{"id":3465,"date":"2020-02-19T21:46:01","date_gmt":"2020-02-19T21:46:01","guid":{"rendered":"https:\/\/www.askpython.com\/?p=3465"},"modified":"2022-08-06T13:12:01","modified_gmt":"2022-08-06T13:12:01","slug":"python-timeit-module","status":"publish","type":"post","link":"https:\/\/www.askpython.com\/python-modules\/python-timeit-module","title":{"rendered":"Python timeit Module"},"content":{"rendered":"\n

The Python timeit<\/strong> module<\/a> is a simple interface to quickly measure the execution time for small blocks of code.<\/p>\n\n\n\n

When you are creating an application, you may wonder how this block of code will perform and would want to test it under different scenarios.<\/p>\n\n\n\n

For this, the timeit<\/code> module provides a very simple solution to this problem. Let’s look at how we can use this to time our code snippets!<\/p>\n\n\n\n

We shall be looking at both the Command Line Interface<\/strong> and a Callable interface<\/strong>.<\/p>\n\n\n\n

\n\n\n\n

Python timeit – Command Line Interface<\/h2>\n\n\n\n

The command-line interface is very similar to that of running a Python program.<\/p>\n\n\n\n

You need to import the external module timeit<\/code> using the -m option and apply it to your code.<\/p>\n\n\n

\npython -m timeit 'print("Hello from AskPython")'\n<\/pre><\/div>\n\n\n
This will run the snippet, passed as a string, using timeit<\/code>.<\/p>\n\n\n\n
By default, this will run the code 1 million times on Linux and 20 million times on Windows, and measure the best time among those values. The below outputs are from my Linux system.<\/p>\n\n\n\n
\"Python<\/figure><\/div>\n\n\n\n
Note that if you already have a for loop in your snippet, the module will ensure that the total number of iterations is close to 1 million, so your entire loop will not run for 1 million times! <\/p>\n\n\n\n
\"Python<\/figure><\/div>\n\n\n\n
Python timeit – Python Interface<\/h2>\n\n\n\n
We can also use timeit<\/code> via the Python interpreter, and import it using:<\/p>\n\n\n
\nimport timeit\n<\/pre><\/div>\n\n\n
To find the execution time, pass the code as a string to timeit.timeit()<\/code>.<\/p>\n\n\n
\nexecution_time = timeit.timeit(code, number)\n<\/pre><\/div>\n\n\n
We can control the number of iterations using the number<\/code> parameter.<\/p>\n\n\n
\n>>> import timeit\n>>> timeit.timeit('"-".join(str(n) for n in range(100))', number=10000)\n0.19053685299877543\n>>> timeit.timeit('"-".join([str(n) for n in range(100)])', number=10000)\n0.172546762998536\n>>> timeit.timeit('"-".join(map(str, range(100)))', number=10000)\n0.13625987299747067\n>>> \n<\/pre><\/div>\n\n\n
\n\n\n\n
Using the timeit Module<\/h2>\n\n\n\n
Let’s now look at how we can use timeit<\/code> to time a snippet, inside our program.<\/p>\n\n\n\n
But before that, you may be wondering about something. What if your code required some previous setup to be done? And if you need to import certain modules too?<\/p>\n\n\n\n
Well, the solution to this problem is to use a setup<\/strong> code block, which will do all the necessary work in setting up all required modules and variables.<\/p>\n\n\n
\nsetup_code = "import math"\n<\/pre><\/div>\n\n\n
Writing a setup block is very simple. You simply write whatever code you need, and pass it as a string, into a variable.<\/p>\n\n\n\n
After this, you can write your main code block and pass it to timeit.timeit()<\/code>, using the setup<\/code> and stmt<\/code> parameters.<\/p>\n\n\n
\nexecution_time = timeit.timeit(setup = setup_code, stmt = main_block, number = 100)\n<\/pre><\/div>\n\n\n
timeit<\/code> will ensure that the setup is done before your main loop is measured, so it is only executed once!<\/p>\n\n\n\n
Let’s quickly look at an example now.<\/p>\n\n\n\n
This code tries to get all sub-arrays from the starting element of a numpy<\/strong> array. Notice that the setup block is run only once. <\/p>\n\n\n
\nimport timeit\n\n# Setup is run only once\nsetup_code = '''\nimport numpy as np\na = np.arange(0, 1000)\nprint(a.shape)\ndef print_subarrays(a):\n    op = []\n    for i in range(a.shape[0]):\n        op.append(a[:i])\n'''\n\nmain_block = '''\nprint_subarrays(a)\n'''\n\n# Main Block is run 1000 times\nprint('Best execution Time among 1000 iterations:', timeit.timeit(setup=setup_code, stmt=main_block, number=1000))\n\n<\/pre><\/div>\n\n\n
Output<\/strong><\/p>\n\n\n
\n(1000,)\nBest execution Time among 1000 iterations: 0.3830194959991786\n<\/pre><\/div>\n\n\n
\n\n\n\n
Compare the Performance of Code Blocks<\/h2>\n\n\n\n
We can easily compare the performance of multiple blocks of code using timeit<\/code>.<\/p>\n\n\n\n
We will use a timer for this purpose, using timeit.default_timer()<\/code>.<\/p>\n\n\n\n
The time taken for a code block will be the current time minus the initial time taken as reference, which you can pass through variables.<\/p>\n\n\n
\nimport timeit\n\nstart_time = timeit.default_timer()\nfunction_1()\ntime_1 = timeit.default_timer() - start_time\n\nstart_time = timeit.default_timer()\nfunction_2()\ntime_2 = timeit.default_timer() - start_time\n\nprint('Function 1 took', time_1)\nprint('Function 2 took', time_2)\n<\/pre><\/div>\n\n\n
Let’s test 2 functions on a numpy array range()<\/code> and np.arange()<\/code> and see how they compare.<\/p>\n\n\n
\nimport timeit\nimport numpy as np\n\ndef time_range(size):\n    for i in range(size):\n        pass\n\ndef time_arange(size):\n    np.arange(size)\n\nif __name__ == '__main__':\n    # For smaller arrays\n    print('Array size: 1000')\n\n    start_time = timeit.default_timer();\n    time_range(1000)\n    range_time_1 = timeit.default_timer() - start_time\n\n    start_time = timeit.default_timer();\n    time_arange(1000)\n    arange_time_1 = timeit.default_timer() - start_time\n\n    # For large arrays\n    print('Array size: 1000000')\n\n    start_time = timeit.default_timer();\n    time_range(1000000)\n    range_time_2 = timeit.default_timer() - start_time\n\n    start_time = timeit.default_timer();\n    time_arange(1000000)\n    arange_time_2 = timeit.default_timer() - start_time\n\n    print(f'size 1000: range() took {range_time_1}')\n    print(f'size 1000: arange() took {arange_time_1}')\n    print(f'size 1000000: range() took {range_time_2}')\n    print(f'size 1000000: arange() took {arange_time_2}')\n\n<\/pre><\/div>\n\n\n
Output<\/strong><\/p>\n\n\n
\nArray size: 1000\nArray size: 1000000\nsize 1000: range() took 2.2970001737121493e-05\nsize 1000: arange() took 8.393999451072887e-06\nsize 1000000: range() took 0.02567379199899733\nsize 1000000: arange() took 0.0031752489994687494\n<\/pre><\/div>\n\n\n
Thus, we could easily use timeit<\/code> to compare the performance of different functions.<\/p>\n\n\n\n
\n\n\n\n
Time a Particular Function<\/h2>\n\n\n\n
We can also only time the performance of a specific function in a script, without running the other code blocks.<\/p>\n\n\n\n
If the earlier file was called numpy_compare.py<\/strong>, we can find the execution time using the below invocation:<\/p>\n\n\n
\npython -m timeit -s 'import numpy_compare' 'numpy_compare.time_range(1000)'\n<\/pre><\/div>\n\n\n
Output<\/strong><\/p>\n\n\n\n
\"Python
Python Timeit Test On Function<\/figcaption><\/figure><\/div>\n\n\n\n
\n\n\n\n
Conclusion<\/h2>\n\n\n\n
Hopefully, you’re now familiar with the basics of this module, so it’s time for you to start using it and increase your productivity!<\/p>\n\n\n\n
\n\n\n\n
References<\/h2>\n\n\n\n