sampleBy (DataFrame)

Returns a stratified sample without replacement based on the fraction given on each stratum.

Syntax

sampleBy(col: "ColumnOrName", fractions: Dict[Any, float], seed: Optional[int] = None)

Parameters

Parameter	Type	Description
col	Column or str	column that defines strata.
fractions	dict	sampling fraction for each stratum. If a stratum is not specified, we treat its fraction as zero.
seed	int, optional	random seed.

Returns

a new DataFrame that represents the stratified sample.

Examples

from pyspark.sql import functions as sf
dataset = spark.range(0, 100, 1, 5).select((sf.col("id") % 3).alias("key"))
sampled = dataset.sampleBy("key", fractions={0: 0.1, 1: 0.2}, seed=0)
sampled.groupBy("key").count().orderBy("key").show()
# +---+-----+
# |key|count|
# +---+-----+
# |  0|    4|
# |  1|    9|
# +---+-----+

dataset.sampleBy(sf.col("key"), fractions={2: 1.0}, seed=0).count()
# 33