prateekmehta/flatten

Flatten JSON in Python

flatten_json

Flattens JSON objects in Python. flatten_json flattens the hierarchy in your object which can be useful if you want to force your objects into a table.

Installation

pip install flatten_json

flatten

Usage

Let's say you have the following object:

dic = {
    "a": 1,
    "b": 2,
    "c": [{"d": [2, 3, 4], "e": [{"f": 1, "g": 2}]}]
}

which you want to flatten. Just apply flatten:

from flatten_json import flatten
flatten(dic)

Results:

{'a': '1',
 'b': '2',
 'c_0_d_0': '2',
 'c_0_d_1': '3',
 'c_0_d_2': '4',
 'c_0_e_0_f': '1',
 'c_0_e_0_g': '2'}

Usage with Pandas

For the following object:

dic = [
    {"a": 1, "b": 2, "c": {"d": 3, "e": 4}},
    {"a": 0.5, "c": {"d": 3.2}},
    {"a": 0.8, "b": 1.8},
]

We can apply flatten to each element in the array and then use pandas to capture the output as a dataframe:

dic_flattened = (flatten(d) for d in dic)

which creates an array of flattened objects:

[{'a': '1', 'b': '2', 'c_d': '3', 'c_e': '4'},
 {'a': '0.5', 'c_d': '3.2'},
 {'a': '0.8', 'b': '1.8'}]

Finally you can use pd.DataFrame to capture the flattened array:

import pandas as pd
df = pd.DataFrame(dic_flattened)

The final result as a Pandas dataframe:

	a	b	c_d	c_e
0	1	2	3	4
1	0.5	NaN	3.2	NaN
2	0.8	1.8	NaN	NaN

Custom separator

By default _ is used to separate nested element. You can change this by passing the desired character:

flatten({"a": [1]}, '|')

returns:

{'a|0': 1}

Thanks to @jvalhondo, @drajen, and @azaitsev for contributing to this feature.

Ignore root keys

By default flatten goes through all the keys in the object. If you are not interested in output from a set of keys you can pass this set as an argument to root_keys_to_ignore:

dic = {
    'a': {'a': [1, 2, 3]},
    'b': {'b': 'foo', 'c': 'bar'},
    'c': {'c': [{'foo': 5, 'bar': 6, 'baz': [1, 2, 3]}]}
}
flatten(dic, root_keys_to_ignore={'b', 'c'})

returns:

{
    'a_a_0': 1,
    'a_a_1': 2,
    'a_a_2': 3
}

This feature can prevent unnecessary processing which is a concern with deeply nested objects.

Thanks to @mcarans and @aquilax for requesting and helping with fleshing out this feature.

unflatten

Reverses the flattening process. Example usage:

from flatten_json import unflatten

dic = {
    'a': 1,
    'b_a': 2,
    'b_b': 3,
    'c_a_b': 5
}
unflatten(dic)

returns:

{
    'a': 1,
    'b': {'a': 2, 'b': 3},
    'c': {'a': {'b': 5}}
}

Unflatten with lists

flatten encodes key for list values with integer indices which makes it ambiguous for reversing the process. Consider this flattened dictionary:

a = {'a': 1, 'b_0': 5}

Both {'a': 1, 'b': [5]} and {'a': 1, 'b': {0: 5}} are legitimate answers.

Calling unflatten_list the dictionary is first unflattened and then in a post-processing step the function looks for a list pattern (zero-indexed consecutive integer keys) and transforms the matched values into a list.

Here's an example:

from flatten_json import unflatten_list
dic = {
    'a': 1,
    'b_0': 1,
    'b_1': 2,
    'c_a': 'a',
    'c_b_0': 1,
    'c_b_1': 2,
    'c_b_2': 3
}
unflatten_list(dic)

returns:

{
    'a': 1,
    'b': [1, 2],
    'c': {'a': 'a', 'b': [1, 2, 3]}
}

Thanks to @nmaas87 for requesting this feature and @aquilax for making it actually work.