python/exercises/practice/reverse-string/.articles/performance/content.md

Performance

In this article, we'll find out how to most efficiently reverse a string in Python.

The approaches introduction lists six groups of approaches:

1. Sequence Slice with Negative Step
2. Iteration with String Concatenation
3. Reverse Iteration with Range()
4. Make a list and Use str.join()
5. Make a list and use list.reverse()
6. Use the built-in reversed()
7. Other interesting approaches

For our performance investigations, we will compare the most performant from each group and a seventh approach using map().

Benchmarks

To benchmark these functions, we wrote a small benchmarking script using the timeit module along with third-party libraries numpy and pandas.

The reverse slice is by far the most performant, followed by the built-ins list.reverse() and reversed(). Iteration and concatenation is next, due to the CPython string optimization (see the iteration and concatenation approach for all the details), but this approach slows radically for strings longer than 142 characters.

With more than 142 characters, using a list, swapping positions, and joining via join() is the most performant method that doesn't use built-ins. Using map() with join() was the least performant approach overall.

string length >>>>	5	11	22	52	66	86	142	1420	14200	142000
reverse slice	1.71e-07	1.73e-07	1.86e-07	2.07e-07	2.19e-07	2.36e-07	3.49e-07	1.51e-06	1.19e-05	1.18e-04
list reverse	3.29e-07	4.28e-07	5.73e-07	8.92e-07	1.20e-06	1.51e-06	2.34e-06	1.94e-05	1.90e-04	1.91e-03
reversed builtin	3.68e-07	4.83e-07	6.98e-07	1.20e-06	1.62e-06	2.03e-06	2.71e-06	2.42e-05	2.35e-04	2.36e-03
iterate & concatenate	4.18e-07	8.10e-07	1.49e-06	3.49e-06	4.35e-06	6.18e-06	4.12e-06	2.03e-04	3.31e-03	4.61e-01
list swap	6.43e-07	4.00e-07	1.54e-06	3.01e-06	2.06e-06	4.71e-06	7.47e-06	8.97e-05	2.52e-03	1.02e-02
iterate with range	9.19e-07	1.35e-06	2.12e-06	4.15e-06	5.23e-06	6.60e-06	1.10e-05	1.05e-04	1.02e-03	1.07e-02
map and join	9.56e-07	1.72e-06	3.08e-06	6.27e-06	7.96e-06	1.03e-05	1.71e-05	1.70e-04	1.68e-03	1.70e-02

Measurements were taken on a 3.1 GHz Quad-Core Intel Core i7 Mac running MacOS Ventura. Tests used timeit.Timer.autorange(), repeated 3 times. Time is reported in seconds taken per string after calculating the 'best of' time. The timeit module docs have more details, and note.nkmk.me has a nice summary of methods.