Pyspark Dataframe – Map Strings to Numeric

In this article, we are going to see how to convert map strings to numeric.

Creating dataframe for demonstration:

Here we are creating a row of data for college names and then pass the createdataframe() method and then we are displaying the dataframe.

Python3

# importing module
import pyspark

# importing sparksession from pyspark.sql module and Row module
from pyspark.sql import SparkSession,Row

# creating sparksession and giving an app name
spark = SparkSession.builder.appName('sparkdf').getOrCreate()

# list  of college data
dataframe = spark.createDataFrame([Row("vignan"),
                                   Row("rvrjc"),
                                   Row("klu"),
                                   Row("rvrjc"),
                                   Row("klu"),
                                   Row("vignan"),
                                   Row("iit")],
                                  ["college"])

# display dataframe
dataframe.show()

Output:

Method 1: Using map() function

Here we created a function to convert string to numeric through a lambda expression

Syntax: dataframe.select(“string_column_name”).rdd.map(lambda x: string_to_numeric(x[0])).map(lambda x: Row(x)).toDF([“numeric_column_name”]).show()

where,

• dataframe is the pyspark dataframe
• string_column_name is the actual column to be mapped to numeric_column_name
• string_to_numericis the function used to take numeric data
• lambda expression is to call the function such that numeric value is returned

Here we are going to create a college spark dataframe using the Row method and then we are going to map the numeric value by using the lambda function and rename college name as college_number. For that, we are going to create a function and check the condition and return numeric value 1 if college is IIT, return numeric value 2 if college is vignan, return numeric value 3 if college is rvrjc, return numeric value 4 if college is other than above three

Python3

# function that converts string to numeric
def string_to_numeric(x):

      # return numeric value 1 if college is iit
    if(x == 'iit'):
       return 1
    elif(x == "vignan"):

    # return numeric value 2 if college is vignan
       return 2
    elif(x == "rvrjc"):

      # return numeric value 3 if college is rvrjc
       return 3
    else:

    # return numeric value 4 if college
    # is other than above three
       return 4

# map the  numeric value by using lambda
# function and rename college name as college_number
dataframe.select("college").
rdd.map(lambda x: string_to_numeric(x[0])).
map(lambda x: Row(x)).toDF(["college_number"]).show()

Output:

Method 2: Using withColumn() method.

Here we are using withColumn() method to select the columns.

Syntax: dataframe.withColumn(“string_column”, when(col(“column”)==’value’, 1)).otherwise(value))

Where

• dataframe is the pyspark dataframe
• string_column is the column to be mapped to numeric
• value is the numeric value

Example: Here we are going to create a college spark dataframe using Row method and  map college name with college number using with column method along with when().

Python3

# import col and when modules
from pyspark.sql.functions import col, when

# map college name with college number
# using with column method along with when module
dataframe.withColumn("college_number",
                     when(col("college")=='iit', 1)
                     .when(col("college")=='vignan', 2)
                     .when(col("college")=='rvrjc', 3)
                     .otherwise(4)).show()

Output: