spark-hyperloglog

Algebird's HyperLogLog support for Apache Spark. This package can be used in concert with presto-hyperloglog to share HyperLogLog sets between Spark and Presto.

Example usage

import com.mozilla.spark.sql.hyperloglog.aggregates._
import com.mozilla.spark.sql.hyperloglog.functions._

val hllMerge = new HyperLogLogMerge
sqlContext.udf.register("hll_merge", hllMerge)
sqlContext.udf.register("hll_create", hllCreate _)
sqlContext.udf.register("hll_cardinality", hllCardinality _)

val frame = sc.parallelize(List("a", "b", "c", "c")).toDF("id")
val count = frame
  .select(expr("hll_create(id, 12) as hll"))
  .groupBy()
  .agg(expr("hll_cardinality(hll_merge(hll)) as count"))
  .show()

yields:

+-----+
|count|
+-----+
|    3|
+-----+

Deployment

1. Configure your credentials for the Spark Packages repository in ~/.ivy2/.sbtcredentials, e.g:

   realm=Sonatype Nexus Repository Manager
   host=oss.sonatype.org
   user=foo
   password=bar
   

2. Publish a new release with sbt publishSigned