kind_of? vs is_a?

[anthonyryan1]

My understanding is that the kind_of? and is_a? methods are exactly the same.

I would like to propose defining a guideline for the use of theses two methods. I recently encountered code that used both of these in random places, and felt like it harmed the quality of the code. Upon consulting this guide I felt this was something that could be improved upon.

I would propose:
kind_of? makes sense when extending classes: cat.kind_of?(Animal)
is_a? is preferable with all built-in types: input.is_a?(Hash)

[eclubb]

is_a? matches on an exact class name, while kind_of? will match the given class and any sub-classes. Because of this, I tend to use kind_of? almost exclusively. This lets me use a child class in place of its parent class without breaking the test.

[anthonyryan1]

Are you certain? The documentation appears to indicate that this behavior is identical: http://ruby-doc.org/core-2.1.1/Object.html#method-i-is_a-3F

[eclubb]

Hrm... I was certain, but it appears my memory has failed me. I ran into an issue a long time ago that caused me to prefer kind_of?. Might have been trying to use instance_of? or some such. Anyhow I take back my original comment. Apologies.

[bbatsov]

I guess someone should measure what's more popular in the wild. I have a feeling it will be is_a?.

[NARKOZ]

The method is_a? is used before class names starting with a vowel sound

0.is_a? Fixnum

Where the class name begins with a consonant sound, kind_of? is used

0.kind_of? Array

[sos4nt]

Comment from http://stackoverflow.com/questions/3893278/ruby-kind-of-vs-instance-of-vs-is-a#comment4147181_3893305

It just reads better sometimes. Think @honda.kind_of? Car and @person.is_a? Administrator, Ruby's all about the aesthetics. In fact, notice the grammatical error... with active support you can write @person.is_an? Administrator :)... That might have made it into Ruby core by now, actually.

[bbatsov]

@NARKOZ I guess you meant to say the opposite. Even so, given Ruby's authors total disregard for rules of English such suggestions seem pretty far fetched.

[bbatsov]

Has anyone looked into this further?

[fuadsaud]

I agree with the comment quoted by @sos4nt. It's the same as with inject/reduce. I'm not sure if a guideline can be defined easily in this case.

[jdickey]

I'd tend to agree with @fuadsaud (and hence presumably with @sos4nt). Given that, it's hard to make a rule that's both unambiguous and comprehensively correct, isn't it, and that would cause problems for automation. Not saying the Style Guide shouldn't say something about it, but enforcing the convention with RuboCop would be an interesting challenge, yes?

[anthonyryan1]

I have looked into this more and the move I've looked the more I'm observing different projects using different rules to distinguish when one or the other is used. Sometimes there's no pattern (which is what bothered me initially into creating this).

Given the validity of all the examples and rules brought forward in this issue, I agree it would be too hard to define a rule that would universally benefit people.

[bbatsov]

@jdickey Doing a cop for that should be pretty straight-forward. Guess you can open a ticket so we won't forget about this.