Style/CommentAnnotation reports only first of multiple offending lines

[jdno]

Good morning everyone,

I stumbled upon some unexpected behavior when working on a project. I created a minimal working example here.

I am working on a class that has multiple # TODO ... annotations. Something like this:

class ToBeDone
  ITEMS = [
    '', # TODO Item 1
    '', # TODO Item 2

    '', # TODO Item 3
  ].freeze
end

When running Rubocop, it reports a Style/CommentAnnotation offense, because the proper style would be # TODO: .... But it does that only for the first and third occurrence, and not for the second one.

jandavid@STS ~/D/j/rubocop-todo-bug> rubocop
Inspecting 2 files
C.

Offenses:

todo.rb:5:11: C: Style/CommentAnnotation: Annotation keywords like TODO should be all upper case, followed by a colon, and a space, then a note describing the problem.
    '', # TODO Item 1
          ^^^^^
todo.rb:8:11: C: Style/CommentAnnotation: Annotation keywords like TODO should be all upper case, followed by a colon, and a space, then a note describing the problem.
    '', # TODO Item 3
          ^^^^^

2 files inspected, 2 offenses detected

Even after fixing the first and third line, the second line is never reported as an offense. This also means that rubicon -a will never pick that line up and fix it automatically.

Expected behavior

When there are Style/CommentAnnotation in multiple lines that follow each other, Rubocop should report a violation for each individual line.

Actual behavior

When there are Style/CommentAnnotation in multiple lines that follow each other, only the first offense is reported.

Steps to reproduce the problem

Check out the minimal working example. Download the Gist as a ZIP, extract the archive, execute bundle if necessary and run rubocop in it.

RuboCop version

Include the output of rubocop -V. Here's an example:

jandavid@STS ~/D/j/rubocop-todo-bug> rubocop -V
0.54.0 (using Parser 2.5.0.4, running on ruby 2.5.0 x86_64-darwin17)

[smackesey]

I can see this is closed but it appears to still be in effect. On rubocop 0.86:

# TODO a
# TODO b
# TODO c

Only marks the first line as problematic.

[dmytro-savochkin]

I can see this is closed but it appears to still be in effect. On rubocop 0.86:

# TODO a
# TODO b
# TODO c

Only marks the first line as problematic.

I can't see a definitive way to detect whether a TODO on the second line is a new TODO or a continuation of the first line. For example, if we relax the code so it finds every first TODO (or any other keyword) on every line then it would break the following example from the specs:

# Example: There are three reviews, with ranks 1, 2, and 3. A new
# review is saved with rank 2. The two reviews that originally had
# ranks 2 and 3 will have their ranks increased to 3 and 4.

Here, it would correct the second line from review to REVIEW: .

IMHO in this case it's more important to stay safe then find more potential offenses. So I'd vote to keep it as it is.

[smackesey]

That makes sense. I think a note should be added to the docs explaining that the rule will only look at keywords in the first line of a paragraph (consecutive series of comment lines), because otherwise the behavior is confusing.