CompactBlank cop raising an offense when using delete_if

[Benaaaaa]

Currently there is an existing cop that tells us to use compact_blank, compact_blank! instead of reject, reject! respectively when deleting blanks.
There is another option to be used and that is delete_if. So instead of doing this:
{ a: 'a', b: 'b', c: 'c', d: {}, e: {}, f: [] }.delete_if{ |_k, v| v.blank? }
The cop should recommend doing this:
{ a: 'a', b: 'b', c: 'c', d: {}, e: {}, f: [] }.compact_blank!
It works virtually the same as reject! but is not considered in the cop. I have run some local bechmarking to compare reject! and delete_if to see if maybe one should be favored over the other:

    n = 5_000_000
    Benchmark.bm do |x|
      x.report('delete_if') { n.times { { a: 'a', b: 'b', c: 'c', d: {}, e: {}, f: [] }.delete_if { |_k, v| v.blank? } } }
      x.report('reject!') { n.times { { a: 'a', b: 'b', c: 'c', d: {}, e: {}, f: [] }.reject!{ |_k, v|  v.blank? } } }
    end

This is the end result:

       user        system      total        real
  delete_if       5.963637   0.054312   6.017949 (  6.208843)
  reject!         5.812006   0.049361   5.861367 (  6.132459)

As you can see the run time is almost the same ( even when repeating the proccess 10 times the numbers were very similar)
And so because of the results I recommend adding the situation where using delete_if should raise an offense as well.
If need be I could work on the cop myself

This is the output of rubocop -V:

• rubocop-infinum 0.6.0
• rubocop-performance 1.13.2
• rubocop-rails 2.13.2
• rubocop-rspec 2.8.0

[koic]

Thank you for the feedback!
delete_if always returns self, but reject! returns self if one or more elements are deleted, nil if no element is deleted. So they have different behaviors.

And compact_blank! has different implementations for ActionController::Parameters, Array, and Hash. If the cop makes a mistake, auto-corrected code may get unexpected behavior.

I'll take the issue to update the cop implementation and documentation.

[Benaaaaa]

Oh, thanks for this explanation, I had no idea that's the difference in behavior. Great work with dealing the problem so fast 👍