Azeem Akhter Azeem Akhter - 11 months ago 121
Python Question

map on multiple values of one key pyspark

I have an rdd with multiple values(list) against one key, I want to filter the garbage out from each value in a key.

rdd has this data

((key1, [('',val1),('', val2),..]),(key2,[...)

I want to map it to something like this

((key1,[val1, val2,...]), key2[...)

I know a map function is required here but I haven't used map for multiple values against a key.

this is my effort for doing this.

def mapper(x):
values = []
for a in x[1]:
return(x[0], ap) mapper).collect()

but I get a few errors


The main idea is to consider each entry of an RDD as a single collection an process it as so. Meaning, if we consider the following entry

entry = ("key1", [('',"val1"),('',"val2")])

to process this collection into the expected output, we need to understand the structure of the collection

# 'key1'

# [('', 'val1'), ('', 'val2')]

now let's work on this second part :

map(lambda x : x[1],entry[1])
# ['val1', 'val2']

We can now define a function that takes an entry as an input and the resulting output will be a (key,[values...]) tuple. We'll call it mapper. We can apply the mapper on every entry in the rdd.

Putting the code together :

def mapper(entry):
    return (entry[0],map(lambda x : x[1],entry[1]))

data = [("key1", [('',"val1"),('',"val2")]),("key2",[('',"val3"),('',"val2"),('',"val4")])]

rdd = sc.parallelize(data)

rdd2 = x : mapper(x))

# [('key1', ['val1', 'val2']), ('key2', ['val3', 'val2', 'val4'])]