asddfl created SPARK-60657:
------------------------------
Summary: pandas-on-Spark sum does not skip a NaN generated at
runtime by cos(inf)
Key: SPARK-60657
URL: https://issues.apache.org/jira/browse/SPARK-60657
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
The default sum aggregation over np.cos of a float64 Series behaves differently
in Pandas and the pandas API on Spark. For input [inf, 0.0], cos produces a NaN
and the finite value 1.0. Pandas sum skips the NaN and returns 1.0, but
pandas-on-Spark returns NaN.
Both queries complete successfully. The example isolates aggregation of a
computed NaN.
{code:python}
import pandas as pd
import numpy as np
from pyspark.sql import SparkSession
import pyspark.pandas as ps
t0 = pd.DataFrame({'c0': [float('inf'), 0.0]})
print("Pandas:", flush=True)
result = np.cos(t0['c0']).sum()
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': [float('inf'), 0.0]})
print("PySpark Pandas:", flush=True)
result = np.cos(t0['c0']).sum()
print(result, flush=True)
spark.stop()
{code}
{code:bash}
Pandas:
1.0
PySpark Pandas:
nan
{code}
*Expected behavior*
The default sum(skipna=True) behavior should skip the generated NaN and return
1.0, matching Pandas.
*Actual behavior*
Pandas prints 1.0, while pandas-on-Spark prints nan.
--
This message was sent by Atlassian Jira
(v10.3.26#10030026)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
