Pyspark - Aggregation on multiple columns

问题

I have data like below. Filename:babynames.csv.

year    name    percent     sex
1880    John    0.081541    boy
1880    William 0.080511    boy
1880    James   0.050057    boy

I need to sort the input based on year and sex and I want the output aggregated like below (this output is to be assigned to a new RDD).

year    sex   avg(percentage)   count(rows)
1880    boy   0.070703         3

I am not sure how to proceed after the following step in pyspark. Need your help on this

testrdd = sc.textFile("babynames.csv");
rows = testrdd.map(lambda y:y.split(',')).filter(lambda x:"year" not in x[0])
aggregatedoutput = ????

回答1:

1. Follow the instructions from the README to include spark-csv package

2. Load data

   df = (sqlContext.read
       .format("com.databricks.spark.csv")
       .options(inferSchema="true", delimiter=";", header="true")
       .load("babynames.csv"))
   

3. Import required functions

   from pyspark.sql.functions import count, avg
   

4. Group by and aggregate (optionally use Column.alias:

   df.groupBy("year", "sex").agg(avg("percent"), count("*"))
   

Alternatively:

• cast percent to numeric
• reshape to a format ((year, sex), percent)
• aggregateByKey using pyspark.statcounter.StatCounter

来源：https://stackoverflow.com/questions/36251004/pyspark-aggregation-on-multiple-columns