Youssef Hatem

Change representation of aggregate match candidates

7

This proposes a new representation of aggregate plans that gets rid of the first grouping Value in the underlying select expression, and consumes @normen662's fix for an issue with `Value`...

This proposes a new representation of aggregate plans that gets rid of the first grouping `Value` in the underlying `select` expression. Doing so would achieve two goals: - Avoid ambiguous...

It is possible to have a `KeyExpression` whose key is referencing a complex type. For example: ``` [Field { 'A' None}/Field { 'B' None}, Field { 'A' None}/Field { 'C'...

Rollup aggregate

3

This PR enables planning aggregate rollups, it enables Cascades to choose aggregate indexes when planning queries with coarser groupings.

Here is an example of a self-join query that reuses the `TypeFilter` twice (because it scans the same table twice): ```sql select * from t1 as x, t1 as y...