unflatten - convert flat dict to nested dict/array

Description

This package provides a function which can unpack a flat dictionary into a structured dict with nested sub-dicts and/or sub-lists.

Development takes place on github. The package is installable from PyPI.

Synopsis

Nested dicts:

>>> from unflatten import unflatten

>>> unflatten({'foo.bar': 'val'})
{'foo': {'bar': 'val'}}

Nested list:

>>> unflatten({'foo[0]': 'x', 'foo[1]': 'y'})
{'foo': ['x', 'y']}

Nested lists and dicts, intermixed:

>>> unflatten({
...     'foo[0][0]': 'a',
...     'foo[0][1]': 'b',
...     'foo[1].x': 'c',
...      })
{'foo': [['a', 'b'], {'x': 'c'}]}

Notes

Unflatten takes a single argument which should either be a dict (or an object with a dict-like .items() or .iteritems() method) or a sequence of (key, value) pairs. All keys in the dict or sequence must be strings. (Under python 2, keys must be instances of basestring; under python 3, keys just be instances of str.)

Unflatten always returns a dict. By way of example:

>>> unflatten([('[0]', 'x')])
{'': ['x']}

For list-valued nodes, all indexes must be present in the input (flattened) mapping, otherwise a ValueError will be thrown:

>>> unflatten({'a[0]': 'x', 'a[2]': 'y'})
Traceback (most recent call last):
...
ValueError: missing key 'a[1]'

See Also

The morph and flattery packages purport to implement similar functions.

Authors

Jeff Dairiki

History

Release 0.2.0 (2024-09-04)

No substantive code changes from 0.1.1.

Features

• Added type annotations.

Testing

• Test under python 3.10, 3.11, and 3.12.

• Fix tox config to cope with the fact that recent tox/virtualenv does not support EOLed versions of python.

Release 0.1.1 (2021-08-16)

Nits

• Fix backslashes in docstrings

Packaging

• PEP517-ize the packaging
• Use setuptools-scm to maintain version numbers

Testing

• Test under Python 3.7, 3.8, 3.9 and PyPy 3.7. Drop testing for Python 3.4 & 3.5.
• Pin pip version for pypy2 (see pip #8653)
• Clean up and modernize the tox lint and coverage environments

Release 0.1 (2018-01-17)

No code changes.

This package is now deemed "production ready" (though your mileage may vary.)

Release 0.1b1 (2018-01-09)

Initial release.