Pyspark Dataframe - 如何根据作为输入的列数组连接列

Question

I have dataframe of 10 columns and want to do function - concatenation based on Array of columns which come as input:

arr = ["col1", "col2", "col3"]

This is current so far:

newDF = rawDF.select(concat(col("col1"), col("col2"), col("col3") )).exceptAll(updateDF.select( concat(col("col1"), col("col2"), col("col3") ) ) )

Also:

df3 = df2.join(df1, concat( df2.col1, df2.col2, df2.col3, df2.col3 ) == df1.col5 ) 

But I want to make a loop or function to do this based on input array (not hard-coding it as is now).

What is the best way?

Answer 1

You can unpack the cols using (*). In the pyspark.sql docs, if any functions have (*cols), it means that you can unpack the cols. For concat:

pyspark.sql.functions.concat(*cols)

from pyspark.sql import functions as F
arr = ["col1", "col2", "col3"]
newDF = rawDF.select(F.concat(*(F.col(col) for col in arr))).exceptAll(updateDF.select(F.concat(*(F.col(col) for col in arr))))

For joins:

arr=['col1','col2','col3']
df3 = df2.join(df1, F.concat(*(F.col(col) for col in arr)) == df1.col5 )