Jan Urbański <j.urbanski@students.mimuw.edu.pl> writes:
> Come to think of it, the current code is in a way a variant of Lossy
> Counting, it's just doing the pruning after each and every new element,
> isn't it?
Interesting comment. In LC's terms we have w=1 therefore e=1 therefore
the maximum error is as bad as possible?
regards, tom lane