Pandas Mapping: add suffix based on tree structure

Question

1. The Data:

I have the following structure in a pandas DataFrame:

import pandas as pd
df = pd.DataFrame([['A', 'NaN', 'A', 'NaN'],['B', 'A', 'B/A', 'A'], ['B1', 'B', 'B1/B/A', 'B/A'], 
                     ['B2', 'B', 'B2/B/A', 'B/A'], ['C', 'B1', 'C/B1/B/A', 'B1/B/A'], ['D', 'B1', 'D/B1/B/A', 'B1/B/A'], 
                     ['E', 'B2', 'E/B2/B/A', 'B2/B/A']], 
                    columns=['unit_id', 'group_id', 'new_unit_id', 'new_group_id'])

2. The issue and the goal:

I would like to replace the current unit_id and group_id with a value that appends it's parent structure as basically:

= + '/' + parent
and
= parent

as you would see in a file tree structure or similar.

Something like this:

index	unit_id	group_id	new_unit_id	new_group_id
0	A	NaN	A	NaN
1	B	A	B/A	A
2	B1	B	B1/B/A	B/A
3	B2	B	B2/B/A	B/A
4	C	B1	C/B1/B/A	B1/B/A
5	D	B1	D/B1/B/A	B1/B/A
6	E	B2	E/B2/B/A	B2/B/A

3. Attempts and approach:

I have tried mapping in place without creating 'new' columns, but run into issues where when the unit_id of a parent changes, it is not reflected in the group_id of its children.

df['unit_id'] = df['unit_id'] + '/' + df['group_id']

So it seems like I need to iterate row by row so the change on the previous row is factored in. Something like:

df['unit_id'] = df.apply(lambda row : row['unit_id'].replace(str(row['unit_id']), str(row['unit_id'] + '/' + row['group_id'])), axis=1)

This results in the same (inaccurate) values as above, but I think df.apply with the correct anonymous (lambda) function is a lot closer to what I need. Having trouble getting the syntax correct.

user_16343464 · Answer

This can be approach using graph theory.

This is your graph:

You can use networkx to compute your graph and find the shortest_path:

import networkx as nx

# ensure real NaN
df = df.replace('NaN', np.nan)

G = nx.from_pandas_edgelist(df.dropna(subset='group_id'),
                            source='unit_id', target='group_id',
                            create_using=nx.DiGraph)

#get final item
last = list(nx.topological_sort(G))[-1]

# get simple paths
df['new_unit_id'] = ['/'.join(nx.shortest_path(G, s, last))
                      if not pd.isna(s) else float('nan')
                      for s in df['unit_id']]

df['new_group_id'] = df['new_unit_id'].str.extract(r'[^/]+/(.+)')

output:

  unit_id group_id new_unit_id new_group_id
0       A      NaN           A          NaN
1       B        A         B/A            A
2      B1        B      B1/B/A          B/A
3      B2        B      B2/B/A          B/A
4       C       B1    C/B1/B/A       B1/B/A
5       D       B1    D/B1/B/A       B1/B/A
6       E       B2    E/B2/B/A       B2/B/A

Pandas Mapping: add suffix based on tree structure

1 Answers