Pandas flattening nested jsons

user_18036486 0
Viewed 53

so this is probably going to be a duplicate question but i'll make a try since I have not found anything.

I am trying to flatten a json with pandas, normal work. Looking at the examples of the docs here is the closest example for what I am trying to do:

data = [{'state': 'Florida',
         'shortname': 'FL',
         'info': {'governor': 'Rick Scott'},
         'counties': [{'name': 'Dade', 'population': 12345},
                      {'name': 'Broward', 'population': 40000},
                      {'name': 'Palm Beach', 'population': 60000}]},
        {'state': 'Ohio',
         'shortname': 'OH',
         'info': {'governor': 'John Kasich'},
         'counties': [{'name': 'Summit', 'population': 1234},
                      {'name': 'Cuyahoga', 'population': 1337}]}]
result = pd.json_normalize(data, 'counties', ['state', 'shortname',
                                           ['info', 'governor']])
result
         name  population    state shortname info.governor
0        Dade       12345   Florida    FL    Rick Scott
1     Broward       40000   Florida    FL    Rick Scott
2  Palm Beach       60000   Florida    FL    Rick Scott
3      Summit        1234   Ohio       OH    John Kasich
4    Cuyahoga        1337   Ohio       OH    John Kasich

However, this example show us a way to get the data inside counties flatten with alongside the column state and shortname. Let's say that I have n number of columns at the root of each json object ( n number of state or shortname columns in the example above ). How do I include them all, in order to flat the counties but keep everything else that is adjacent?

First I tried things like these:

#None to treat data as a list of records
#Result of counties is still nested, not working
result = pd.json_normalize(data, None, ['counties'])

or


result = pd.json_normalize(data, None, ['counties', 'name'])

Then I Thought of getting the columns with dataframe.columns and reuse it since meta argument of json_normalize can take array of string.

But i'm stuck. and columns appear to return nested json attribute which I don't want to.

#still nested
cols = pd.json_normalize(data).columns.to_list()
#Exclude it because we already have  it 
cols = [index for index in cols if index != 'counties']
#remove nested columns if any
cols = [index for index in cols if "." not in index]

result = pd.json_normalize(data, 'counties', cols, errors="ignore")

#still nested

         name  population    state shortname  ...     other6     other7                                           counties info.governor
0        Dade       12345  Florida        FL  ...  dumb_data  dumb_data  [{'name': 'Dade', 'population': 12345}, {'name...           NaN
1     Broward       40000  Florida        FL  ...  dumb_data  dumb_data  [{'name': 'Dade', 'population': 12345}, {'name...           NaN
2  Palm Beach       60000  Florida        FL  ...  dumb_data  dumb_data  [{'name': 'Dade', 'population': 12345}, {'name...           NaN
3      Summit        1234     Ohio        OH  ...  dumb_data  dumb_data  [{'name': 'Summit', 'population': 1234}, {'nam...           NaN
4    Cuyahoga        1337     Ohio        OH  ...  dumb_data  dumb_data  [{'name': 'Summit', 'population': 1234}, {'nam...           NaN

I would prefere not to just harcode the column names since they change and that for this case I have 64 of them...

For better understanding, this is the real kind of data i'm working on from Woo Rest API. I am not using it here because its really long, but basically I am trying to flat line_items keeping only product_id inside it and of course all the other columns which is adjacent to line_items.

pandas python json
1 Answers

Okay so guys if you want to flatten a json and keeping everything else, you should used pd.Dataframe.explode()

Here is my logic:

import pandas as pd

data = [
        {'state': 'Florida',
        'shortname': 'FL',
        'info': {'governor': 'Rick Scott'},
        'counties': [
                      {'name': 'Dade', 'population': 12345},
                      {'name': 'Broward', 'population': 40000},
                      {'name': 'Palm Beach', 'population': 60000}
        ]
        },
        {'state': 'Ohio',
        'shortname': 'OH',
        'info': {'governor': 'John Kasich'},
        'counties': [{'name': 'Summit', 'population': 1234},
        {'name': 'Cuyahoga', 'population': 1337}]}
]
#No Formating only converting to a Df
result = pd.json_normalize(data)

#Exploding the wanted nested column
exploded = result.explode('counties')

#Keeping the name only - this can be custom
exploded['countie_name'] = exploded['counties'].apply(lambda x: x['name'])

#Drop the used column since we took what interested us inside it.
exploded = exploded.drop(['counties'], axis=1)

print(exploded)

#Duplicate for Florida, as wanted with diferent countie names
     state shortname info.governor countie_name
0  Florida        FL    Rick Scott         Dade
0  Florida        FL    Rick Scott      Broward
0  Florida        FL    Rick Scott   Palm Beach
1     Ohio        OH   John Kasich       Summit
1     Ohio        OH   John Kasich     Cuyahoga

Imagine you have the content of a basket of product as a nested json, to explode the content of the basket while keeping the general basket attributes, you can do this.

user_180364860
Related