Use criterion for benching

This changes the bencher to criterion!

I noticed you don't run benches on CI, so I didn't add/change it. :) I also didn't change/add any benches, I just wanted to show you this neat library @undef1nd ! :)

Criterion is one of my favorite libraries, and I hope you might enjoy it too! If not, I won't be offended if you close this.

This uses the method described in https://bheisler.github.io/criterion.rs/book/user_guide/benchmarking_with_inputs.html to cover the existing benchmarking cases.

From reading your tests and code, it seems like you may wish to have further work deriving the benchmarks from the fixtures in tests/spec_tests?

As for why you might prefer Criterion to bencher:

You can set and compare baselines: https://bheisler.github.io/criterion.rs/book/user_guide/command_line_options.html#baselines
With gnuplot installed you can get beautiful reports in the target/criterion folder:

The output is a bit cleaner:


test result: ok. 0 passed; 0 failed; 74 ignored; 0 measured; 0 filtered out

 Running target\release\deps\bench-272aa01c5cd0567c.exe
parsing_item/c29tZXZlcnlsb25nc3RyaW5ndmFsdWVyZXByZXNlbnRlZGFzYnl0ZXNhbnNvbWVvdGhlcmxvbmdsaW5l
                    time:   [1.8256 us 1.8272 us 1.8290 us]
                    change: [+0.0503% +0.2067% +0.3452%] (p = 0.01 < 0.05)
                    Change within noise threshold.
Found 2 outliers among 100 measurements (2.00%)
2 (2.00%) high mild

parsing_list/a, abcdefghigklmnoprst, 123456785686457, 99999999999.999, (), ("somelongstringvalue" "a... time: [8.9248 us 8.9528 us 8.9827 us] change: [+75.647% +76.620% +77.587%] (p = 0.00 < 0.05) Performance has regressed. Found 5 outliers among 100 measurements (5.00%) 3 (3.00%) high mild 2 (2.00%) high severe

parsing_dict/a, dict_key2=abcdefghigklmnoprst, dict_key3=123456785686457, dict_key4=("inner-list-mem... time: [5.3266 us 5.3333 us 5.3408 us] change: [-0.3320% +0.0254% +0.3779%] (p = 0.89 > 0.05) No change in performance detected. Found 3 outliers among 100 measurements (3.00%) 2 (2.00%) high mild 1 (1.00%) high severe

serializing_item/c29tZXZlcnlsb25nc3RyaW5ndmFsdWVyZXByZXNlbnRlZGFzYnl0ZXNhbnNvbWVvdGhlcmxvbmdsaW5l time: [1.1187 us 1.1198 us 1.1210 us] change: [+0.0001% +0.1534% +0.3028%] (p = 0.05 < 0.05) Change within noise threshold. Found 2 outliers among 100 measurements (2.00%) 2 (2.00%) high mild

serializing_list/a, abcdefghigklmnoprst, 123456785686457, 99999999999.999, (), ("somelongstringvalue... time: [2.6935 us 2.7006 us 2.7074 us] change: [-1.2719% -0.8292% -0.3762%] (p = 0.00 < 0.05) Change within noise threshold.

serializing_dict/a, dict_key2=abcdefghigklmnoprst, dict_key3=123456785686457, dict_key4=("inner-list... time: [1.6210 us 1.6243 us 1.6279 us] change: [-1.4315% -0.9569% -0.5011%] (p = 0.00 < 0.05) Change within noise threshold. Found 3 outliers among 100 measurements (3.00%) 1 (1.00%) high mild 2 (2.00%) high severe



To try for yourself, install `gnuplot` (optional), then run `cargo bench` and explore `target/criterion/report`!

undef1nd / sfv

Use criterion for benching #72