Data manipulation with DataFrames

Introduction to PySpark

Ben Schmidt

Data Engineer

Handling missing data

Use .na.drop() to remove rows with null values

# Drop rows with any nulls
df_cleaned = df.na.drop()


# Filter out nulls
df_cleaned = df.where(col("columnName").isNotNull())

Use .na.fill({"column": value) to replace nulls with a specific value

# Fill nulls in the age column with the value 0
df_filled = df.na.fill({"age": 0})

Column operations

Use .withColumn() to add a new column based on calculations or existing columns

# Create a new column 'age_plus_5'
df = df.withColumn("age_plus_5", df["age"] + 5)

Use withColumnRenamed() to rename columns

# Rename the 'age' column to 'years'
df = df.withColumnRenamed("age", "years")

Use drop() to remove unnecessary columns

# Drop the 'department' column
df = df.drop("department")

Row operations

Use .filter() to select rows based on specific conditions

# Filter rows where salary is greater than 50000
filtered_df = df.filter(df["salary"] > 50000)

Use .groupBy() and aggregate functions (e.g., .sum(), .avg()) to summarize data

# Group by department and calculate the average salary
grouped_df = df.groupBy("department").avg("salary")

Row Operations Outomes

Filtering

+------+---+-----------------+
|salary|age|      occupation |
+------+---+-----------------+
| 60000| 45|Exec-managerial  |
| 70000| 35|Prof-specialty   |
+------+---+-----------------+

GroupBy ` +----------+-----------+ |department|avg(salary)| +----------+-----------+ | HR| 80000.0| | IT| 70000.0| +----------+-----------+

CheatSheet

# Drop rows with any nulls
df_cleaned = df.na.drop()

#Drop nulls on a column
df_cleaned = df.where(col("columnName").isNotNull())

# Fill nulls in the age column with the value 0
df_filled = df.na.fill({"age": 0})

Use .withColumn() to add a new column based on calculations or existing columns. Syntax: .withColumn("new_col_name", "original transformation")
```
# Create a new column 'age_plus_5'
df = df.withColumn("age_plus_5", df["age"] + 5)
```
Use withColumnRenamed() to rename columns Syntax: withColumnRenamed(old column name,new column name`

# Rename the 'age' column to 'years'
df = df.withColumnRenamed("age", "years")

Use drop() to remove unnecessary columns Syntax: .drop(column name)

# Drop the 'department' column
df = df.drop("department")

# Filter rows where salary is greater than 50000
filtered_df = df.filter(df["salary"] > 50000)

Use .groupBy() and aggregate functions (e.g., .sum(), .avg()) to summarize data

# Group by department and calculate the average salary
grouped_df = df.groupBy("department").avg("salary")

Let's practice!

Introduction to PySpark