Weekly update in TiDB

Last week, we landed 44 PRs in the TiDB repository.

Added

• Support the tidb_txn_mode session variable
• Support dump/load correlation of statistics histograms
• Add a make ddltest target to generate ddltest binary data
• Add the tidb_low_resolution_tso session variable for reading data with low resolution TSO
• Support View related privileges in mysql.tables_priv
• Support LOAD DATA...IGNORE/REPLACE in the syntax

Improved

• Make pessimistic lock TTL easier to set and validate the value
• Refine the error message in latches
• Enhance index join for more scenarios
• Update 2019-04-11-indexmerge.md with some edits
• Fix a potential goroutine leak in distSQL
• Add a flag to indicate that the TiDB server is closing
• Validate table information before doing DDL jobs to avoid potential loading infoschema errors
• Merge Window functions with the same specification name
• Only retry needed Regions when getting Region row count during fast Analyze
• Cache the vendor to speed up Docker building

Fixed

• Correct the behavior of the IndexJoin hint for SemiJoin
• Fix notLeader in the Region cache
• Fix the show grants result for RBAC
• Fix the issue that the OOM panic is not recovered currently in the distSQL layer
• Fix a bug in the AlterSchema job and add more tests
• Check the table name for columns in the partition expression
• Fix the issue that ByItem does not filter NULL out
• Fix the issue that canceling range tasks might not return errors
• Fix a GetPartition function bug when adding indexes
• Fix the issue that the maximum value of UINT64 is returned when -1 is used for point-get
• Fix the issue that TiDB cannot be used for a long time under extreme conditions if relying only on the information in PD

Weekly update in TiSpark

Last week, we landed 17 PRs in the TiSpark repository.

Fixed

• Fix index scans on partition tables
• Fix the issue that RegionVerId missed to implement equals()
• Fix the issue that the default value of Integer is not parsed to Long
• Fix the issue that FromSparkType() incorrectly transforms Integer to Long
• Fix the issue that KeyNotInRegion might occur when retrieving rows by handles
• Fix the issue of encodeValue with unsigned comparison

Weekly update in TiKV and PD

Last week, we landed 45 PRs in the TiKV and PD repositories.

Added

• Add the statistics of the store flow
• Add BatchSlowHashAggregationExecutor
• Add configuration items and enable pessimistic-txn by default
• Add RPN functions
  • Int, Real and Decimal
  • IsNull, IsTrue and IsFalse
• Support pessimistic transactions (experimental feature)
• Add the ReadIndex service
• Add hibernate Regions
• Add operator limit for stores

Improved

• Remove the local reader thread
• Retry to connect PD for infinity, and log periodically
• Remove the scheduler thread

Fixed

• Fix the issue that snapshots might lose some applied results

New contributors (Thanks!)

tidb:

• sundl123
• sunxiaoguang

tikv:

• wjhuang2016

parser:

• sundl123