when/otherwise & expressions
Apache Spark · DataFrames
Theory
F.when(condition, value) is Spark's version of SQL CASE WHEN. You chain as many .when() calls as you need and finish with .otherwise(default); Spark checks the conditions top to bottom and returns the value of the first one that is true. Without otherwise, rows that match nothing get NULL. The result is a column expression, so you use it inside withColumn, select, filter, groupBy or orderBy.
The same logic can be written as a SQL string with F.expr("CASE WHEN amount >= 50000 THEN 'High' ... END"). Combine it with other helpers: F.coalesce picks the first non-null value, isin checks a list, and F.greatest/F.least compare columns. For example, banding orders by amount gives 55000 as High, 30000 as Medium and 1500 as Low, and you can then group by the band to count orders in each.
Why we use it
Business rules such as customer tiers, discount slabs, status flags and data clean-up are conditional logic. when/otherwise expresses them as optimized column expressions instead of slow Python UDFs.
When to use it
Use when/otherwise for if/else rules on columns, mapping codes to labels, or building flags for conditional aggregation (sum(when(...))). Order conditions from most specific to least. For long lookup lists, a join with a small mapping table is easier to maintain.
How it works
- For each row
- amount >= 50000? Yes: band = High. No: Check the next condition.
- amount >= 20000? Yes: band = Medium. No: Go to otherwise.
- otherwise: band = Low
- Column added to the output
Syntax
F.when(cond1, v1).when(cond2, v2).otherwise(default)
F.expr("CASE WHEN cond THEN v ELSE d END")
F.sum(F.when(F.col("region") == "North", F.col("amount")).otherwise(0))
F.coalesce(c1, c2, F.lit(0))Example
from pyspark.sql import SparkSession, functions as F
spark = (SparkSession.builder.master("local[1]")
.config("spark.ui.enabled", "false")
.getOrCreate())
orders = spark.createDataFrame([
(101, 1, "North", "Laptop", 1, 55000.0, "2026-01-05"),
(102, 2, "West", "Phone", 2, 30000.0, "2026-01-06"),
(103, 1, "North", "Mouse", 3, 1500.0, "2026-01-06"),
(104, 3, "South", "Laptop", 1, 52000.0, "2026-01-07"),
(105, 4, "West", "Tablet", 1, 25000.0, "2026-01-08"),
(106, 2, "East", "Phone", 1, 15000.0, "2026-01-08"),
(107, 5, "North", "Tablet", 2, 48000.0, "2026-01-09"),
(108, 3, "South", "Mouse", 4, 2000.0, "2026-01-10"),
], "order_id INT, customer_id INT, region STRING, product STRING, "
"qty INT, amount DOUBLE, order_date STRING"
).withColumn("order_date", F.to_date("order_date"))
banded = orders.withColumn(
"band",
F.when(F.col("amount") >= 50000, "High")
.when(F.col("amount") >= 20000, "Medium")
.otherwise("Low"))
banded.select("order_id", "amount", "band").show()
# Conditional aggregation with when()
orders.agg(
F.sum(F.when(F.col("region") == "North", F.col("amount")).otherwise(0)).alias("north_sales"),
F.count(F.when(F.col("qty") > 1, True)).alias("multi_qty_orders")).show()
# Same rule as a SQL CASE expression
orders.select("order_id", F.expr(
"CASE WHEN product IN ('Laptop', 'Tablet') THEN 'Computing' ELSE 'Other' END"
).alias("category")).limit(3).show()
# Output:
# +--------+-------+------+
# |order_id| amount| band|
# +--------+-------+------+
# | 101|55000.0| High|
# | 102|30000.0|Medium|
# | 103| 1500.0| Low|
# | 104|52000.0| High|
# | 105|25000.0|Medium|
# | 106|15000.0| Low|
# | 107|48000.0|Medium|
# | 108| 2000.0| Low|
# +--------+-------+------+
#
# +-----------+----------------+
# |north_sales|multi_qty_orders|
# +-----------+----------------+
# | 104500.0| 4|
# +-----------+----------------+
#
# +--------+---------+
# |order_id| category|
# +--------+---------+
# | 101|Computing|
# | 102| Other|
# | 103| Other|
# +--------+---------+
Key points
- Conditions are checked in order; the first true one wins
- Without .otherwise(), unmatched rows get NULL
- Wrap combined conditions in parentheses: (a > 1) & (b < 5)
- sum(when(...)) and count(when(...)) give conditional aggregates in one pass
- F.expr('CASE WHEN ...') gives the same result as when/otherwise