Convert python pandas iterator and string concat into pyspark

I am attempting to move a process from Pandas into Pyspark, but I am a complete novice in the latter. Note: This is an EDA process so I am not too worried about having it as a loop for now, I can optimise that at a later date.

Set up:

import pandas as pd
import numpy as np
import pyspark.pandas as ps

Dummy Data:

df = ps.DataFrame({'id': ['ID_01', 'ID_02', 'ID_02', 'ID_03', 'ID_03'], 'name': ['Jack', 'John', 'John', 'James', 'Jamie']})
df_pandas = df.to_pandas()
df_spark = df.to_spark()
df
id name
ID_01 Jack
ID_02 John
ID_02 John
ID_03 James
ID_03 Jamie

Pandas code:

unique_ids = df_pandas['id'].unique()
for unique_id in unique_ids:
  names = '; '.join(sorted(df_pandas[df_pandas['id'] == unique_id]['name'].unique()))
  df.loc[df['id'] == unique_id, 'name'] = names
df
id name
ID_01 Jack
ID_02 John
ID_02 John
ID_03 James; Jamie
ID_03 James; Jamie

This last table is the desired output. However, I am having issues achieving this in PySpark. This is where I have got to:

unique_ids = df_spark.select('id').distinct().collect()
for unique_id in unique_ids:
    names = df_spark.filter(df_spark.id == unique_id.id).select('name').distinct()

I am then unsure how to do the next steps; i.e. how to concatenate the resulting single column DataFrame, nor how to ensure the correct replacement.

I have investigated the following sources, with no success (likely due to my inexperience in PySpark):

  • This answer shows how to concatenate columns and not rows
  • This answer might be helpful for the loc conversion (but I have not managed to get there yet
  • This answer initially proved promising, since it would remove the need for the loop as well, but I could not figure out how to do the distinct and sort equivalents on the collect_list output object

>Solution :

Try:

import pyspark.sql.functions as f

new_df = (df_spark.select(['name', 'id'])
                  .distinct()
                  .groupby('id')
                  .agg(f.concat_ws('; ', f.collect_list('name'))
                        .alias('name')))

out_df = (df_spark.join(new_df, df_spark['id'] == new_df['id'], 'left')
                  .drop(df_spark['name']).drop(new_df['id']))

Output:

>>> out_df.show()

+-----+------------+
|   id|        name|
+-----+------------+
|ID_01|        Jack|
|ID_02|        John|
|ID_02|        John|
|ID_03|James; Jamie|
|ID_03|James; Jamie|
+-----+------------+

Leave a Reply