 d-enk
|
68f827d9bf
|
perf: optimize Substring to work directly with strings instead of converting to runes (#822)
* perf: optimize Substring to work directly with strings instead of converting to runes
- Rewrite Substring to iterate over string bytes directly, avoiding full []rune conversion
- Improve performance for long strings by only processing necessary portions
- Add comprehensive test cases for Unicode handling, invalid UTF-8, and edge cases
- Add BenchmarkSubstring to measure performance improvements
- Improve documentation with detailed parameter descriptions
- Handle invalid UTF-8 sequences by converting to []rune when needed
Bencstat:
│ old.txt │ new.txt │
│ sec/op │ sec/op vs base │
Substring/{10_10}-4 558.85n ± 9% 39.75n ± 10% -92.89% (p=0.000 n=8)
Substring/{50_50}-4 783.10n ± 6% 85.15n ± 5% -89.13% (p=0.000 n=8)
Substring/{50_45}-4 773.30n ± 3% 126.5n ± 7% -83.65% (p=0.000 n=8)
Substring/{-50_50}-4 794.00n ± 2% 177.6n ± 7% -77.63% (p=0.000 n=8)
Substring/{-10_10}-4 542.85n ± 20% 41.82n ± 6% -92.30% (p=0.000 n=8)
geomean 680.4n 79.52n -88.31%
│ old.txt │ new.txt │
│ B/op │ B/op vs base │
Substring/{10_10}-4 432.0 ± 0% 0.0 ± 0% -100.00% (p=0.000 n=8)
Substring/{50_50}-4 480.0 ± 0% 0.0 ± 0% -100.00% (p=0.000 n=8)
Substring/{50_45}-4 464.0 ± 0% 0.0 ± 0% -100.00% (p=0.000 n=8)
Substring/{-50_50}-4 480.0 ± 0% 0.0 ± 0% -100.00% (p=0.000 n=8)
Substring/{-10_10}-4 432.0 ± 0% 0.0 ± 0% -100.00% (p=0.000 n=8)
│ old.txt │ new.txt │
│ allocs/op │ allocs/op vs base │
Substring/{10_10}-4 2.000 ± 0% 0.000 ± 0% -100.00% (p=0.000 n=8)
Substring/{50_50}-4 2.000 ± 0% 0.000 ± 0% -100.00% (p=0.000 n=8)
Substring/{50_45}-4 2.000 ± 0% 0.000 ± 0% -100.00% (p=0.000 n=8)
Substring/{-50_50}-4 2.000 ± 0% 0.000 ± 0% -100.00% (p=0.000 n=8)
Substring/{-10_10}-4 2.000 ± 0% 0.000 ± 0% -100.00% (p=0.000 n=8)
* Enhance substring documentation with Unicode details
Returns a substring starting at the given offset with the specified length. Supports negative offsets; out-of-bounds are clamped. Operates on Unicode runes (characters) and is optimized for zero allocations.
---------
Co-authored-by: Samuel Berthe <dev@samuel-berthe.fr>
|
2026-02-27 22:19:20 +01:00 |
|