I did some quick benchmarks of single tuple copies and inserts into an empty and non-empty table and this doesn't appear to make the performance any worse.
It was necessary to make the InMemHashIndex support deleting keys, which maintains the lack of gaps in the in-memory hash index by swapping deleted entries with the last entry.
I did some quick benchmarks of single tuple copies and inserts into an empty and non-empty table and this doesn't appear to make the performance any worse.
It was necessary to make the InMemHashIndex support deleting keys, which maintains the lack of gaps in the in-memory hash index by swapping deleted entries with the last entry.