Weekly update in TiDB

Last week, we landed 63 PRs in the TiDB repository.

Added

• Support the RANK and DENSE_RANK window functions
• Add the CREATE ROLE support
• Support range framed window functions
• Compute and store the column order correlation with a handle

Improved

• Improve the performance of parsing date formatted values
• Improve the compatibility of modifying the column charset with MySQL
• Restrict the adjusting factor for the index feedback
• Control the chunk size for the selection and projection operators
• Handle default values for generated columns when adding an index
• Only cast aggregation's argument types when necessary
• Add the error count limit for DDL jobs
• Disallow generated columns to refer to columns with the AUTO_INCREMENT flag
• Reduce lock contention of the statistics collector
• Write the binlog record when the DDL job state is done
• Check the column field length when adding a column

Fixed

• Fix the rollback logic for the LOAD DATA statement
• Fix the wrong default value for time columns
• Only show valid columns in SHOW stats_histograms
• Fix a timezone misusage bug when converting the timestamp constant expression to the protobuf
• Fix the infinite retry issue when the sleep time is larger than max-txn-time
• Remove correlated columns from the items of the ORDER BY clause
• Only read the transaction buffer for dirty rows in UnionScan
• Fix a bug of modifying the column flag from null to not null
• Refine Compare methods of Merge Join to avoid incorrect results
• Fix wrong behaviors of PREPARE and EXECUTE
• Make Semi Join null- and empty-aware
• Fix a wrong result of UnionScan on partitioned tables

Weekly update in TiSpark

Last week, we landed 2 PRs in the TiSpark repository.

Improved

• Skip LockResolverTest when the PD client is not present

Weekly update in TiKV and PD

Last week, we landed 36 PRs in the TiKV and PD repositories.

Added

• Add thread context switch metrics
• Add a command line parameter to specify the metrics push address
• Add the encryption support for Importer
• Add make audit
• Add metrics for the key exceeding bound
• Add the batch executor interface

Improved

• Use the dyn keyword for traits objects
• Direct the local reader for raw read
• Remove the Tokio runtime
• Implement encoding functions for VectorValue and LazyBatchColumn(Vec)

Fixed

• Fix the encrypted file generated issue
• Fix the scatter range NPE when getting storeInfo of a non-existing store
• Fix the issue of getting tombstone stores
• Fix the merge panic when receiving multiple snapshots
• Fix range properties

New contributors (Thanks!)

tidb:

• guilhermehubner

tikv:

• crlf0710