Merged PRs

dolt

• 9236: go: sqle/statspro: prollyStats: FlushFromMem: Fix long-held lock when flushing stats to disk.
  For sql-server instances with many databases, many branches or many tables and indexes, flushing stats could cause queries to block until stats were finished flushing. StatsController attempted to rate limited how much I/O bandwidth would be used to flush stats, but this is currently counter-productive because all queries are blocked from analyzing successfully until stats are flushed.
  This changes it back to flush stats to disk as quickly as possible when we need to do it. Computing stats is already rate limited, so for large flushes this should not happen too often.
  We will continue to improve things here so that stats writes have less impact on server resource utilization and do not block user-facing work as they proceed.
• 9229: When performing a three-way merge, Use a patchBuffer to build the new primary index instead of a MutableMap.
  MutableMaps are designed for caching point modifications to a table. But during merge, the new primary index is computed sequentially. There's no benefit to using a MutableMap here.
  MutableMaps are built on top of the ApplyMutations function, which takes a sequential stream of modifications (called a PatchBuffer) and applies them to a chunker. This has the added benefit of being parallelizable: the patches are produced in one goroutine and consumed in another.
  Instead of using the MutableMap, we can extract the underlying PatchBuffer and use it directly. This should be both more performant, and more correct as it avoids a failure case with schema merges where the MutableMap flushes changes to disk and writes a chunk containing rows with different schemas.
• 9215: go/store/nbs: table_index.go: Fix uint32 overflow in findPrefix when working with large table files.

go-mysql-server

• 2985: fix default value for non null enum columns
  In MySQL non null enum columns default to the very first enum instead of throwing a missing default error.
  fixes: #8617
• 2983: fix explain plan on server engine
  fixes: #8882
• 2982: Bug fix: use ExtendedType serialization if present
  The grouping key wasn't getting created correctly for DoltgresTypes, since they implement sql.StringType, but aren't supported in GMS' types.ConvertToString function. Tests for grouping with DoltgresTypes added in Doltgres.
• 2979: fix indexing for GROUP BYs and WINDOWs in INSERT and REPLACE statements in TRIGGERS
  Using aggregation and window functions inside a select statement inside an insert source inside a trigger was causing problems. For example, a trigger defined like so:

  create trigger trig before insert on t1
  for each row
  begin
  insert into t2
  select
  max(id),
  first_value(id) over (partition by id order by id),
  ...
  from
  t3;
  end;
  

  The issue involved the Projections over the Group Bys. The scope for the group bys already contained the trigger's columns and are indexed uniquely, so we shouldn't include the trigger/parent scope.

Closed Issues

• 8617: INSERT to ENUM column with NOT NULL errors
• 8882: Error in DESCRIBE PLAN SELECT COUNT(*) FROM table on running server
• 5317: dolt FixedWidthTableWriter panic when calling stored proc
• 5530: PREPARE / EXECUTE in stored procedures is buggy
• 7068: first call of a stored procedure that uses session user variable always return NULL
• 9213: Connection concurrency issues with DDL

Performance