• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1:mod:`timeit` --- Measure execution time of small code snippets
2===============================================================
3
4.. module:: timeit
5   :synopsis: Measure the execution time of small code snippets.
6
7**Source code:** :source:`Lib/timeit.py`
8
9.. index::
10   single: Benchmarking
11   single: Performance
12
13--------------
14
15This module provides a simple way to time small bits of Python code. It has both
16a :ref:`timeit-command-line-interface` as well as a :ref:`callable <python-interface>`
17one.  It avoids a number of common traps for measuring execution times.
18See also Tim Peters' introduction to the "Algorithms" chapter in the *Python
19Cookbook*, published by O'Reilly.
20
21
22Basic Examples
23--------------
24
25The following example shows how the :ref:`timeit-command-line-interface`
26can be used to compare three different expressions:
27
28.. code-block:: sh
29
30   $ python3 -m timeit '"-".join(str(n) for n in range(100))'
31   10000 loops, best of 3: 30.2 usec per loop
32   $ python3 -m timeit '"-".join([str(n) for n in range(100)])'
33   10000 loops, best of 3: 27.5 usec per loop
34   $ python3 -m timeit '"-".join(map(str, range(100)))'
35   10000 loops, best of 3: 23.2 usec per loop
36
37This can be achieved from the :ref:`python-interface` with::
38
39   >>> import timeit
40   >>> timeit.timeit('"-".join(str(n) for n in range(100))', number=10000)
41   0.3018611848820001
42   >>> timeit.timeit('"-".join([str(n) for n in range(100)])', number=10000)
43   0.2727368790656328
44   >>> timeit.timeit('"-".join(map(str, range(100)))', number=10000)
45   0.23702679807320237
46
47
48Note however that :mod:`timeit` will automatically determine the number of
49repetitions only when the command-line interface is used.  In the
50:ref:`timeit-examples` section you can find more advanced examples.
51
52
53.. _python-interface:
54
55Python Interface
56----------------
57
58The module defines three convenience functions and a public class:
59
60
61.. function:: timeit(stmt='pass', setup='pass', timer=<default timer>, number=1000000, globals=None)
62
63   Create a :class:`Timer` instance with the given statement, *setup* code and
64   *timer* function and run its :meth:`.timeit` method with *number* executions.
65   The optional *globals* argument specifies a namespace in which to execute the
66   code.
67
68   .. versionchanged:: 3.5
69      The optional *globals* parameter was added.
70
71
72.. function:: repeat(stmt='pass', setup='pass', timer=<default timer>, repeat=3, number=1000000, globals=None)
73
74   Create a :class:`Timer` instance with the given statement, *setup* code and
75   *timer* function and run its :meth:`.repeat` method with the given *repeat*
76   count and *number* executions.  The optional *globals* argument specifies a
77   namespace in which to execute the code.
78
79   .. versionchanged:: 3.5
80      The optional *globals* parameter was added.
81
82.. function:: default_timer()
83
84   The default timer, which is always :func:`time.perf_counter`.
85
86   .. versionchanged:: 3.3
87      :func:`time.perf_counter` is now the default timer.
88
89
90.. class:: Timer(stmt='pass', setup='pass', timer=<timer function>, globals=None)
91
92   Class for timing execution speed of small code snippets.
93
94   The constructor takes a statement to be timed, an additional statement used
95   for setup, and a timer function.  Both statements default to ``'pass'``;
96   the timer function is platform-dependent (see the module doc string).
97   *stmt* and *setup* may also contain multiple statements separated by ``;``
98   or newlines, as long as they don't contain multi-line string literals.  The
99   statement will by default be executed within timeit's namespace; this behavior
100   can be controlled by passing a namespace to *globals*.
101
102   To measure the execution time of the first statement, use the :meth:`.timeit`
103   method.  The :meth:`.repeat` and :meth:`.autorange` methods are convenience
104   methods to call :meth:`.timeit` multiple times.
105
106   The execution time of *setup* is excluded from the overall timed execution run.
107
108   The *stmt* and *setup* parameters can also take objects that are callable
109   without arguments.  This will embed calls to them in a timer function that
110   will then be executed by :meth:`.timeit`.  Note that the timing overhead is a
111   little larger in this case because of the extra function calls.
112
113   .. versionchanged:: 3.5
114      The optional *globals* parameter was added.
115
116   .. method:: Timer.timeit(number=1000000)
117
118      Time *number* executions of the main statement.  This executes the setup
119      statement once, and then returns the time it takes to execute the main
120      statement a number of times, measured in seconds as a float.
121      The argument is the number of times through the loop, defaulting to one
122      million.  The main statement, the setup statement and the timer function
123      to be used are passed to the constructor.
124
125      .. note::
126
127         By default, :meth:`.timeit` temporarily turns off :term:`garbage
128         collection` during the timing.  The advantage of this approach is that
129         it makes independent timings more comparable.  This disadvantage is
130         that GC may be an important component of the performance of the
131         function being measured.  If so, GC can be re-enabled as the first
132         statement in the *setup* string.  For example::
133
134            timeit.Timer('for i in range(10): oct(i)', 'gc.enable()').timeit()
135
136
137   .. method:: Timer.autorange(callback=None)
138
139      Automatically determine how many times to call :meth:`.timeit`.
140
141      This is a convenience function that calls :meth:`.timeit` repeatedly
142      so that the total time >= 0.2 second, returning the eventual
143      (number of loops, time taken for that number of loops). It calls
144      :meth:`.timeit` with *number* set to successive powers of ten (10,
145      100, 1000, ...) up to a maximum of one billion, until the time taken
146      is at least 0.2 second, or the maximum is reached.
147
148      If *callback* is given and is not ``None``, it will be called after
149      each trial with two arguments: ``callback(number, time_taken)``.
150
151      .. versionadded:: 3.6
152
153
154   .. method:: Timer.repeat(repeat=3, number=1000000)
155
156      Call :meth:`.timeit` a few times.
157
158      This is a convenience function that calls the :meth:`.timeit` repeatedly,
159      returning a list of results.  The first argument specifies how many times
160      to call :meth:`.timeit`.  The second argument specifies the *number*
161      argument for :meth:`.timeit`.
162
163      .. note::
164
165         It's tempting to calculate mean and standard deviation from the result
166         vector and report these.  However, this is not very useful.
167         In a typical case, the lowest value gives a lower bound for how fast
168         your machine can run the given code snippet; higher values in the
169         result vector are typically not caused by variability in Python's
170         speed, but by other processes interfering with your timing accuracy.
171         So the :func:`min` of the result is probably the only number you
172         should be interested in.  After that, you should look at the entire
173         vector and apply common sense rather than statistics.
174
175
176   .. method:: Timer.print_exc(file=None)
177
178      Helper to print a traceback from the timed code.
179
180      Typical use::
181
182         t = Timer(...)       # outside the try/except
183         try:
184             t.timeit(...)    # or t.repeat(...)
185         except Exception:
186             t.print_exc()
187
188      The advantage over the standard traceback is that source lines in the
189      compiled template will be displayed.  The optional *file* argument directs
190      where the traceback is sent; it defaults to :data:`sys.stderr`.
191
192
193.. _timeit-command-line-interface:
194
195Command-Line Interface
196----------------------
197
198When called as a program from the command line, the following form is used::
199
200   python -m timeit [-n N] [-r N] [-u U] [-s S] [-t] [-c] [-h] [statement ...]
201
202Where the following options are understood:
203
204.. program:: timeit
205
206.. cmdoption:: -n N, --number=N
207
208   how many times to execute 'statement'
209
210.. cmdoption:: -r N, --repeat=N
211
212   how many times to repeat the timer (default 3)
213
214.. cmdoption:: -s S, --setup=S
215
216   statement to be executed once initially (default ``pass``)
217
218.. cmdoption:: -p, --process
219
220   measure process time, not wallclock time, using :func:`time.process_time`
221   instead of :func:`time.perf_counter`, which is the default
222
223   .. versionadded:: 3.3
224
225.. cmdoption:: -t, --time
226
227   use :func:`time.time` (deprecated)
228
229.. cmdoption:: -u, --unit=U
230
231    specify a time unit for timer output; can select usec, msec, or sec
232
233   .. versionadded:: 3.5
234
235.. cmdoption:: -c, --clock
236
237   use :func:`time.clock` (deprecated)
238
239.. cmdoption:: -v, --verbose
240
241   print raw timing results; repeat for more digits precision
242
243.. cmdoption:: -h, --help
244
245   print a short usage message and exit
246
247A multi-line statement may be given by specifying each line as a separate
248statement argument; indented lines are possible by enclosing an argument in
249quotes and using leading spaces.  Multiple :option:`-s` options are treated
250similarly.
251
252If :option:`-n` is not given, a suitable number of loops is calculated by trying
253successive powers of 10 until the total time is at least 0.2 seconds.
254
255:func:`default_timer` measurements can be affected by other programs running on
256the same machine, so the best thing to do when accurate timing is necessary is
257to repeat the timing a few times and use the best time.  The :option:`-r`
258option is good for this; the default of 3 repetitions is probably enough in
259most cases.  You can use :func:`time.process_time` to measure CPU time.
260
261.. note::
262
263   There is a certain baseline overhead associated with executing a pass statement.
264   The code here doesn't try to hide it, but you should be aware of it.  The
265   baseline overhead can be measured by invoking the program without arguments,
266   and it might differ between Python versions.
267
268
269.. _timeit-examples:
270
271Examples
272--------
273
274It is possible to provide a setup statement that is executed only once at the beginning:
275
276.. code-block:: sh
277
278   $ python -m timeit -s 'text = "sample string"; char = "g"'  'char in text'
279   10000000 loops, best of 3: 0.0877 usec per loop
280   $ python -m timeit -s 'text = "sample string"; char = "g"'  'text.find(char)'
281   1000000 loops, best of 3: 0.342 usec per loop
282
283::
284
285   >>> import timeit
286   >>> timeit.timeit('char in text', setup='text = "sample string"; char = "g"')
287   0.41440500499993504
288   >>> timeit.timeit('text.find(char)', setup='text = "sample string"; char = "g"')
289   1.7246671520006203
290
291The same can be done using the :class:`Timer` class and its methods::
292
293   >>> import timeit
294   >>> t = timeit.Timer('char in text', setup='text = "sample string"; char = "g"')
295   >>> t.timeit()
296   0.3955516149999312
297   >>> t.repeat()
298   [0.40193588800002544, 0.3960157959998014, 0.39594301399984033]
299
300
301The following examples show how to time expressions that contain multiple lines.
302Here we compare the cost of using :func:`hasattr` vs. :keyword:`try`/:keyword:`except`
303to test for missing and present object attributes:
304
305.. code-block:: sh
306
307   $ python -m timeit 'try:' '  str.__bool__' 'except AttributeError:' '  pass'
308   100000 loops, best of 3: 15.7 usec per loop
309   $ python -m timeit 'if hasattr(str, "__bool__"): pass'
310   100000 loops, best of 3: 4.26 usec per loop
311
312   $ python -m timeit 'try:' '  int.__bool__' 'except AttributeError:' '  pass'
313   1000000 loops, best of 3: 1.43 usec per loop
314   $ python -m timeit 'if hasattr(int, "__bool__"): pass'
315   100000 loops, best of 3: 2.23 usec per loop
316
317::
318
319   >>> import timeit
320   >>> # attribute is missing
321   >>> s = """\
322   ... try:
323   ...     str.__bool__
324   ... except AttributeError:
325   ...     pass
326   ... """
327   >>> timeit.timeit(stmt=s, number=100000)
328   0.9138244460009446
329   >>> s = "if hasattr(str, '__bool__'): pass"
330   >>> timeit.timeit(stmt=s, number=100000)
331   0.5829014980008651
332   >>>
333   >>> # attribute is present
334   >>> s = """\
335   ... try:
336   ...     int.__bool__
337   ... except AttributeError:
338   ...     pass
339   ... """
340   >>> timeit.timeit(stmt=s, number=100000)
341   0.04215312199994514
342   >>> s = "if hasattr(int, '__bool__'): pass"
343   >>> timeit.timeit(stmt=s, number=100000)
344   0.08588060699912603
345
346
347To give the :mod:`timeit` module access to functions you define, you can pass a
348*setup* parameter which contains an import statement::
349
350   def test():
351       """Stupid test function"""
352       L = [i for i in range(100)]
353
354   if __name__ == '__main__':
355       import timeit
356       print(timeit.timeit("test()", setup="from __main__ import test"))
357
358Another option is to pass :func:`globals` to the  *globals* parameter, which will cause the code
359to be executed within your current global namespace.  This can be more convenient
360than individually specifying imports::
361
362   def f(x):
363       return x**2
364   def g(x):
365       return x**4
366   def h(x):
367       return x**8
368
369   import timeit
370   print(timeit.timeit('[func(42) for func in (f,g,h)]', globals=globals()))
371