Strange behavior for inline blocks about rubocop HOT 12 CLOSED

I got the the same problem using String interpolation inside a block. So I guess there should be made some changes on how interpolation braces are identified.

Ehm? What comma? I've already had that comma here.

Commas in [1,2,3], It should be [1, 2, 3]

from rubocop.

@ismaelga Were you also using Ruby 2.0? My suspicion is that these errors are all related to differences in the output from Ripper in 1.9 and 2.0. They are bugs nonetheless, but it would be interesting to see if they appear in Ruby 1.9 or not.

from rubocop.

@jonas054 Yeah. I just tested on 1.9.3 and it worked there. So this seems to be only on 2.0.0

from rubocop.

And the output from Ripper is the same on both. So it should be something else

from rubocop.

Minimal ruby source:

foo { "#{n}" }

I pry'ed blocks.rb, and this is what I get for @reverse_correlations on 1.9.3:

{70102431113960=>[0], 70102431112780=>[2, 10], 70102431121220=>[6]}

Here's the matching @correlations:

{0=>[:program, :method_add_block, :method_add_arg, :fcall, :@ident],
 2=>[:program, :method_add_block, :brace_block],
 10=>[:program, :method_add_block, :brace_block],
 6=>
  [:program,
   :method_add_block,
   :brace_block,
   :string_literal,
   :string_content,
   :string_embexpr,
   :vcall,
   :@ident]}

while this is what I get for @reverse_correlations on 2.0:

{70225979303160=>[0], 70225979322180=>[2], 70225979338680=>[6]}

and the matching @correlations

{0=>[:program, :method_add_block, :method_add_arg, :fcall, :@ident],
 2=>[:program, :method_add_block, :brace_block],
 6=>
  [:program,
   :method_add_block,
   :brace_block,
   :string_literal,
   :string_content,
   :string_embexpr,
   :vcall,
   :@ident]}

So #inspect and #check look correct, just fed with bad data.

Next stop, Cop::Grammar#correlate found via cli.rb#L61 and #L91

from rubocop.

In Grammar#add_matching_rbrace, @tokens_without_pos contains:

[[:on_ident, "foo"],
 [:on_sp, " "],
 [:on_lbrace, "{"],
 [:on_sp, " "],
 [:on_tstring_beg, "\""],
 [:on_embexpr_beg, "\#{"],
 [:on_ident, "n"],
 [:on_embexpr_end, "}"],
 [:on_tstring_end, "\""],
 [:on_sp, " "],
 [:on_embexpr_end, "}"]]

Since it misses :on_rbrace, it won't be able to match.

in Grammar#initialize, @tokens_without_pos has :on_rbrace before the call to process_embedded_expressions and lacks :on_rbrace right after. This is, I gather, the crux of the problem.

from rubocop.

I focused too much on the last brace and finally found the cause:

On ruby 2.0, this is what @tokens_without_pos looks like right after being set:

[[:on_ident, "foo"],
 [:on_sp, " "],
 [:on_lbrace, "{"],
 [:on_sp, " "],
 [:on_tstring_beg, "\""],
 [:on_embexpr_beg, "\#{"],
 [:on_ident, "n"],
 [:on_embexpr_end, "}"],
 [:on_tstring_end, "\""],
 [:on_sp, " "],
 [:on_rbrace, "}"]]

Notice how @tokens_without_pos[-4] is not a :on_rbrace, as it is supposed to be according to the process_embedded_expressions code and comment.

It follows that:

• when :on_rbrace can be changed to when :on_rbrace, :on_embexpr_end
• the call to process_embedded_expressions is only necessary for 1.9.3 and below.

from rubocop.

Fixed in PR #46.

from rubocop.

That's some impressive bug hunting you did, @lloeki! You fixed this issue just as I was about to take a closer look at it. Thanks!

from rubocop.

@lloeki awesome! 👍

from rubocop.

Good job!

from rubocop.

RuboCop 0.4.1 is now out and features that fix!

from rubocop.