🐛 utils: fix EqualFold and docs

[jfcg]

My original assumptions were wrong, but along the way I improved the tests/benchmarks and fixed a bug in EqualFold.

[ReneWerner87]

Please provide before and after benchmark results.

[ReneWerner87]

sorry but cannot accept
before(master):

Benchmark_EqualFoldBytes/fiber-12         	35822534	        34.25 ns/op	       0 B/op	       0 allocs/op
Benchmark_EqualFoldBytes/fiber-12         	33736502	        34.00 ns/op	       0 B/op	       0 allocs/op
Benchmark_EqualFoldBytes/fiber-12         	35774396	        34.25 ns/op	       0 B/op	       0 allocs/op
Benchmark_EqualFoldBytes/fiber-12         	35762599	        33.62 ns/op	       0 B/op	       0 allocs/op
Benchmark_EqualFoldBytes/default-12       	11068182	       108.7 ns/op	       0 B/op	       0 allocs/op
Benchmark_EqualFoldBytes/default-12       	10881543	       106.4 ns/op	       0 B/op	       0 allocs/op
Benchmark_EqualFoldBytes/default-12       	11284395	       119.6 ns/op	       0 B/op	       0 allocs/op
Benchmark_EqualFoldBytes/default-12       	 9314128	       112.8 ns/op	       0 B/op	       0 allocs/op
Benchmark_EqualFold/fiber-12              	35246314	        38.88 ns/op	       0 B/op	       0 allocs/op
Benchmark_EqualFold/fiber-12              	30589214	        36.25 ns/op	       0 B/op	       0 allocs/op
Benchmark_EqualFold/fiber-12              	31582516	        34.62 ns/op	       0 B/op	       0 allocs/op
Benchmark_EqualFold/fiber-12              	35663390	        33.86 ns/op	       0 B/op	       0 allocs/op
Benchmark_EqualFold/default-12            	12866125	        95.12 ns/op	       0 B/op	       0 allocs/op
Benchmark_EqualFold/default-12            	12907418	        94.02 ns/op	       0 B/op	       0 allocs/op
Benchmark_EqualFold/default-12            	12663051	        94.05 ns/op	       0 B/op	       0 allocs/op
Benchmark_EqualFold/default-12            	12583516	        93.54 ns/op	       0 B/op	       0 allocs/op

after(your branch):

Benchmark_EqualFoldBytes/fiber-12         	13826067	        76.53 ns/op	       0 B/op	       0 allocs/op
Benchmark_EqualFoldBytes/fiber-12         	15962703	        76.39 ns/op	       0 B/op	       0 allocs/op
Benchmark_EqualFoldBytes/fiber-12         	15975374	        75.84 ns/op	       0 B/op	       0 allocs/op
Benchmark_EqualFoldBytes/fiber-12         	15986911	        75.67 ns/op	       0 B/op	       0 allocs/op
Benchmark_EqualFoldBytes/default-12       	 5385681	       223.6 ns/op	       0 B/op	       0 allocs/op
Benchmark_EqualFoldBytes/default-12       	 4988062	       234.4 ns/op	       0 B/op	       0 allocs/op
Benchmark_EqualFoldBytes/default-12       	 5206946	       225.8 ns/op	       0 B/op	       0 allocs/op
Benchmark_EqualFoldBytes/default-12       	 5001892	       226.4 ns/op	       0 B/op	       0 allocs/op
Benchmark_EqualFold/fiber-12              	19036320	        63.21 ns/op	       0 B/op	       0 allocs/op
Benchmark_EqualFold/fiber-12              	19063516	        62.18 ns/op	       0 B/op	       0 allocs/op
Benchmark_EqualFold/fiber-12              	18857341	        62.93 ns/op	       0 B/op	       0 allocs/op
Benchmark_EqualFold/fiber-12              	19338664	        62.21 ns/op	       0 B/op	       0 allocs/op
Benchmark_EqualFold/default-12            	 6329773	       212.9 ns/op	       0 B/op	       0 allocs/op
Benchmark_EqualFold/default-12            	 5200621	       199.7 ns/op	       0 B/op	       0 allocs/op
Benchmark_EqualFold/default-12            	 5531955	       224.3 ns/op	       0 B/op	       0 allocs/op
Benchmark_EqualFold/default-12            	 6175377	       202.6 ns/op	       0 B/op	       0 allocs/op

[ReneWerner87]

i like the comment change and the test change, but not the performance drop

if you can customize that i would merge it

next time please share the benchmark results directly

since fiber stands for speed, we always have to look at the benchmarks, because the users trust us to always get the best possible out of it

[jfcg]

@ReneWerner87, see the definition of largeStr. It is about twice length of the previous input. That's the reason, I have not changed ToUpper and ToLower functions, see the updated patch. The change in EqualFold is mandatory because there was a bug in it. Also see the extra tests for EqualFold.
Thanks..

[ReneWerner87]

okay i see, so in any case 7ns for the fix of the bug come on it

tested it again with the original strings

is there a reason why you do the comparison from back to front ?

[jfcg]

Reverse iteration produces one less opcode on x86 for a good compiler, old trick, but irrelevant for modern CPUs today. It could be either way.

[ReneWerner87]

Can you also create a PR for https://github.com/gofiber/utils

[jfcg]

Hi @ReneWerner87, is gofiber/utils repo an outdated version of fiber/utils directory?

[ReneWerner87]

I had the idea to use the functions which have nothing to do with the web server even if you are in a project without a web server.

[ReneWerner87]

Hi @ReneWerner87, is gofiber/utils repo an outdated version of fiber/utils directory?

Yes

[jfcg]

@ReneWerner87 see the discussion here.