⚡️ perf(ctx.Range): reduce allocations

[nickajacks1]

Description

strings.Split was causing extra allocations where strings.Cut can suffice. Also switch from strconv.Atoi because it causes an allocation when parsing a non-integer, which is common for Ranges.

The remaining allocation (see benchmarks) comes from the returned slice.

Checklist:

• I have performed a self-review of my own code
• I have commented my code, particularly in hard-to-understand areas
• New and existing unit tests pass locally with my changes
• I tried to make my code as fast as possible with as few allocations as possible
• For new code I have written benchmarks so that they can be analyzed and improved

Bench:

                            │  master.txt  │             branch.txt              │
                            │    sec/op    │   sec/op     vs base                │
_Ctx_Range/bytes=-700-8       197.80n ± 0%   65.06n ± 0%  -67.11% (p=0.000 n=20)
_Ctx_Range/bytes=500--8       198.10n ± 0%   63.70n ± 0%  -67.84% (p=0.000 n=20)
_Ctx_Range/bytes=500-1000-8   162.30n ± 0%   64.76n ± 0%  -60.10% (p=0.000 n=20)
_Ctx_Range/bytes=500-700-8    161.75n ± 0%   64.48n ± 0%  -60.14% (p=0.000 n=20)
geomean                        179.1n        64.50n       -63.98%

                            │ master.txt  │             branch.txt             │
                            │    B/op     │    B/op     vs base                │
_Ctx_Range/bytes=-700-8       144.00 ± 0%   16.00 ± 0%  -88.89% (p=0.000 n=20)
_Ctx_Range/bytes=500--8       144.00 ± 0%   16.00 ± 0%  -88.89% (p=0.000 n=20)
_Ctx_Range/bytes=500-1000-8    96.00 ± 0%   16.00 ± 0%  -83.33% (p=0.000 n=20)
_Ctx_Range/bytes=500-700-8     96.00 ± 0%   16.00 ± 0%  -83.33% (p=0.000 n=20)
geomean                        117.6        16.00       -86.39%

                            │ master.txt │             branch.txt             │
                            │ allocs/op  │ allocs/op   vs base                │
_Ctx_Range/bytes=-700-8       5.000 ± 0%   1.000 ± 0%  -80.00% (p=0.000 n=20)
_Ctx_Range/bytes=500--8       5.000 ± 0%   1.000 ± 0%  -80.00% (p=0.000 n=20)
_Ctx_Range/bytes=500-1000-8   4.000 ± 0%   1.000 ± 0%  -75.00% (p=0.000 n=20)
_Ctx_Range/bytes=500-700-8    4.000 ± 0%   1.000 ± 0%  -75.00% (p=0.000 n=20)
geomean                       4.472        1.000       -77.64%

[ReneWerner87]

thanks for the optimization

unfortunately we can't do it like this
because of

maybe you can do the exchange of the cut function only for the golang version from 1.18 ? here an example
https://github.com/gofiber/fiber/blob/master/utils/convert_s2b_new.go#L1
https://github.com/gofiber/fiber/blob/master/utils/convert_s2b_old.go#L1

[efectn]

thanks for the optimization

unfortunately we can't do it like this because of

maybe you can do the exchange of the cut function only for the golang version from 1.18 ? here an example https://github.com/gofiber/fiber/blob/master/utils/convert_s2b_new.go#L1 https://github.com/gofiber/fiber/blob/master/utils/convert_s2b_old.go#L1

Or we can use strings.Index instead of strings.Cut. It might worth for performance gain on 1.17

[ReneWerner87]

ok sure, lets test this

[nickajacks1]

Sure, I'll try out strings.IndexByte

[nickajacks1]

oops...I'll fix this too...

[gaby]

@nickajacks1

Error: ./ctx.go:1398:26: undefined: strings.Cut
Error: ./ctx.go:1410:40: undefined: strings.Cut
note: module requires Go 1.20

[ReneWerner87]

@nickajacks1 any progress?

[nickajacks1]

Yes, sorry, been busy. Just need to find time to sit down and finish this. Should be ready within the next couple days.

[nickajacks1]

Removed strings.Cut, added more tests, updated Benchmark to include a case with multiple ranges.
New benchmarks:

                                  │  master.txt  │             branch.txt              │
                                  │    sec/op    │   sec/op     vs base                │
_Ctx_Range/bytes=-700-8             198.20n ± 1%   55.46n ± 1%  -72.02% (p=0.000 n=20)
_Ctx_Range/bytes=500--8             197.50n ± 0%   55.08n ± 0%  -72.11% (p=0.000 n=20)
_Ctx_Range/bytes=500-1000-8         160.60n ± 0%   55.93n ± 1%  -65.18% (p=0.000 n=20)
_Ctx_Range/bytes=0-700,800-1000-8   245.10n ± 0%   93.25n ± 0%  -61.95% (p=0.000 n=20)
geomean                              198.1n        63.18n       -68.11%

                                  │ master.txt  │             branch.txt             │
                                  │    B/op     │    B/op     vs base                │
_Ctx_Range/bytes=-700-8             144.00 ± 0%   16.00 ± 0%  -88.89% (p=0.000 n=20)
_Ctx_Range/bytes=500--8             144.00 ± 0%   16.00 ± 0%  -88.89% (p=0.000 n=20)
_Ctx_Range/bytes=500-1000-8          96.00 ± 0%   16.00 ± 0%  -83.33% (p=0.000 n=20)
_Ctx_Range/bytes=0-700,800-1000-8   176.00 ± 0%   48.00 ± 0%  -72.73% (p=0.000 n=20)
geomean                              136.8        21.06       -84.61%

                                  │ master.txt │             branch.txt             │
                                  │ allocs/op  │ allocs/op   vs base                │
_Ctx_Range/bytes=-700-8             5.000 ± 0%   1.000 ± 0%  -80.00% (p=0.000 n=20)
_Ctx_Range/bytes=500--8             5.000 ± 0%   1.000 ± 0%  -80.00% (p=0.000 n=20)
_Ctx_Range/bytes=500-1000-8         4.000 ± 0%   1.000 ± 0%  -75.00% (p=0.000 n=20)
_Ctx_Range/bytes=0-700,800-1000-8   6.000 ± 0%   2.000 ± 0%  -66.67% (p=0.000 n=20)
geomean                             4.949        1.189       -75.97%