Maybe? Sure, mutating data sets might be a useful use case. But, inserting thousands of items at once one at a time in a tight loop, then asking for all of th

Maybe? Sure, mutating data sets might be a useful use case. But, inserting thousands of items at once one at a time in a tight loop, then asking for all of them is testing an unusual use case in my opinion. My point was that we're comparing apples and oranges. By default, I think, Python's dbm implementation doesn't do any sort of transaction or even sync after every insert, where as SQLite does have a pretty hefty atomic guarantee after each INSERT, so they're quite different actions.