Spark - (Reduce|Aggregate) function

1 - About

Spark permits to reduce a data set through:

3 - Reduce

The reduce function of the map reduce framework


Reduce is a spark action that aggregates a data set (RDD) element using a function.

That function takes two arguments and returns one.

The function must be parallel enabled.

reduce can return a single value such as an int.

3..1 - Reduce a List

rdd = sc.parallelize([1, 2, 3]) 
rdd.reduce(lambda a, b: a * b)
Value: 6 

3..2 - Reduce a List of Tuple

3..2.1 - Numeric value

reduceByKey(function|func) return a new distributed dataset of (K, V) pairs where the values for each key are aggregated using the given reduce function func, which must be of type (V,V) → V

rdd = sc.parallelize([(1,2), (3,4), (3,6)])
rdd.reduceByKey(lambda a, b: a + b) 
RDD: [(1,2), (3,4), (3,6)] → [(1,2), (3,10)] 

If the value is a string, you can use the groupByKey() to reduce it. See below

3..2.2 - String value

groupByKey() return a new dataset of (K, Iterable<V>) pairs Key-Value Transformations

rdd2 = sc.parallelize([(1,'a'), (2,'c'), (1,'b')])
RDD: [(1,'a'), (1,'b'), (2,'c')] -> [(1,['a','b']), (2,['c'])] 

Be careful using groupByKey() as it can cause a lot of data movement across the network and create large Iterables at workers

Imagine you have an RDD where you have 1 million pairs that have the key 1. All of the values will have to fit in a single worker if you use group by key. So instead of a group by key, consider using reduced by key or a different key value transformation.

3.1 - Aggregate Function

  • count()
  • countApprox
  • countApproxDistinct
  • sum()
  • max
  • mean
  • meanApprox
  • min
db/spark/reduce.txt · Last modified: 2017/09/13 20:06 by gerardnico