps(), ST_Disjoint(),
and ST_Crosses() functions.
- Improve the serialization performance of geometry values.
- Improve the performance of functions that return maps.
- Improve the performance of joins and aggregations that include map columns.
Server RPM Changes
- Add support for installing on machines with OpenJDK.
Security Changes
- Add support for authentication with JWT access token.
JDBC Driver Changes
- Make driver compatible with Java 9+. It previously failed with
IncompatibleClassChangeError.
Hive Changes
- Fix ORC writer failure when writing
NULL values into columns of type ROW, MAP, or ARRAY.
- Fix ORC writers incorrectly writing non-null values as
NULL for all types.
- Support reading Hive partitions that have a different bucket count than the table,
as long as the ratio is a power of two (
1:2^n or 2^n:1).
- Add support for the
skip.header.line.count table property.
- Prevent reading from tables with the
skip.footer.line.count table property.
- Partitioned tables now have a hidden system table that contains the partition values.
A table named
example will have a partitions table named example$partitions.
This provides the same functionality and data as SHOW PARTITIONS.
- Partition name listings, both via the
$partitions table and using
SHOW PARTITIONS, are no longer subject to the limit defined by the
hive.max-partitions-per-scan config option.
- Allow marking partitions as offline via the
presto_offline partition property.
Thrift Connector Changes
- Most of the config property names are different due to replacing the
underlying Thrift client implementation. Please see Thrift Connector
for details on the new properties.
SPI Changes
- Allow connectors to provide system tables dynamically.
- Add
resourceGroupId and queryType fields to SessionConfigurationContext.
- Simplify the constructor of
RowBlock.
Block.writePositionTo() now closes the current entry.- Replace the
writeObject() method in BlockBuilder with appendStructure().
