Release 0.199#
General#
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()
, andhmac_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
andUNION ALL
, making it possible to execute aggregations with less peak memory usage.Change the signature of
round(x, d)
andtruncate(x, d)
functions so thatd
is of typeINTEGER
. Previously,d
could be of typeBIGINT
. This behavior can be restored with thedeprecated.legacy-round-n-bigint
config option or thelegacy_round_n_bigint
session property.Accessing anonymous row fields via
.field0
,.field1
, etc., is no longer allowed. This behavior can be restored with thedeprecated.legacy-row-field-ordinal-access
config option or thelegacy_row_field_ordinal_access
session property.Optimize the
ST_Intersection()
function for rectangles aligned with coordinate axes (e.g., polygons produced by theST_Envelope()
andbing_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 ofinformation_schema
tables.Improve the performance of map-to-map cast.
Improve the performance of
ST_Touches()
,ST_Within()
,ST_Overlaps()
,ST_Disjoint()
, andST_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#
Add support for installing on machines with OpenJDK.
Security#
Add support for authentication with JWT access token.
JDBC driver#
Make driver compatible with Java 9+. It previously failed with
IncompatibleClassChangeError
.
Hive#
Fix ORC writer failure when writing
NULL
values into columns of typeROW
,MAP
, orARRAY
.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
or2^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 namedexample$partitions
. This provides the same functionality and data asSHOW PARTITIONS
.Partition name listings, both via the
$partitions
table and usingSHOW PARTITIONS
, are no longer subject to the limit defined by thehive.max-partitions-per-scan
config option.Allow marking partitions as offline via the
presto_offline
partition property.
Thrift connector#
Most of the config property names are different due to replacing the underlying Thrift client implementation. Please see Thrift коннектор for details on the new properties.
SPI#
Allow connectors to provide system tables dynamically.
Add
resourceGroupId
andqueryType
fields toSessionConfigurationContext
.Simplify the constructor of
RowBlock
.Block.writePositionTo()
now closes the current entry.Replace the
writeObject()
method inBlockBuilder
withappendStructure()
.