如何使用pyspark计算出现次数

Question

I'm trying to use pyspark to count the number of occurrences.

Suppose I have data like this:

data = sc.parallelize([(1,[u'a',u'b',u'd']),
                       (2,[u'a',u'c',u'd']),
                       (3,[u'a']) ])

count = sc.parallelize([(u'a',0),(u'b',0),(u'c',0),(u'd',0)])

Is possible to count the number of occurrences in data and update in count ?

The result should be like [(u'a',3),(u'b',1),(u'c',1),(u'd',2)] .

Answer 1

I would use Counter :

>>> from collections import Counter
>>>
>>> data.values().map(Counter).reduce(lambda x, y: x + y)
Counter({'a': 3, 'b': 1, 'c': 1, 'd': 2})

Answer 2

RDDs are immutable and thus cannot be updated. Instead, you compute the count based on your data as:

count = (rdd
         .flatMap(lambda (k, data): data)
         .map(lambda w: (w,1))
         .reduceByKey(lambda a, b: a+b))

Then, if the result can fit in master main memory feel free to .collect() from count .

Answer 3

You wouldn't update count since RDDs are immutable. Just run the calculation you want and then save directly to any variable you want:

In [17]: data.flatMap(lambda x: x[1]).map(lambda x: (x, 1)).reduceByKey(lambda x, y: x + y).collect()
Out[17]: [('b', 1), ('c', 1), ('d', 2), ('a', 3)]