This package implements some counters on top of redis.
The count, countUnique and trending functions use buckets to approximate the value. The countUnique function also uses the HyperLogLog algorithm to approximate the unique set count for each bucket.
Future work could include switching to the Sliding HyperLogLog algorithm for countUnique. This would enable it to count all time windows together with better accuracy (no need for bucketing) and smaller space requirements.
Note: Only the counts that are present in the source file are tracked. This means that all the counts will begin at zero when each counter is first introduced to your source.
addLabel(entity | entity list, label)
Adds the provided label to the specified entities
count(BY Feature[, …] [WHERE Condition] [LAST Timespan])
Returns the streaming count for the given window
Timespans: LAST [X] SECONDS/MINUTES/HOURS/DAYS/WEEKS
DAY OVER DAY, DAY OVER WEEK, WEEK OVER WEEK
TOTAL
countUnique(Feature[, …] [GROUP BY Feature[, …]] [WHERE Condition] [LAST Duration] [BEFORE ACTION])
Performs a sliding window unique set count
entity(type, key)
Create an entity of the given type
entityList(type, keys)
Create a list of entities of the given type
hasLabel(entity, label)
Returns true if the provided entity has the given label
rateLimit(BY Feature[, …] [MAX Tokens] EVERY Duration [REFILL Count] [TAKE Count] [STRICT] [WHERE Condition])
Returns the number of tokens left in the token bucket ratelimiter before decrementing
rateLimited(BY Feature[, …] [MAX Tokens] EVERY Duration [REFILL Count] [TAKE Count] [STRICT] [WHERE Condition])
Returns true if the token bucket rate limiter has no tokens left, false otherwise
rateLimitedValues(BY Feature[, …] [MAX Tokens] EVERY Duration [REFILL Count] [TAKE Count] [STRICT] [WHERE Condition])
Returns the values that were rate limited by the token bucket rate limiter
removeLabel(entity | entity list, label)
Removes the provided label to the specified entities
sessionize(BY Feature[, …] [MAX Tokens] EVERY Duration [REFILL Count] [TAKE Count] [STRICT] [WHERE Condition])
Creates a new session using a token bucket rate limiter
trending(Feature[, …] [WHERE Condition] [WITH MIN Count EVENTS] (Timespan))
Returns values whose counts have gone up by an order of magnitude
Timespans: DAY OVER DAY, DAY OVER WEEK, DAY OVER FULL WEEK