Compute document's relative_path faster

[ashmaroli]

Use String#sub instead of Pathutil#relative_path_from which uses regexes internally

Benchmark script:

#!/usr/bin/env ruby

require 'benchmark/ips'
require 'pathutil'

DOC_PATH = File.join(File.expand_path(__dir__), "_puppies", "rover.md")
COL_PATH = File.join(File.expand_path(__dir__), "_puppies")


def pathutil_relative
  Pathutil.new(DOC_PATH).relative_path_from(COL_PATH).to_s
end

def native_relative
  DOC_PATH.sub("#{COL_PATH}/", "")
end

if pathutil_relative == native_relative
  Benchmark.ips do |x|
    x.report("pathutil") { pathutil_relative }
    x.report("native")   { native_relative }
    x.compare!
  end
else
  print "PATHUTIL: "
  puts pathutil_relative
  print "NATIVE:   "
  puts native_relative
end

Note: Since Document#relative_path is a memoized function, the difference is noticeable only on large sites that containing a huge number of Documents

[pathawks]

Seems reasonable

[parkr]

@ashmaroli Would you mind placing the benchmark script in the PR body into the repo?

~/jekyll/jekyll#master$ be ruby benchmark/pathutil-relative-vs-gsub.rb
Warming up --------------------------------------
            pathutil     5.992k i/100ms
              native    50.261k i/100ms
Calculating -------------------------------------
            pathutil     61.075k (± 8.6%) i/s -    305.592k in   5.042290s
              native    584.400k (± 5.4%) i/s -      2.915M in   5.001884s

Comparison:
              native:   584399.5 i/s
            pathutil:    61074.8 i/s - 9.57x  slower

[ashmaroli]

Would you mind placing the benchmark script in the PR body into the repo?

Sure.. not a problem.. 😃

[DirtyF]

Good catch!

[DirtyF]

@jekyllbot: merge +minor