Merged PRs

dolt

• 4948: go/cmd/dolt: sqlserver: yaml_config.go: Make some structs public to better support programmatic config generation.
• 4940: go/cmd/commands: added --datasets flag to dolt branch to list all dat…
  …asets
• 4939: ambyii's escape commit tag
  re: #4920
• 4937: support push.autoSetupRemote config variable
• 4929: Adds skipped bats for silent merge failures
  5 skipped bats showing how merge silently fails today.
  After this PR, next steps is to throw an error if we are three-way merging these kinds of schema changes. Still thinking about how to detect these issues and fail-fast.
• 4924: found another minor bug in results.go
  Found the same kind of bug as before this time in results.go. I changed the comparison to be correct as before.
  Question about line 36. We are requiring exact order and not just content, is that correct or should we sort the args before comparing them?
• 4923: found minor bug in validate.go
  In validate.go I found a small bug where there is a comparison on the same object (len(new) != len(new)). I assume this was meant to compare len(new) with len(old) so I fixed that here.
• 4913: Remove GetGCKeepers
• 4911: go/cmd/dolt/commands/sqlserver: Allow a server in a cluster to configure which address it listens on for replication traffic.
• 4907: Added bats tests for dolt_branch(-c...)

go-mysql-server

• 1458: Reorder joins before finalizing subqueries
  We cannot set field indexes for subquery expressions until the parent scope has been finalized.
  The query that surfaced this bug was a subquery expression with a dependency on a join child, whose order is rearranged after we finalize subquery. Ex: the query below swaps the schema from under the subquery expression; y and is_one should match, but we reorder the join for a lookup, changing the scope schema from [xy,uv] -> [uv,xy]. The subquery expression now selects the wrong column for comparison:

  > select y, (select 1 where y = 1) is_one
  from xy join uv on x = v
  order by y;
  +---+--------+
  | y | is_one |
  +---+--------+
  | 0 | 1      |
  | 0 | 1      |
  | 1 | NULL   |
  | 1 | NULL   |
  +---+--------+

  We select the correct field now when the subquery is finalized after the parent scope is fixed:

  > select y, (select 1 where y = 1) is_one
  from xy join uv on x = v
  order by y;
  +---+--------+
  | y | is_one |
  +---+--------+
  | 0 | NULL   |
  | 0 | NULL   |
  | 1 | 1      |
  | 1 | 1      |
  +---+--------+

• 1457: simplify LIKE expressions to take advantage of indexes
  I needed this for my blog :)
  It might give performance benefits 🤷
  Certain LIKE expressions over string columns should take advantage of prefix indexes.
  TODO: there are a few more optimization we could make, but they'd take a bit more effort (I'd also need to confirm that MySQL does them as well).
• 1456: writing tests with non-determinism
• 1455: fix incorrect result from arithmetic ops
  The result of the query select 2000.0 / 250000000.0 * (24.0 * 6.0 * 6.25 * 10.0); is fixed, but the decimal scale is incorrect (made issue for it)
• 1454: enforce test ordering
• 1452: Order the test results.
  This fixes the currently-failing dolt tests: https://github.com/dolthub/dolt/actions/runs/3622688428/jobs/6107705037
• 1451: fix visibility for on duplicate key update with cte
  fix for: #4562
  adding a special case in analyzer when insert.Source is a *plan.With.
• 1450: If a window does not have "ORDER BY" specified, use a default frame.
  If a window does not have "ORDER BY" specified, use a default frame.
  This fixes dolthub/go-mysql-server#1449
• 1447: Disambiguate aggregation expressions
  Disambiguate aggregation expressions so different columns with same sources don't end up clobbering each other.
  Fixes #4140
• 1445: fix type handling for derived table columns
  derived table column is now created from the most permissive type and the values are casted into that type at Eval.
  have skipped Prepared test as sort field type does not match non-prepared execution.
• 1440: Implemented DECLARE support for variables and cursors
  This adds support for DECLARE variables and cursors. I implemented nesting support by giving all variables a procedure-unique ID, with the scope determining which ID is resolved. A flat slice of IDs are far simpler to track than a proper tree structure.
  I did skip one test for GMS, as a RowIter reads the current state of the table, rather than the state that the RowIter generated with.
• 1437: add any_value() and option to enforce strict group by
  Added support for another option in @@sql_mode called ONLY_FULL_GROUP_BY, which makes it so select statements over group by must contain columns that are part of the group by. Before, dolt was always strict about group by selects, now (to match MySQL) we are not strict by default.
  Additionally, there is now support for the "aggregate" function any_value() which bypasses the ONLY_FULL_GROUP_BY if it is set.
  Also, changes GetGlobalVariable and GetSessionVariable to convert sql.SetTypes to strings.
  MySQL docs for reference:
  • any_value(): https://dev.mysql.com/doc/refman/8.0/en/miscellaneous-functions.html
  • ONLY_FULL_GROUP_BY: https://dev.mysql.com/doc/refman/5.7/en/sql-mode.html#sqlmode_only_full_group_by
    fix for: #1448
    and: #4604

Closed Issues

• 4919: Percent symbol (%) in commit strings is interpreted a format placeholder
• 3631: Command Line working sets become a working set for every branch after starting a server
• 4792: Add the git config variable to always set the upstream on push
• 1448: Support for selecting non-aggregate expressions that don't appear in GROUP BY clause
• 4604: Support for any_value
• 4562: on duplicate key clause missing visibility into tables in subqueries and ctes
• 4298: DISTINCT returns incorrect result for decimal types
• 4140: Result from first window function reused by other window functions
• 1449: Windows without ORDER BY return wrong results

Latency