Weekly update in TiDB

Last three weeks, we landed 74 PRs in the TiDB repository.

Added

• Add a session variable tidb_enable_cascades_planner
• Make explain support explain analyze
• Calculate radix bit number before building a hash table
• Add a benchmark test case of a complex query for the parser
• Support autoAnalyze for partitioned tables
• Add a force-priority option to the configuration file
• Add metrics for slow query logs
• Add a built-in function json_length
• Redesign the latch scheduler
• Add the admin show slow command

Improved

• Support IndexJoin over UnionScan
• Improve the precision for the Avg aggregation function
• Refine chunk.SwapColumn to rebuild the column reference
• Build the anti semi join plan for NOT EXISTS
• Refine chunk.SwapColumn to rebuild the column reference
• Make the sysdate built-in function unfoldable
• Improve the INFORMATION_SCHEMA.ENGINES table compatibility
• Improve the warning and error information for the group_concat() built-in function
• Change the default value of the performance_schema session variable to OFF
• Support show create table with the Compression option
• Optimize regionErr, notLeader, and staleEpoch error handling
• Speed up the recovery when the PD leader is down
• Support building TiDB on Windows operation system
• Allow alter table to append values to an enum typed column
• Return a table dual logical plan when the filter is false or null
• Refine the handle check to use the point get plan for a prepared statement
• Change FLUSH TABLES WITH READ LOCK to produce an error
• Remove some useless code and avoid some redundancy check
• Use the chunk grow for a simple executor

Fixed

• Fix a parser bug that an empty string in the ESCAPED BY subclause of FIELDS causes a panic
• Fix the statistics updating bug when there is no stats worker
• Fix the memory leak issue in statistics
• Check the privilege for show processlist
• Fix the log format for gc worker
• Fix a panic by avoiding the Send operation on the closed slow query channel
• Fix the results of the Command and Time fields in show processlist
• Fix the issue that the float type does not take effect in AddDate and SubDate built-in functions
• Exclude IsNull from the constant propagation
• Fix a bug in admin check table
• Return an error when the argument of the VALUES() function is not a column
• Fix the combined index low-bound check in statistics
• Fix a panic in the substring_index built-in function
• Fix the show create table result when the table has a Compression option
• Fix an issue of casting an unsigned integer to a decimal
• Fix a panic when the values for the PointGet operator are all null
• Fix the compatibility issue that updating the subquery table should be forbidden
• Fix some corner cases when parsing a string to a datetime
• Fix alter table add index on a virtual column bug

Weekly update in TiSpark

Last three weeks, we landed 4 PRs in the TiSpark repository.

Added

• Update Spark support to Spark 2.3

Fixed

• Fix the date timezone in tests
• Fix an issue that Explain statements do not work in Spark 2.3

Weekly update in TiKV and PD

Last three weeks, we landed 42 PRs in the TiKV and PD repositories.

Added

• Support more functions in Coprocessor:

  • InetAton
  • InetNtoa
  • Concat
  • Right
  • Sha2
  • TruncateDecimal
  • TruncateReal
  • Year
  • RouldReal
  • RoundDec
  • RoundWithFracDec
  • RoundWithFracInt
  • RoundWithFracReal
  • TruncateInt

• Fuzz codec::number

• Add RegionChangeObserver

• Get Regions with the top size

Fixed

• Fix the hanging problem when etcd fails to start
• Fix an issue of parsing a datetime from a string

Improved

• Improve the performance of Coprocessors md5 and sha1
• Unify the PD logger
• Check the level 0 files before ingesting
• Integrate the error code with pkg/errors to provide error tracing
• Upgrade Golang gRPC to 1.12.2
• Upgrade Rust gRPC to 0.4.0

New contributors (Thanks!)

tikv:

• Kingwl
• colinback

docs-cn:

• mccxj
• sweetIan