How to get L2 norm of an array type column in PySpark?

I have a PySpark dataframe.

df1 = spark.createDataFrame([
    ("u1", [0, 1, 2]),
    ("u1", [1, 2, 3]),
    ("u2", [2, 3, 4]),

    ],
    ['user_id', 'features'])

print(df1.printSchema())
df1.show(truncate=False)

Output-

root
 |-- user_id: string (nullable = true)
 |-- features: array (nullable = true)
 |    |-- element: long (containsNull = true)

None
+-------+---------+
|user_id|features |
+-------+---------+
|u1     |[0, 1, 2]|
|u1     |[1, 2, 3]|
|u2     |[2, 3, 4]|
+-------+---------+

I want to get the L2 norm of the features, so I wrote a UDF-

def norm_2_func(features):
    return features/np.linalg.norm(features, 2)

norm_2_udf = udf(norm_2_func, ArrayType(FloatType()))
df2 = df1.withColumn('l2_features', norm_2_udf(F.col('features')))

But it is throwing some errors. How can I achieve this?

The expected output is -

+-------+---------+----------------------+
|user_id|features |               L2_norm|
+-------+---------+----------------------+
|u1     |[0, 1, 2]| [0.000, 0.447, 0.894]|
|u1     |[1, 2, 3]| [0.267, 0.534, 0.801]|
|u2     |[2, 3, 4]| [0.371, 0.557, 0.742]|
+-------+---------+----------------------+

Numpy arrays contain numpy dtypes which needs to be cast to normal Python dtypes (float/int etc.) before returning:

import numpy as np
import pyspark.sql.functions as F
from pyspark.sql.types import ArrayType, FloatType

def norm_2_func(features):
    return [float(i) for i in features/np.linalg.norm(features, 2)]
    # you can also use
    # return list(map(float, features/np.linalg.norm(features, 2)))

norm_2_udf = F.udf(norm_2_func, ArrayType(FloatType()))
df2 = df1.withColumn('l2_features', norm_2_udf(F.col('features')))

df2.show(truncate=False)
+-------+---------+-----------------------------------+
|user_id|features |l2_features                        |
+-------+---------+-----------------------------------+
|u1     |[0, 1, 2]|[0.0, 0.4472136, 0.8944272]        |
|u1     |[1, 2, 3]|[0.26726124, 0.5345225, 0.80178374]|
|u2     |[2, 3, 4]|[0.37139067, 0.557086, 0.74278134] |
+-------+---------+-----------------------------------+