将 PySpark DenseVector 转换为数组

Question

我正在尝试将 DenseVector 的 pyspark 数据帧列转换为数组，但我总是遇到错误。

data = [(Vectors.dense([8.0, 1.0, 3.0, 2.0, 5.0]),),
(Vectors.dense([2.0, 0.0, 3.0, 4.0, 5.0]),),
(Vectors.dense([4.0, 0.0, 0.0, 6.0, 7.0]),)]

df = spark.createDataFrame(data,["features"])

我试图定义一个 UDF 并使用 toArray()

to_array = udf(lambda x: x.toArray(), ArrayType(FloatType()))
df = df.withColumn('features', to_array('features'))

但是，如果我执行 df.collect()，我会收到以下错误

org.apache.spark.SparkException: Job aborted due to stage failure: Task 1 in stage 17.0 failed 4 times, 
most recent failure: Lost task 1.3 in stage 17.0 (TID 100, 10.139.64.6, executor 0): 
net.razorvine.pickle.PickleException: expected zero arguments for construction of ClassDict 
(for numpy.core.multiarray._reconstruct)

关于如何实现这一目标的任何想法？

Answer 1

toArray()返回一个不能ArrayType(FloatType())隐式转换的 numpy.ndarray 。另外使用.tolist()来转换它：

import pyspark.sql.functions as F
import pyspark.sql.types as T

#or: to_array = F.udf(lambda v: list([float(x) for x in v]), T.ArrayType(T.FloatType()))
to_array = F.udf(lambda v: v.toArray().tolist(), T.ArrayType(T.FloatType()))
df = df.withColumn('features', to_array('features'))

如果您使用的是 Pyspark >=3.0.0，您可以使用新的vector_to_array函数：

from pyspark.ml.functions import vector_to_array
df = df.withColumn('features', vector_to_array('features'))