Why data quality prevents downstream failures in Apache Spark - See It in Action

Why data quality prevents downstream failures

📖 Scenario: You work as a data analyst in a company that collects sales data from different stores. Sometimes, the data has missing or wrong values, which causes errors in reports and decisions.Your job is to check the quality of the sales data before it is used for analysis. You will identify and filter out bad data to prevent problems later.

🎯 Goal: Build a Spark program that loads sales data, sets a quality threshold, filters out bad data based on that threshold, and shows the clean data. This will help avoid failures in later steps.

📋 What You'll Learn

Create a Spark DataFrame with sales data including some bad entries

Set a threshold for minimum valid sales amount

Filter the DataFrame to keep only rows with sales amount above the threshold

Print the filtered clean data

💡 Why This Matters

🌍 Real World

In real companies, data often has errors or missing values. Checking data quality early stops bad data from causing wrong reports or system crashes.

💼 Career

Data analysts and engineers must clean and validate data before analysis or machine learning to ensure reliable results and avoid costly mistakes.

1

Create sales data DataFrame

Create a Spark DataFrame called sales_df with these exact rows: ("Store1", 100), ("Store2", -50), ("Store3", 200), ("Store4", 0), ("Store5", 150). The columns should be store and sales.

Apache Spark

from pyspark.sql import SparkSession

spark = SparkSession.builder.appName("DataQuality").getOrCreate()
# Create sales_df DataFrame with given rows and columns
# Your code here

Hint

Use spark.createDataFrame() with a list of tuples and specify the column names.

2

Set quality threshold

Create a variable called min_sales and set it to 1. This will be the minimum valid sales amount to keep.

Apache Spark

from pyspark.sql import SparkSession

spark = SparkSession.builder.appName("DataQuality").getOrCreate()
sales_data = [("Store1", 100), ("Store2", -50), ("Store3", 200), ("Store4", 0), ("Store5", 150)]
sales_df = spark.createDataFrame(sales_data, ["store", "sales"])
# Create min_sales variable and set it to 1
# Your code here

Hint

Just create a variable named min_sales and assign the value 1.

3

Filter bad data using threshold

Create a new DataFrame called clean_sales_df by filtering sales_df to keep only rows where the sales column is greater than or equal to min_sales.

Apache Spark

from pyspark.sql import SparkSession

spark = SparkSession.builder.appName("DataQuality").getOrCreate()
sales_data = [("Store1", 100), ("Store2", -50), ("Store3", 200), ("Store4", 0), ("Store5", 150)]
sales_df = spark.createDataFrame(sales_data, ["store", "sales"])
min_sales = 1
# Filter sales_df to keep rows with sales >= min_sales and assign to clean_sales_df
# Your code here

Hint

Use the filter() method on sales_df with the condition sales_df.sales >= min_sales.

4

Show clean sales data

Use print() to display the rows of clean_sales_df by collecting them and printing the list.

Apache Spark

from pyspark.sql import SparkSession

spark = SparkSession.builder.appName("DataQuality").getOrCreate()
sales_data = [("Store1", 100), ("Store2", -50), ("Store3", 200), ("Store4", 0), ("Store5", 150)]
sales_df = spark.createDataFrame(sales_data, ["store", "sales"])
min_sales = 1
clean_sales_df = sales_df.filter(sales_df.sales >= min_sales)
# Print the collected rows of clean_sales_df
# Your code here

Hint

Use clean_sales_df.collect() inside print() to show the filtered rows.

Practice

1. Why is maintaining good data quality important in Apache Spark pipelines?

easy

A. It prevents errors and failures in later processing steps

B. It makes the data files smaller in size

C. It speeds up the Spark installation process

D. It automatically creates visualizations

Solution

1. Step 1: Understand the role of data quality

   Good data quality means data is accurate, complete, and consistent, which helps avoid problems later.

2. Step 2: Connect data quality to pipeline stability

   If data has errors or missing values, Spark jobs can fail or produce wrong results downstream.

3. Final Answer:

   It prevents errors and failures in later processing steps -> Option A

4. Quick Check:

   Good data quality = fewer downstream errors [OK]

Hint: Good data quality stops errors later [OK]

Common Mistakes:

• Thinking data size reduction is main goal
• Confusing data quality with installation speed
• Assuming data quality creates visualizations

2. Which Spark DataFrame method is used to remove rows with missing values to improve data quality?

easy

A. fillna()

B. dropna()

C. select()

D. groupBy()

Solution

1. Step 1: Identify method to remove missing data

   dropna() removes rows with null or missing values from a DataFrame.

2. Step 2: Differentiate from other methods

   fillna() fills missing values but does not remove rows; select() chooses columns; groupBy() aggregates data.

3. Final Answer:

   dropna() -> Option B

4. Quick Check:

   dropna() removes missing rows [OK]

Hint: dropna() removes missing rows [OK]

Common Mistakes:

• Confusing fillna() with dropna()
• Using select() to remove rows
• Thinking groupBy() cleans missing data

3. Given this Spark code snippet:

df = spark.createDataFrame([(1, 'Alice'), (None, 'Bob'), (3, None)], ['id', 'name'])
df_clean = df.dropna()
df_clean.show()

What will be the output?

medium

A. +----+-----+
| id | name|
+----+-----+
|null| Bob |
+----+-----+

B. +----+-----+
| id | name|
+----+-----+
| 1 |Alice|
| 3 | null|
+----+-----+

C. +---+-----+
| id| name|
+---+-----+
| 1|Alice|
+---+-----+

D. Error: dropna() is not a valid method

Solution

1. Step 1: Understand dropna() effect on DataFrame

   dropna() removes any row with null in any column.

2. Step 2: Apply dropna() to given data

   Rows (None, 'Bob') and (3, None) have nulls, so only (1, 'Alice') remains.

3. Final Answer:

   +---+-----+ | id| name| +---+-----+ | 1|Alice| +---+-----+ -> Option C

4. Quick Check:

   dropna() removes rows with nulls [OK]

Hint: dropna() removes any row with null [OK]

Common Mistakes:

• Expecting rows with null to remain
• Confusing dropna() with fillna()
• Thinking dropna() causes error

4. You run this code but get an error:

df = spark.createDataFrame([(1, 'Alice'), (2, None)], ['id', 'name'])
df_clean = df.dropna(how='anyy')
df_clean.show()

What is the cause of the error?

medium

A. The 'how' parameter value 'anyy' is invalid

B. dropna() cannot be used on DataFrames

C. The DataFrame has no missing values

D. show() method is not available on DataFrames

Solution

1. Step 1: Check dropna() parameters

   dropna() accepts 'how' parameter with values 'any' or 'all' only.

2. Step 2: Identify invalid parameter usage

   Using 'anyy' causes a runtime error because it's not a valid option.

3. Final Answer:

   The 'how' parameter value 'anyy' is invalid -> Option A

4. Quick Check:

   Invalid parameter value causes error [OK]

Hint: Check parameter spelling carefully [OK]

Common Mistakes:

• Thinking dropna() is invalid on DataFrames
• Ignoring typo in 'how' parameter
• Assuming show() causes error

5. You have a Spark DataFrame with many missing and invalid values. Which approach best prevents downstream failures in your data pipeline?

hard

A. Convert all data to strings to avoid type errors

B. Ignore missing data and handle errors only when they occur downstream

C. Run transformations first, then clean data at the end

D. Remove or fix bad data early using dropna() and fillna() before transformations

Solution

1. Step 1: Understand impact of bad data early

   Bad data causes errors or wrong results if not handled before transformations.

2. Step 2: Choose best practice for data quality

   Removing or fixing missing/invalid data early with dropna() and fillna() prevents failures downstream.

3. Step 3: Evaluate other options

   Ignoring errors delays failure; cleaning last risks wasted work; converting all to strings loses data meaning.

4. Final Answer:

   Remove or fix bad data early using dropna() and fillna() before transformations -> Option D

5. Quick Check:

   Early cleaning prevents downstream failures [OK]

Hint: Clean data early to avoid later errors [OK]

Common Mistakes:

• Delaying cleaning until after transformations
• Ignoring missing data causes hidden errors
• Converting all data to strings loses type info