这里有一个解决方案。

import pandas as pd
from pyspark.sql.functions import concat_ws, udf
from pyspark.sql.types import StringType

pandas_df = pd.DataFrame({
    'column_1': [None, '1', None, None, '1', '1', '1'],
    'column_2': [None, None, '1', None, '1', '1', None],
    'column_3': [None, None, None, '1', None, '1', None]
})

df = spark.createDataFrame(pandas_df)
df.show()

# +--------+--------+--------+

# |column_1|column_2|column_3|

# +--------+--------+--------+

# |    null|    null|    null|

# |       1|    null|    null|

# |    null|       1|    null|

# |    null|    null|       1|

# |       1|       1|    null|

# |       1|       1|       1|

# |       1|    null|    null|

# +--------+--------+--------+

def non_null_to_column_name(name):
    return udf(lambda value: None if value is None else name, StringType())

atleast_two_udf = udf(lambda s: None if (s is None) or (',' not in s) else s, 
                      StringType())

cols = []
for name in df.columns:
    f = non_null_to_column_name(name)
    cols += [f(df[name])]

df = df.withColumn('collection', atleast_two_udf(concat_ws(',', *cols)))
df.show()

# +--------+--------+--------+--------------------+

# |column_1|column_2|column_3|          collection|

# +--------+--------+--------+--------------------+

# |    null|    null|    null|                null|

# |       1|    null|    null|                null|

# |    null|       1|    null|                null|

# |    null|    null|       1|                null|

# |       1|       1|    null|   column_1,column_2|

# |       1|       1|       1|column_1,column_2...|

# |       1|    null|    null|                null|

# +--------+--------+--------+--------------------+