How to count unique ID after groupBy in pyspark

user_8674940 0
Viewed 139871

I'm using the following code to agregate students per year. The purpose is to know the total number of student for each year.

from pyspark.sql.functions import col
import pyspark.sql.functions as fn
gr = Df2.groupby(['Year'])
df_grouped = 
gr.agg(fn.count(col('Student_ID')).alias('total_student_by_year'))

The problem that I discovered that so many ID's are repeated, so the result is wrong and huge.

I want to agregate the students by year, count the total number of student by year and avoid the repetition of ID's.

pyspark apache-spark-sql python
5 Answers

Use countDistinct function

from pyspark.sql.functions import countDistinct
x = [("2001","id1"),("2002","id1"),("2002","id1"),("2001","id1"),("2001","id2"),("2001","id2"),("2002","id2")]
y = spark.createDataFrame(x,["year","id"])

gr = y.groupBy("year").agg(countDistinct("id"))
gr.show()

output

+----+------------------+
|year|count(DISTINCT id)|
+----+------------------+
|2002|                 2|
|2001|                 2|
+----+------------------+
user_41173310

You can also do:

gr.groupBy("year", "id").count().groupBy("year").count()

This query will return the unique students per year.

user_63957700

countDistinct() and multiple aggr both are not supported in streaming.

user_58078260

If you are working with an older Spark version and don't have the countDistinct function, you can replicate it using the combination of size and collect_set functions like so:

gr = gr.groupBy("year").agg(fn.size(fn.collect_set("id")).alias("distinct_count"))

In case you have to count distinct over multiple columns, simply concatenate the columns into a new one using concat and perform the same as above.

user_49587170

By using Spark/PySpark SQL

y.createOrReplaceTempView("STUDENT")
    
spark.sql("SELECT year, count(DISTINCT id) as count" + \
"FROM STUDENT group by year").show()
user_107290350
Related