pyspark multiple column join

How to join multiple columns in PySpark Azure Databricks?

by

Are you looking to find out how to join multiple columns in PySpark Azure Databricks cloud or maybe you are looking for a solution, to join two DataFrames on multiple columns without duplication? If you are looking for any of these problem solutions, you have landed on the correct page. I will also show you how to use both PySpark and Spark SQL way of doing a multiple-column join in Azure Databricks. I will explain it with a practical example. So please don’t waste time let’s start with a step-by-step guide to understand how to join multiple columns in PySpark Azure Databricks.

In this blog, I will teach you the following with practical examples:

  • Syntax of join()
  • Joining multiple columns
  • Joining multiple columns using where()
  • Eliminate duplicate columns while joining DataFrames
  • Multiple column join SQL expression

join() method is used to join two Dataframes together based on condition specified in PySpark Azure Databricks.

Syntax: dataframe_name.join()

What is the syntax of the join() function in PySpark Azure Databricks?

The syntax is as follows:

dataframe_name.join(other, on, how)
Parameter NameRequiredDescription
other (Dataframe)YesIt represents the second column to be joined.
on (str, list, or Column)YesA string for the join column name, a list of column names, a join expression (Column), or a list of Columns. If on is a string or a list of strings indicating the name of the join column(s), the column(s) must exist on both sides.
how (str)OptionalIt represents join type, by default how=”inner”.
Table 1: join() Method in PySpark Databricks Parameter list with Details

Apache Spark Official documentation link: join()

Create a simple DataFrame

Gentle reminder:

In Databricks,

  • sparkSession made available as spark
  • sparkContext made available as sc

In case, you want to create it manually, use the below code.

from pyspark.sql.session import SparkSession

spark = SparkSession.builder 
    .master("local[*]") 
    .appName("azurelib.com") 
    .getOrCreate()

sc = spark.sparkContext

a) Create manual PySpark DataFrame

# 1. Student Dataset

student_data = [
    (1,"Karee",1,"A"),
    (2,"Jobie",2,"B"),
    (3,"Kyle",3,"A"),
    (4,"Georges",1,"B"),
    (5,"Tracey",2,"A")
]

std_df = spark.createDataFrame(student_data, schema=["id","name","dept_id","section"])
std_df.printSchema()
std_df.show(truncate=False)

"""
root
 |-- id: long (nullable = true)
 |-- name: string (nullable = true)
 |-- dept_id: long (nullable = true)
 |-- section: string (nullable = true)

+---+-------+-------+-------+
|id |name   |dept_id|section|
+---+-------+-------+-------+
|1  |Karee  |1      |A      |
|2  |Jobie  |2      |B      |
|3  |Kyle   |3      |A      |
|4  |Georges|1      |B      |
|5  |Tracey |2      |A      |
+---+-------+-------+-------+
"""

# 2. Department Dataset

dept_data = [
    (1,"civil","A",301),
    (1,"civil","B",302),
    (2,"mech","A",401),
    (2,"mech","B",402),
    (3,"ece","A",501),
    (3,"ece","B",502)
]

columns = ["dept_id","dept_name","section","hall_no"]
dept_df = spark.createDataFrame(data = dept_data, schema=columns)
dept_df.printSchema()
dept_df.show(truncate=False)

"""
root
 |-- dept_id: long (nullable = true)
 |-- dept_name: string (nullable = true)
 |-- section: string (nullable = true)
 |-- hall_no: long (nullable = true)

+-------+---------+-------+-------+
|dept_id|dept_name|section|hall_no|
+-------+---------+-------+-------+
|1      |civil    |A      |301    |
|1      |civil    |B      |302    |
|2      |mech     |A      |401    |
|2      |mech     |B      |402    |
|3      |ece      |A      |501    |
|3      |ece      |B      |502    |
+-------+---------+-------+-------+
"""

b) Creating a DataFrame by reading files

Download and use the below source file.

Download Sample Student.c
Download Sample Department.csv
# replace the file_paths with the source file location which you have downloaded.

std_df_2 = spark.read.format("csv").option("header", True).load(student_file_path)
std_df_2.printSchema()

dept_df_2 = spark.read.format("csv").option("header", True).load(department_file_path)
dept_df_2.printSchema()

"""
root
 |-- id: long (nullable = true)
 |-- name: string (nullable = true)
 |-- dept_id: long (nullable = true)
 |-- section: string (nullable = true)

root
 |-- dept_id: long (nullable = true)
 |-- dept_name: string (nullable = true)
 |-- section: string (nullable = true)
 |-- hall_no: long (nullable = true)
"""

Note: Here, I will be using the manually created DataFrame.

How to join multiple columns in PySpark Azure Databricks?

Before joining DataFrames based on multiple columns, let’s discuss a bit about the dataset we have. We have a student dataset that includes their id, name, department id, and section. We have a department dataset with the department id, name, section, and hall number. Now, let’s try joining both DataFrames using multiple columns and display each student and their hall number.

Example:

Note: In order to join DataFrames based on multiple conditions, multiple join conditions has been specified.

# 1. Multiple join
multiple_join_condition = (std_df.dept_id == dept_df.dept_id) & (std_df.section == dept_df.section)
std_dept_df = std_df.join(dept_df, multiple_join_condition , how="inner")

# 2. How to select a duplicate column
std_dept_df.select("id","name", "dept_name", dept_df.section, "hall_no").show()

"""
Output:

+---+-------+---------+-------+-------+
| id|   name|dept_name|section|hall_no|
+---+-------+---------+-------+-------+
|  1|  Karee|    civil|      A|    301|
|  4|Georges|    civil|      B|    302|
|  5| Tracey|     mech|      A|    401|
|  2|  Jobie|     mech|      B|    402|
|  3|   Kyle|      ece|      A|    501|
+---+-------+---------+-------+-------+
"""

Remember, always specific condition inside ‘()’ brackets, because ‘==' has lower precedence than bitwise AND and OR.

Since we have duplicate columns in the std_dept_df DataFrame, we can’t call straight away, which leads to an error. Therefore, we have used the department DataFrame to call the specific column.

How to join multiple columns in PySpark Azure Databricks using the where() function?

In this section, let’s see how to join multiple columns using the where() function in PySpark Azure Databricks by solving the above-mentioned problem.

Example:

# 1. Multiple join using where
first_condition = std_df.dept_id == dept_df.dept_id
second_condition = std_df.section == dept_df.section

std_dept_df_2 = std_df.join(dept_df, first_condition, how="inner") # <- first condition in join()
std_dept_df_2 = std_dept_df_2.where(second_condition) # <- Second condition in where() or filter()

# 2. How to select a duplicate column
std_dept_df_2.select("id","name", "dept_name", dept_df.section, "hall_no").show()

"""
Output:

+---+-------+---------+-------+-------+
| id|   name|dept_name|section|hall_no|
+---+-------+---------+-------+-------+
|  1|  Karee|    civil|      A|    301|
|  5| Tracey|     mech|      A|    401|
|  3|   Kyle|      ece|      A|    501|
|  4|Georges|    civil|      B|    302|
|  2|  Jobie|     mech|      B|    402|
+---+-------+---------+-------+-------+
"""

In this example, we have specified two joining conditions, by applying the first condition in the join() function and the second condition in the where() function.

How to remove duplicate columns while joining DataFrames in PySpark Azure Databricks?

In this section, let’s see how to remove duplicate columns while joining DataFrames in PySpark Azure Databricks with an example.

Example:

std_df.join(dept_df, ["dept_id", "section"], how="inner").show()

"""
Output:

+-------+-------+---+-------+---------+-------+
|dept_id|section| id|   name|dept_name|hall_no|
+-------+-------+---+-------+---------+-------+
|      1|      A|  1|  Karee|    civil|    301|
|      1|      B|  4|Georges|    civil|    302|
|      2|      A|  5| Tracey|     mech|    401|
|      2|      B|  2|  Jobie|     mech|    402|
|      3|      A|  3|   Kyle|      ece|    501|
+-------+-------+---+-------+---------+-------+
"""

In this example, we have no duplicate columns in the output DataFrame. By specifying the column names in array format, we can remove the duplicate columns.

How to join multiple columns in PySpark Azure Databricks using SQL expression?

In this section, let’s try to join multiple columns using student and department datasets using SQL expressions. In order to use a raw SQL expression, we have to convert our DataFrame into a SQL view.

std_df.createOrReplaceTempView("student")
dept_df.createOrReplaceTempView("department")

Example 1:

# 1. With duplicate joining columns

spark.sql('''
    SELECT *
    FROM student AS std
    JOIN department AS dept
    ON std.dept_id = dept.dept_id AND std.section = dept.section
''').show()

"""
Output:

+---+-------+-------+-------+-------+---------+-------+-------+
| id|   name|dept_id|section|dept_id|dept_name|section|hall_no|
+---+-------+-------+-------+-------+---------+-------+-------+
|  1|  Karee|      1|      A|      1|    civil|      A|    301|
|  4|Georges|      1|      B|      1|    civil|      B|    302|
|  5| Tracey|      2|      A|      2|     mech|      A|    401|
|  2|  Jobie|      2|      B|      2|     mech|      B|    402|
|  3|   Kyle|      3|      A|      3|      ece|      A|    501|
+---+-------+-------+-------+-------+---------+-------+-------+
"""

Example 2:

# 2: Without duplicate joining columns

spark.sql('''
    SELECT *
    FROM student AS std
    JOIN department AS dept
    USING (dept_id, section)
''').show()

"""
Output:

+-------+-------+---+-------+---------+-------+
|dept_id|section| id|   name|dept_name|hall_no|
+-------+-------+---+-------+---------+-------+
|      1|      A|  1|  Karee|    civil|    301|
|      1|      B|  4|Georges|    civil|    302|
|      2|      A|  5| Tracey|     mech|    401|
|      2|      B|  2|  Jobie|     mech|    402|
|      3|      A|  3|   Kyle|      ece|    501|
+-------+-------+---+-------+---------+-------+
"""

I have attached the complete code used in this blog in notebook format to this GitHub link. You can download and import this notebook in databricks, jupyter notebook, etc.

GitHub Link

When should you use multiple column join in PySpark using Azure Databricks?

This could be the possible reason:

To extract all the Left DataFrame and Right DataFrame records based n multiple columns matching values.

Real World Use Case Scenarios for using multiple columns join in PySpark Azure Databricks?

Assume that you have a student and department data set. The student dataset has the student id, name, department id, and section. And the department dataset has the id, name, section, and hall number. Now, you were given a requirement to combine both DataFrames to show student’s hall numbers to avoid classroom confusion. You can those techniques mentioned in the above section to achieve the requirement.

What are the alternatives for multiple-column joining in PySpark using Azure Databricks?

There are multiple alternatives for multiple-column joining in PySpark DataFrame, which are as follows:

  • DataFrame.join(): used for combining DataFrames
  • Using PySpark SQL expressions

Final Thoughts

In this article, we have learned about how to join multiple columns in PySpark Azure Databricks along with the examples explained clearly. I have also covered different scenarios with practical examples that could be possible. I hope the information that was provided helped in gaining knowledge.

Please share your comments and suggestions in the comment section below and I will try to answer all your queries as time permits.

Join Telegram Group
Download Free Azure Videos
Connect on LinkedIn

PySpark in Azure Databricks, as explained by Arud Seka Berne S on azurelib.com.
Arud Seka Berne S

As a big data engineer, I design and build scalable data processing systems and integrate them with various data sources and databases. I have a strong background in Python and am proficient in big data technologies such as Hadoop, Hive, Spark, Databricks, and Azure. My interest lies in working with large datasets and deriving actionable insights to support informed business decisions.