[temuze]

HyperLogLog is an algorithm that allows you to find an estimate of the distinct elements of a dataset with very little memory. Here's a simple explanation:

http://opensourceconnections.com/blog/2015/02/04/its-log-its...

HyperLogSandwich solves a similar problem. From the README:

> Unlike the HyperLogLog, which estimates the cardinality of unique items in a set, and unlike a CountMinSketch, which estimates the frequency of a specific item, the HyperLogSandwich estimates the number of unique items that have occurred for any frequency.

Real world examples include:

- How many people viewed my Tweets 4+ times this month?

- How many users watched this video 2 times this week?

- How many users visited my website 3 times during an arbitrary time range?

- How many log exceptions happened 10+ times last week?