Skip to content

A unique id counter implemented with redis bitmaps. Count the unique users across a date range.

Notifications You must be signed in to change notification settings

cmdrkeene/bitmap-counter

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

7 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Bitmap Counter

An efficient id-based counter. Current implementation uses redis.

Usage

require "bitmap/counter"
counter = Bitmap::Counter.new("name")

counter.add(id = 1)   # a numeric user_id for example
counter.count         # 2
counter.includes?(1)  # true
counter.includes?(2)  # false

Redis

You can use your own redis connection.

Bitmap::Counter.redis = your_redis_here

DateCounter

Count ids for a series of dates.

require "bitmap/date_counter"
counter = Bitmap::DateCounter.new("name")
counter.add(1, Date.today)
counter.add(2, Date.today)
counter.add(1, Date.today - 1)

You can inspect a specific date:

counter.count(Date.today)     # 2
counter.count(Date.today - 1) # 1

counter.includes?(1, Date.today)      # true
counter.includes?(2, Date.today)      # true
counter.includes?(1, Date.today - 1)  # true
counter.includes?(2, Date.today - 1)  # false

Or a range of dates:

counter.count((Date.today - 1)..Date.today) # 2

counter.includes?(1, (Date.today - 1)..Date.today) # true
counter.includes?(2, (Date.today - 1)..Date.today) # true
counter.includes?(3, (Date.today - 1)..Date.today) # false

Benchmarks

Memory

Redis doesn't expose easy statistics on memory/disk usage :(

TODO

Intersections

TODO

Cardinality

Performance is acceptable, but worse than native redis sets.

Sending the entire bitstring across the wire is not all that efficient.

Some people suggested implementing this a Lua routine inside redis. I suspect the performance will be better, but is unlikely to beat native sets.

Totally unscientific, but here are my results:

Bitmap::Counter.unique (10_000_000 ids): 76.39ms
Redis.scard            (10_000_000 ids): 2.38ms

It's not any better for small datasets:

unique (1000): 0.26ms
unique (10000): 0.59ms
unique (100000): 0.94ms
scard  (1000): 0.19ms
scard  (10000): 0.23ms
scard  (100000): 0.26ms

Credits

Inspired by the awesome post by Chandra Patni.

Ruby Version

This was developed on ruby 1.9.2p180 and not tested on other versions...yet :)

TODO

  • Add different backends (File, Postgres, etc)

About

A unique id counter implemented with redis bitmaps. Count the unique users across a date range.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages