Weekly update in TiDB

Last week, we landed 59 PRs in the TiDB repository.

Added

• Support drop session binding
• Support show session bindings
• Support add session binding
• Support show analyze status
• Support building CMSketch with Top-N
• Support incremental Analyze for indexes without feedback updates

Improved

• Add cop-tasks and memory information into the slow_query table
• Support building statistics on samples from fast Analyze
• Add memory tables to store TiKV status
• Check inconsistent indexes in IndexLookupExecutor
• Remove some unused gRPC metrics in the TiKV client
• Predefine metrics labels in the conn/session package to improve the performance
• Predefine metrics labels in the plan/executor package to improve the performance
• Predefine metrics labels in the store package to improve the performance
• Reduce alloc and lock-hold-time caused by counting errors and warnings
• Make SHOW COLLATIONS show supported collations only
• Lazily evaluate explainID and the memory tracker label to improve the performance
• Add pre_split_regions when creating a table with shard_row_id_bits
• Take more prefix columns into account when building ranges
• Support the column collate syntax create table t(col type collate...)
• Improve the row count estimation by using column order correlation
• Make Join reorder by dynamic programming work

Fixed

• Fix the panic when moving Analyze jobs to history
• Fix wrong PointGet judgement conditions in some cases
• Fix the issue that the slow log parser parses logs wrongly when a slow log includes #
• Fix the issue that distinct on multiple strings returns wrong results in some cases
• Fix the wrong behavior of group_concat(distinct) in some cases

Weekly update in TiSpark

Last week, we landed 14 PRs in the TiSpark repository.

Added

• Support Spark 2.4.x with the maven profile

Fixed

• Fix the retry method in Scan when encountering locks
• Fix the next() implementation in the point key
• Fix inconsistent timestamps of different tables in the same plan
• Fix the issue that the supported Spark version is not updated in the maven profile
• Fix the issue that the desc formatted table command for hive tables might lead to ArrayIndexOutOfBoundException

Weekly update in TiKV and PD

Last week, we landed 25 PRs in the TiKV and PD repositories.

Added

• Add a speed limit for tikv-importer
• Implement the period_add and period_diff Coprocessor functions
• Add the check command for operators
• Add the GetOperator service
• Add BatchSelectionExecutor
• Add the COUNT batch aggregate function

Improvement

• Disable unused features of dependencies
• Add storeID to metrics back
• Utilize the iter_batched_ref new feature of Criterion
• Add a timeout for the HTTP client
• Improve some logs
• Add jq and cache dependencies for the docker image
• Use the memory to cache split SST instead of disk files
• Upgrade the Rust compiler version to 2019-04-25

Fixed

• Fix the problem of not considering the learner in the merge process
• Fix the sending error failure in heartbeat stream

New contributors (Thanks!)

tikv:

• Seact
• jswh
• psinghal20