Roadmap v0.5

[Fedomn]

Introduce distinct and subquery:

-- distinct
select distinct state from employee;
select count(distinct(b)) from t2;
-- alias
select a as c1 from t1 order by c1 desc limit 1;
select t.a from t1 t where t.b > 1 order by t.a desc limit 1;
-- subquery
select t.* from (select * from t1 where a > 1) t where t.b > 7;
select a, (select max(b) from t1) max_b from t1;

• distinct
  • support distinct select #62
  • support distinct aggregation #64
• alias
  • support select alias expression #70
  • support table alias #72
• subquery
  • support subquery as a source #66
  • support subquery alias #74
  • support cross product join for scalar subquery #76
  • refactor binder alias #78
  • support multi-layer binder context #79
  • support uncorrelated scalar subquery expression #67
  • support predicate subquery expression #68
• bugfix
  • fix PushProjectThroughChild rule generate empty exprs in Project #82
  • fix PushPredicateThroughNonJoin rule not work with column alias #83
• refactor
  • refactor PlanNode output columns #87
  • refactor BoundSubquery to Expr and TableRef #89

[Fedomn]

Currently, our binder and planner need to refactor and redesign to support more complex SQL, such as scalar and predicate subquery.

After some investigation, DuckDB is a good example to refer to. So let's start by refactoring.

[Fedomn]

Through recent thinking, the project needs to redesign to be more like a database query engine and also to rethink its position that not just a learning project, but an OLAP query engine embedded in-process that sounds like duckdb, but leverages rust ecosystem and Apache Arrow.
Therefore, I renamed the project to sqlrs and started a new milestone.