Autocorrection of Lint/LiteralInInterpolation changes code behaviour in regexes

[patrick-gleeson]

Rubocop generally auto-replaces "#{'a'}" with "a" as a response to the "Lint/LiteralInInterpolation: Literal interpolation detected." offence. However, when it comes to regex capture group literals, this changes the behaviour of the code, making this auto-correction unsafe

---

Expected behavior

When running rubocop -a on this line of code:

x = "ABC".gsub(/(A)(B)(C)/, "D#{'\2'}F")

The resulting code should look like:

x = "ABC".gsub(/(A)(B)(C)/, "D\\2F")

In both cases, the value of x is "DBF"

Actual behavior

When running rubocop -a on this line of code:

x = "ABC".gsub(/(A)(B)(C)/, "D#{'\2'}F")

The resulting code actually looks like:

x = "ABC".gsub(/(A)(B)(C)/, "D\2F")

The value of x is now "D\u0002F"

Steps to reproduce the problem

Put this in a file:

"ABC".gsub(/(A)(B)(C)/, "D#{'\2'}F")

run rubocop -a /path/to/file.rb

RuboCop version

$ [bundle exec] rubocop -V
0.65.0 (using Parser 2.6.0.0, running on ruby 2.5.3 x86_64-darwin17)

[rrosenblum]

Does anyone have thoughts on the best way to fix this? Ignore the offenses if they are an argument to sub, gsub, and similar methods? Do not register an offense if the string contains \<number>?

[hoshinotsuyoshi]

@rrosenblum

I think that there is no need to treat #gsub, any methods, or \<number> specially.

The problem is that auto-correcting results in wrong value replacing; like "\\2" -> "\2".

---

I think that RuboCop::Cop::Util#escape_string will fit to fix that...

--- a/lib/rubocop/cop/lint/literal_in_interpolation.rb
+++ b/lib/rubocop/cop/lint/literal_in_interpolation.rb
@@ -55,7 +55,7 @@ module RuboCop
           when :float
             node.children.last.to_f.to_s
           when :str
-            node.children.last
+            escape_string(node.children.last)
           when :sym
             autocorrected_value_for_symbol(node)
           else

I will investigate later.

Escaping is difficult...

[hoshinotsuyoshi]

I think that RuboCop::Cop::Util#escape_string will fit to fix that...

Sorry, I found that it breaks other behaviors. 😭

Forget that...

memo:

"foo#{"\n"}bar"

now latest rubocop (v0.67.2) auto-corrects to that

"foo
bar"

but using RuboCop::Cop::Util#escape_string breaks this:

"foo\nbar"

[hoshinotsuyoshi]

But I believe that the current behavior below is simply a bug.

# a.rb
"foo#{'\n'}bar"

rubocop (v0.67.2) corrects

rubocop -a a.rb --only Lint/LiteralInInterpolation
Inspecting 1 file
W

Offenses:

a.rb:3:7: W: [Corrected] Lint/LiteralInInterpolation: Literal interpolation detected.
"foo#{'\n'}bar"
      ^^^^

1 file inspected, 1 offense detected, 1 offense corrected

results in:

# a.rb
"foo\nbar"

---

I want to fix this problem.

[patrick-gleeson]

@hoshinotsuyoshi you are a true pillar of the community. Thank you for fixing this!