Add defined? -> instance_variable_defined? mutation

[dgollahon]

• Mutates expressions where defined? is being used to check specifically for an instance variable such as defined?(@a) to the more narrow instance_variable_defined?(:@a) instead.
• Closes Add mutation from defined(@ivar) to instance_variable_defined?(:ivar). #1138.

[dgollahon]

I would do this as a separate PR, but should the defined? mutator also emit false? It's interesting that it is only true. Is there a reason for that?

[mbj]

I think this was a deliberate decision as singleton_mutations never emits false:

irb(main):001:0> defined?(@a)
=> nil
irb(main):002:0> @a = "foo"
=> "foo"
irb(main):003:0> defined?(@a)
=> "instance-variable"

emitting true is also wrong in that light ;) I'll add a commit to improve this.

[dgollahon]

Ah, I guess since we are emitting nil we already have something falsey.

@mbj I actually think true was a good mutation--don't we normally replace predicates with true and false? (or maybe there is an issue for this somewhere.

Or are you thinking it's just too hard to cover normal defined? mutations because it's hard to have both states?

[dgollahon]

oh! nevermind. I see that defined? never returns true normally. Ok, cool, good call, ignore me. I misunderstood what you were saying for a moment. Nice improvement.

[mbj]

no worries, defined? is weird in other ways.

It actually gives information about the AST, but when the AST references a constant, than it also checks the presence of that constant. Ruby....

[dgollahon]

My editor trimmed some trailing whitespace. I can remove this from the commit if you want but also seems like it is worth fixing.

[mbj]

Added an extra commit, and my editors whitespace trimming broke ;) Good catch.