CODEWITHSUNDEEP
apache-sparkdataframepyspark
Mpizos Dimitris
Mpizos Dimitris

Reputation: 4991

Spark assign value if null to column (python)

Assuming that I have the following data

+--------------------+-----+--------------------+
|              values|count|             values2|
+--------------------+-----+--------------------+
|              aaaaaa|  249|                null|
|              bbbbbb|  166|                  b2|
|              cccccc| 1680|           something|
+--------------------+-----+--------------------+

So if there is a null value in values2 column how to assign the values1 column to it? So the result should be:

+--------------------+-----+--------------------+
|              values|count|             values2|
+--------------------+-----+--------------------+
|              aaaaaa|  249|              aaaaaa|
|              bbbbbb|  166|                  b2|
|              cccccc| 1680|           something|
+--------------------+-----+--------------------+

I thought of something of the following but it doesnt work:

df.na.fill({"values2":df['values']}).show()

I found this way to solve it but there should be something more clear forward:

def change_null_values(a,b):
    if b:
        return b
    else:
        return a

udf_change_null = udf(change_null_values,StringType())

df.withColumn("values2",udf_change_null("values","values2")).show()

Upvotes: 3

Views: 26741

Answers (3)

ZettaP
ZettaP

Reputation: 1359

Following up on @shadow_dev's method :

    df.withColumn("values2", 
                  when(col("values2").isNull(), col("values1"))
                  .otherwise(col("values2")))

Dmytro Popovych's solution is still the cleanest.

If one needs more fancy when/otherwise logic :

df.withColumn("values2", when(col("values2").isNull() | col("values3").isNull(), col("values1"))
.when(col("values1") == col("values2"), 1)
.otherwise(0))

Upvotes: 4

shadow_dev
shadow_dev

Reputation: 130

You can use the column attribute .isNull().

df.where(col("dt_mvmt").isNull())

df.where(col("dt_mvmt").isNotNull())

This answer comes from this answer - I just don't have enough reputation to add a comment.

Upvotes: -3

Dmytro Popovych
Dmytro Popovych

Reputation: 954

You can use https://spark.apache.org/docs/1.6.2/api/python/pyspark.sql.html#pyspark.sql.functions.coalesce

df.withColumn('values2', coalesce(df.values2, df.values)).show()

Upvotes: 5

Related Questions