How to create a nested JSON from pandas DataFrame?

Question

I am trying to generate a nested JSON from a DataFrame, where attributes of a car are distributed in several rows.

DataFrame

cars = {'brand': ['Honda','Toyota','Ford','Audi','Honda','Toyota','Ford','Audi'],
        'model': ['Civic','Corolla','Focus','A4','Civic','Corolla','Focus','A4'],
        'attributeName': ['color','color','color','color','doors','doors','doors','doors'],
        'attributeValue': ['red','blue','black','red',2,4,4,2]
        }

df = pd.DataFrame(cars)

What I tried

At first I grouped the rows and tried to apply the nesting:

df.groupby(['brand','model'])\
             .apply(lambda x: x[['attributeName','attributeValue']].to_dict('records'))\
             .to_json(orient='records')

Result

[[{"attributeName":"color","attributeValue":"red"},{"attributeName":"doors","attributeValue":2}],[{"attributeName":"color","attributeValue":"black"},{"attributeName":"doors","attributeValue":4}],[{"attributeName":"color","attributeValue":"red"},{"attributeName":"doors","attributeValue":2}],[{"attributeName":"color","attributeValue":"blue"},{"attributeName":"doors","attributeValue":4}]]

Expected result

[
    {
        'brand':'Honda',
        'model':'Civic',
        'attributes':[
            {
                'name':'color',
                'value':'red'
            }
        ]
    },
    {...}
]

So what can I do to get also the other records and not only the attributes?

jezrael · Accepted Answer · 2021-03-24 09:06:03Z

In your solution is added rename with reset_index():

d = {'attributeName':'name','attributeValue':'value'}
j = df.rename(columns=d).groupby(['brand','model']).apply(lambda x: x[['name','value']].to_dict('records')).reset_index(name='attributes').to_json(orient='records')
print (j)
[{"brand":"Audi","model":"A4","attributes":[{"name":"color","value":"red"},{"name":"doors","value":2}]},{"brand":"Ford","model":"Focus","attributes":[{"name":"color","value":"black"},{"name":"doors","value":4}]},{"brand":"Honda","model":"Civic","attributes":[{"name":"color","value":"red"},{"name":"doors","value":2}]},{"brand":"Toyota","model":"Corolla","attributes":[{"name":"color","value":"blue"},{"name":"doors","value":4}]}]

Or:

d = {'attributeName':'name','attributeValue':'value'}
j = df.rename(columns=d).groupby(['brand','model']).apply(lambda x: x[['name','value']].to_dict('records')).explode().apply(lambda x: [x]).reset_index(name='attributes').to_json(orient='records')
print (j)
[{"brand":"Audi","model":"A4","attributes":[{"name":"color","value":"red"}]},{"brand":"Audi","model":"A4","attributes":[{"name":"doors","value":2}]},{"brand":"Ford","model":"Focus","attributes":[{"name":"color","value":"black"}]},{"brand":"Ford","model":"Focus","attributes":[{"name":"doors","value":4}]},{"brand":"Honda","model":"Civic","attributes":[{"name":"color","value":"red"}]},{"brand":"Honda","model":"Civic","attributes":[{"name":"doors","value":2}]},{"brand":"Toyota","model":"Corolla","attributes":[{"name":"color","value":"blue"}]},{"brand":"Toyota","model":"Corolla","attributes":[{"name":"doors","value":4}]}]

df['attributes'] = df.apply(lambda x: [{'name': x['attributeName'], 'value': x['attributeValue']}], axis=1)
df = df.drop(['attributeName','attributeValue'], axis=1)
print (df)
    brand    model                             attributes
0   Honda    Civic    [{'name': 'color', 'value': 'red'}]
1  Toyota  Corolla   [{'name': 'color', 'value': 'blue'}]
2    Ford    Focus  [{'name': 'color', 'value': 'black'}]
3    Audi       A4    [{'name': 'color', 'value': 'red'}]
4   Honda    Civic        [{'name': 'doors', 'value': 2}]
5  Toyota  Corolla        [{'name': 'doors', 'value': 4}]
6    Ford    Focus        [{'name': 'doors', 'value': 4}]
7    Audi       A4        [{'name': 'doors', 'value': 2}]

j = df.to_json(orient='records')
print (j)
[{"brand":"Honda","model":"Civic","attributes":[{"name":"color","value":"red"}]},{"brand":"Toyota","model":"Corolla","attributes":[{"name":"color","value":"blue"}]},{"brand":"Ford","model":"Focus","attributes":[{"name":"color","value":"black"}]},{"brand":"Audi","model":"A4","attributes":[{"name":"color","value":"red"}]},{"brand":"Honda","model":"Civic","attributes":[{"name":"doors","value":2}]},{"brand":"Toyota","model":"Corolla","attributes":[{"name":"doors","value":4}]},{"brand":"Ford","model":"Focus","attributes":[{"name":"doors","value":4}]},{"brand":"Audi","model":"A4","attributes":[{"name":"doors","value":2}]}]

That looks great and is working - Going for the first option, cause I am not sure how the explode() works in the moment, but will research it more closely.
@HedgeHog - I was a bit not sure, how looks expcted ouput, first solution get different ouput like secnd and third.

Henry Ecker · Accepted Answer · 2022-10-30 19:17:54Z

0

This should give you the desired output:

d = {'attributeName':'name','attributeValue':'value'}
df_cars= (df.rename(columns=d).groupby(['brand','model']).apply(lambda x:x[['name','value']].drop_duplicates().to_dict('records')).reset_index().rename(columns={0:'attributes'}))

df_cars.head(10)

edited Oct 30, 2022 at 19:17

Henry Ecker♦

35.8k19 gold badges48 silver badges67 bronze badges

answered Aug 23, 2022 at 19:18

user13356552

1

1 Comment

Mike_H Over a year ago

Please mark code as such by using code fencing

Collectives™ on Stack Overflow

How to create a nested JSON from pandas DataFrame?

DataFrame

What I tried

Result

Expected result

2 Answers 2

2 Comments

1 Comment

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

DataFrame

What I tried

Result

Expected result

2 Answers 2

2 Comments

1 Comment

Your Answer

Sign up or log in

Post as a guest

Linked

Related