Skip to content

Query: we can uniquely identify keyless entities that have Distinct operation applied on them #26078

@maumar

Description

@maumar

Currently we have logic for Distinct/GroupBy to "salvage" identifiers, in case they are not part of the distinct projection / grouping key. In case of Distinct we simply can use the entire projection (unless it contains some complex expressions, see #26077), since it's now guaranteed to be unique, and in case of group by we use entire grouping key.

We should also be able to do this for keyless entities - they don't have identifiers so anytime we have a nested collection scenario involving them, we throw due to lack of identifiers needed to perform correct bucketing of the results. For now, we only do this trick if there were any identifiers to begin with.

Metadata

Metadata

Assignees

No one assigned

    Projects

    No projects

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions