Here's how the new bits of the visual spec look with this branch:

This is unfortunately another one of those bizarre cases where ruby's parsing behaviour inexplicably depends on what variables happen to be in scope...?

[0: jneen@cinnamon rouge ] -> bugfix.ruby-modulo-spacing $
; ruby -e 'def x(arr); arr; end; puts (x %w]reverse])'
reverse

[0: jneen@cinnamon rouge ] -> bugfix.ruby-modulo-spacing $
; ruby -e 'x=1; w=2; puts (x %w]reverse])'
-e:1: syntax error, unexpected ']', expecting ')'
x=1; w=2; puts (x %w]reverse])

Those examples cleaned up a bit:

def x(arr)
  arr
end

puts (x %w]reverse])
# => prints "reverse"

and

x = 1
w = 2
puts (x %w]reverse])
# => syntax error

A really good read re: the zaniness of ruby's parser: https://whitequark.org/blog/2013/04/01/ruby-hacking-guide-ch-11-finite-state-lexer/

/me buckles up

I FOUND IT BY GOD I FOUND IT

https://github.com/ruby/ruby/blob/master/parse.y#L8966

This is the line where Ruby checks what variables are in scope and changes its parsing behaviour. Found by running ruby -yce on two very similar examples:

x=1; puts (x %w]reverse])

# vs

w=1; puts (x %w]reverse])

and diffing the output.

Anyways, the moral of the story is:

• Ruby lexing will never be perfect. It just won't. We can't afford to keep track of every nuance of block scoping enough to decide whether a variable is in scope without sacrificing performance.
• This patch improves the lexer's behaviour in some important real-world cases, so it helps.

@0x7ffc Echoing @jneen's thanks for submitting the bug report. This will be part of v3.23.0 of Rouge. That's actually scheduled for release later today (or early tomorrow depending on your time zone) 🎉