Merged PRs

dolt

3308: Revert table_reader changes
This revert the changes made to table_reader.go in #3223
This was done to fix a performance regression to oltp_delete caused by the above PR.
In summary, the regression was caused by a change to how chunks are located. A tableReader from table_reader.go, previously held a copy of the chunk hash prefixes while also holding the table file index itself. In the above PR, the copy of the chunk hash prefixes were removed and instead the accesses were relayed to the table file index.
As a consequence, each access to a prefix does additional work. onHeapTableIndex holds the serialized bytes of the table file index, so each access requires deserialization work. Namely, a multiplication and the deserialization of 8 bytes into a uint64 prefix value.
Reads were not affected for two reasons. First, I/O masks any read regressions in this area because we read many chunks at a time. And secondly, writes have patterns that frequently access this code path.
Reverting these changes restored the performance seen in Dolt v0.39.0.
3298: Fix ErrBranchNotFound
This reverts commit 8cd5f16, reversing
changes made to f857be6.
3286: skip -race enginetests until merge

980: Add tests for handling join in parentheses
Tests regarding this issue, #2203
Some cases give incorrect result in Dolt.
979: Fixed overly aggressive type change check with foreign keys
Previously PR dolthub/go-mysql-server#977
The bug has existed since the GMS foreign key implementation, but only surfaced after the previous PR. When changing a column's type, I was only checking that the table was used in foreign keys, when I needed to check that the column is used. This fixes that bug.
975: Fixing return types for spatial functions
Fix for: #3279
Adding tests that check the return type for functions
- st_asgeojson
- st_aswkb
- st_aswkt
- st_astext
- st_dimension
- st_latitude
- st_longitude
- st_swapxy
- st_x
- st_y