blob: f8f2aac70f9b063aad79ebc3f35d31dd09c1184a [file] [log] [blame]
Georg Brandl116aa622007-08-15 14:28:22 +00001********************************
2 Functional Programming HOWTO
3********************************
4
Christian Heimesfe337bf2008-03-23 21:54:12 +00005:Author: A. M. Kuchling
Andrew Kuchling2a9c8e82013-05-20 10:14:53 -04006:Release: 0.32
Georg Brandl116aa622007-08-15 14:28:22 +00007
Georg Brandl116aa622007-08-15 14:28:22 +00008In this document, we'll take a tour of Python's features suitable for
9implementing programs in a functional style. After an introduction to the
10concepts of functional programming, we'll look at language features such as
Georg Brandl9afde1c2007-11-01 20:32:30 +000011:term:`iterator`\s and :term:`generator`\s and relevant library modules such as
12:mod:`itertools` and :mod:`functools`.
Georg Brandl116aa622007-08-15 14:28:22 +000013
14
15Introduction
16============
17
Andrew Kuchling2a9c8e82013-05-20 10:14:53 -040018This section explains the basic concept of functional programming; if
19you're just interested in learning about Python language features,
20skip to the next section on :ref:`functional-howto-iterators`.
Georg Brandl116aa622007-08-15 14:28:22 +000021
22Programming languages support decomposing problems in several different ways:
23
24* Most programming languages are **procedural**: programs are lists of
25 instructions that tell the computer what to do with the program's input. C,
26 Pascal, and even Unix shells are procedural languages.
27
28* In **declarative** languages, you write a specification that describes the
29 problem to be solved, and the language implementation figures out how to
30 perform the computation efficiently. SQL is the declarative language you're
31 most likely to be familiar with; a SQL query describes the data set you want
32 to retrieve, and the SQL engine decides whether to scan tables or use indexes,
33 which subclauses should be performed first, etc.
34
35* **Object-oriented** programs manipulate collections of objects. Objects have
36 internal state and support methods that query or modify this internal state in
37 some way. Smalltalk and Java are object-oriented languages. C++ and Python
38 are languages that support object-oriented programming, but don't force the
39 use of object-oriented features.
40
41* **Functional** programming decomposes a problem into a set of functions.
42 Ideally, functions only take inputs and produce outputs, and don't have any
43 internal state that affects the output produced for a given input. Well-known
44 functional languages include the ML family (Standard ML, OCaml, and other
45 variants) and Haskell.
46
Christian Heimes0449f632007-12-15 01:27:15 +000047The designers of some computer languages choose to emphasize one
48particular approach to programming. This often makes it difficult to
49write programs that use a different approach. Other languages are
50multi-paradigm languages that support several different approaches.
51Lisp, C++, and Python are multi-paradigm; you can write programs or
52libraries that are largely procedural, object-oriented, or functional
53in all of these languages. In a large program, different sections
54might be written using different approaches; the GUI might be
55object-oriented while the processing logic is procedural or
56functional, for example.
Georg Brandl116aa622007-08-15 14:28:22 +000057
58In a functional program, input flows through a set of functions. Each function
Christian Heimes0449f632007-12-15 01:27:15 +000059operates on its input and produces some output. Functional style discourages
Georg Brandl116aa622007-08-15 14:28:22 +000060functions with side effects that modify internal state or make other changes
61that aren't visible in the function's return value. Functions that have no side
62effects at all are called **purely functional**. Avoiding side effects means
63not using data structures that get updated as a program runs; every function's
64output must only depend on its input.
65
66Some languages are very strict about purity and don't even have assignment
67statements such as ``a=3`` or ``c = a + b``, but it's difficult to avoid all
68side effects. Printing to the screen or writing to a disk file are side
Georg Brandl0df79792008-10-04 18:33:26 +000069effects, for example. For example, in Python a call to the :func:`print` or
70:func:`time.sleep` function both return no useful value; they're only called for
71their side effects of sending some text to the screen or pausing execution for a
Georg Brandl116aa622007-08-15 14:28:22 +000072second.
73
74Python programs written in functional style usually won't go to the extreme of
75avoiding all I/O or all assignments; instead, they'll provide a
76functional-appearing interface but will use non-functional features internally.
77For example, the implementation of a function will still use assignments to
78local variables, but won't modify global variables or have other side effects.
79
80Functional programming can be considered the opposite of object-oriented
81programming. Objects are little capsules containing some internal state along
82with a collection of method calls that let you modify this state, and programs
83consist of making the right set of state changes. Functional programming wants
84to avoid state changes as much as possible and works with data flowing between
85functions. In Python you might combine the two approaches by writing functions
86that take and return instances representing objects in your application (e-mail
87messages, transactions, etc.).
88
89Functional design may seem like an odd constraint to work under. Why should you
90avoid objects and side effects? There are theoretical and practical advantages
91to the functional style:
92
93* Formal provability.
94* Modularity.
95* Composability.
96* Ease of debugging and testing.
97
Christian Heimesfe337bf2008-03-23 21:54:12 +000098
Georg Brandl116aa622007-08-15 14:28:22 +000099Formal provability
100------------------
101
102A theoretical benefit is that it's easier to construct a mathematical proof that
103a functional program is correct.
104
105For a long time researchers have been interested in finding ways to
106mathematically prove programs correct. This is different from testing a program
107on numerous inputs and concluding that its output is usually correct, or reading
108a program's source code and concluding that the code looks right; the goal is
109instead a rigorous proof that a program produces the right result for all
110possible inputs.
111
112The technique used to prove programs correct is to write down **invariants**,
113properties of the input data and of the program's variables that are always
114true. For each line of code, you then show that if invariants X and Y are true
115**before** the line is executed, the slightly different invariants X' and Y' are
116true **after** the line is executed. This continues until you reach the end of
117the program, at which point the invariants should match the desired conditions
118on the program's output.
119
120Functional programming's avoidance of assignments arose because assignments are
121difficult to handle with this technique; assignments can break invariants that
122were true before the assignment without producing any new invariants that can be
123propagated onward.
124
125Unfortunately, proving programs correct is largely impractical and not relevant
126to Python software. Even trivial programs require proofs that are several pages
127long; the proof of correctness for a moderately complicated program would be
128enormous, and few or none of the programs you use daily (the Python interpreter,
129your XML parser, your web browser) could be proven correct. Even if you wrote
130down or generated a proof, there would then be the question of verifying the
131proof; maybe there's an error in it, and you wrongly believe you've proved the
132program correct.
133
Christian Heimesfe337bf2008-03-23 21:54:12 +0000134
Georg Brandl116aa622007-08-15 14:28:22 +0000135Modularity
136----------
137
138A more practical benefit of functional programming is that it forces you to
139break apart your problem into small pieces. Programs are more modular as a
140result. It's easier to specify and write a small function that does one thing
141than a large function that performs a complicated transformation. Small
142functions are also easier to read and to check for errors.
143
144
Georg Brandl48310cd2009-01-03 21:18:54 +0000145Ease of debugging and testing
Georg Brandl116aa622007-08-15 14:28:22 +0000146-----------------------------
147
148Testing and debugging a functional-style program is easier.
149
150Debugging is simplified because functions are generally small and clearly
151specified. When a program doesn't work, each function is an interface point
152where you can check that the data are correct. You can look at the intermediate
153inputs and outputs to quickly isolate the function that's responsible for a bug.
154
155Testing is easier because each function is a potential subject for a unit test.
156Functions don't depend on system state that needs to be replicated before
157running a test; instead you only have to synthesize the right input and then
158check that the output matches expectations.
159
160
Georg Brandl116aa622007-08-15 14:28:22 +0000161Composability
162-------------
163
164As you work on a functional-style program, you'll write a number of functions
165with varying inputs and outputs. Some of these functions will be unavoidably
166specialized to a particular application, but others will be useful in a wide
167variety of programs. For example, a function that takes a directory path and
168returns all the XML files in the directory, or a function that takes a filename
169and returns its contents, can be applied to many different situations.
170
171Over time you'll form a personal library of utilities. Often you'll assemble
172new programs by arranging existing functions in a new configuration and writing
173a few functions specialized for the current task.
174
175
Andrew Kuchling2a9c8e82013-05-20 10:14:53 -0400176.. _functional-howto-iterators:
177
Georg Brandl116aa622007-08-15 14:28:22 +0000178Iterators
179=========
180
181I'll start by looking at a Python language feature that's an important
182foundation for writing functional-style programs: iterators.
183
184An iterator is an object representing a stream of data; this object returns the
185data one element at a time. A Python iterator must support a method called
Ezio Melotti45a101d2012-10-12 12:42:51 +0300186:meth:`~iterator.__next__` that takes no arguments and always returns the next
187element of the stream. If there are no more elements in the stream,
188:meth:`~iterator.__next__` must raise the :exc:`StopIteration` exception.
189Iterators don't have to be finite, though; it's perfectly reasonable to write
190an iterator that produces an infinite stream of data.
Georg Brandl116aa622007-08-15 14:28:22 +0000191
192The built-in :func:`iter` function takes an arbitrary object and tries to return
193an iterator that will return the object's contents or elements, raising
194:exc:`TypeError` if the object doesn't support iteration. Several of Python's
195built-in data types support iteration, the most common being lists and
Ezio Melotti45a101d2012-10-12 12:42:51 +0300196dictionaries. An object is called :term:`iterable` if you can get an iterator
197for it.
Georg Brandl116aa622007-08-15 14:28:22 +0000198
Christian Heimesfe337bf2008-03-23 21:54:12 +0000199You can experiment with the iteration interface manually:
Georg Brandl116aa622007-08-15 14:28:22 +0000200
Sergey Fedoseevdb8707c2018-08-08 02:38:00 +0500201 >>> L = [1, 2, 3]
Georg Brandl116aa622007-08-15 14:28:22 +0000202 >>> it = iter(L)
Ezio Melotti35cbf162012-10-12 13:24:19 +0300203 >>> it #doctest: +ELLIPSIS
Christian Heimesfe337bf2008-03-23 21:54:12 +0000204 <...iterator object at ...>
Ezio Melotti45a101d2012-10-12 12:42:51 +0300205 >>> it.__next__() # same as next(it)
Georg Brandl116aa622007-08-15 14:28:22 +0000206 1
Benjamin Petersone7c78b22008-07-03 20:28:26 +0000207 >>> next(it)
Georg Brandl116aa622007-08-15 14:28:22 +0000208 2
Benjamin Petersone7c78b22008-07-03 20:28:26 +0000209 >>> next(it)
Georg Brandl116aa622007-08-15 14:28:22 +0000210 3
Benjamin Petersone7c78b22008-07-03 20:28:26 +0000211 >>> next(it)
Georg Brandl116aa622007-08-15 14:28:22 +0000212 Traceback (most recent call last):
UltimateCoder88569402017-05-03 22:16:45 +0530213 File "<stdin>", line 1, in <module>
Georg Brandl116aa622007-08-15 14:28:22 +0000214 StopIteration
Georg Brandl48310cd2009-01-03 21:18:54 +0000215 >>>
Georg Brandl116aa622007-08-15 14:28:22 +0000216
217Python expects iterable objects in several different contexts, the most
Ezio Melotti45a101d2012-10-12 12:42:51 +0300218important being the :keyword:`for` statement. In the statement ``for X in Y``,
219Y must be an iterator or some object for which :func:`iter` can create an
220iterator. These two statements are equivalent::
Georg Brandl116aa622007-08-15 14:28:22 +0000221
Georg Brandl116aa622007-08-15 14:28:22 +0000222
Christian Heimesfe337bf2008-03-23 21:54:12 +0000223 for i in iter(obj):
Neal Norwitz752abd02008-05-13 04:55:24 +0000224 print(i)
Christian Heimesfe337bf2008-03-23 21:54:12 +0000225
226 for i in obj:
Neal Norwitz752abd02008-05-13 04:55:24 +0000227 print(i)
Georg Brandl116aa622007-08-15 14:28:22 +0000228
229Iterators can be materialized as lists or tuples by using the :func:`list` or
Christian Heimesfe337bf2008-03-23 21:54:12 +0000230:func:`tuple` constructor functions:
Georg Brandl116aa622007-08-15 14:28:22 +0000231
Sergey Fedoseevdb8707c2018-08-08 02:38:00 +0500232 >>> L = [1, 2, 3]
Georg Brandl116aa622007-08-15 14:28:22 +0000233 >>> iterator = iter(L)
234 >>> t = tuple(iterator)
235 >>> t
236 (1, 2, 3)
237
238Sequence unpacking also supports iterators: if you know an iterator will return
Christian Heimesfe337bf2008-03-23 21:54:12 +0000239N elements, you can unpack them into an N-tuple:
Georg Brandl116aa622007-08-15 14:28:22 +0000240
Sergey Fedoseevdb8707c2018-08-08 02:38:00 +0500241 >>> L = [1, 2, 3]
Georg Brandl116aa622007-08-15 14:28:22 +0000242 >>> iterator = iter(L)
Sergey Fedoseevdb8707c2018-08-08 02:38:00 +0500243 >>> a, b, c = iterator
244 >>> a, b, c
Georg Brandl116aa622007-08-15 14:28:22 +0000245 (1, 2, 3)
246
247Built-in functions such as :func:`max` and :func:`min` can take a single
248iterator argument and will return the largest or smallest element. The ``"in"``
249and ``"not in"`` operators also support iterators: ``X in iterator`` is true if
250X is found in the stream returned by the iterator. You'll run into obvious
Ezio Melotti45a101d2012-10-12 12:42:51 +0300251problems if the iterator is infinite; :func:`max`, :func:`min`
Georg Brandl116aa622007-08-15 14:28:22 +0000252will never return, and if the element X never appears in the stream, the
Sandro Tosidd7c5522012-08-15 21:37:35 +0200253``"in"`` and ``"not in"`` operators won't return either.
Georg Brandl116aa622007-08-15 14:28:22 +0000254
255Note that you can only go forward in an iterator; there's no way to get the
256previous element, reset the iterator, or make a copy of it. Iterator objects
257can optionally provide these additional capabilities, but the iterator protocol
Ezio Melotti45a101d2012-10-12 12:42:51 +0300258only specifies the :meth:`~iterator.__next__` method. Functions may therefore
259consume all of the iterator's output, and if you need to do something different
260with the same stream, you'll have to create a new iterator.
Georg Brandl116aa622007-08-15 14:28:22 +0000261
262
263
264Data Types That Support Iterators
265---------------------------------
266
267We've already seen how lists and tuples support iterators. In fact, any Python
268sequence type, such as strings, will automatically support creation of an
269iterator.
270
271Calling :func:`iter` on a dictionary returns an iterator that will loop over the
Ezio Melotti35cbf162012-10-12 13:24:19 +0300272dictionary's keys::
Georg Brandl116aa622007-08-15 14:28:22 +0000273
274 >>> m = {'Jan': 1, 'Feb': 2, 'Mar': 3, 'Apr': 4, 'May': 5, 'Jun': 6,
275 ... 'Jul': 7, 'Aug': 8, 'Sep': 9, 'Oct': 10, 'Nov': 11, 'Dec': 12}
Stig Johan Berggren5e5bbbe2018-07-11 11:53:07 +0200276 >>> for key in m:
Georg Brandl6911e3c2007-09-04 07:15:32 +0000277 ... print(key, m[key])
Stig Johan Berggren5e5bbbe2018-07-11 11:53:07 +0200278 Jan 1
Georg Brandl116aa622007-08-15 14:28:22 +0000279 Feb 2
Stig Johan Berggren5e5bbbe2018-07-11 11:53:07 +0200280 Mar 3
Christian Heimesfe337bf2008-03-23 21:54:12 +0000281 Apr 4
Stig Johan Berggren5e5bbbe2018-07-11 11:53:07 +0200282 May 5
Georg Brandl116aa622007-08-15 14:28:22 +0000283 Jun 6
284 Jul 7
Stig Johan Berggren5e5bbbe2018-07-11 11:53:07 +0200285 Aug 8
286 Sep 9
287 Oct 10
Georg Brandl116aa622007-08-15 14:28:22 +0000288 Nov 11
289 Dec 12
Georg Brandl116aa622007-08-15 14:28:22 +0000290
Stig Johan Berggren5e5bbbe2018-07-11 11:53:07 +0200291Note that starting with Python 3.7, dictionary iteration order is guaranteed
292to be the same as the insertion order. In earlier versions, the behaviour was
293unspecified and could vary between implementations.
Georg Brandl116aa622007-08-15 14:28:22 +0000294
Fred Drake2e748782007-09-04 17:33:11 +0000295Applying :func:`iter` to a dictionary always loops over the keys, but
296dictionaries have methods that return other iterators. If you want to iterate
297over values or key/value pairs, you can explicitly call the
Chris Jerdonek006d9072012-10-12 20:28:26 -0700298:meth:`~dict.values` or :meth:`~dict.items` methods to get an appropriate
299iterator.
Georg Brandl116aa622007-08-15 14:28:22 +0000300
301The :func:`dict` constructor can accept an iterator that returns a finite stream
Christian Heimesfe337bf2008-03-23 21:54:12 +0000302of ``(key, value)`` tuples:
Georg Brandl116aa622007-08-15 14:28:22 +0000303
304 >>> L = [('Italy', 'Rome'), ('France', 'Paris'), ('US', 'Washington DC')]
Stig Johan Berggren5e5bbbe2018-07-11 11:53:07 +0200305 >>> dict(iter(L))
306 {'Italy': 'Rome', 'France': 'Paris', 'US': 'Washington DC'}
Georg Brandl116aa622007-08-15 14:28:22 +0000307
Ezio Melotti45a101d2012-10-12 12:42:51 +0300308Files also support iteration by calling the :meth:`~io.TextIOBase.readline`
309method until there are no more lines in the file. This means you can read each
310line of a file like this::
Georg Brandl116aa622007-08-15 14:28:22 +0000311
312 for line in file:
313 # do something for each line
314 ...
315
316Sets can take their contents from an iterable and let you iterate over the set's
317elements::
318
Georg Brandlf6945182008-02-01 11:56:49 +0000319 S = {2, 3, 5, 7, 11, 13}
Georg Brandl116aa622007-08-15 14:28:22 +0000320 for i in S:
Georg Brandl6911e3c2007-09-04 07:15:32 +0000321 print(i)
Georg Brandl116aa622007-08-15 14:28:22 +0000322
323
324
325Generator expressions and list comprehensions
326=============================================
327
328Two common operations on an iterator's output are 1) performing some operation
329for every element, 2) selecting a subset of elements that meet some condition.
330For example, given a list of strings, you might want to strip off trailing
331whitespace from each line or extract all the strings containing a given
332substring.
333
334List comprehensions and generator expressions (short form: "listcomps" and
335"genexps") are a concise notation for such operations, borrowed from the
Georg Brandl5d941342016-02-26 19:37:12 +0100336functional programming language Haskell (https://www.haskell.org/). You can strip
Georg Brandl116aa622007-08-15 14:28:22 +0000337all the whitespace from a stream of strings with the following code::
338
Christian Heimesfe337bf2008-03-23 21:54:12 +0000339 line_list = [' line 1\n', 'line 2 \n', ...]
Georg Brandl116aa622007-08-15 14:28:22 +0000340
Christian Heimesfe337bf2008-03-23 21:54:12 +0000341 # Generator expression -- returns iterator
342 stripped_iter = (line.strip() for line in line_list)
Georg Brandl116aa622007-08-15 14:28:22 +0000343
Christian Heimesfe337bf2008-03-23 21:54:12 +0000344 # List comprehension -- returns list
345 stripped_list = [line.strip() for line in line_list]
Georg Brandl116aa622007-08-15 14:28:22 +0000346
347You can select only certain elements by adding an ``"if"`` condition::
348
Christian Heimesfe337bf2008-03-23 21:54:12 +0000349 stripped_list = [line.strip() for line in line_list
350 if line != ""]
Georg Brandl116aa622007-08-15 14:28:22 +0000351
352With a list comprehension, you get back a Python list; ``stripped_list`` is a
353list containing the resulting lines, not an iterator. Generator expressions
354return an iterator that computes the values as necessary, not needing to
355materialize all the values at once. This means that list comprehensions aren't
356useful if you're working with iterators that return an infinite stream or a very
357large amount of data. Generator expressions are preferable in these situations.
358
359Generator expressions are surrounded by parentheses ("()") and list
360comprehensions are surrounded by square brackets ("[]"). Generator expressions
361have the form::
362
Georg Brandl48310cd2009-01-03 21:18:54 +0000363 ( expression for expr in sequence1
Georg Brandl116aa622007-08-15 14:28:22 +0000364 if condition1
365 for expr2 in sequence2
366 if condition2
367 for expr3 in sequence3 ...
368 if condition3
369 for exprN in sequenceN
370 if conditionN )
371
372Again, for a list comprehension only the outside brackets are different (square
373brackets instead of parentheses).
374
375The elements of the generated output will be the successive values of
376``expression``. The ``if`` clauses are all optional; if present, ``expression``
377is only evaluated and added to the result when ``condition`` is true.
378
379Generator expressions always have to be written inside parentheses, but the
380parentheses signalling a function call also count. If you want to create an
381iterator that will be immediately passed to a function you can write::
382
Christian Heimesfe337bf2008-03-23 21:54:12 +0000383 obj_total = sum(obj.count for obj in list_all_objects())
Georg Brandl116aa622007-08-15 14:28:22 +0000384
385The ``for...in`` clauses contain the sequences to be iterated over. The
386sequences do not have to be the same length, because they are iterated over from
387left to right, **not** in parallel. For each element in ``sequence1``,
388``sequence2`` is looped over from the beginning. ``sequence3`` is then looped
389over for each resulting pair of elements from ``sequence1`` and ``sequence2``.
390
391To put it another way, a list comprehension or generator expression is
392equivalent to the following Python code::
393
394 for expr1 in sequence1:
395 if not (condition1):
396 continue # Skip this element
397 for expr2 in sequence2:
398 if not (condition2):
Serhiy Storchakadba90392016-05-10 12:01:23 +0300399 continue # Skip this element
Georg Brandl116aa622007-08-15 14:28:22 +0000400 ...
401 for exprN in sequenceN:
Serhiy Storchakadba90392016-05-10 12:01:23 +0300402 if not (conditionN):
403 continue # Skip this element
Georg Brandl116aa622007-08-15 14:28:22 +0000404
Serhiy Storchakadba90392016-05-10 12:01:23 +0300405 # Output the value of
406 # the expression.
Georg Brandl116aa622007-08-15 14:28:22 +0000407
408This means that when there are multiple ``for...in`` clauses but no ``if``
409clauses, the length of the resulting output will be equal to the product of the
410lengths of all the sequences. If you have two lists of length 3, the output
Christian Heimesfe337bf2008-03-23 21:54:12 +0000411list is 9 elements long:
Georg Brandl116aa622007-08-15 14:28:22 +0000412
Christian Heimesfe337bf2008-03-23 21:54:12 +0000413 >>> seq1 = 'abc'
Sergey Fedoseevdb8707c2018-08-08 02:38:00 +0500414 >>> seq2 = (1, 2, 3)
Ezio Melotti35cbf162012-10-12 13:24:19 +0300415 >>> [(x, y) for x in seq1 for y in seq2] #doctest: +NORMALIZE_WHITESPACE
Georg Brandl48310cd2009-01-03 21:18:54 +0000416 [('a', 1), ('a', 2), ('a', 3),
417 ('b', 1), ('b', 2), ('b', 3),
Georg Brandl116aa622007-08-15 14:28:22 +0000418 ('c', 1), ('c', 2), ('c', 3)]
419
420To avoid introducing an ambiguity into Python's grammar, if ``expression`` is
421creating a tuple, it must be surrounded with parentheses. The first list
422comprehension below is a syntax error, while the second one is correct::
423
424 # Syntax error
Ezio Melotti45a101d2012-10-12 12:42:51 +0300425 [x, y for x in seq1 for y in seq2]
Georg Brandl116aa622007-08-15 14:28:22 +0000426 # Correct
Ezio Melotti45a101d2012-10-12 12:42:51 +0300427 [(x, y) for x in seq1 for y in seq2]
Georg Brandl116aa622007-08-15 14:28:22 +0000428
429
430Generators
431==========
432
433Generators are a special class of functions that simplify the task of writing
434iterators. Regular functions compute a value and return it, but generators
435return an iterator that returns a stream of values.
436
437You're doubtless familiar with how regular function calls work in Python or C.
438When you call a function, it gets a private namespace where its local variables
439are created. When the function reaches a ``return`` statement, the local
440variables are destroyed and the value is returned to the caller. A later call
441to the same function creates a new private namespace and a fresh set of local
442variables. But, what if the local variables weren't thrown away on exiting a
443function? What if you could later resume the function where it left off? This
444is what generators provide; they can be thought of as resumable functions.
445
Christian Heimesfe337bf2008-03-23 21:54:12 +0000446Here's the simplest example of a generator function:
447
Ezio Melotti35cbf162012-10-12 13:24:19 +0300448 >>> def generate_ints(N):
449 ... for i in range(N):
450 ... yield i
Georg Brandl116aa622007-08-15 14:28:22 +0000451
Ezio Melotti45a101d2012-10-12 12:42:51 +0300452Any function containing a :keyword:`yield` keyword is a generator function;
453this is detected by Python's :term:`bytecode` compiler which compiles the
454function specially as a result.
Georg Brandl116aa622007-08-15 14:28:22 +0000455
456When you call a generator function, it doesn't return a single value; instead it
457returns a generator object that supports the iterator protocol. On executing
458the ``yield`` expression, the generator outputs the value of ``i``, similar to a
459``return`` statement. The big difference between ``yield`` and a ``return``
460statement is that on reaching a ``yield`` the generator's state of execution is
461suspended and local variables are preserved. On the next call to the
Ezio Melotti45a101d2012-10-12 12:42:51 +0300462generator's :meth:`~generator.__next__` method, the function will resume
463executing.
Georg Brandl116aa622007-08-15 14:28:22 +0000464
Christian Heimesfe337bf2008-03-23 21:54:12 +0000465Here's a sample usage of the ``generate_ints()`` generator:
Georg Brandl116aa622007-08-15 14:28:22 +0000466
467 >>> gen = generate_ints(3)
Ezio Melotti35cbf162012-10-12 13:24:19 +0300468 >>> gen #doctest: +ELLIPSIS
Benjamin Peterson25c95f12009-05-08 20:42:26 +0000469 <generator object generate_ints at ...>
Benjamin Petersone7c78b22008-07-03 20:28:26 +0000470 >>> next(gen)
Georg Brandl116aa622007-08-15 14:28:22 +0000471 0
Benjamin Petersone7c78b22008-07-03 20:28:26 +0000472 >>> next(gen)
Georg Brandl116aa622007-08-15 14:28:22 +0000473 1
Benjamin Petersone7c78b22008-07-03 20:28:26 +0000474 >>> next(gen)
Georg Brandl116aa622007-08-15 14:28:22 +0000475 2
Benjamin Petersone7c78b22008-07-03 20:28:26 +0000476 >>> next(gen)
Georg Brandl116aa622007-08-15 14:28:22 +0000477 Traceback (most recent call last):
UltimateCoder88569402017-05-03 22:16:45 +0530478 File "stdin", line 1, in <module>
Georg Brandl116aa622007-08-15 14:28:22 +0000479 File "stdin", line 2, in generate_ints
480 StopIteration
481
Sergey Fedoseevdb8707c2018-08-08 02:38:00 +0500482You could equally write ``for i in generate_ints(5)``, or ``a, b, c =
Georg Brandl116aa622007-08-15 14:28:22 +0000483generate_ints(3)``.
484
Yury Selivanov8170e8c2015-05-09 11:44:30 -0400485Inside a generator function, ``return value`` causes ``StopIteration(value)``
486to be raised from the :meth:`~generator.__next__` method. Once this happens, or
487the bottom of the function is reached, the procession of values ends and the
488generator cannot yield any further values.
Georg Brandl116aa622007-08-15 14:28:22 +0000489
490You could achieve the effect of generators manually by writing your own class
491and storing all the local variables of the generator as instance variables. For
492example, returning a list of integers could be done by setting ``self.count`` to
Ezio Melotti45a101d2012-10-12 12:42:51 +03004930, and having the :meth:`~iterator.__next__` method increment ``self.count`` and
494return it.
Georg Brandl116aa622007-08-15 14:28:22 +0000495However, for a moderately complicated generator, writing a corresponding class
496can be much messier.
497
Ezio Melotti45a101d2012-10-12 12:42:51 +0300498The test suite included with Python's library,
499:source:`Lib/test/test_generators.py`, contains
Georg Brandl116aa622007-08-15 14:28:22 +0000500a number of more interesting examples. Here's one generator that implements an
Christian Heimesfe337bf2008-03-23 21:54:12 +0000501in-order traversal of a tree using generators recursively. ::
Georg Brandl116aa622007-08-15 14:28:22 +0000502
503 # A recursive generator that generates Tree leaves in in-order.
504 def inorder(t):
505 if t:
506 for x in inorder(t.left):
507 yield x
508
509 yield t.label
510
511 for x in inorder(t.right):
512 yield x
513
514Two other examples in ``test_generators.py`` produce solutions for the N-Queens
515problem (placing N queens on an NxN chess board so that no queen threatens
516another) and the Knight's Tour (finding a route that takes a knight to every
517square of an NxN chessboard without visiting any square twice).
518
519
520
521Passing values into a generator
522-------------------------------
523
524In Python 2.4 and earlier, generators only produced output. Once a generator's
525code was invoked to create an iterator, there was no way to pass any new
526information into the function when its execution is resumed. You could hack
527together this ability by making the generator look at a global variable or by
528passing in some mutable object that callers then modify, but these approaches
529are messy.
530
531In Python 2.5 there's a simple way to pass values into a generator.
532:keyword:`yield` became an expression, returning a value that can be assigned to
533a variable or otherwise operated on::
534
535 val = (yield i)
536
537I recommend that you **always** put parentheses around a ``yield`` expression
538when you're doing something with the returned value, as in the above example.
539The parentheses aren't always necessary, but it's easier to always add them
540instead of having to remember when they're needed.
541
Ezio Melotti45a101d2012-10-12 12:42:51 +0300542(:pep:`342` explains the exact rules, which are that a ``yield``-expression must
Georg Brandl116aa622007-08-15 14:28:22 +0000543always be parenthesized except when it occurs at the top-level expression on the
544right-hand side of an assignment. This means you can write ``val = yield i``
545but have to use parentheses when there's an operation, as in ``val = (yield i)
546+ 12``.)
547
Ezio Melotti45a101d2012-10-12 12:42:51 +0300548Values are sent into a generator by calling its :meth:`send(value)
549<generator.send>` method. This method resumes the generator's code and the
550``yield`` expression returns the specified value. If the regular
551:meth:`~generator.__next__` method is called, the ``yield`` returns ``None``.
Georg Brandl116aa622007-08-15 14:28:22 +0000552
553Here's a simple counter that increments by 1 and allows changing the value of
554the internal counter.
555
Christian Heimesfe337bf2008-03-23 21:54:12 +0000556.. testcode::
Georg Brandl116aa622007-08-15 14:28:22 +0000557
Ezio Melotti45a101d2012-10-12 12:42:51 +0300558 def counter(maximum):
Georg Brandl116aa622007-08-15 14:28:22 +0000559 i = 0
560 while i < maximum:
561 val = (yield i)
562 # If value provided, change counter
563 if val is not None:
564 i = val
565 else:
566 i += 1
567
568And here's an example of changing the counter:
569
Ezio Melotti35cbf162012-10-12 13:24:19 +0300570 >>> it = counter(10) #doctest: +SKIP
571 >>> next(it) #doctest: +SKIP
Georg Brandl116aa622007-08-15 14:28:22 +0000572 0
Ezio Melotti35cbf162012-10-12 13:24:19 +0300573 >>> next(it) #doctest: +SKIP
Georg Brandl116aa622007-08-15 14:28:22 +0000574 1
Ezio Melotti35cbf162012-10-12 13:24:19 +0300575 >>> it.send(8) #doctest: +SKIP
Georg Brandl116aa622007-08-15 14:28:22 +0000576 8
Ezio Melotti35cbf162012-10-12 13:24:19 +0300577 >>> next(it) #doctest: +SKIP
Georg Brandl116aa622007-08-15 14:28:22 +0000578 9
Ezio Melotti35cbf162012-10-12 13:24:19 +0300579 >>> next(it) #doctest: +SKIP
Georg Brandl116aa622007-08-15 14:28:22 +0000580 Traceback (most recent call last):
UltimateCoder88569402017-05-03 22:16:45 +0530581 File "t.py", line 15, in <module>
Georg Brandl6911e3c2007-09-04 07:15:32 +0000582 it.next()
Georg Brandl116aa622007-08-15 14:28:22 +0000583 StopIteration
584
585Because ``yield`` will often be returning ``None``, you should always check for
586this case. Don't just use its value in expressions unless you're sure that the
Zachary Ware0aecc182014-06-16 11:13:01 -0500587:meth:`~generator.send` method will be the only method used to resume your
Ezio Melotti45a101d2012-10-12 12:42:51 +0300588generator function.
Georg Brandl116aa622007-08-15 14:28:22 +0000589
Ezio Melotti45a101d2012-10-12 12:42:51 +0300590In addition to :meth:`~generator.send`, there are two other methods on
591generators:
Georg Brandl116aa622007-08-15 14:28:22 +0000592
Ezio Melotti45a101d2012-10-12 12:42:51 +0300593* :meth:`throw(type, value=None, traceback=None) <generator.throw>` is used to
594 raise an exception inside the generator; the exception is raised by the
595 ``yield`` expression where the generator's execution is paused.
Georg Brandl116aa622007-08-15 14:28:22 +0000596
Ezio Melotti45a101d2012-10-12 12:42:51 +0300597* :meth:`~generator.close` raises a :exc:`GeneratorExit` exception inside the
598 generator to terminate the iteration. On receiving this exception, the
599 generator's code must either raise :exc:`GeneratorExit` or
600 :exc:`StopIteration`; catching the exception and doing anything else is
601 illegal and will trigger a :exc:`RuntimeError`. :meth:`~generator.close`
602 will also be called by Python's garbage collector when the generator is
603 garbage-collected.
Georg Brandl116aa622007-08-15 14:28:22 +0000604
605 If you need to run cleanup code when a :exc:`GeneratorExit` occurs, I suggest
606 using a ``try: ... finally:`` suite instead of catching :exc:`GeneratorExit`.
607
608The cumulative effect of these changes is to turn generators from one-way
609producers of information into both producers and consumers.
610
611Generators also become **coroutines**, a more generalized form of subroutines.
612Subroutines are entered at one point and exited at another point (the top of the
613function, and a ``return`` statement), but coroutines can be entered, exited,
614and resumed at many different points (the ``yield`` statements).
615
616
617Built-in functions
618==================
619
620Let's look in more detail at built-in functions often used with iterators.
621
Georg Brandlf6945182008-02-01 11:56:49 +0000622Two of Python's built-in functions, :func:`map` and :func:`filter` duplicate the
623features of generator expressions:
Georg Brandl116aa622007-08-15 14:28:22 +0000624
Ezio Melotti45a101d2012-10-12 12:42:51 +0300625:func:`map(f, iterA, iterB, ...) <map>` returns an iterator over the sequence
Georg Brandlf6945182008-02-01 11:56:49 +0000626 ``f(iterA[0], iterB[0]), f(iterA[1], iterB[1]), f(iterA[2], iterB[2]), ...``.
Georg Brandl116aa622007-08-15 14:28:22 +0000627
Christian Heimesfe337bf2008-03-23 21:54:12 +0000628 >>> def upper(s):
629 ... return s.upper()
Georg Brandl116aa622007-08-15 14:28:22 +0000630
Georg Brandla3deea12008-12-15 08:29:32 +0000631 >>> list(map(upper, ['sentence', 'fragment']))
Christian Heimesfe337bf2008-03-23 21:54:12 +0000632 ['SENTENCE', 'FRAGMENT']
633 >>> [upper(s) for s in ['sentence', 'fragment']]
634 ['SENTENCE', 'FRAGMENT']
Georg Brandl116aa622007-08-15 14:28:22 +0000635
Georg Brandl48310cd2009-01-03 21:18:54 +0000636You can of course achieve the same effect with a list comprehension.
Georg Brandl116aa622007-08-15 14:28:22 +0000637
Ezio Melotti45a101d2012-10-12 12:42:51 +0300638:func:`filter(predicate, iter) <filter>` returns an iterator over all the
639sequence elements that meet a certain condition, and is similarly duplicated by
640list comprehensions. A **predicate** is a function that returns the truth
641value of some condition; for use with :func:`filter`, the predicate must take a
642single value.
Georg Brandl116aa622007-08-15 14:28:22 +0000643
Christian Heimesfe337bf2008-03-23 21:54:12 +0000644 >>> def is_even(x):
645 ... return (x % 2) == 0
Georg Brandl116aa622007-08-15 14:28:22 +0000646
Georg Brandla3deea12008-12-15 08:29:32 +0000647 >>> list(filter(is_even, range(10)))
Christian Heimesfe337bf2008-03-23 21:54:12 +0000648 [0, 2, 4, 6, 8]
Georg Brandl116aa622007-08-15 14:28:22 +0000649
Georg Brandl116aa622007-08-15 14:28:22 +0000650
Christian Heimesfe337bf2008-03-23 21:54:12 +0000651This can also be written as a list comprehension:
Georg Brandl116aa622007-08-15 14:28:22 +0000652
Georg Brandlf6945182008-02-01 11:56:49 +0000653 >>> list(x for x in range(10) if is_even(x))
Georg Brandl116aa622007-08-15 14:28:22 +0000654 [0, 2, 4, 6, 8]
655
Georg Brandl116aa622007-08-15 14:28:22 +0000656
csabella9be4ff32017-06-04 13:39:21 -0400657:func:`enumerate(iter, start=0) <enumerate>` counts off the elements in the
658iterable returning 2-tuples containing the count (from *start*) and
659each element. ::
Georg Brandl116aa622007-08-15 14:28:22 +0000660
Christian Heimesfe337bf2008-03-23 21:54:12 +0000661 >>> for item in enumerate(['subject', 'verb', 'object']):
Neal Norwitz752abd02008-05-13 04:55:24 +0000662 ... print(item)
Christian Heimesfe337bf2008-03-23 21:54:12 +0000663 (0, 'subject')
664 (1, 'verb')
665 (2, 'object')
Georg Brandl116aa622007-08-15 14:28:22 +0000666
667:func:`enumerate` is often used when looping through a list and recording the
668indexes at which certain conditions are met::
669
670 f = open('data.txt', 'r')
671 for i, line in enumerate(f):
672 if line.strip() == '':
Georg Brandl6911e3c2007-09-04 07:15:32 +0000673 print('Blank line at line #%i' % i)
Georg Brandl116aa622007-08-15 14:28:22 +0000674
Ezio Melotti45a101d2012-10-12 12:42:51 +0300675:func:`sorted(iterable, key=None, reverse=False) <sorted>` collects all the
676elements of the iterable into a list, sorts the list, and returns the sorted
Andrew Kuchling2a9c8e82013-05-20 10:14:53 -0400677result. The *key* and *reverse* arguments are passed through to the
Ezio Melotti45a101d2012-10-12 12:42:51 +0300678constructed list's :meth:`~list.sort` method. ::
Christian Heimesfe337bf2008-03-23 21:54:12 +0000679
680 >>> import random
681 >>> # Generate 8 random numbers between [0, 10000)
682 >>> rand_list = random.sample(range(10000), 8)
Ezio Melotti35cbf162012-10-12 13:24:19 +0300683 >>> rand_list #doctest: +SKIP
Christian Heimesfe337bf2008-03-23 21:54:12 +0000684 [769, 7953, 9828, 6431, 8442, 9878, 6213, 2207]
Ezio Melotti35cbf162012-10-12 13:24:19 +0300685 >>> sorted(rand_list) #doctest: +SKIP
Christian Heimesfe337bf2008-03-23 21:54:12 +0000686 [769, 2207, 6213, 6431, 7953, 8442, 9828, 9878]
Ezio Melotti35cbf162012-10-12 13:24:19 +0300687 >>> sorted(rand_list, reverse=True) #doctest: +SKIP
Christian Heimesfe337bf2008-03-23 21:54:12 +0000688 [9878, 9828, 8442, 7953, 6431, 6213, 2207, 769]
Georg Brandl116aa622007-08-15 14:28:22 +0000689
Ezio Melotti45a101d2012-10-12 12:42:51 +0300690(For a more detailed discussion of sorting, see the :ref:`sortinghowto`.)
Georg Brandl116aa622007-08-15 14:28:22 +0000691
Georg Brandl4216d2d2008-11-22 08:27:24 +0000692
Ezio Melotti45a101d2012-10-12 12:42:51 +0300693The :func:`any(iter) <any>` and :func:`all(iter) <all>` built-ins look at the
Serhiy Storchakafbc1c262013-11-29 12:17:13 +0200694truth values of an iterable's contents. :func:`any` returns ``True`` if any element
695in the iterable is a true value, and :func:`all` returns ``True`` if all of the
Ezio Melotti45a101d2012-10-12 12:42:51 +0300696elements are true values:
Georg Brandl116aa622007-08-15 14:28:22 +0000697
Sergey Fedoseevdb8707c2018-08-08 02:38:00 +0500698 >>> any([0, 1, 0])
Christian Heimesfe337bf2008-03-23 21:54:12 +0000699 True
Sergey Fedoseevdb8707c2018-08-08 02:38:00 +0500700 >>> any([0, 0, 0])
Christian Heimesfe337bf2008-03-23 21:54:12 +0000701 False
Sergey Fedoseevdb8707c2018-08-08 02:38:00 +0500702 >>> any([1, 1, 1])
Christian Heimesfe337bf2008-03-23 21:54:12 +0000703 True
Sergey Fedoseevdb8707c2018-08-08 02:38:00 +0500704 >>> all([0, 1, 0])
Christian Heimesfe337bf2008-03-23 21:54:12 +0000705 False
Sergey Fedoseevdb8707c2018-08-08 02:38:00 +0500706 >>> all([0, 0, 0])
Christian Heimesfe337bf2008-03-23 21:54:12 +0000707 False
Sergey Fedoseevdb8707c2018-08-08 02:38:00 +0500708 >>> all([1, 1, 1])
Christian Heimesfe337bf2008-03-23 21:54:12 +0000709 True
Georg Brandl116aa622007-08-15 14:28:22 +0000710
711
Ezio Melotti45a101d2012-10-12 12:42:51 +0300712:func:`zip(iterA, iterB, ...) <zip>` takes one element from each iterable and
Georg Brandl4216d2d2008-11-22 08:27:24 +0000713returns them in a tuple::
Georg Brandl116aa622007-08-15 14:28:22 +0000714
Georg Brandl4216d2d2008-11-22 08:27:24 +0000715 zip(['a', 'b', 'c'], (1, 2, 3)) =>
716 ('a', 1), ('b', 2), ('c', 3)
Georg Brandl116aa622007-08-15 14:28:22 +0000717
Georg Brandl4216d2d2008-11-22 08:27:24 +0000718It doesn't construct an in-memory list and exhaust all the input iterators
719before returning; instead tuples are constructed and returned only if they're
720requested. (The technical term for this behaviour is `lazy evaluation
Georg Brandl5d941342016-02-26 19:37:12 +0100721<https://en.wikipedia.org/wiki/Lazy_evaluation>`__.)
Georg Brandl116aa622007-08-15 14:28:22 +0000722
Georg Brandl4216d2d2008-11-22 08:27:24 +0000723This iterator is intended to be used with iterables that are all of the same
724length. If the iterables are of different lengths, the resulting stream will be
725the same length as the shortest iterable. ::
Georg Brandl116aa622007-08-15 14:28:22 +0000726
Georg Brandl4216d2d2008-11-22 08:27:24 +0000727 zip(['a', 'b'], (1, 2, 3)) =>
728 ('a', 1), ('b', 2)
Georg Brandl116aa622007-08-15 14:28:22 +0000729
Georg Brandl4216d2d2008-11-22 08:27:24 +0000730You should avoid doing this, though, because an element may be taken from the
731longer iterators and discarded. This means you can't go on to use the iterators
732further because you risk skipping a discarded element.
Georg Brandl116aa622007-08-15 14:28:22 +0000733
734
735The itertools module
736====================
737
738The :mod:`itertools` module contains a number of commonly-used iterators as well
739as functions for combining several iterators. This section will introduce the
740module's contents by showing small examples.
741
742The module's functions fall into a few broad classes:
743
744* Functions that create a new iterator based on an existing iterator.
745* Functions for treating an iterator's elements as function arguments.
746* Functions for selecting portions of an iterator's output.
747* A function for grouping an iterator's output.
748
749Creating new iterators
750----------------------
751
csabella9be4ff32017-06-04 13:39:21 -0400752:func:`itertools.count(start, step) <itertools.count>` returns an infinite
753stream of evenly spaced values. You can optionally supply the starting number,
754which defaults to 0, and the interval between numbers, which defaults to 1::
Georg Brandl116aa622007-08-15 14:28:22 +0000755
Christian Heimesfe337bf2008-03-23 21:54:12 +0000756 itertools.count() =>
757 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, ...
758 itertools.count(10) =>
759 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, ...
csabella9be4ff32017-06-04 13:39:21 -0400760 itertools.count(10, 5) =>
761 10, 15, 20, 25, 30, 35, 40, 45, 50, 55, ...
Georg Brandl116aa622007-08-15 14:28:22 +0000762
Ezio Melotti45a101d2012-10-12 12:42:51 +0300763:func:`itertools.cycle(iter) <itertools.cycle>` saves a copy of the contents of
764a provided iterable and returns a new iterator that returns its elements from
765first to last. The new iterator will repeat these elements infinitely. ::
Georg Brandl116aa622007-08-15 14:28:22 +0000766
Sergey Fedoseevdb8707c2018-08-08 02:38:00 +0500767 itertools.cycle([1, 2, 3, 4, 5]) =>
Christian Heimesfe337bf2008-03-23 21:54:12 +0000768 1, 2, 3, 4, 5, 1, 2, 3, 4, 5, ...
Georg Brandl116aa622007-08-15 14:28:22 +0000769
Ezio Melotti45a101d2012-10-12 12:42:51 +0300770:func:`itertools.repeat(elem, [n]) <itertools.repeat>` returns the provided
771element *n* times, or returns the element endlessly if *n* is not provided. ::
Georg Brandl116aa622007-08-15 14:28:22 +0000772
773 itertools.repeat('abc') =>
774 abc, abc, abc, abc, abc, abc, abc, abc, abc, abc, ...
775 itertools.repeat('abc', 5) =>
776 abc, abc, abc, abc, abc
777
Ezio Melotti45a101d2012-10-12 12:42:51 +0300778:func:`itertools.chain(iterA, iterB, ...) <itertools.chain>` takes an arbitrary
779number of iterables as input, and returns all the elements of the first
780iterator, then all the elements of the second, and so on, until all of the
781iterables have been exhausted. ::
Georg Brandl116aa622007-08-15 14:28:22 +0000782
783 itertools.chain(['a', 'b', 'c'], (1, 2, 3)) =>
784 a, b, c, 1, 2, 3
785
Ezio Melotti45a101d2012-10-12 12:42:51 +0300786:func:`itertools.islice(iter, [start], stop, [step]) <itertools.islice>` returns
787a stream that's a slice of the iterator. With a single *stop* argument, it
788will return the first *stop* elements. If you supply a starting index, you'll
789get *stop-start* elements, and if you supply a value for *step*, elements
790will be skipped accordingly. Unlike Python's string and list slicing, you can't
791use negative values for *start*, *stop*, or *step*. ::
Georg Brandl116aa622007-08-15 14:28:22 +0000792
793 itertools.islice(range(10), 8) =>
794 0, 1, 2, 3, 4, 5, 6, 7
795 itertools.islice(range(10), 2, 8) =>
796 2, 3, 4, 5, 6, 7
797 itertools.islice(range(10), 2, 8, 2) =>
798 2, 4, 6
799
Ezio Melotti45a101d2012-10-12 12:42:51 +0300800:func:`itertools.tee(iter, [n]) <itertools.tee>` replicates an iterator; it
801returns *n* independent iterators that will all return the contents of the
802source iterator.
803If you don't supply a value for *n*, the default is 2. Replicating iterators
Georg Brandl116aa622007-08-15 14:28:22 +0000804requires saving some of the contents of the source iterator, so this can consume
805significant memory if the iterator is large and one of the new iterators is
Christian Heimesfe337bf2008-03-23 21:54:12 +0000806consumed more than the others. ::
Georg Brandl116aa622007-08-15 14:28:22 +0000807
808 itertools.tee( itertools.count() ) =>
809 iterA, iterB
810
811 where iterA ->
812 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, ...
813
814 and iterB ->
815 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, ...
816
817
818Calling functions on elements
819-----------------------------
820
Ezio Melotti45a101d2012-10-12 12:42:51 +0300821The :mod:`operator` module contains a set of functions corresponding to Python's
822operators. Some examples are :func:`operator.add(a, b) <operator.add>` (adds
823two values), :func:`operator.ne(a, b) <operator.ne>` (same as ``a != b``), and
824:func:`operator.attrgetter('id') <operator.attrgetter>`
825(returns a callable that fetches the ``.id`` attribute).
Georg Brandl116aa622007-08-15 14:28:22 +0000826
Ezio Melotti45a101d2012-10-12 12:42:51 +0300827:func:`itertools.starmap(func, iter) <itertools.starmap>` assumes that the
828iterable will return a stream of tuples, and calls *func* using these tuples as
829the arguments::
Georg Brandl116aa622007-08-15 14:28:22 +0000830
Georg Brandl48310cd2009-01-03 21:18:54 +0000831 itertools.starmap(os.path.join,
Ezio Melotti45a101d2012-10-12 12:42:51 +0300832 [('/bin', 'python'), ('/usr', 'bin', 'java'),
833 ('/usr', 'bin', 'perl'), ('/usr', 'bin', 'ruby')])
Georg Brandl116aa622007-08-15 14:28:22 +0000834 =>
Ezio Melotti45a101d2012-10-12 12:42:51 +0300835 /bin/python, /usr/bin/java, /usr/bin/perl, /usr/bin/ruby
Georg Brandl116aa622007-08-15 14:28:22 +0000836
837
838Selecting elements
839------------------
840
841Another group of functions chooses a subset of an iterator's elements based on a
842predicate.
843
Ezio Melotti45a101d2012-10-12 12:42:51 +0300844:func:`itertools.filterfalse(predicate, iter) <itertools.filterfalse>` is the
Andrew Kuchling2a9c8e82013-05-20 10:14:53 -0400845opposite of :func:`filter`, returning all elements for which the predicate
846returns false::
Georg Brandl116aa622007-08-15 14:28:22 +0000847
Georg Brandl4216d2d2008-11-22 08:27:24 +0000848 itertools.filterfalse(is_even, itertools.count()) =>
Georg Brandl116aa622007-08-15 14:28:22 +0000849 1, 3, 5, 7, 9, 11, 13, 15, ...
850
Ezio Melotti45a101d2012-10-12 12:42:51 +0300851:func:`itertools.takewhile(predicate, iter) <itertools.takewhile>` returns
852elements for as long as the predicate returns true. Once the predicate returns
853false, the iterator will signal the end of its results. ::
Georg Brandl116aa622007-08-15 14:28:22 +0000854
855 def less_than_10(x):
Ezio Melotti45a101d2012-10-12 12:42:51 +0300856 return x < 10
Georg Brandl116aa622007-08-15 14:28:22 +0000857
858 itertools.takewhile(less_than_10, itertools.count()) =>
859 0, 1, 2, 3, 4, 5, 6, 7, 8, 9
860
861 itertools.takewhile(is_even, itertools.count()) =>
862 0
863
Ezio Melotti45a101d2012-10-12 12:42:51 +0300864:func:`itertools.dropwhile(predicate, iter) <itertools.dropwhile>` discards
865elements while the predicate returns true, and then returns the rest of the
866iterable's results. ::
Georg Brandl116aa622007-08-15 14:28:22 +0000867
868 itertools.dropwhile(less_than_10, itertools.count()) =>
869 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, ...
870
871 itertools.dropwhile(is_even, itertools.count()) =>
872 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, ...
873
Andrew Kuchling2a9c8e82013-05-20 10:14:53 -0400874:func:`itertools.compress(data, selectors) <itertools.compress>` takes two
875iterators and returns only those elements of *data* for which the corresponding
876element of *selectors* is true, stopping whenever either one is exhausted::
877
Sergey Fedoseevdb8707c2018-08-08 02:38:00 +0500878 itertools.compress([1, 2, 3, 4, 5], [True, True, False, False, True]) =>
Andrew Kuchling2a9c8e82013-05-20 10:14:53 -0400879 1, 2, 5
880
881
882Combinatoric functions
883----------------------
884
885The :func:`itertools.combinations(iterable, r) <itertools.combinations>`
886returns an iterator giving all possible *r*-tuple combinations of the
887elements contained in *iterable*. ::
888
889 itertools.combinations([1, 2, 3, 4, 5], 2) =>
890 (1, 2), (1, 3), (1, 4), (1, 5),
891 (2, 3), (2, 4), (2, 5),
892 (3, 4), (3, 5),
893 (4, 5)
894
895 itertools.combinations([1, 2, 3, 4, 5], 3) =>
896 (1, 2, 3), (1, 2, 4), (1, 2, 5), (1, 3, 4), (1, 3, 5), (1, 4, 5),
897 (2, 3, 4), (2, 3, 5), (2, 4, 5),
898 (3, 4, 5)
899
900The elements within each tuple remain in the same order as
901*iterable* returned them. For example, the number 1 is always before
9022, 3, 4, or 5 in the examples above. A similar function,
903:func:`itertools.permutations(iterable, r=None) <itertools.permutations>`,
904removes this constraint on the order, returning all possible
905arrangements of length *r*::
906
907 itertools.permutations([1, 2, 3, 4, 5], 2) =>
908 (1, 2), (1, 3), (1, 4), (1, 5),
909 (2, 1), (2, 3), (2, 4), (2, 5),
910 (3, 1), (3, 2), (3, 4), (3, 5),
911 (4, 1), (4, 2), (4, 3), (4, 5),
912 (5, 1), (5, 2), (5, 3), (5, 4)
913
914 itertools.permutations([1, 2, 3, 4, 5]) =>
915 (1, 2, 3, 4, 5), (1, 2, 3, 5, 4), (1, 2, 4, 3, 5),
916 ...
917 (5, 4, 3, 2, 1)
918
919If you don't supply a value for *r* the length of the iterable is used,
920meaning that all the elements are permuted.
921
922Note that these functions produce all of the possible combinations by
923position and don't require that the contents of *iterable* are unique::
924
925 itertools.permutations('aba', 3) =>
926 ('a', 'b', 'a'), ('a', 'a', 'b'), ('b', 'a', 'a'),
927 ('b', 'a', 'a'), ('a', 'a', 'b'), ('a', 'b', 'a')
928
929The identical tuple ``('a', 'a', 'b')`` occurs twice, but the two 'a'
930strings came from different positions.
931
932The :func:`itertools.combinations_with_replacement(iterable, r) <itertools.combinations_with_replacement>`
933function relaxes a different constraint: elements can be repeated
934within a single tuple. Conceptually an element is selected for the
935first position of each tuple and then is replaced before the second
936element is selected. ::
937
938 itertools.combinations_with_replacement([1, 2, 3, 4, 5], 2) =>
939 (1, 1), (1, 2), (1, 3), (1, 4), (1, 5),
940 (2, 2), (2, 3), (2, 4), (2, 5),
941 (3, 3), (3, 4), (3, 5),
942 (4, 4), (4, 5),
943 (5, 5)
944
Georg Brandl116aa622007-08-15 14:28:22 +0000945
946Grouping elements
947-----------------
948
Ezio Melotti45a101d2012-10-12 12:42:51 +0300949The last function I'll discuss, :func:`itertools.groupby(iter, key_func=None)
950<itertools.groupby>`, is the most complicated. ``key_func(elem)`` is a function
951that can compute a key value for each element returned by the iterable. If you
952don't supply a key function, the key is simply each element itself.
Georg Brandl116aa622007-08-15 14:28:22 +0000953
Ezio Melotti45a101d2012-10-12 12:42:51 +0300954:func:`~itertools.groupby` collects all the consecutive elements from the
955underlying iterable that have the same key value, and returns a stream of
9562-tuples containing a key value and an iterator for the elements with that key.
Georg Brandl116aa622007-08-15 14:28:22 +0000957
958::
959
Georg Brandl48310cd2009-01-03 21:18:54 +0000960 city_list = [('Decatur', 'AL'), ('Huntsville', 'AL'), ('Selma', 'AL'),
Georg Brandl116aa622007-08-15 14:28:22 +0000961 ('Anchorage', 'AK'), ('Nome', 'AK'),
Georg Brandl48310cd2009-01-03 21:18:54 +0000962 ('Flagstaff', 'AZ'), ('Phoenix', 'AZ'), ('Tucson', 'AZ'),
Georg Brandl116aa622007-08-15 14:28:22 +0000963 ...
964 ]
965
Ezio Melotti45a101d2012-10-12 12:42:51 +0300966 def get_state(city_state):
Georg Brandl0df79792008-10-04 18:33:26 +0000967 return city_state[1]
Georg Brandl116aa622007-08-15 14:28:22 +0000968
969 itertools.groupby(city_list, get_state) =>
970 ('AL', iterator-1),
971 ('AK', iterator-2),
972 ('AZ', iterator-3), ...
973
974 where
975 iterator-1 =>
976 ('Decatur', 'AL'), ('Huntsville', 'AL'), ('Selma', 'AL')
Georg Brandl48310cd2009-01-03 21:18:54 +0000977 iterator-2 =>
Georg Brandl116aa622007-08-15 14:28:22 +0000978 ('Anchorage', 'AK'), ('Nome', 'AK')
979 iterator-3 =>
980 ('Flagstaff', 'AZ'), ('Phoenix', 'AZ'), ('Tucson', 'AZ')
981
Ezio Melotti45a101d2012-10-12 12:42:51 +0300982:func:`~itertools.groupby` assumes that the underlying iterable's contents will
983already be sorted based on the key. Note that the returned iterators also use
984the underlying iterable, so you have to consume the results of iterator-1 before
Georg Brandl116aa622007-08-15 14:28:22 +0000985requesting iterator-2 and its corresponding key.
986
987
988The functools module
989====================
990
991The :mod:`functools` module in Python 2.5 contains some higher-order functions.
992A **higher-order function** takes one or more functions as input and returns a
993new function. The most useful tool in this module is the
994:func:`functools.partial` function.
995
996For programs written in a functional style, you'll sometimes want to construct
997variants of existing functions that have some of the parameters filled in.
998Consider a Python function ``f(a, b, c)``; you may wish to create a new function
999``g(b, c)`` that's equivalent to ``f(1, b, c)``; you're filling in a value for
1000one of ``f()``'s parameters. This is called "partial function application".
1001
Ezio Melotti45a101d2012-10-12 12:42:51 +03001002The constructor for :func:`~functools.partial` takes the arguments
1003``(function, arg1, arg2, ..., kwarg1=value1, kwarg2=value2)``. The resulting
1004object is callable, so you can just call it to invoke ``function`` with the
1005filled-in arguments.
Georg Brandl116aa622007-08-15 14:28:22 +00001006
1007Here's a small but realistic example::
1008
1009 import functools
1010
Ezio Melotti45a101d2012-10-12 12:42:51 +03001011 def log(message, subsystem):
1012 """Write the contents of 'message' to the specified subsystem."""
Georg Brandl6911e3c2007-09-04 07:15:32 +00001013 print('%s: %s' % (subsystem, message))
Georg Brandl116aa622007-08-15 14:28:22 +00001014 ...
1015
1016 server_log = functools.partial(log, subsystem='server')
1017 server_log('Unable to open socket')
1018
Ezio Melotti45a101d2012-10-12 12:42:51 +03001019:func:`functools.reduce(func, iter, [initial_value]) <functools.reduce>`
1020cumulatively performs an operation on all the iterable's elements and,
1021therefore, can't be applied to infinite iterables. *func* must be a function
1022that takes two elements and returns a single value. :func:`functools.reduce`
1023takes the first two elements A and B returned by the iterator and calculates
1024``func(A, B)``. It then requests the third element, C, calculates
1025``func(func(A, B), C)``, combines this result with the fourth element returned,
1026and continues until the iterable is exhausted. If the iterable returns no
1027values at all, a :exc:`TypeError` exception is raised. If the initial value is
1028supplied, it's used as a starting point and ``func(initial_value, A)`` is the
1029first calculation. ::
Georg Brandl4216d2d2008-11-22 08:27:24 +00001030
1031 >>> import operator, functools
1032 >>> functools.reduce(operator.concat, ['A', 'BB', 'C'])
1033 'ABBC'
1034 >>> functools.reduce(operator.concat, [])
1035 Traceback (most recent call last):
1036 ...
1037 TypeError: reduce() of empty sequence with no initial value
Sergey Fedoseevdb8707c2018-08-08 02:38:00 +05001038 >>> functools.reduce(operator.mul, [1, 2, 3], 1)
Georg Brandl4216d2d2008-11-22 08:27:24 +00001039 6
1040 >>> functools.reduce(operator.mul, [], 1)
1041 1
1042
1043If you use :func:`operator.add` with :func:`functools.reduce`, you'll add up all the
1044elements of the iterable. This case is so common that there's a special
1045built-in called :func:`sum` to compute it:
1046
Zachary Ware378a1d72016-08-09 16:47:04 -05001047 >>> import functools, operator
Sergey Fedoseevdb8707c2018-08-08 02:38:00 +05001048 >>> functools.reduce(operator.add, [1, 2, 3, 4], 0)
Georg Brandl4216d2d2008-11-22 08:27:24 +00001049 10
Sergey Fedoseevdb8707c2018-08-08 02:38:00 +05001050 >>> sum([1, 2, 3, 4])
Georg Brandl4216d2d2008-11-22 08:27:24 +00001051 10
1052 >>> sum([])
1053 0
1054
Ezio Melotti45a101d2012-10-12 12:42:51 +03001055For many uses of :func:`functools.reduce`, though, it can be clearer to just
1056write the obvious :keyword:`for` loop::
Georg Brandl4216d2d2008-11-22 08:27:24 +00001057
1058 import functools
1059 # Instead of:
Sergey Fedoseevdb8707c2018-08-08 02:38:00 +05001060 product = functools.reduce(operator.mul, [1, 2, 3], 1)
Georg Brandl4216d2d2008-11-22 08:27:24 +00001061
1062 # You can write:
1063 product = 1
Sergey Fedoseevdb8707c2018-08-08 02:38:00 +05001064 for i in [1, 2, 3]:
Georg Brandl4216d2d2008-11-22 08:27:24 +00001065 product *= i
1066
csabella9be4ff32017-06-04 13:39:21 -04001067A related function is :func:`itertools.accumulate(iterable, func=operator.add)
1068<itertools.accumulate>`. It performs the same calculation, but instead of
1069returning only the final result, :func:`accumulate` returns an iterator that
1070also yields each partial result::
Andrew Kuchling2a9c8e82013-05-20 10:14:53 -04001071
Sergey Fedoseevdb8707c2018-08-08 02:38:00 +05001072 itertools.accumulate([1, 2, 3, 4, 5]) =>
Andrew Kuchling2a9c8e82013-05-20 10:14:53 -04001073 1, 3, 6, 10, 15
1074
Sergey Fedoseevdb8707c2018-08-08 02:38:00 +05001075 itertools.accumulate([1, 2, 3, 4, 5], operator.mul) =>
Andrew Kuchling2a9c8e82013-05-20 10:14:53 -04001076 1, 2, 6, 24, 120
1077
Georg Brandl116aa622007-08-15 14:28:22 +00001078
1079The operator module
1080-------------------
1081
1082The :mod:`operator` module was mentioned earlier. It contains a set of
1083functions corresponding to Python's operators. These functions are often useful
1084in functional-style code because they save you from writing trivial functions
1085that perform a single operation.
1086
1087Some of the functions in this module are:
1088
Georg Brandlf6945182008-02-01 11:56:49 +00001089* Math operations: ``add()``, ``sub()``, ``mul()``, ``floordiv()``, ``abs()``, ...
Georg Brandl116aa622007-08-15 14:28:22 +00001090* Logical operations: ``not_()``, ``truth()``.
1091* Bitwise operations: ``and_()``, ``or_()``, ``invert()``.
1092* Comparisons: ``eq()``, ``ne()``, ``lt()``, ``le()``, ``gt()``, and ``ge()``.
1093* Object identity: ``is_()``, ``is_not()``.
1094
1095Consult the operator module's documentation for a complete list.
1096
1097
Georg Brandl4216d2d2008-11-22 08:27:24 +00001098Small functions and the lambda expression
1099=========================================
1100
1101When writing functional-style programs, you'll often need little functions that
1102act as predicates or that combine elements in some way.
1103
1104If there's a Python built-in or a module function that's suitable, you don't
1105need to define a new function at all::
1106
1107 stripped_lines = [line.strip() for line in lines]
1108 existing_files = filter(os.path.exists, file_list)
1109
1110If the function you need doesn't exist, you need to write it. One way to write
Serhiy Storchaka2b57c432018-12-19 08:09:46 +02001111small functions is to use the :keyword:`lambda` expression. ``lambda`` takes a
Ezio Melotti45a101d2012-10-12 12:42:51 +03001112number of parameters and an expression combining these parameters, and creates
1113an anonymous function that returns the value of the expression::
Georg Brandl4216d2d2008-11-22 08:27:24 +00001114
1115 adder = lambda x, y: x+y
1116
Ezio Melotti45a101d2012-10-12 12:42:51 +03001117 print_assign = lambda name, value: name + '=' + str(value)
1118
Georg Brandl4216d2d2008-11-22 08:27:24 +00001119An alternative is to just use the ``def`` statement and define a function in the
1120usual way::
1121
Ezio Melotti45a101d2012-10-12 12:42:51 +03001122 def adder(x, y):
1123 return x + y
Georg Brandl4216d2d2008-11-22 08:27:24 +00001124
1125 def print_assign(name, value):
1126 return name + '=' + str(value)
1127
Georg Brandl4216d2d2008-11-22 08:27:24 +00001128Which alternative is preferable? That's a style question; my usual course is to
1129avoid using ``lambda``.
1130
1131One reason for my preference is that ``lambda`` is quite limited in the
1132functions it can define. The result has to be computable as a single
1133expression, which means you can't have multiway ``if... elif... else``
1134comparisons or ``try... except`` statements. If you try to do too much in a
1135``lambda`` statement, you'll end up with an overly complicated expression that's
Ezio Melotti45a101d2012-10-12 12:42:51 +03001136hard to read. Quick, what's the following code doing? ::
Georg Brandl4216d2d2008-11-22 08:27:24 +00001137
1138 import functools
1139 total = functools.reduce(lambda a, b: (0, a[1] + b[1]), items)[1]
1140
1141You can figure it out, but it takes time to disentangle the expression to figure
1142out what's going on. Using a short nested ``def`` statements makes things a
1143little bit better::
1144
1145 import functools
Ezio Melotti45a101d2012-10-12 12:42:51 +03001146 def combine(a, b):
Georg Brandl4216d2d2008-11-22 08:27:24 +00001147 return 0, a[1] + b[1]
1148
1149 total = functools.reduce(combine, items)[1]
1150
1151But it would be best of all if I had simply used a ``for`` loop::
1152
1153 total = 0
1154 for a, b in items:
1155 total += b
1156
1157Or the :func:`sum` built-in and a generator expression::
1158
Sergey Fedoseevdb8707c2018-08-08 02:38:00 +05001159 total = sum(b for a, b in items)
Georg Brandl4216d2d2008-11-22 08:27:24 +00001160
1161Many uses of :func:`functools.reduce` are clearer when written as ``for`` loops.
1162
1163Fredrik Lundh once suggested the following set of rules for refactoring uses of
1164``lambda``:
1165
Ezio Melotti45a101d2012-10-12 12:42:51 +030011661. Write a lambda function.
11672. Write a comment explaining what the heck that lambda does.
11683. Study the comment for a while, and think of a name that captures the essence
Georg Brandl4216d2d2008-11-22 08:27:24 +00001169 of the comment.
Ezio Melotti45a101d2012-10-12 12:42:51 +030011704. Convert the lambda to a def statement, using that name.
11715. Remove the comment.
Georg Brandl4216d2d2008-11-22 08:27:24 +00001172
Georg Brandl48310cd2009-01-03 21:18:54 +00001173I really like these rules, but you're free to disagree
Georg Brandl4216d2d2008-11-22 08:27:24 +00001174about whether this lambda-free style is better.
1175
1176
Georg Brandl116aa622007-08-15 14:28:22 +00001177Revision History and Acknowledgements
1178=====================================
1179
1180The author would like to thank the following people for offering suggestions,
1181corrections and assistance with various drafts of this article: Ian Bicking,
1182Nick Coghlan, Nick Efford, Raymond Hettinger, Jim Jewett, Mike Krell, Leandro
1183Lameiro, Jussi Salmela, Collin Winter, Blake Winton.
1184
1185Version 0.1: posted June 30 2006.
1186
1187Version 0.11: posted July 1 2006. Typo fixes.
1188
1189Version 0.2: posted July 10 2006. Merged genexp and listcomp sections into one.
1190Typo fixes.
1191
1192Version 0.21: Added more references suggested on the tutor mailing list.
1193
1194Version 0.30: Adds a section on the ``functional`` module written by Collin
1195Winter; adds short section on the operator module; a few other edits.
1196
1197
1198References
1199==========
1200
1201General
1202-------
1203
1204**Structure and Interpretation of Computer Programs**, by Harold Abelson and
1205Gerald Jay Sussman with Julie Sussman. Full text at
Georg Brandl5d941342016-02-26 19:37:12 +01001206https://mitpress.mit.edu/sicp/. In this classic textbook of computer science,
Georg Brandl116aa622007-08-15 14:28:22 +00001207chapters 2 and 3 discuss the use of sequences and streams to organize the data
1208flow inside a program. The book uses Scheme for its examples, but many of the
1209design approaches described in these chapters are applicable to functional-style
1210Python code.
1211
1212http://www.defmacro.org/ramblings/fp.html: A general introduction to functional
1213programming that uses Java examples and has a lengthy historical introduction.
1214
Georg Brandl5d941342016-02-26 19:37:12 +01001215https://en.wikipedia.org/wiki/Functional_programming: General Wikipedia entry
Georg Brandl116aa622007-08-15 14:28:22 +00001216describing functional programming.
1217
Georg Brandl5d941342016-02-26 19:37:12 +01001218https://en.wikipedia.org/wiki/Coroutine: Entry for coroutines.
Georg Brandl116aa622007-08-15 14:28:22 +00001219
Georg Brandl5d941342016-02-26 19:37:12 +01001220https://en.wikipedia.org/wiki/Currying: Entry for the concept of currying.
Georg Brandl116aa622007-08-15 14:28:22 +00001221
1222Python-specific
1223---------------
1224
1225http://gnosis.cx/TPiP/: The first chapter of David Mertz's book
1226:title-reference:`Text Processing in Python` discusses functional programming
1227for text processing, in the section titled "Utilizing Higher-Order Functions in
1228Text Processing".
1229
1230Mertz also wrote a 3-part series of articles on functional programming
Georg Brandl48310cd2009-01-03 21:18:54 +00001231for IBM's DeveloperWorks site; see
Serhiy Storchaka6dff0202016-05-07 10:49:07 +03001232`part 1 <https://www.ibm.com/developerworks/linux/library/l-prog/index.html>`__,
1233`part 2 <https://www.ibm.com/developerworks/linux/library/l-prog2/index.html>`__, and
1234`part 3 <https://www.ibm.com/developerworks/linux/library/l-prog3/index.html>`__,
Georg Brandl116aa622007-08-15 14:28:22 +00001235
1236
1237Python documentation
1238--------------------
1239
1240Documentation for the :mod:`itertools` module.
1241
csabella9be4ff32017-06-04 13:39:21 -04001242Documentation for the :mod:`functools` module.
1243
Georg Brandl116aa622007-08-15 14:28:22 +00001244Documentation for the :mod:`operator` module.
1245
1246:pep:`289`: "Generator Expressions"
1247
1248:pep:`342`: "Coroutines via Enhanced Generators" describes the new generator
1249features in Python 2.5.
1250
1251.. comment
1252
Georg Brandl116aa622007-08-15 14:28:22 +00001253 Handy little function for printing part of an iterator -- used
1254 while writing this document.
1255
1256 import itertools
1257 def print_iter(it):
1258 slice = itertools.islice(it, 10)
1259 for elem in slice[:-1]:
1260 sys.stdout.write(str(elem))
1261 sys.stdout.write(', ')
Georg Brandl6911e3c2007-09-04 07:15:32 +00001262 print(elem[-1])