array assign to memo in reduce errors out

[jhellerstein]

(Courtesy Ben Sklaroff:)

The typical array-assignment syntax for ruby seems to cause errors within the block of a Bloom collection's reduce.

Given that the syntax seems to be perfectly legal Ruby, it's tempting to think that the bug may occur as a result of rewriting somewhere...

require 'rubygems'
require 'bud'

class Test
  include Bud

  state do
    table :a, [:idx]
  end

  bloom do
    a <= [[23], [47]]
    stdio <~ a.reduce({}) do |memo, t|
      memo[[0, 0, 'joe']] ||= 0
      # this works:
      memo[[0,0,'joe']] += 1
      # this doesn't work
      # memo[[0, 0, 'joe']] = memo[[0, 0, 'joe']] + 1
      # nor does the following
      memo[[0,1,'joe']] = 7
      memo
    end.inspected
  end

end

x = Test.new
x.tick

[neilconway]

At first glance looks like a bug in either RubyParser or Ruby2Ruby:

require 'rubygems'
require 'ruby_parser'
require 'ruby2ruby'

src="a = [[23], [47]]
a.reduce({}) do |memo, t|
  memo[[0, 0, 'joe']] ||= 0
  memo[[0, 0, 'joe']] = memo[[0, 0, 'joe']] + 1
  memo
end
"

p = RubyParser.for_current_ruby
ast = p.parse(src)
new_src = Ruby2Ruby.new.process(ast)
puts new_src

Output:

a = [[23], [47]]
a.reduce({}) do |memo, t|
  memo[[0, 0, "joe"]] ||= 0
  memo[0, 0, "joe"] = (memo[[0, 0, "joe"]] + 1)
  memo
end

From looking at the AST it seems more likely to be a RubyParser bug. I'll file a ticket with the upstream folks.