Weekly update in TiDB

Last week, we landed 61 PRs in the TiDB repository.

Improved

• Improve the unit test coverage of util/set from 0% to 100%
• Support a fast Analyze session control variable
• Improve the error message for connection closing
• Remove the errors.Trace call in some high-level packages to improve the performance
• Correct the estimated row count for the inner plan of IndexJoin
• Replace lock operations in MemTracker with atomic operations to improve the performance
• Improve the unit test coverage of executor/aggfuncs from 47.7% to 70%
• Improve the unit test coverage of global_vars_cache from 0% to 100%
• Raise a warning for unmatched Join hint in some cases
• Improve the unit test coverage of infoschema from 65.20% to 85.2%
• Push down constant filters over Join properly
• Allow kill tidb [session id] to stop executors and release resources quickly
• Improve the performance of computing FieldType.Flen of int64/uint64
• Make TableReader&IndexReader&IndexLookup support the chunk size control

Fixed

• Fix the issue that Desc table is incompatible with MySQL in some cases
• Fix the issue that JSON_contains_path is incompatible with MySQL
• Fix the issue that the dayname function is incompatible with MySQL when doing arithmetic
• Fix wrong results of group_concat in some cases
• Fix the data race issue in config.TreatOldVersionUTF8AsUTF8MB4
• Fix incorrect time fraction parsing method
• Fix wrong results of count(distinct)/sum(distinct) of decimal columns in some cases
• Make the Alter and Rename privileges the same with MySQL
• Fix the Alter table charset bug and compatibility in some cases
• Fix the read-only check for the Prepare/Execute statement
• Fix improper using of the time zone in checking default column values

Weekly update in TiSpark

Last week, we landed 9 PRs in the TiSpark repository.

Added

• Support range partition pruning

Fixed

• Fix the issue that ManagedChannel is not closed in the test

Weekly update in TiKV and PD

Last week, we landed 21 PRs in the TiKV and PD repositories.

Added

• Add the execution summary framework
• Introduce the vectorized evaluation framework
• Add one more configuration option for very fast builds
• Make end-point-enable-batch-if-possible configurable
• Use static metrics for KV_COMMAND_COUNTER
• Enlarge the ingest buffer

Fixed

• Fix the performance bug
• Improve the Region heartbeat
• Fix the issue that the Region scatter might transfer the leader to a removed peer

New contributors (Thanks!)

tidb:

• Debiancc

tikv:

• myguyy
• seansu4you87
• yaalsn

pd:

• yaalsn

parser:

• wuudjac