nice implementation Parker 👍 I did come across this transformation while reading up for another PR, but never bothered to bench an implementation..

If you have the time can you please benchmark using apples[0] and oranges[0] too?
Array[0] is a tad faster than Array.first

If you have the time can you please benchmark using apples[0] and oranges[0] too?
Array[0] is a tad faster than Array.first

Updated. I had to ensure I was copying the docs array for each new try (otherwise everything would be faster after the first iteration which sorted in place). Now showing much higher gains: 7x faster and 1.75x faster.

Now showing much higher gains: 7x faster and 1.75x faster

WOW!! Très cool!!!

@parkr IMO, this change should ideally be made into a public Utility method which can then be adapted for your suggestion to #5904

# lib/jekyll/utils.rb

def schwartzian_transform(input, property, order)
  [...]
end

this change should ideally be made into a public Utility method

I don't think we can do that, because each implementation will pull the property from input through different means each time, e.g. using input.data[property] or input[property] or item_property(input, property) in the filters.

What happens if we switch to Enumerable#sort_by directly? What's the reason that I'm missing that keeps us from doing:

input.sort_by { |item| item_property(item, property) }

What happens if we switch to Enumerable#sort_by directly?

@mattr- : @parkr added a comment about this in the benchmark script, sort_by doesn't like nils

["foo", nil, "bar"].sort_by{|el| el}
# => ArgumentError: comparison of String with nil failed

@parkr added a comment about this in the benchmark script

Always good to know that I can't read very well. 🤣 Thanks @Crunch09!

@jekyllbot: merge +minor