Pyspark 数据框过滤器 OR 条件

Question

I am trying to filter my pyspark dataframe based on an OR condition like so:

filtered_df = file_df.filter(file_df.dst_name == "ntp.obspm.fr").filter(file_df.fw == "4940" | file_df.fw == "4960")

I want to return only rows where file_df.fw == "4940" OR file_df.fw == "4960" However when I try this I get this error:

Py4JError: An error occurred while calling o157.or. Trace:
py4j.Py4JException: Method or([class java.lang.String]) does not exist

What am I doing wrong?

Without the OR condition it works when I try to filter only on one condition ( file_df.fw=="4940" )

Answer 1

The error message is caused by the different priorities of the operators. The | (OR) has a higher priority as the comparison operator == . Spark tries to apply the OR on
"4940" and file_df.fw and not like you want it on (file_df.fw == "4940") and (file_df.fw == "4960") . You can change the priorities by using brackets. Have a look at the following example:

columns = ['dst_name','fw']

file_df=spark.createDataFrame([('ntp.obspm.fr','3000'),
                               ('ntp.obspm.fr','4940'),
                               ('ntp.obspm.fr','4960'),
                               ('ntp.obspm.de', '4940' )],
                              columns)

#here I have added the brackets
filtered_df = file_df.filter(file_df.dst_name == "ntp.obspm.fr").filter((file_df.fw == "4940") | (file_df.fw == "4960"))
filtered_df.show()

Output:

+------------+----+ 
|    dst_name|  fw| 
+------------+----+ 
|ntp.obspm.fr|4940| 
|ntp.obspm.fr|4960| 
+------------+----+