GROUP BY support

[Samyak2]

Overview

This is a tracking issue for adding support for the GROUP BY clause of a SELECT statement. This allows aggregations over unique values of a column.

Tasks

• add instructions for GROUP BY and aggregations
  • See the GroupBy instruction
  • Aggregations are defined in the expr for Project or Filter just like normal non-grouped aggregations
  • Note: this is most likely subject to change
• add GROUP BY codegen to the generic dialect
  • See this part of the generic codegen
• evaluating GroupBy instructions and aggregations
  • from How should aggregate work in the VM #16, we saw that
    • GroupBy should probably have input and output registers like Project
    • we need a different table with an efficient index implementation to look up the grouped columns
    • the HAVING clause can re-use aggregations that are already computed and it might be difficult to perform common sub-expression elimination here