Remove paths part size hash from Matcher

[matthewmcgarvey]

Summary

Refer to #27 for a more detailed explanation of why this change is being made, but TLDR is that this speeds up the case of finding a matching request without sacrificing behavior.

This benchmark update is to better compare the code when it is warmed up. It runs the router much more but fewer times.

Benchmark (with --release)

• Before: 224 ms
• After: 212 ms
• Difference: 12 ms
• Improvement: 5%

[paulcsmith]

Will take a look at this soon! One thing to note is that I think those benchmarks are without the --release flag. Sometimes in release mode things can change a lot. So I think it'd be good to check the previous merged change and this one with --release to make sure the numbers really are better. Maybe they're even bigger than we previously thought!

EDIT: One more thought, maybe it should run fewer times, but for much longer. That way it has some time to warm up and may produce more accurate results:

elapsed_times = [] of Time::Span
10.times do
  elapsed = Time.measure do
    100_000.times do
      router.match!("post", "/users")
      router.match!("get", "/users/1")
      router.match!("delete", "/users/1")
      router.match!("put", "/users/1")
      router.match!("get", "/users/1/edit")
      router.match!("get", "/users/1/new")
    end
  end
  elapsed_times << elapsed
end

[matthewmcgarvey]

@paulcsmith Good call! I updated the benchmark and re-ran with release flag and am seeing very different results. I updated the description with the changes.

[matthewmcgarvey]

@paulcsmith Is this ready to be merged?

[jwoertink]

This looks good to me. I think we'll just have to try it out some apps to make sure we're catching all of the different use cases.

[paulcsmith]

I think this is good to go. Would love to try this out on some apps, but I think the specs cover everything so this should be good!