PySpark: StructField(..., ..., False) always returns `nullable=true` instead of `nullable=false`

Question

I'm new to PySpark and am facing a strange problem. I'm trying to set some column to non-nullable while loading a CSV dataset. I can reproduce my case with a very small dataset (test.csv):

col1,col2,col3
11,12,13
21,22,23
31,32,33
41,42,43
51,,53

There is a null value at row 5, column 2 and I don't want to get that row inside my DF. I set all fields as non-nullable (nullable=false) but I get a schema with all the three columns having nullable=true. This happens even if I set all the three columns as non-nullable! I'm running the latest available version of Spark, 2.0.1.

Here's the code:

from pyspark.sql import SparkSession
from pyspark.sql.functions import *
from pyspark.sql.types import *

spark = SparkSession \
    .builder \
    .appName("Python Spark SQL basic example") \
    .config("spark.some.config.option", "some-value") \
    .getOrCreate()

struct = StructType([   StructField("col1", StringType(), False), \
                        StructField("col2", StringType(), False), \
                        StructField("col3", StringType(), False) \
                    ])

df = spark.read.load("test.csv", schema=struct, format="csv", header="true")

df.printSchema() returns:

root
 |-- col1: string (nullable = true)
 |-- col2: string (nullable = true)
 |-- col3: string (nullable = true)

and df.show() returns:

+----+----+----+
|col1|col2|col3|
+----+----+----+
|  11|  12|  13|
|  21|  22|  23|
|  31|  32|  33|
|  41|  42|  43|
|  51|null|  53|
+----+----+----+

while I expect this:

root
 |-- col1: string (nullable = false)
 |-- col2: string (nullable = false)
 |-- col3: string (nullable = false)

+----+----+----+
|col1|col2|col3|
+----+----+----+
|  11|  12|  13|
|  21|  22|  23|
|  31|  32|  33|
|  41|  42|  43|
+----+----+----+

Answer 1

While Spark behavior (switch from False to True here is confusing there is nothing fundamentally wrong going on here. nullable argument is not a constraint but a reflection of the source and type semantics which enables certain types of optimization

You state that you want to avoid null values in your data. For this you should use na.drop method.

df.na.drop()

For other ways of handling nulls please take a look at the DataFrameNaFunctions (exposed using DataFrame.na property) documentation.

CSV format doesn't provide any tools which allow you to specify data constraints so by definition reader cannot assume that input is not null and your data indeed contains nulls.