count_min_sketch

Returns a count-min sketch of a column with the given esp, confidence and seed. The result is an array of bytes, which can be deserialized to a CountMinSketch before usage. Count-min sketch is a probabilistic data structure used for cardinality estimation using sub-linear space.

Syntax

Python

from pyspark.sql import functions as sf

sf.count_min_sketch(col, eps, confidence, seed=None)

Parameters

Parameter	Type	Description
col	pyspark.sql.Column or str	Target column to compute on.
eps	pyspark.sql.Column or float	Relative error, must be positive.
confidence	pyspark.sql.Column or float	Confidence, must be positive and less than 1.0.
seed	pyspark.sql.Column or int, optional	Random seed.

Returns

pyspark.sql.Column: count-min sketch of the column

Examples

Example 1: Using columns as arguments

Python

from pyspark.sql import functions as sf
spark.range(100).select(
    sf.hex(sf.count_min_sketch(sf.col("id"), sf.lit(3.0), sf.lit(0.1), sf.lit(1)))
).show(truncate=False)

Output

+------------------------------------------------------------------------+
|hex(count_min_sketch(id, 3.0, 0.1, 1))                                  |
+------------------------------------------------------------------------+
|0000000100000000000000640000000100000001000000005D8D6AB90000000000000064|
+------------------------------------------------------------------------+

Example 2: Using numbers as arguments

Python

from pyspark.sql import functions as sf
spark.range(100).select(
    sf.hex(sf.count_min_sketch("id", 1.0, 0.3, 2))
).show(truncate=False)

Output

+----------------------------------------------------------------------------------------+
|hex(count_min_sketch(id, 1.0, 0.3, 2))                                                  |
+----------------------------------------------------------------------------------------+
|0000000100000000000000640000000100000002000000005D96391C00000000000000320000000000000032|
+----------------------------------------------------------------------------------------+

Example 3: Using a long seed

Python

from pyspark.sql import functions as sf
spark.range(100).select(
    sf.hex(sf.count_min_sketch("id", sf.lit(1.5), 0.2, 1111111111111111111))
).show(truncate=False)

Output

+----------------------------------------------------------------------------------------+
|hex(count_min_sketch(id, 1.5, 0.2, 1111111111111111111))                                |
+----------------------------------------------------------------------------------------+
|00000001000000000000006400000001000000020000000044078BA100000000000000320000000000000032|
+----------------------------------------------------------------------------------------+