Update: prefer-rest-params relax for member accesses (fixes #5990)

[mysticatea]

Fixes #5990.

This PR makes prefer-rest-params rule relaxing.
New behavior ignores arguments which is accessing non-computed property (e.g. arguments.length).

• Ignores:
  • arguments.length
  • arguments.callee
• Still errors:
  • arguments
  • foo(arguments)
  • arguments[0]
  • arguments[i]
  • arguments[Symbol.iterator]

[mention-bot]

@mysticatea, thanks for your PR! By analyzing the annotation information on this pull request, we identified @vitorbal, @nzakas and @lencioni to be potential reviewers

[eslintbot]

LGTM

[platinumazure]

What about arguments["length"] (with a string literal)? Should we relax the rule in that case?

[mysticatea]

This will warn arguments["length"] since this distinguishes by computed or not. We can avoid warning by arguments.length, so I thought this does not need a complex check (e.g. arguments["0"] is same as arguments[0]).

[platinumazure]

LGTM except:

• Left a question about whether arguments["length"] (computed property, but with string literal) should also be relaxed under this rule.
• AppVeyor failed-- probably due to the too many arguments issue that gyandeeps fixed. Hopefully if this is rebased or updated, we'll see AppVeyor pass.

Thanks for submitting this PR!

[eslintbot]

LGTM

[mysticatea]

Rebased for #6870.

[vitorbal]

LGTM. I dont think its a big deal that it wont relax the case arguments["length"], eapecially if it lets us avoid some added code complexity.
Waiting for other to review though, as that's just my opinion!

[platinumazure]

I don't feel strongly about it, and this is a positive incremental change, so LGTM as well.

[ilyavolodin]

LGTM