asddfl created SPARK-60656:
------------------------------
Summary: pandas-on-Spark np.maximum fails to propagate a missing
operand generated by log
Key: SPARK-60656
URL: https://issues.apache.org/jira/browse/SPARK-60656
Project: Spark
Issue Type: Bug
Components: Pandas API on Spark, PySpark
Affects Versions: 4.2.0
Environment: Platform: Ubuntu 24.04 Linux-6.17.0-29-generic x86_64
Python: 3.11.17
pyspark 4.2.0
findspark 2.0.1
pandas 3.0.6
numpy 2.4.6
Reporter: asddfl
When using the pandas API on Spark (pyspark.pandas), np.maximum does not
preserve the missing result of np.log(-1.0). Pandas/NumPy returns NaN, while
pandas-on-Spark returns the other operand, -1.0.
The example contains one float64 value and computes the missing operand at
runtime. Both queries complete successfully, but the result changes from
missing to a finite number.
{code:python}
import pandas as pd
import numpy as np
from pyspark.sql import SparkSession
import pyspark.pandas as ps
t0 = pd.DataFrame({'c0': [-1.0]})
print("Pandas:", flush=True)
result = np.maximum(np.log(t0['c0']), t0['c0'])
print(result, flush=True)
spark = (
SparkSession.builder
.master("local[2]")
.config("spark.sql.ansi.enabled", "false")
.config("spark.sql.shuffle.partitions", "2")
.getOrCreate()
)
ps.set_option("compute.ops_on_diff_frames", True)
t0 = ps.DataFrame({'c0': [-1.0]})
print("PySpark Pandas:", flush=True)
result = np.maximum(np.log(t0['c0']), t0['c0'])
print(result, flush=True)
spark.stop()
{code}
{code:bash}
Pandas:
0 NaN
Name: c0, dtype: float64
PySpark Pandas:
0 -1.0
Name: c0, dtype: float64
{code}
*Expected behavior*
np.maximum(np.log(-1.0), -1.0) should propagate the missing operand and return
NaN, matching Pandas/NumPy.
*Actual behavior*
Pandas prints NaN. pandas-on-Spark prints -1.0. Both result Series have dtype
float64.
--
This message was sent by Atlassian Jira
(v10.3.26#10030026)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
