Some optimizations

[anatoly-kussul]

v4.0.0:

cpu: 12th Gen Intel(R) Core(TM) i5-12400F
BenchmarkUUID-12                  355501              3152 ns/op            2650 B/op         98 allocs/op
BenchmarkEncoding-12              414798              2828 ns/op            2640 B/op         98 allocs/op
BenchmarkDecoding-12             2422132               470.2 ns/op           152 B/op          5 allocs/op

with these optimizations:

BenchmarkUUID-12                 3777259               314.0 ns/op            48 B/op          2 allocs/op
BenchmarkEncoding-12             6792055               184.5 ns/op            32 B/op          1 allocs/op
BenchmarkDecoding-12             5202288               230.9 ns/op            64 B/op          2 allocs/op

[anatoly-kussul]

fixed decoding benchmark, as it was returning error, and not fully decoding.

last commit benchmarks:

cpu: 12th Gen Intel(R) Core(TM) i5-12400F
BenchmarkUUID-12                 4058689               283.6 ns/op            40 B/op          2 allocs/op
BenchmarkEncoding-12             8116075               149.6 ns/op            24 B/op          1 allocs/op
BenchmarkDecoding-12             4644693               258.5 ns/op             0 B/op          0 allocs/op

v4.0.0 benchmarks with fixed decoding test:

cpu: 12th Gen Intel(R) Core(TM) i5-12400F
BenchmarkUUID-12                  363300              3179 ns/op            2650 B/op         98 allocs/op
BenchmarkEncoding-12              431427              2851 ns/op            2640 B/op         98 allocs/op
BenchmarkDecoding-12             1000000              1005 ns/op             192 B/op          8 allocs/op

[lithammer]

Very impressing work! 👏 You've managed to drastically reduce the number of allocations. Here's the output benchstat which shows the difference in percentage.

Nice find out error in the benchmark test case as well.

goos: darwin
goarch: arm64
pkg: github.com/lithammer/shortuuid/v4
cpu: Apple M2
           │   old.txt    │               new.txt               │
           │    sec/op    │   sec/op     vs base                │
UUID-8       2782.5n ± 1%   479.6n ± 0%  -82.77% (p=0.000 n=10)
Encoding-8   2438.5n ± 2%   216.5n ± 2%  -91.12% (p=0.000 n=10)
Decoding-8    368.6n ± 0%   266.9n ± 0%  -27.58% (p=0.000 n=10)
geomean       1.357µ        302.6n       -77.71%

           │   old.txt    │                 new.txt                 │
           │     B/op     │    B/op     vs base                     │
UUID-8       2650.00 ± 0%   40.00 ± 0%   -98.49% (p=0.000 n=10)
Encoding-8   2640.00 ± 5%   24.00 ± 0%   -99.09% (p=0.000 n=10)
Decoding-8     152.0 ± 0%     0.0 ± 0%  -100.00% (p=0.000 n=10)
geomean       1020.7                    ?                       ¹ ²
¹ summaries must be >0 to compute geomean
² ratios must be >0 to compute geomean

           │   old.txt   │                 new.txt                 │
           │  allocs/op  │ allocs/op   vs base                     │
UUID-8       98.000 ± 0%   2.000 ± 0%   -97.96% (p=0.000 n=10)
Encoding-8   98.000 ± 4%   1.000 ± 0%   -98.98% (p=0.000 n=10)
Decoding-8    5.000 ± 0%   0.000 ± 0%  -100.00% (p=0.000 n=10)
geomean       36.35                    ?                       ¹ ²
¹ summaries must be >0 to compute geomean
² ratios must be >0 to compute geomean

[lithammer]

Included in v4.1.0. Thanks for your contribution!

[anatoly-kussul]

Thanks for such quick approval and release!