tidb-serverv0.0.0

Compatibility Changes

• Change drop partition and truncate partition's job arguments to support the ID array of multiple partitions #18930
• Add the delete-only state for checking add partition replicas #18865

Improvements

• Optimize the performance of DecodePlan for big union queries #18941
• Reduce the number of GC lock scans when meeting the Region cache miss error #18876
• Ease the impact of statistical feedback on cluster performance #18772
• Support canceling operations before the RPC response is returned #18580
• Add the HTTP API to generate the TiDB metric profile #18531
• Support scattering partitioned tables #17863
• Add detailed memory usage of each instance in Grafana #18679
• Show the detailed runtime information of the BatchPointGet executor in the result of EXPLAIN #18892
• Show the detailed runtime information of the PointGet executor in the result of EXPLAIN #18817
• Warn the potential deadlock for Consume in remove() #18395
• Support the Action when memory exceed quota for TableReader Executor #18392
• Refine the behaviors of StrToInt and StrToFloat and support converting JSON to the date, time, and timestamp types #18159
• Support limiting the memory usage of the TableReader executor #18392
• Avoid too many times of backoff when retrying the batch cop request #18999
• Improve compatibility for ALTER TABLE algorithms #19270
• Make the single partitioned table support IndexJoin on the inner side #19151
• Support searching the log file even when the log includes invalid lines #18579

Bug Fixes

• Fix the should ensure all columns have the same length error that occurs because the ErrTruncate/Overflow error is incorrectly handled in the builtinCastRealAsDecimalSig function #18967
• Fix the issue that the pre_split_regions table option does not work in the partitioned table #18837
• Fixe the issue that might cause a large transaction to be terminated prematurely #18813
• Fix the issue that using the collation functions get wrong query results #18735
• Fix the bug that the getAutoIncrementID() function does not consider the tidb_snapshot session variable, which might cause the dumper tool to fail with the table not exist error #18692
• Fix the unknown column error for SQL statement like select a from t having t.a #18434
• Fix the panic issue that writing the 64-bit unsigned type into the hash partitioned table causes overflow and gets an unexpected negative number when the partition key is the integer type #18186
• Fix the wrong behavior of the char function #18122
• Fix the issue that the ADMIN REPAIR TABLE statement cannot parse integer in the expressions on the range partition #17988
• Fix the wrong behavior of the SET CHARSET statement #17289
• Fix the bug caused by the wrong collation setting which leads to the wrong result of the collation function #17231
• Fix the issue that STR_TO_DATE's handling of the format tokens '%r', '%h' is inconsistent with that of MySQL #18727
• Fix issues that the TiDB version information is inconsistent with that of PD/TiKV in the cluster_info table #18413
• Fix the existent checks for pessimistic transactions #19004
• Fix the issue that executing union select for update might cause concurrent race #19006
• Fix the wrong query result when apply has a child of the PointGet operator #19046
• Fix the incorrect result that occurs when IndexLookUp is in the inner side of the Apply operator #19496
• Fix the incorrect result of anti-semi-join queries #19472
• Fix the incorrect result caused by the mistaken usage of BatchPointGet #19456
• Fix the incorrect result that occurs when UnionScan is in the inner side of the Apply operator #19447
• Fix the panic caused by using the EXECUTE statement to print an expensive query log #17419
• Fix the index join error when the join key is ENUM or SET #19235
• Fix the issue that the query range cannot be built when the NULL value exists on the index column #19358
• Fix the data race issue caused by updating the global configuration #17964
• Fix the panic issue caused when modifying the character set in an uppercase schema #19286
• Fix an unexpected error caused by changing the temporary directory during the disk spill action #18970
• Fix the wrong hash key for the decimal type #19131
• Fix the issue that the PointGet and BatchPointGet operators do not consider the partition selection syntax and get incorrect results #19141
• Fix the incorrect results when using the Apply operator together with the UnionScan operator #19104
• Fix the bug that causes the indexed virtual generated column to return wrong value #17989
• Add the lock for runtime statistics to fix a panic caused by concurrent execution #18983
• Check ErrTruncate/Overflow locally for builtinCastRealAsDecimalSig to fix the should ensure all columns have the same length error #18961