20.37. Release 0.185
- Fix incorrect column names in
- Fix excessive CPU usage in coordinator for queries that have
large string literals containing non-ASCII characters.
- Fix potential infinite loop during query optimization when constant
expressions fail during evaluation.
- Fix incorrect ordering when the same field appears multiple times
with different ordering specifications in a window function
clause. For example:
OVER (ORDER BY x ASC, x DESC).
- Do not allow dropping or renaming hidden columns.
- When preparing to drop a column, ignore hidden columns when
checking if the table only has one column.
- Improve performance of joins where the condition is a range over a function.
a JOIN b ON b.x < f(a.x) AND b.x > g(a.x)
- Improve performance of certain window functions (e.g.,
LAG) with similar specifications.
substr() function to work on
VARBINARY in addition to
- Add cast from
- Allow usage of
TRY within lambda expressions.
- Improve ORC reader efficiency by only reading small ORC streams when accessed in the query.
- Improve RCFile IO efficiency by increasing the buffer size from 1 to 8 MB.
- Fix native memory leak for optimized RCFile writer.
- Fix potential native memory leak for optimized ORC writer.