.. image:: https://github.com/ianlini/flatten-dict/actions/workflows/main.yml/badge.svg :target: https://github.com/ianlini/flatten-dict/actions .. image:: https://img.shields.io/pypi/v/flatten-dict.svg :target: https://pypi.org/project/flatten-dict/ .. image:: https://img.shields.io/pypi/l/flatten-dict.svg :target: https://github.com/ianlini/flatten-dict/blob/master/LICENSE .. image:: https://img.shields.io/github/stars/ianlini/flatten-dict.svg?style=social :target: https://github.com/ianlini/flatten-dict

A flexible utility for flattening and unflattening dict-like objects in Python.

This package provides a function flatten() for flattening dict-like objects in Python (tested with 3.10-3.15). It also provides some key joining methods (reducer), and you can choose the reducer you want or even implement your own reducer. You can also invert the resulting flat dict using unflatten().

.. code-block:: bash

pip install flatten-dict

Flatten

.. code-block:: python

   def flatten(d, reducer='tuple', inverse=False, enumerate_types=(), keep_empty_types=()):
       """Flatten `Mapping` object.

       Parameters
       ----------
       d : dict-like object
           The dict that will be flattened.
       reducer : {'tuple', 'path', 'underscore', 'dot', Callable}
           The key joining method. If a `Callable` is given, the `Callable` will be
           used to reduce.
           'tuple': The resulting key will be tuple of the original keys.
           'path': Use `os.path.join` to join keys.
           'underscore': Use underscores to join keys.
           'dot': Use dots to join keys.
       inverse : bool
           Whether you want invert the resulting key and value.
       max_flatten_depth : int
           Maximum depth to merge.
       enumerate_types : Sequence[type]
           Flatten these types using `enumerate`.
           For example, if we set `enumerate_types` to ``(list,)``,
           `list` indices become keys: ``{'a': ['b', 'c']}`` -> ``{('a', 0): 'b', ('a', 1): 'c'}``.
       keep_empty_types : Sequence[type]
           By default, ``flatten({1: 2, 3: {}})`` will give you ``{(1,): 2}``, that is, the key ``3``
           will disappear.
           This is also applied for the types in `enumerate_types`, that is,
           ``flatten({1: 2, 3: []}, enumerate_types=(list,))`` will give you ``{(1,): 2}``.
           If you want to keep those empty values, you can specify the types in `keep_empty_types`:

           >>> flatten({1: 2, 3: {}}, keep_empty_types=(dict,))
           {(1,): 2, (3,): {}}

       Returns
       -------
       flat_dict : dict
       """

Examples
::::::::

>>> from flatten_dict import flatten
>>> from pprint import pprint
>>> normal_dict = {
...     'a': '0',
...     'b': {
...         'a': '1.0',
...         'b': '1.1',
...     },
...     'c': {
...         'a': '2.0',
...         'b': {
...             'a': '2.1.0',
...             'b': '2.1.1',
...         },
...     },
... }
>>> pprint(flatten(normal_dict))
{('a',): '0',
 ('b', 'a'): '1.0',
 ('b', 'b'): '1.1',
 ('c', 'a'): '2.0',
 ('c', 'b', 'a'): '2.1.0',
 ('c', 'b', 'b'): '2.1.1'}
>>> pprint(flatten(normal_dict, reducer='path'))
{'a': '0',
 'b/a': '1.0',
 'b/b': '1.1',
 'c/a': '2.0',
 'c/b/a': '2.1.0',
 'c/b/b': '2.1.1'}
>>> pprint(flatten(normal_dict, reducer='path', inverse=True))
{'0': 'a',
 '1.0': 'b/a',
 '1.1': 'b/b',
 '2.0': 'c/a',
 '2.1.0': 'c/b/a',
 '2.1.1': 'c/b/b'}
>>> pprint(flatten(normal_dict, reducer='path', max_flatten_depth=2))
{'a': '0',
 'b/a': '1.0',
 'b/b': '1.1',
 'c/a': '2.0',
 'c/b': {'a': '2.1.0', 'b': '2.1.1'}}

The `reducer` parameter supports ``'tuple'``, ``'path'``, ``'underscore'``, ``'dot'`` and `Callable`. We can customize the reducer using a function:

>>> def underscore_reducer(k1, k2):
...     if k1 is None:
...         return k2
...     else:
...         return k1 + "_" + k2
...
>>> pprint(flatten(normal_dict, reducer=underscore_reducer))
{'a': '0',
 'b_a': '1.0',
 'b_b': '1.1',
 'c_a': '2.0',
 'c_b_a': '2.1.0',
 'c_b_b': '2.1.1'}

There is also a factory function `make_reducer()` to help you create customized reducer. The function currently only supports customized delimiter:

>>> from flatten_dict.reducers import make_reducer
>>> pprint(flatten(normal_dict, reducer=make_reducer(delimiter='_')))
{'a': '0',
 'b_a': '1.0',
 'b_b': '1.1',
 'c_a': '2.0',
 'c_b_a': '2.1.0',
 'c_b_b': '2.1.1'}

If we have some iterable (e.g., `list`) in the `dict`, we will normally get this:

>>> flatten({'a': [1, 2, 3], 'b': 'c'})
{('a',): [1, 2, 3], ('b',): 'c'}

If we want to use its indices as keys, then we can use the parameter `enumerate_types`:

>>> flatten({'a': [1, 2, 3], 'b': 'c'}, enumerate_types=(list,))
{('a', 0): 1, ('a', 1): 2, ('a', 2): 3, ('b',): 'c'}

We can even flatten a `list` directly:

>>> flatten([1, 2, 3], enumerate_types=(list,))
{(0,): 1, (1,): 2, (2,): 3}

If there is an empty dict in the values, by default, it will disappear after flattened:

>>> flatten({1: 2, 3: {}})
{(1,): 2}

We can keep the empty dict in the result using ``keep_empty_types=(dict,)``:

>>> flatten({1: 2, 3: {}}, keep_empty_types=(dict,))
{(1,): 2, (3,): {}}

Unflatten

.. code-block:: python

def unflatten(d, splitter='tuple', inverse=False): """Unflatten dict-like object.

   Parameters
   ----------
   d : dict-like object
       The dict that will be unflattened.
   splitter : {'tuple', 'path', 'underscore', 'dot', Callable}
       The key splitting method. If a Callable is given, the Callable will be
       used to split `d`.
       'tuple': Use each element in the tuple key as the key of the unflattened dict.
       'path': Use `pathlib.Path.parts` to split keys.
       'underscore': Use underscores to split keys.
       'dot': Use dots to split keys.
   inverse : bool
       Whether you want to invert the key and value before flattening.

   Returns
   -------
   unflattened_dict : dict
   """

Examples ::::::::

from pprint import pprint from flatten_dict import unflatten flat_dict = { ... ('a',): '0', ... ('b', 'a'): '1.0', ... ('b', 'b'): '1.1', ... ('c', 'a'): '2.0', ... ('c', 'b', 'a'): '2.1.0', ... ('c', 'b', 'b'): '2.1.1', ... } pprint(unflatten(flat_dict)) {'a': '0', 'b': {'a': '1.0', 'b': '1.1'}, 'c': {'a': '2.0', 'b': {'a': '2.1.0', 'b': '2.1.1'}}} flat_dict = { ... 'a': '0', ... 'b/a': '1.0', ... 'b/b': '1.1', ... 'c/a': '2.0', ... 'c/b/a': '2.1.0', ... 'c/b/b': '2.1.1', ... } pprint(unflatten(flat_dict, splitter='path')) {'a': '0', 'b': {'a': '1.0', 'b': '1.1'}, 'c': {'a': '2.0', 'b': {'a': '2.1.0', 'b': '2.1.1'}}} flat_dict = { ... '0': 'a', ... '1.0': 'b/a', ... '1.1': 'b/b', ... '2.0': 'c/a', ... '2.1.0': 'c/b/a', ... '2.1.1': 'c/b/b', ... } pprint(unflatten(flat_dict, splitter='path', inverse=True)) {'a': '0', 'b': {'a': '1.0', 'b': '1.1'}, 'c': {'a': '2.0', 'b': {'a': '2.1.0', 'b': '2.1.1'}}}

The splitter parameter supports 'tuple', 'path', 'underscore', 'dot' and Callable. We can customize the reducer using a function:

def underscore_splitter(flat_key): ... return flat_key.split("_") ... flat_dict = { ... 'a': '0', ... 'b_a': '1.0', ... 'b_b': '1.1', ... 'c_a': '2.0', ... 'c_b_a': '2.1.0', ... 'c_b_b': '2.1.1', ... } pprint(unflatten(flat_dict, splitter=underscore_splitter)) {'a': '0', 'b': {'a': '1.0', 'b': '1.1'}, 'c': {'a': '2.0', 'b': {'a': '2.1.0', 'b': '2.1.1'}}}

There is also a factory function make_splitter() to help you create customized splitter. The function currently only supports customized delimiter:

from flatten_dict.splitters import make_splitter pprint(unflatten(flat_dict, splitter=make_splitter(delimiter='_'))) {'a': '0', 'b': {'a': '1.0', 'b': '1.1'}, 'c': {'a': '2.0', 'b': {'a': '2.1.0', 'b': '2.1.1'}}}