Style/RedundantRegexpCharacterClass flags single characters in interpolations

[beechtom]

Expected behavior

The Style/RedundantRegexpCharacterClass should not flag or fix single characters in brackets in interpolations. Interpolations with bracket notation are typically used to access objects and should not be treated as single-element classes.

Actual behavior

The Style/RedundantRegexpCharacterClass flags and fixes single characters in brackets in interpolations. Interpolations with bracket notation is viewed as a single-element class and has the brackets removed. This results in the Lint/Syntax cop flagging the change due to unrecognized characters in the interpolation. See below.

Steps to reproduce the problem

Take this trivial script:

string = "one.two.three"
regexp = /#{string.split('.')[0]}/

puts regexp.match('one')

The expected behavior is that the regexp should resolve to /one/. However, running the Style/RedundantRegexpCharacterClass will flag [0] as a single-element class even though it is valid notation for accessing the first element of an array. Attempting to auto-correct with this cop results in the following change to the file:

string = "one.two.three"
regexp = /#{string.split('.')0}/

puts regexp.match('one')

The interpolation string.split('.')0 has invalid syntax and results in the Lint/Syntax cop flagging the change due to an unexpected integer:

E: Lint/Syntax: unexpected token tINTEGER
(Using Ruby 2.5 parser; configure using TargetRubyVersion parameter, under AllCops)
      regex = /#{string.split('.')0}/

RuboCop version

$ rubocop -V
0.85.0 (using Parser 2.7.1.3, rubocop-ast 0.0.3, running on ruby 2.6.3 x86_64-darwin19)