I was wondering a bit about a use case to have old_idx and new_idx in the Change_Entry structure. Obviously, they're only being written in {node, way, relation}_updater, but never read again anywhere else. The "changed" statement as single consumer of changelog tables extracts the element id, and ignores old_idx and new_idx.
Those extra 8 bytes consume quite a bit of space given that we have more than 8 billion changelog entries for nodes. Is there some special use case for those two fields, like debugging or troubleshooting? Are they really needed at all?
I was wondering a bit about a use case to have old_idx and new_idx in the Change_Entry structure. Obviously, they're only being written in {node, way, relation}_updater, but never read again anywhere else. The "changed" statement as single consumer of changelog tables extracts the element id, and ignores old_idx and new_idx.
Those extra 8 bytes consume quite a bit of space given that we have more than 8 billion changelog entries for nodes. Is there some special use case for those two fields, like debugging or troubleshooting? Are they really needed at all?