PySpark DataFrame - Convert JSON Column to Row using json_tuple

Code description

PySpark SQL functions json_tuple can be used to convert DataFrame JSON string columns to tuples (new rows in the DataFrame). 

Syntax of this function looks like the following:

pyspark.sql.functions.json_tuple(col, *fields)

The first parameter is the JSON string column name in the DataFrame and the second is the filed name list to extract.

If you need to extract complex JSON documents like JSON arrays, you can follow this article - PySpark: Convert JSON String Column to Array of Object (StructType) in DataFrame.

Output

StructType([StructField('id', LongType(), True), StructField('c0', StringType(), True), StructField('c1', StringType(), True), StructField('c2', StringType(), True)])

+---+---+------+----------+
| id| c0|    c1|        c2|
+---+---+------+----------+
|  1|  1|10.201|2021-01-01|
|  2|  2|20.201|2022-01-01|
+---+---+------+----------+

Code snippet

from pyspark.sql import SparkSession
from pyspark.sql.functions import json_tuple

app_name = "PySpark json_tuple sql functions"
master = "local"

spark = SparkSession.builder \
    .appName(app_name) \
    .master(master) \
    .getOrCreate()

spark.sparkContext.setLogLevel("WARN")

# Create a DataFrame
df = spark.createDataFrame(
    [[1, '{"Attr_INT":1, "ATTR_DOUBLE":10.201, "ATTR_DATE": "2021-01-01"}'],
     [2, '{"Attr_INT":2, "ATTR_DOUBLE":20.201, "ATTR_DATE": "2022-01-01"}']], ['id', 'json_col'])

# Extract JSON values
df = df.select(df.id, json_tuple(
    df.json_col, 'Attr_INT', 'ATTR_DOUBLE', 'ATTR_DATE'))
print(df.schema)
df.show()