Merged PRs
dolt
- 11185: go/store/nbs: boostrapJournal: Improve journal bootstraping resiliency in the face of errors in journal.idx.
journal.idx is an optimization to make opening the database faster. Errors in reading it should never cause opening the database to fail. Make it so that we gracefully handle data and I/O errors in accessing journal.idx so that database opening proceeds as expected.
There is still one fatal error associated with journal.idx: when we are in read-write mode, we must be able to open journal.idx for writing. - 11184: bug fix for merging tables with vector indexes
Fixes: #11180
Related Doltgres update to skip the unsupportedVECTORindex test: dolthub/doltgresql#2827
go-mysql-server
- 3578: fix: retain table comment when primary key is a separate clause
CREATE TABLE ... COMMENT='c'dropped the comment whenPRIMARY KEYwas given as a separate clause.- Add a
commentparameter toIndexedTableCreator.CreateIndexedTableand thread it through the executor.
Block #11170
- Add a
- 3572: Rewrite
CachedResultsnode andcachedResultsIterto not useMemoryManager
This PR is a follow-up to #3561 and further addresses the memory leak mentioned in #3560. While the initial fix in #3561 made it seem like the issue was that we were not properly disposing of theCachedResultsnode, the real issue was that acachedResultsIterthat never wrote to theCachedResultsnode was never disposed of in theMemoryManagerif it was never closed.
This PR rewrites theCachedResultsnode andcachedResultsIterto never actually make use of theMemoryManager. Since we do not make use of parallel partitions, there's no need to have a locking cache. Instead, the results of thecachedResultsIterare stored directly in theCachedResultsnode. This greatly simplifies both theCachedResultsnode and thecachedResultsIter.
This PR also moves theemptyCacheIterandEmptyIterout of theplanpackage and de-dupes any code related to empty iters.
Closed Issues
- 11180: 3-way DOLT_MERGE fails on table with VECTOR INDEX: expected prollyIndex, found: durable.proximityIndex