GroupBy and aggregations in Apache Spark - Step-by-Step Execution

Concept Flow - GroupBy and aggregations

Start with DataFrame

↓

Choose column(s) to group by

↓

Apply groupBy operation

↓

Select aggregation function(s)

↓

Compute aggregated results

↓

Return new DataFrame with grouped and aggregated data

GroupBy splits data into groups by column values, then aggregation computes summary stats per group.

Execution Sample

Apache Spark

from pyspark.sql import SparkSession
spark = SparkSession.builder.getOrCreate()
data = [("A", 10), ("B", 20), ("A", 30), ("B", 40)]
df = spark.createDataFrame(data, ["Category", "Value"])
grouped = df.groupBy("Category").sum("Value")
grouped.show()

This code groups rows by 'Category' and sums the 'Value' column for each group.

Execution Table

Step	Action	Input Data	GroupBy Column	Aggregation	Output DataFrame
1	Create DataFrame	[('A',10),('B',20),('A',30),('B',40)]	-	-	[('A',10),('B',20),('A',30),('B',40)]
2	Apply groupBy on 'Category'	[('A',10),('B',20),('A',30),('B',40)]	Category	-	Groups: {'A': [('A',10),('A',30)], 'B': [('B',20),('B',40)]}
3	Aggregate sum on 'Value'	Groups: {'A': [...], 'B': [...]}	Category	sum(Value)	[('A',40),('B',60)]
4	Show result	[('A',40),('B',60)]	-	-	Output displayed as table with Category and sum(Value) columns

💡 Aggregation complete, grouped sums computed for each category.

Variable Tracker

Variable	Start	After Step 1	After Step 2	After Step 3	Final
df	None	[('A',10),('B',20),('A',30),('B',40)]	[('A',10),('B',20),('A',30),('B',40)]	[('A',10),('B',20),('A',30),('B',40)]	[('A',10),('B',20),('A',30),('B',40)]
grouped	None	None	Groups: {'A': [('A',10),('A',30)], 'B': [('B',20),('B',40)]}	[('A',40),('B',60)]	[('A',40),('B',60)]

Key Moments - 3 Insights

Why does the output DataFrame only have one row per category?

Because groupBy collects all rows with the same category, then aggregation (sum) combines their values into one summary row per group, as shown in execution_table step 3.

What happens if we group by a column but don't apply aggregation?

The groupBy operation alone creates groups but does not compute results. Aggregation functions like sum() must be applied to get summarized output, as seen between steps 2 and 3.

Can we group by multiple columns?

Yes, grouping by multiple columns creates groups based on unique combinations of those columns. The flow is the same but groups are more specific.

Visual Quiz - 3 Questions

Look at the execution_table at step 3, what is the sum of 'Value' for category 'B'?

A 60

B 40

C 20

D 80

💡 Hint

Check the 'Output DataFrame' column in step 3 of execution_table.

At which step does the DataFrame get split into groups?

A Step 1

B Step 2

C Step 3

D Step 4

💡 Hint

Look at the 'Action' and 'GroupBy Column' columns in execution_table.

If we change the aggregation from sum to count, what changes in the output DataFrame?

A Output DataFrame stays the same

B Values show maximum value per group

C Values show total count of rows per group instead of sum

D DataFrame will have more columns

💡 Hint

Aggregation function changes the summary metric computed per group.

Concept Snapshot

GroupBy and aggregations in Spark:
- Use df.groupBy('col') to group data by column(s)
- Apply aggregation like sum(), count(), avg() on grouped data
- Result is a new DataFrame with one row per group
- Aggregations summarize data within each group
- Useful for summarizing large datasets by categories

Full Transcript

This visual execution shows how to use groupBy and aggregation in Apache Spark. First, a DataFrame is created with categories and values. Then, groupBy splits the data into groups by the 'Category' column. Next, sum aggregation adds the 'Value' numbers within each group. The final output DataFrame shows one row per category with the sum of values. Key points include that groupBy alone does not summarize data until aggregation is applied, and that multiple columns can be used to group data. The execution table traces each step clearly, and the variable tracker shows how the DataFrame and grouped data change. This helps beginners see how grouping and aggregation work step-by-step.

Practice

1. What does the groupBy function do in Apache Spark?

easy

A. It splits the data into groups based on column values.

B. It sorts the data in ascending order.

C. It filters rows based on a condition.

D. It joins two DataFrames together.

Solution

1. Step 1: Understand the purpose of groupBy

   The groupBy function is used to split data into groups based on one or more column values.

2. Step 2: Compare with other functions

   Sorting, filtering, and joining are different operations and do not split data into groups.

3. Final Answer:

   It splits the data into groups based on column values. -> Option A

4. Quick Check:

   groupBy = splits data into groups [OK]

Hint: Remember: groupBy groups rows by column values [OK]

Common Mistakes:

• Confusing groupBy with filter or sort
• Thinking groupBy changes row order
• Assuming groupBy merges DataFrames

2. Which of the following is the correct syntax to calculate the average of column sales after grouping by region in Spark DataFrame df?

easy

A. df.groupBy("region").sum("sales")

B. df.groupBy("region").agg(avg("sales"))

C. df.groupBy("region").filter(avg("sales"))

D. df.groupBy("region").select(avg("sales"))

Solution

1. Step 1: Identify correct aggregation syntax

   To calculate average after grouping, use groupBy followed by agg(avg(column)).

2. Step 2: Check each option

   df.groupBy("region").agg(avg("sales")) uses agg(avg("sales")) correctly. df.groupBy("region").sum("sales") uses sum, not average. df.groupBy("region").filter(avg("sales")) uses filter incorrectly. df.groupBy("region").select(avg("sales")) uses select which is not valid after groupBy for aggregation.

3. Final Answer:

   df.groupBy("region").agg(avg("sales")) -> Option B

4. Quick Check:

   groupBy + agg(avg) = df.groupBy("region").agg(avg("sales")) [OK]

Hint: Use agg() with avg() inside groupBy for averages [OK]

Common Mistakes:

• Using sum instead of avg for average
• Using filter instead of agg
• Using select after groupBy without agg

3. Given the following Spark code, what will be the output schema and result?

from pyspark.sql import SparkSession
from pyspark.sql.functions import sum

spark = SparkSession.builder.getOrCreate()
data = [("East", 100), ("West", 200), ("East", 150)]
df = spark.createDataFrame(data, ["region", "sales"])
result = df.groupBy("region").agg(sum("sales").alias("total_sales"))
result.show()

medium

A. region | total_sales
East | 250
West | 200

B. region | sales
East | 100
West | 200
East | 150

C. region | total_sales
East | 150
West | 200

D. region | total_sales
East | 100
West | 200

Solution

1. Step 1: Understand groupBy and sum aggregation

   The code groups rows by "region" and sums the "sales" values for each group.

2. Step 2: Calculate sums for each region

   For "East": 100 + 150 = 250; For "West": 200.

3. Final Answer:

   region | total_sales East | 250 West | 200 -> Option A

4. Quick Check:

   Sum sales by region = 250 (East), 200 (West) [OK]

Hint: Sum sales per group to get total_sales [OK]

Common Mistakes:

• Not summing all values per group
• Confusing alias name with original column
• Expecting original rows without aggregation

4. Identify the error in the following Spark code snippet:

df.groupBy("category").agg(sum("amount"), avg("amount"))

medium

A. Missing alias for aggregated columns causes error.

B. agg() requires a dictionary, not multiple arguments.

C. groupBy cannot be used with agg.

D. sum and avg functions are not imported.

Solution

1. Step 1: Check function imports

   The code uses sum and avg functions but does not show imports from pyspark.sql.functions.

2. Step 2: Validate other options

   agg() accepts multiple arguments; aliasing is optional; groupBy works with agg.

3. Final Answer:

   sum and avg functions are not imported. -> Option D

4. Quick Check:

   Missing imports cause NameError [OK]

Hint: Always import aggregation functions before use [OK]

Common Mistakes:

• Assuming agg needs dict, but it accepts multiple args
• Forgetting to import sum, avg from pyspark.sql.functions
• Thinking groupBy can't be chained with agg

5. You have a Spark DataFrame sales_df with columns store, product, and revenue. You want to find the maximum revenue per store and the average revenue per product across all stores in a single query. Which code snippet correctly achieves this?

hard

A. sales_df.groupBy("store").agg(max("revenue").alias("max_revenue"), avg("revenue").alias("avg_revenue"))

B. sales_df.agg(max("revenue").alias("max_revenue"), avg("revenue").alias("avg_revenue"))

C. sales_df.groupBy("store").agg(max("revenue").alias("max_revenue")).crossJoin(sales_df.groupBy("product").agg(avg("revenue").alias("avg_revenue")))

D. sales_df.groupBy("store").agg(max("revenue").alias("max_revenue")).groupBy("product").agg(avg("revenue").alias("avg_revenue"))

Solution

1. Step 1: Understand the requirements

   We need max revenue per store and average revenue per product in one result.

2. Step 2: Analyze options

   sales_df.groupBy("store").agg(max("revenue").alias("max_revenue"), avg("revenue").alias("avg_revenue")) aggregates both by store only, so avg revenue per product is incorrect. sales_df.agg(max("revenue").alias("max_revenue"), avg("revenue").alias("avg_revenue")) aggregates globally, losing grouping. sales_df.groupBy("store").agg(max("revenue").alias("max_revenue")).groupBy("product").agg(avg("revenue").alias("avg_revenue")) tries chaining groupBy but overwrites previous aggregation. sales_df.groupBy("store").agg(max("revenue").alias("max_revenue")).crossJoin(sales_df.groupBy("product").agg(avg("revenue").alias("avg_revenue"))) computes both groupings separately and joins results, correctly combining both summaries.

3. Final Answer:

   sales_df.groupBy("store").agg(max("revenue").alias("max_revenue")).crossJoin(sales_df.groupBy("product").agg(avg("revenue").alias("avg_revenue"))) -> Option C

4. Quick Check:

   Separate groupBy + crossJoin combines both aggregations [OK]

Hint: Use separate groupBy then join to combine different aggregations [OK]

Common Mistakes:

• Trying to aggregate different groupings in one groupBy
• Overwriting aggregation by chaining groupBy
• Forgetting to join separate aggregation results