1
:mod:`collections` --- High-performance container datatypes
2
===========================================================
4
.. module:: collections
5
:synopsis: High-performance datatypes
6
.. moduleauthor:: Raymond Hettinger <python@rcn.com>
7
.. sectionauthor:: Raymond Hettinger <python@rcn.com>
13
from collections import *
15
__name__ = '<doctest>'
17
**Source code:** :source:`Lib/collections.py` and :source:`Lib/_abcoll.py`
21
This module implements specialized container datatypes providing alternatives to
22
Python's general purpose built-in containers, :class:`dict`, :class:`list`,
23
:class:`set`, and :class:`tuple`.
25
===================== ==================================================================== ===========================
26
:func:`namedtuple` factory function for creating tuple subclasses with named fields .. versionadded:: 2.6
27
:class:`deque` list-like container with fast appends and pops on either end .. versionadded:: 2.4
28
:class:`Counter` dict subclass for counting hashable objects .. versionadded:: 2.7
29
:class:`OrderedDict` dict subclass that remembers the order entries were added .. versionadded:: 2.7
30
:class:`defaultdict` dict subclass that calls a factory function to supply missing values .. versionadded:: 2.5
31
===================== ==================================================================== ===========================
33
In addition to the concrete container classes, the collections module provides
34
:ref:`abstract base classes <collections-abstract-base-classes>` that can be
35
used to test whether a class provides a particular interface, for example,
36
whether it is hashable or a mapping.
39
:class:`Counter` objects
40
------------------------
42
A counter tool is provided to support convenient and rapid tallies.
45
>>> # Tally occurrences of words in a list
47
>>> for word in ['red', 'blue', 'red', 'green', 'blue', 'blue']:
50
Counter({'blue': 3, 'red': 2, 'green': 1})
52
>>> # Find the ten most common words in Hamlet
54
>>> words = re.findall(r'\w+', open('hamlet.txt').read().lower())
55
>>> Counter(words).most_common(10)
56
[('the', 1143), ('and', 966), ('to', 762), ('of', 669), ('i', 631),
57
('you', 554), ('a', 546), ('my', 514), ('hamlet', 471), ('in', 451)]
59
.. class:: Counter([iterable-or-mapping])
61
A :class:`Counter` is a :class:`dict` subclass for counting hashable objects.
62
It is an unordered collection where elements are stored as dictionary keys
63
and their counts are stored as dictionary values. Counts are allowed to be
64
any integer value including zero or negative counts. The :class:`Counter`
65
class is similar to bags or multisets in other languages.
67
Elements are counted from an *iterable* or initialized from another
68
*mapping* (or counter):
70
>>> c = Counter() # a new, empty counter
71
>>> c = Counter('gallahad') # a new counter from an iterable
72
>>> c = Counter({'red': 4, 'blue': 2}) # a new counter from a mapping
73
>>> c = Counter(cats=4, dogs=8) # a new counter from keyword args
75
Counter objects have a dictionary interface except that they return a zero
76
count for missing items instead of raising a :exc:`KeyError`:
78
>>> c = Counter(['eggs', 'ham'])
79
>>> c['bacon'] # count of a missing element is zero
82
Setting a count to zero does not remove an element from a counter.
83
Use ``del`` to remove it entirely:
85
>>> c['sausage'] = 0 # counter entry with a zero count
86
>>> del c['sausage'] # del actually removes the entry
91
Counter objects support three methods beyond those available for all
94
.. method:: elements()
96
Return an iterator over elements repeating each as many times as its
97
count. Elements are returned in arbitrary order. If an element's count
98
is less than one, :meth:`elements` will ignore it.
100
>>> c = Counter(a=4, b=2, c=0, d=-2)
101
>>> list(c.elements())
102
['a', 'a', 'a', 'a', 'b', 'b']
104
.. method:: most_common([n])
106
Return a list of the *n* most common elements and their counts from the
107
most common to the least. If *n* is omitted or ``None``,
108
:func:`most_common` returns *all* elements in the counter.
109
Elements with equal counts are ordered arbitrarily:
111
>>> Counter('abracadabra').most_common(3)
112
[('a', 5), ('r', 2), ('b', 2)]
114
.. method:: subtract([iterable-or-mapping])
116
Elements are subtracted from an *iterable* or from another *mapping*
117
(or counter). Like :meth:`dict.update` but subtracts counts instead
118
of replacing them. Both inputs and outputs may be zero or negative.
120
>>> c = Counter(a=4, b=2, c=0, d=-2)
121
>>> d = Counter(a=1, b=2, c=3, d=4)
124
Counter({'a': 3, 'b': 0, 'c': -3, 'd': -6})
126
The usual dictionary methods are available for :class:`Counter` objects
127
except for two which work differently for counters.
129
.. method:: fromkeys(iterable)
131
This class method is not implemented for :class:`Counter` objects.
133
.. method:: update([iterable-or-mapping])
135
Elements are counted from an *iterable* or added-in from another
136
*mapping* (or counter). Like :meth:`dict.update` but adds counts
137
instead of replacing them. Also, the *iterable* is expected to be a
138
sequence of elements, not a sequence of ``(key, value)`` pairs.
140
Common patterns for working with :class:`Counter` objects::
142
sum(c.values()) # total of all counts
143
c.clear() # reset all counts
144
list(c) # list unique elements
145
set(c) # convert to a set
146
dict(c) # convert to a regular dictionary
147
c.items() # convert to a list of (elem, cnt) pairs
148
Counter(dict(list_of_pairs)) # convert from a list of (elem, cnt) pairs
149
c.most_common()[:-n-1:-1] # n least common elements
150
c += Counter() # remove zero and negative counts
152
Several mathematical operations are provided for combining :class:`Counter`
153
objects to produce multisets (counters that have counts greater than zero).
154
Addition and subtraction combine counters by adding or subtracting the counts
155
of corresponding elements. Intersection and union return the minimum and
156
maximum of corresponding counts. Each operation can accept inputs with signed
157
counts, but the output will exclude results with counts of zero or less.
159
>>> c = Counter(a=3, b=1)
160
>>> d = Counter(a=1, b=2)
161
>>> c + d # add two counters together: c[x] + d[x]
162
Counter({'a': 4, 'b': 3})
163
>>> c - d # subtract (keeping only positive counts)
165
>>> c & d # intersection: min(c[x], d[x])
166
Counter({'a': 1, 'b': 1})
167
>>> c | d # union: max(c[x], d[x])
168
Counter({'a': 3, 'b': 2})
172
Counters were primarily designed to work with positive integers to represent
173
running counts; however, care was taken to not unnecessarily preclude use
174
cases needing other types or negative values. To help with those use cases,
175
this section documents the minimum range and type restrictions.
177
* The :class:`Counter` class itself is a dictionary subclass with no
178
restrictions on its keys and values. The values are intended to be numbers
179
representing counts, but you *could* store anything in the value field.
181
* The :meth:`most_common` method requires only that the values be orderable.
183
* For in-place operations such as ``c[key] += 1``, the value type need only
184
support addition and subtraction. So fractions, floats, and decimals would
185
work and negative values are supported. The same is also true for
186
:meth:`update` and :meth:`subtract` which allow negative and zero values
187
for both inputs and outputs.
189
* The multiset methods are designed only for use cases with positive values.
190
The inputs may be negative or zero, but only outputs with positive values
191
are created. There are no type restrictions, but the value type needs to
192
support addition, subtraction, and comparison.
194
* The :meth:`elements` method requires integer counts. It ignores zero and
199
* `Counter class <http://code.activestate.com/recipes/576611/>`_
200
adapted for Python 2.5 and an early `Bag recipe
201
<http://code.activestate.com/recipes/259174/>`_ for Python 2.4.
203
* `Bag class <http://www.gnu.org/software/smalltalk/manual-base/html_node/Bag.html>`_
206
* Wikipedia entry for `Multisets <http://en.wikipedia.org/wiki/Multiset>`_.
208
* `C++ multisets <http://www.demo2s.com/Tutorial/Cpp/0380__set-multiset/Catalog0380__set-multiset.htm>`_
209
tutorial with examples.
211
* For mathematical operations on multisets and their use cases, see
212
*Knuth, Donald. The Art of Computer Programming Volume II,
213
Section 4.6.3, Exercise 19*.
215
* To enumerate all distinct multisets of a given size over a given set of
216
elements, see :func:`itertools.combinations_with_replacement`.
218
map(Counter, combinations_with_replacement('ABC', 2)) --> AA AB AC BB BC CC
221
:class:`deque` objects
222
----------------------
224
.. class:: deque([iterable[, maxlen]])
226
Returns a new deque object initialized left-to-right (using :meth:`append`) with
227
data from *iterable*. If *iterable* is not specified, the new deque is empty.
229
Deques are a generalization of stacks and queues (the name is pronounced "deck"
230
and is short for "double-ended queue"). Deques support thread-safe, memory
231
efficient appends and pops from either side of the deque with approximately the
232
same O(1) performance in either direction.
234
Though :class:`list` objects support similar operations, they are optimized for
235
fast fixed-length operations and incur O(n) memory movement costs for
236
``pop(0)`` and ``insert(0, v)`` operations which change both the size and
237
position of the underlying data representation.
239
.. versionadded:: 2.4
241
If *maxlen* is not specified or is *None*, deques may grow to an
242
arbitrary length. Otherwise, the deque is bounded to the specified maximum
243
length. Once a bounded length deque is full, when new items are added, a
244
corresponding number of items are discarded from the opposite end. Bounded
245
length deques provide functionality similar to the ``tail`` filter in
246
Unix. They are also useful for tracking transactions and other pools of data
247
where only the most recent activity is of interest.
249
.. versionchanged:: 2.6
250
Added *maxlen* parameter.
252
Deque objects support the following methods:
255
.. method:: append(x)
257
Add *x* to the right side of the deque.
260
.. method:: appendleft(x)
262
Add *x* to the left side of the deque.
267
Remove all elements from the deque leaving it with length 0.
272
Count the number of deque elements equal to *x*.
274
.. versionadded:: 2.7
276
.. method:: extend(iterable)
278
Extend the right side of the deque by appending elements from the iterable
282
.. method:: extendleft(iterable)
284
Extend the left side of the deque by appending elements from *iterable*.
285
Note, the series of left appends results in reversing the order of
286
elements in the iterable argument.
291
Remove and return an element from the right side of the deque. If no
292
elements are present, raises an :exc:`IndexError`.
295
.. method:: popleft()
297
Remove and return an element from the left side of the deque. If no
298
elements are present, raises an :exc:`IndexError`.
301
.. method:: remove(value)
303
Removed the first occurrence of *value*. If not found, raises a
306
.. versionadded:: 2.5
308
.. method:: reverse()
310
Reverse the elements of the deque in-place and then return ``None``.
312
.. versionadded:: 2.7
314
.. method:: rotate(n)
316
Rotate the deque *n* steps to the right. If *n* is negative, rotate to
317
the left. Rotating one step to the right is equivalent to:
318
``d.appendleft(d.pop())``.
321
Deque objects also provide one read-only attribute:
323
.. attribute:: maxlen
325
Maximum size of a deque or *None* if unbounded.
327
.. versionadded:: 2.7
330
In addition to the above, deques support iteration, pickling, ``len(d)``,
331
``reversed(d)``, ``copy.copy(d)``, ``copy.deepcopy(d)``, membership testing with
332
the :keyword:`in` operator, and subscript references such as ``d[-1]``. Indexed
333
access is O(1) at both ends but slows to O(n) in the middle. For fast random
334
access, use lists instead.
340
>>> from collections import deque
341
>>> d = deque('ghi') # make a new deque with three items
342
>>> for elem in d: # iterate over the deque's elements
343
... print elem.upper()
348
>>> d.append('j') # add a new entry to the right side
349
>>> d.appendleft('f') # add a new entry to the left side
350
>>> d # show the representation of the deque
351
deque(['f', 'g', 'h', 'i', 'j'])
353
>>> d.pop() # return and remove the rightmost item
355
>>> d.popleft() # return and remove the leftmost item
357
>>> list(d) # list the contents of the deque
359
>>> d[0] # peek at leftmost item
361
>>> d[-1] # peek at rightmost item
364
>>> list(reversed(d)) # list the contents of a deque in reverse
366
>>> 'h' in d # search the deque
368
>>> d.extend('jkl') # add multiple elements at once
370
deque(['g', 'h', 'i', 'j', 'k', 'l'])
371
>>> d.rotate(1) # right rotation
373
deque(['l', 'g', 'h', 'i', 'j', 'k'])
374
>>> d.rotate(-1) # left rotation
376
deque(['g', 'h', 'i', 'j', 'k', 'l'])
378
>>> deque(reversed(d)) # make a new deque in reverse order
379
deque(['l', 'k', 'j', 'i', 'h', 'g'])
380
>>> d.clear() # empty the deque
381
>>> d.pop() # cannot pop from an empty deque
382
Traceback (most recent call last):
383
File "<pyshell#6>", line 1, in -toplevel-
385
IndexError: pop from an empty deque
387
>>> d.extendleft('abc') # extendleft() reverses the input order
389
deque(['c', 'b', 'a'])
392
:class:`deque` Recipes
393
^^^^^^^^^^^^^^^^^^^^^^
395
This section shows various approaches to working with deques.
397
Bounded length deques provide functionality similar to the ``tail`` filter
400
def tail(filename, n=10):
401
'Return the last n lines of a file'
402
return deque(open(filename), n)
404
Another approach to using deques is to maintain a sequence of recently
405
added elements by appending to the right and popping to the left::
407
def moving_average(iterable, n=3):
408
# moving_average([40, 30, 50, 46, 39, 44]) --> 40.0 42.0 45.0 43.0
409
# http://en.wikipedia.org/wiki/Moving_average
411
d = deque(itertools.islice(it, n-1))
415
s += elem - d.popleft()
419
The :meth:`rotate` method provides a way to implement :class:`deque` slicing and
420
deletion. For example, a pure Python implementation of ``del d[n]`` relies on
421
the :meth:`rotate` method to position elements to be popped::
423
def delete_nth(d, n):
428
To implement :class:`deque` slicing, use a similar approach applying
429
:meth:`rotate` to bring a target element to the left side of the deque. Remove
430
old entries with :meth:`popleft`, add new entries with :meth:`extend`, and then
431
reverse the rotation.
432
With minor variations on that approach, it is easy to implement Forth style
433
stack manipulations such as ``dup``, ``drop``, ``swap``, ``over``, ``pick``,
434
``rot``, and ``roll``.
437
:class:`defaultdict` objects
438
----------------------------
440
.. class:: defaultdict([default_factory[, ...]])
442
Returns a new dictionary-like object. :class:`defaultdict` is a subclass of the
443
built-in :class:`dict` class. It overrides one method and adds one writable
444
instance variable. The remaining functionality is the same as for the
445
:class:`dict` class and is not documented here.
447
The first argument provides the initial value for the :attr:`default_factory`
448
attribute; it defaults to ``None``. All remaining arguments are treated the same
449
as if they were passed to the :class:`dict` constructor, including keyword
452
.. versionadded:: 2.5
454
:class:`defaultdict` objects support the following method in addition to the
455
standard :class:`dict` operations:
457
.. method:: __missing__(key)
459
If the :attr:`default_factory` attribute is ``None``, this raises a
460
:exc:`KeyError` exception with the *key* as argument.
462
If :attr:`default_factory` is not ``None``, it is called without arguments
463
to provide a default value for the given *key*, this value is inserted in
464
the dictionary for the *key*, and returned.
466
If calling :attr:`default_factory` raises an exception this exception is
467
propagated unchanged.
469
This method is called by the :meth:`__getitem__` method of the
470
:class:`dict` class when the requested key is not found; whatever it
471
returns or raises is then returned or raised by :meth:`__getitem__`.
473
Note that :meth:`__missing__` is *not* called for any operations besides
474
:meth:`__getitem__`. This means that :meth:`get` will, like normal
475
dictionaries, return ``None`` as a default rather than using
476
:attr:`default_factory`.
479
:class:`defaultdict` objects support the following instance variable:
482
.. attribute:: default_factory
484
This attribute is used by the :meth:`__missing__` method; it is
485
initialized from the first argument to the constructor, if present, or to
489
:class:`defaultdict` Examples
490
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
492
Using :class:`list` as the :attr:`default_factory`, it is easy to group a
493
sequence of key-value pairs into a dictionary of lists:
495
>>> s = [('yellow', 1), ('blue', 2), ('yellow', 3), ('blue', 4), ('red', 1)]
496
>>> d = defaultdict(list)
501
[('blue', [2, 4]), ('red', [1]), ('yellow', [1, 3])]
503
When each key is encountered for the first time, it is not already in the
504
mapping; so an entry is automatically created using the :attr:`default_factory`
505
function which returns an empty :class:`list`. The :meth:`list.append`
506
operation then attaches the value to the new list. When keys are encountered
507
again, the look-up proceeds normally (returning the list for that key) and the
508
:meth:`list.append` operation adds another value to the list. This technique is
509
simpler and faster than an equivalent technique using :meth:`dict.setdefault`:
513
... d.setdefault(k, []).append(v)
516
[('blue', [2, 4]), ('red', [1]), ('yellow', [1, 3])]
518
Setting the :attr:`default_factory` to :class:`int` makes the
519
:class:`defaultdict` useful for counting (like a bag or multiset in other
522
>>> s = 'mississippi'
523
>>> d = defaultdict(int)
528
[('i', 4), ('p', 2), ('s', 4), ('m', 1)]
530
When a letter is first encountered, it is missing from the mapping, so the
531
:attr:`default_factory` function calls :func:`int` to supply a default count of
532
zero. The increment operation then builds up the count for each letter.
534
The function :func:`int` which always returns zero is just a special case of
535
constant functions. A faster and more flexible way to create constant functions
536
is to use :func:`itertools.repeat` which can supply any constant value (not just
539
>>> def constant_factory(value):
540
... return itertools.repeat(value).next
541
>>> d = defaultdict(constant_factory('<missing>'))
542
>>> d.update(name='John', action='ran')
543
>>> '%(name)s %(action)s to %(object)s' % d
544
'John ran to <missing>'
546
Setting the :attr:`default_factory` to :class:`set` makes the
547
:class:`defaultdict` useful for building a dictionary of sets:
549
>>> s = [('red', 1), ('blue', 2), ('red', 3), ('blue', 4), ('red', 1), ('blue', 4)]
550
>>> d = defaultdict(set)
555
[('blue', set([2, 4])), ('red', set([1, 3]))]
558
:func:`namedtuple` Factory Function for Tuples with Named Fields
559
----------------------------------------------------------------
561
Named tuples assign meaning to each position in a tuple and allow for more readable,
562
self-documenting code. They can be used wherever regular tuples are used, and
563
they add the ability to access fields by name instead of position index.
565
.. function:: namedtuple(typename, field_names, [verbose=False], [rename=False])
567
Returns a new tuple subclass named *typename*. The new subclass is used to
568
create tuple-like objects that have fields accessible by attribute lookup as
569
well as being indexable and iterable. Instances of the subclass also have a
570
helpful docstring (with typename and field_names) and a helpful :meth:`__repr__`
571
method which lists the tuple contents in a ``name=value`` format.
573
The *field_names* are a sequence of strings such as ``['x', 'y']``.
574
Alternatively, *field_names* can be a single string with each fieldname
575
separated by whitespace and/or commas, for example ``'x y'`` or ``'x, y'``.
577
Any valid Python identifier may be used for a fieldname except for names
578
starting with an underscore. Valid identifiers consist of letters, digits,
579
and underscores but do not start with a digit or underscore and cannot be
580
a :mod:`keyword` such as *class*, *for*, *return*, *global*, *pass*, *print*,
583
If *rename* is true, invalid fieldnames are automatically replaced
584
with positional names. For example, ``['abc', 'def', 'ghi', 'abc']`` is
585
converted to ``['abc', '_1', 'ghi', '_3']``, eliminating the keyword
586
``def`` and the duplicate fieldname ``abc``.
588
If *verbose* is true, the class definition is printed just before being built.
590
Named tuple instances do not have per-instance dictionaries, so they are
591
lightweight and require no more memory than regular tuples.
593
.. versionadded:: 2.6
595
.. versionchanged:: 2.7
596
added support for *rename*.
601
:options: +NORMALIZE_WHITESPACE
603
>>> Point = namedtuple('Point', ['x', 'y'], verbose=True)
611
def __new__(_cls, x, y):
612
'Create a new instance of Point(x, y)'
613
return _tuple.__new__(_cls, (x, y))
616
def _make(cls, iterable, new=tuple.__new__, len=len):
617
'Make a new Point object from a sequence or iterable'
618
result = new(cls, iterable)
620
raise TypeError('Expected 2 arguments, got %d' % len(result))
624
'Return a nicely formatted representation string'
625
return 'Point(x=%r, y=%r)' % self
628
'Return a new OrderedDict which maps field names to their values'
629
return OrderedDict(zip(self._fields, self))
631
def _replace(_self, **kwds):
632
'Return a new Point object replacing specified fields with new values'
633
result = _self._make(map(kwds.pop, ('x', 'y'), _self))
635
raise ValueError('Got unexpected field names: %r' % kwds.keys())
638
def __getnewargs__(self):
639
'Return self as a plain tuple. Used by copy and pickle.'
642
__dict__ = _property(_asdict)
644
def __getstate__(self):
645
'Exclude the OrderedDict from pickling'
648
x = _property(_itemgetter(0), doc='Alias for field number 0')
650
y = _property(_itemgetter(1), doc='Alias for field number 1')
653
>>> p = Point(11, y=22) # instantiate with positional or keyword arguments
654
>>> p[0] + p[1] # indexable like the plain tuple (11, 22)
656
>>> x, y = p # unpack like a regular tuple
659
>>> p.x + p.y # fields also accessible by name
661
>>> p # readable __repr__ with a name=value style
664
Named tuples are especially useful for assigning field names to result tuples returned
665
by the :mod:`csv` or :mod:`sqlite3` modules::
667
EmployeeRecord = namedtuple('EmployeeRecord', 'name, age, title, department, paygrade')
670
for emp in map(EmployeeRecord._make, csv.reader(open("employees.csv", "rb"))):
671
print emp.name, emp.title
674
conn = sqlite3.connect('/companydata')
675
cursor = conn.cursor()
676
cursor.execute('SELECT name, age, title, department, paygrade FROM employees')
677
for emp in map(EmployeeRecord._make, cursor.fetchall()):
678
print emp.name, emp.title
680
In addition to the methods inherited from tuples, named tuples support
681
three additional methods and one attribute. To prevent conflicts with
682
field names, the method and attribute names start with an underscore.
684
.. classmethod:: somenamedtuple._make(iterable)
686
Class method that makes a new instance from an existing sequence or iterable.
694
.. method:: somenamedtuple._asdict()
696
Return a new :class:`OrderedDict` which maps field names to their corresponding
699
>>> p = Point(x=11, y=22)
701
OrderedDict([('x', 11), ('y', 22)])
703
.. versionchanged:: 2.7
704
Returns an :class:`OrderedDict` instead of a regular :class:`dict`.
706
.. method:: somenamedtuple._replace(kwargs)
708
Return a new instance of the named tuple replacing specified fields with new
711
>>> p = Point(x=11, y=22)
715
>>> for partnum, record in inventory.items():
716
inventory[partnum] = record._replace(price=newprices[partnum], timestamp=time.now())
718
.. attribute:: somenamedtuple._fields
720
Tuple of strings listing the field names. Useful for introspection
721
and for creating new named tuple types from existing named tuples.
725
>>> p._fields # view the field names
728
>>> Color = namedtuple('Color', 'red green blue')
729
>>> Pixel = namedtuple('Pixel', Point._fields + Color._fields)
730
>>> Pixel(11, 22, 128, 255, 0)
731
Pixel(x=11, y=22, red=128, green=255, blue=0)
733
To retrieve a field whose name is stored in a string, use the :func:`getattr`
739
To convert a dictionary to a named tuple, use the double-star-operator
740
(as described in :ref:`tut-unpacking-arguments`):
742
>>> d = {'x': 11, 'y': 22}
746
Since a named tuple is a regular Python class, it is easy to add or change
747
functionality with a subclass. Here is how to add a calculated field and
748
a fixed-width print format:
750
>>> class Point(namedtuple('Point', 'x y')):
754
return (self.x ** 2 + self.y ** 2) ** 0.5
756
return 'Point: x=%6.3f y=%6.3f hypot=%6.3f' % (self.x, self.y, self.hypot)
758
>>> for p in Point(3, 4), Point(14, 5/7.):
760
Point: x= 3.000 y= 4.000 hypot= 5.000
761
Point: x=14.000 y= 0.714 hypot=14.018
763
The subclass shown above sets ``__slots__`` to an empty tuple. This helps
764
keep memory requirements low by preventing the creation of instance dictionaries.
766
Subclassing is not useful for adding new, stored fields. Instead, simply
767
create a new named tuple type from the :attr:`_fields` attribute:
769
>>> Point3D = namedtuple('Point3D', Point._fields + ('z',))
771
Default values can be implemented by using :meth:`_replace` to
772
customize a prototype instance:
774
>>> Account = namedtuple('Account', 'owner balance transaction_count')
775
>>> default_account = Account('<owner name>', 0.0, 0)
776
>>> johns_account = default_account._replace(owner='John')
778
Enumerated constants can be implemented with named tuples, but it is simpler
779
and more efficient to use a simple class declaration:
781
>>> Status = namedtuple('Status', 'open pending closed')._make(range(3))
782
>>> Status.open, Status.pending, Status.closed
785
open, pending, closed = range(3)
789
`Named tuple recipe <http://code.activestate.com/recipes/500261/>`_
790
adapted for Python 2.4.
793
:class:`OrderedDict` objects
794
----------------------------
796
Ordered dictionaries are just like regular dictionaries but they remember the
797
order that items were inserted. When iterating over an ordered dictionary,
798
the items are returned in the order their keys were first added.
800
.. class:: OrderedDict([items])
802
Return an instance of a dict subclass, supporting the usual :class:`dict`
803
methods. An *OrderedDict* is a dict that remembers the order that keys
804
were first inserted. If a new entry overwrites an existing entry, the
805
original insertion position is left unchanged. Deleting an entry and
806
reinserting it will move it to the end.
808
.. versionadded:: 2.7
810
.. method:: OrderedDict.popitem(last=True)
812
The :meth:`popitem` method for ordered dictionaries returns and removes
813
a (key, value) pair. The pairs are returned in LIFO order if *last* is
814
true or FIFO order if false.
816
In addition to the usual mapping methods, ordered dictionaries also support
817
reverse iteration using :func:`reversed`.
819
Equality tests between :class:`OrderedDict` objects are order-sensitive
820
and are implemented as ``list(od1.items())==list(od2.items())``.
821
Equality tests between :class:`OrderedDict` objects and other
822
:class:`Mapping` objects are order-insensitive like regular
823
dictionaries. This allows :class:`OrderedDict` objects to be substituted
824
anywhere a regular dictionary is used.
826
The :class:`OrderedDict` constructor and :meth:`update` method both accept
827
keyword arguments, but their order is lost because Python's function call
828
semantics pass-in keyword arguments using a regular unordered dictionary.
832
`Equivalent OrderedDict recipe <http://code.activestate.com/recipes/576693/>`_
833
that runs on Python 2.4 or later.
835
:class:`OrderedDict` Examples and Recipes
836
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
838
Since an ordered dictionary remembers its insertion order, it can be used
839
in conjuction with sorting to make a sorted dictionary::
841
>>> # regular unsorted dictionary
842
>>> d = {'banana': 3, 'apple':4, 'pear': 1, 'orange': 2}
844
>>> # dictionary sorted by key
845
>>> OrderedDict(sorted(d.items(), key=lambda t: t[0]))
846
OrderedDict([('apple', 4), ('banana', 3), ('orange', 2), ('pear', 1)])
848
>>> # dictionary sorted by value
849
>>> OrderedDict(sorted(d.items(), key=lambda t: t[1]))
850
OrderedDict([('pear', 1), ('orange', 2), ('banana', 3), ('apple', 4)])
852
>>> # dictionary sorted by length of the key string
853
>>> OrderedDict(sorted(d.items(), key=lambda t: len(t[0])))
854
OrderedDict([('pear', 1), ('apple', 4), ('orange', 2), ('banana', 3)])
856
The new sorted dictionaries maintain their sort order when entries
857
are deleted. But when new keys are added, the keys are appended
858
to the end and the sort is not maintained.
860
It is also straight-forward to create an ordered dictionary variant
861
that remembers the order the keys were *last* inserted.
862
If a new entry overwrites an existing entry, the
863
original insertion position is changed and moved to the end::
865
class LastUpdatedOrderedDict(OrderedDict):
866
'Store items in the order the keys were last added'
868
def __setitem__(self, key, value):
871
OrderedDict.__setitem__(self, key, value)
873
An ordered dictionary can be combined with the :class:`Counter` class
874
so that the counter remembers the order elements are first encountered::
876
class OrderedCounter(Counter, OrderedDict):
877
'Counter that remembers the order elements are first encountered'
880
return '%s(%r)' % (self.__class__.__name__, OrderedDict(self))
882
def __reduce__(self):
883
return self.__class__, (OrderedDict(self),)
886
.. _collections-abstract-base-classes:
888
Collections Abstract Base Classes
889
---------------------------------
891
The collections module offers the following :term:`ABCs <abstract base class>`:
893
========================= ===================== ====================== ====================================================
894
ABC Inherits from Abstract Methods Mixin Methods
895
========================= ===================== ====================== ====================================================
896
:class:`Container` ``__contains__``
897
:class:`Hashable` ``__hash__``
898
:class:`Iterable` ``__iter__``
899
:class:`Iterator` :class:`Iterable` ``next`` ``__iter__``
900
:class:`Sized` ``__len__``
901
:class:`Callable` ``__call__``
903
:class:`Sequence` :class:`Sized`, ``__getitem__``, ``__contains__``, ``__iter__``, ``__reversed__``,
904
:class:`Iterable`, ``__len__`` ``index``, and ``count``
907
:class:`MutableSequence` :class:`Sequence` ``__getitem__``, Inherited :class:`Sequence` methods and
908
``__setitem__``, ``append``, ``reverse``, ``extend``, ``pop``,
909
``__delitem__``, ``remove``, and ``__iadd__``
913
:class:`Set` :class:`Sized`, ``__contains__``, ``__le__``, ``__lt__``, ``__eq__``, ``__ne__``,
914
:class:`Iterable`, ``__iter__``, ``__gt__``, ``__ge__``, ``__and__``, ``__or__``,
915
:class:`Container` ``__len__`` ``__sub__``, ``__xor__``, and ``isdisjoint``
917
:class:`MutableSet` :class:`Set` ``__contains__``, Inherited :class:`Set` methods and
918
``__iter__``, ``clear``, ``pop``, ``remove``, ``__ior__``,
919
``__len__``, ``__iand__``, ``__ixor__``, and ``__isub__``
923
:class:`Mapping` :class:`Sized`, ``__getitem__``, ``__contains__``, ``keys``, ``items``, ``values``,
924
:class:`Iterable`, ``__iter__``, ``get``, ``__eq__``, and ``__ne__``
925
:class:`Container` ``__len__``
927
:class:`MutableMapping` :class:`Mapping` ``__getitem__``, Inherited :class:`Mapping` methods and
928
``__setitem__``, ``pop``, ``popitem``, ``clear``, ``update``,
929
``__delitem__``, and ``setdefault``
934
:class:`MappingView` :class:`Sized` ``__len__``
935
:class:`ItemsView` :class:`MappingView`, ``__contains__``,
936
:class:`Set` ``__iter__``
937
:class:`KeysView` :class:`MappingView`, ``__contains__``,
938
:class:`Set` ``__iter__``
939
:class:`ValuesView` :class:`MappingView` ``__contains__``, ``__iter__``
940
========================= ===================== ====================== ====================================================
948
ABCs for classes that provide respectively the methods :meth:`__contains__`,
949
:meth:`__hash__`, :meth:`__len__`, and :meth:`__call__`.
953
ABC for classes that provide the :meth:`__iter__` method.
954
See also the definition of :term:`iterable`.
958
ABC for classes that provide the :meth:`~iterator.__iter__` and
959
:meth:`~iterator.next` methods. See also the definition of :term:`iterator`.
964
ABCs for read-only and mutable :term:`sequences <sequence>`.
969
ABCs for read-only and mutable sets.
974
ABCs for read-only and mutable :term:`mappings <mapping>`.
976
.. class:: MappingView
981
ABCs for mapping, items, keys, and values :term:`views <dictionary view>`.
984
These ABCs allow us to ask classes or instances if they provide
985
particular functionality, for example::
988
if isinstance(myvar, collections.Sized):
991
Several of the ABCs are also useful as mixins that make it easier to develop
992
classes supporting container APIs. For example, to write a class supporting
993
the full :class:`Set` API, it only necessary to supply the three underlying
994
abstract methods: :meth:`__contains__`, :meth:`__iter__`, and :meth:`__len__`.
995
The ABC supplies the remaining methods such as :meth:`__and__` and
996
:meth:`isdisjoint` ::
998
class ListBasedSet(collections.Set):
999
''' Alternate set implementation favoring space over speed
1000
and not requiring the set elements to be hashable. '''
1001
def __init__(self, iterable):
1002
self.elements = lst = []
1003
for value in iterable:
1004
if value not in lst:
1007
return iter(self.elements)
1008
def __contains__(self, value):
1009
return value in self.elements
1011
return len(self.elements)
1013
s1 = ListBasedSet('abcdef')
1014
s2 = ListBasedSet('defghi')
1015
overlap = s1 & s2 # The __and__() method is supported automatically
1017
Notes on using :class:`Set` and :class:`MutableSet` as a mixin:
1020
Since some set operations create new sets, the default mixin methods need
1021
a way to create new instances from an iterable. The class constructor is
1022
assumed to have a signature in the form ``ClassName(iterable)``.
1023
That assumption is factored-out to an internal classmethod called
1024
:meth:`_from_iterable` which calls ``cls(iterable)`` to produce a new set.
1025
If the :class:`Set` mixin is being used in a class with a different
1026
constructor signature, you will need to override :meth:`_from_iterable`
1027
with a classmethod that can construct new instances from
1028
an iterable argument.
1031
To override the comparisons (presumably for speed, as the
1032
semantics are fixed), redefine :meth:`__le__` and :meth:`__ge__`,
1033
then the other operations will automatically follow suit.
1036
The :class:`Set` mixin provides a :meth:`_hash` method to compute a hash value
1037
for the set; however, :meth:`__hash__` is not defined because not all sets
1038
are hashable or immutable. To add set hashabilty using mixins,
1039
inherit from both :meth:`Set` and :meth:`Hashable`, then define
1040
``__hash__ = Set._hash``.
1044
* `OrderedSet recipe <http://code.activestate.com/recipes/576694/>`_ for an
1045
example built on :class:`MutableSet`.
1047
* For more about ABCs, see the :mod:`abc` module and :pep:`3119`.