How to select rows from list in PySpark

Question

Suppose we have two dataframes df1 and df2 where df1 has columns [a, b, c, p, q, r] and df2 has columns [d, e, f, a, b, c]. Suppose the common columns are stored in a list common_cols = ['a', 'b', 'c'].

How do you join the two dataframes using the common_cols list within a sql command? The code below attempts to do this.

common_cols = ['a', 'b', 'c']
filter_df = spark.sql("""
    select * from df1 inner join df2
    on df1.common_cols = df2.common_cols
""")

Answer 1

Demo setup

df1 = spark.createDataFrame([(1,2,3,4,5,6)],['a','b','c','p','q','r'])
df2 = spark.createDataFrame([(7,8,9,1,2,3)],['d','e','f','a','b','c'])
common_cols = ['a','b','c']

df1.show()

+---+---+---+---+---+---+
|  a|  b|  c|  p|  q|  r|
+---+---+---+---+---+---+
|  1|  2|  3|  4|  5|  6|
+---+---+---+---+---+---+


df2.show()

+---+---+---+---+---+---+
|  d|  e|  f|  a|  b|  c|
+---+---+---+---+---+---+
|  7|  8|  9|  1|  2|  3|
+---+---+---+---+---+---+

Solution, based on using (SQL syntax for join)

df1.createOrReplaceTempView('df1')
df2.createOrReplaceTempView('df2')
common_cols_csv = ','.join(common_cols)

query = f'''\
select  * 
from    df1 inner join df2 using ({common_cols_csv})
'''

       

---

print(query)

select  * 
from    df1 inner join df2 using (a,b,c)

---

filter_df = spark.sql(query)

filter_df.show()

+---+---+---+---+---+---+---+---+---+
|  a|  b|  c|  p|  q|  r|  d|  e|  f|
+---+---+---+---+---+---+---+---+---+
|  1|  2|  3|  4|  5|  6|  7|  8|  9|
+---+---+---+---+---+---+---+---+---+

Answer 2

You can do so with using instead of on. See documentation.

common_cols = ['a', 'b', 'c']

spark.sql(
    f'''
    SELECT *
    FROM
    (SELECT 1 a, 2 b, 3 c, 10 val1)
    JOIN
    (SELECT 1 a, 2 b, 3 c, 20 val2)
    USING ({','.join(common_cols)})
    '''
).show()

+---+---+---+----+----+
|  a|  b|  c|val1|val2|
+---+---+---+----+----+
|  1|  2|  3|  10|  20|
+---+---+---+----+----+

Answer 3

Adding to @David ???? Markovitz's answer in order to get the columns in a dynamic fashion you could do something like below -

Input Data

df1 = spark.createDataFrame([(1,2,3,4,5,6)],['a','b','c','p','q','r'])
df2 = spark.createDataFrame([(7,8,9,1,2,3)],['d','e','f','a','b','c'])

df1.createOrReplaceTempView("df1")
df2.createOrReplaceTempView("df2")

Finding the common columns using set intersection

common_cols = set(df1.columns).intersection(set(df2.columns))
print(common_cols)

{'a', 'b', 'c'}

Creating query string -

query = '''
select  * 
from    df1 inner join df2 using ({common_cols})
'''.format(common_cols=', '.join(map(str, common_cols)))

print(query)

select  * 
from    df1 inner join df2 using (a, b, c)

Finally, execute the query within spark.sql -

spark.sql(query).show()

+---+---+---+---+---+---+---+---+---+
|  a|  b|  c|  p|  q|  r|  d|  e|  f|
+---+---+---+---+---+---+---+---+---+
|  1|  2|  3|  4|  5|  6|  7|  8|  9|
+---+---+---+---+---+---+---+---+---+