Improve performance of queries of the form SELECT *, CASE ... END #18056

Is your feature request related to a problem or challenge?

A query of the form SELECT *, CASE ... END adds an extra column to an existing relation. The current implementation of case will take the entire input relation (i.e. all columns) and filter the entire records batch and all its columns for each when branch. This is done regardless of whether the when, then or else expressions use any of these columns or not. The implication is that the implementation may end up needlessly filtering columns that will never be read. For wide queries with many columns this has a significant impact on case evaluation.

As an example assume a query SELECT *, CASE c1 WHEN 1 THEN 'foo' WHEN 2 THEN 'bar' ELSE 'baz' END FROM t where t has 100 columns named c1 to c100. Since the case expression is only dependent on c1 it is pointless and wasteful to filter columns c2 through c100.

Describe the solution you'd like

Investigate if there are ways to optimize case evaluation.

Describe alternatives you've considered

No response

Additional context

No response