Timeline for Correct way to append to string in python
Current License: CC BY-SA 4.0
14 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Feb 14, 2023 at 16:50 | answer | added | JaMa | timeline score: 2 | |
| Feb 2, 2023 at 17:11 | comment | added | JonSG |
@HelpfulHelper your assumption is incorrect. join() is by far the fasted method. Feel free to try the code below.
|
|
| Feb 2, 2023 at 17:10 | comment | added | HelpfulHelper |
@JonSG yes, join() is slower because it does more operations
|
|
| Feb 1, 2023 at 21:54 | vote | accept | Étienne | ||
| Feb 1, 2023 at 21:37 | comment | added | Étienne |
@chepner s2 indeed uses += and is getting the optimization. This is fully intentional, since s2 benchmarks the optimization. s1 is the completely unoptimized case.
|
|
| Feb 1, 2023 at 17:00 | comment | added | JonSG |
@HelpfulHelper Are you suggesting we should expect join() to be slow as a result of extra operations?
|
|
| Feb 1, 2023 at 16:43 | comment | added | JNevill | I think it's worth mentioning that the PEP recommendation isn't to ensure you have the fastest code, but rather that your code doesn't experience non-linear timing growths depending on which implementation (cpython, jython, etc) you use. It's a middle ground between non-linear timings across implementations that may lead to unexpected slow performance, and the fastest possible execution because of a particular implementations efficiencies. | |
| Feb 1, 2023 at 16:38 | comment | added | HelpfulHelper |
join internally does something like arg[0]+<value>+arg[1]+<value>+.... So instead of joining 3 items together like with a+b+c, 5 items are joined together, like a+''+b+''+c, therefore, when using join instead of the + operator, it internally does len(arg)-1 additional calculations. When doing a single operation, it doesn't really matter, but when doing 100000 iterations like in your program, 200000 additional operations are done, and it makes a difference.
|
|
| Feb 1, 2023 at 16:37 | comment | added | chepner |
s2 uses +=, so you are still getting the optimization. The idea with join is not to call it many times with small lists, but to build up one list and pass it to one call of join.
|
|
| Feb 1, 2023 at 16:37 | comment | added | Friedrich | Replying to my own question: on my machine, list creation accounted for about 15% of the time. Not enough to explain the differences away. | |
| Feb 1, 2023 at 16:35 | comment | added | Barmar |
BTW, you should look into using the timeit module for benchmarking.
|
|
| Feb 1, 2023 at 16:34 | answer | added | JonSG | timeline score: 7 | |
| Feb 1, 2023 at 16:28 | comment | added | Friedrich |
How expensive is it to create [s3, str(i), '3'] alone?
|
|
| Feb 1, 2023 at 16:22 | history | asked | Étienne | CC BY-SA 4.0 |