Skip to main content

A set that remembers its order, and allows looking up its items by their index in that order.

Project description

Travis Codecov Pypi

An OrderedSet is a mutable data structure that is a hybrid of a list and a set. It remembers the order of its entries, and every entry has an index number that can be looked up.

Usage examples

An OrderedSet is created and used like a set:

>>> from ordered_set import OrderedSet

>>> letters = OrderedSet('abracadabra')

>>> letters
OrderedSet(['a', 'b', 'r', 'c', 'd'])

>>> 'r' in letters

It is efficient to find the index of an entry in an OrderedSet, or find an entry by its index. To help with this use case, the .add() method returns the index of the added item, whether it was already in the set or not.

>>> letters.index('r')

>>> letters[2]

>>> letters.add('r')

>>> letters.add('x')

OrderedSets implement the union (|), intersection (&), and difference (-) operators like sets do.

>>> letters |= OrderedSet('shazam')

>>> letters
OrderedSet(['a', 'b', 'r', 'c', 'd', 'x', 's', 'h', 'z', 'm'])

>>> letters & set('aeiou')

>>> letters -= 'abcd'

>>> letters
OrderedSet(['r', 'x', 's', 'h', 'z', 'm'])

The __getitem__() and index() methods have been extended to accept any iterable except a string, returning a list, to perform NumPy-like "fancy indexing".

>>> letters = OrderedSet('abracadabra')

>>> letters[[0, 2, 3]]
['a', 'r', 'c']

>>> letters.index(['a', 'r', 'c'])
[0, 2, 3]

OrderedSet implements __getstate__ and __setstate__ so it can be pickled, and implements the abstract base classes collections.MutableSet and collections.Sequence.

OrderedSet can be used as a generic collection type, similar to the collections in the typing module like List, Dict, and Set. For example, you can annotate a variable as having the type OrderedSet[str] or OrderedSet[Tuple[int, str]].

OrderedSet in data science applications

An OrderedSet can be used as a bi-directional mapping between a sparse vocabulary and dense index numbers. As of version 3.1, it accepts NumPy arrays of index numbers as well as lists.

This combination of features makes OrderedSet a simple implementation of many of the things that pandas.Index is used for, and many of its operations are faster than the equivalent pandas operations.

For further compatibility with pandas.Index, get_loc (the pandas method for looking up a single index) and get_indexer (the pandas method for fancy indexing in reverse) are both aliases for index (which handles both cases in OrderedSet).


OrderedSet was implemented by Robyn Speer. Jon Crall contributed changes and tests to make it fit the Python set API. Roman Inflianskas added the original type annotations.


The original implementation of OrderedSet was a recipe posted to ActiveState Recipes by Raymond Hettiger, released under the MIT license.

Hettiger's implementation kept its content in a doubly-linked list referenced by a dict. As a result, looking up an item by its index was an O(N) operation, while deletion was O(1).

This version makes different trade-offs for the sake of efficient lookups. Its content is a standard Python list instead of a doubly-linked list. This provides O(1) lookups by index at the expense of O(N) deletion, as well as slightly faster iteration.

In Python 3.6 and later, the built-in dict type is inherently ordered. If you ignore the dictionary values, that also gives you a simple ordered set, with fast O(1) insertion, deletion, iteration and membership testing. However, dict does not provide the list-like random access features of OrderedSet. You would have to convert it to a list in O(N) to look up the index of an entry or look up an entry by its index.

Project details

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Files for ordered-set, version 4.0.2
Filename, size File type Python version Upload date Hashes
Filename, size ordered-set-4.0.2.tar.gz (10.7 kB) File type Source Python version None Upload date Hashes View

Supported by

AWS AWS Cloud computing Datadog Datadog Monitoring DigiCert DigiCert EV certificate Facebook / Instagram Facebook / Instagram PSF Sponsor Fastly Fastly CDN Google Google Object Storage and Download Analytics Pingdom Pingdom Monitoring Salesforce Salesforce PSF Sponsor Sentry Sentry Error logging StatusPage StatusPage Status page