# Attr\_\*, \*variables\_\* **URL:** \ **Category:** Ruby\ **Created:** [September 6, 2008, 5:03pm UTC](https://www.ruby-forum.com/t/attr-variables/147890 "2008-09-06T17:03:21Z")\ **Posts on this page:** 5\ **Page:** 1
**Author:** ![pedro-silva](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/p/ee7513/32.png) [@pedro-silva](https://www.ruby-forum.com/u/pedro-silva)\ **Post date:** [September 6, 2008, 5:03pm UTC](https://www.ruby-forum.com/t/attr-variables/147890/1 "2008-09-06T17:03:21Z")
It wouldn’t be nice if all these methods respect a unique arguments form. attr\_\* methods receive a symbol without the @ prefix, but _variables_ methods receive @ prefixed symbols. Even class\_variables or instance\_variables return an array of strings, wouldn’t be better if they return an array of symbols? I prefer the attr\* notation, but what really matters is to have a unique invocation form. Does anyone think the same? Pedro. P.S - I’ve consulted [RDoc Documentation](http://www.ruby-doc.org/core-1.9/index.html) for Ruby 1.9 core reference. ---
**Author:** ![pedro-silva](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/p/ee7513/32.png) [@pedro-silva](https://www.ruby-forum.com/u/pedro-silva)\ **Post date:** [September 6, 2008, 5:28pm UTC](https://www.ruby-forum.com/t/attr-variables/147890/2 "2008-09-06T17:28:22Z")
HI – On Sat, 6 Sep 2008, Pedro S. wrote: > It wouldn’t be nice if all these methods respect a unique arguments > form. attr\_\* methods receive a symbol without the @ prefix, but > _variables_ methods receive @ prefixed symbols. Even class\_variables > or instance\_variables return an array of strings, wouldn’t be better if > they return an array of symbols? I guess so. I don’t really care too much. > I prefer the attr\* notation, but what really matters is to have a unique > invocation form. Does anyone think the same? I’m not sure what problem you’re trying to solve. Why should methods that manipulate instance variables (@var), class variables (@@var), and method-name roots (attr\_accessor :m) all conform to the same format, when the identifiers themselves don’t? David ---
**Author:** ![pedro-silva](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/p/ee7513/32.png) [@pedro-silva](https://www.ruby-forum.com/u/pedro-silva)\ **Post date:** [September 6, 2008, 6:02pm UTC](https://www.ruby-forum.com/t/attr-variables/147890/3 "2008-09-06T18:02:52Z")
Hi David, > I’m not sure what problem you’re trying to solve. Why should methods > that manipulate instance variables (@var), class variables (@@var), > and method-name roots (attr\_accessor :m) all conform to the same > format, when the identifiers themselves don’t? Of course this is a minor detail in how methods process their arguments. What I was trying to say is that when you use for instance class\_variable\_set, you know that you’re dealing with class variables so why it’s necessary to prefix the regular name with @@. The same applies for instance\_variables. A convention avoids the need to ask yourself if any particular method has or hasn’t the variable name prefixed by anything. I know that this is just a small detail but I think that the result with be more predictable if there’s a convention for kinds of methods. Pedro. ---
**Author:** ![pedro-silva](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/p/ee7513/32.png) [@pedro-silva](https://www.ruby-forum.com/u/pedro-silva)\ **Post date:** [September 6, 2008, 7:41pm UTC](https://www.ruby-forum.com/t/attr-variables/147890/4 "2008-09-06T19:41:52Z")
Pedro S. wrote: > I prefer the attr\* notation, but what really matters is to have a unique > invocation form. Does anyone think the same? > (Did you mean ‘uniform’? Or I don’t understand this sentence.) I agree that it is good to have one standard, but I think that this is the case here, more or less. And the standard is to give prefixes when we perform an action on a prefixed variable, and in attr\_\* arguments are not prefixed simply because this action (defining accessors) involves both the variable, and the newly created method, of which the latter is not prefixed. So they chose that you sort of specify the method name and not the attribute name. I agree that writing attr\_accessor :@a could be regarded as correct as well, because this would let anybody use the form they find more coherent with the rest of the methods, and it would seem more clear to a beginner (I remember my confusion when I first saw it). TPR. ---
**Author:** ![pedro-silva](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/p/ee7513/32.png) [@pedro-silva](https://www.ruby-forum.com/u/pedro-silva)\ **Post date:** [September 6, 2008, 7:53pm UTC](https://www.ruby-forum.com/t/attr-variables/147890/5 "2008-09-06T19:53:01Z")
Hi – On Sun, 7 Sep 2008, Thomas B. wrote: > not prefixed. So they chose that you sort of specify the method name and > not the attribute name. > > I agree that writing attr\_accessor :@a could be regarded as correct as > well, because this would let anybody use the form they find more > coherent with the rest of the methods, and it would seem more clear to a > beginner (I remember my confusion when I first saw it). attr\_accessor :@a is ultimately confusing, though, because the attribute is not the instance variable itself; it’s implemented with the instance variable. I know that sounds like splitting hairs, but it’s important. Attribute is a higher-level concept than instance variable, and it’s pieced together in Ruby from instance variables and methods – so :@a would be mixing in an implementation detail (as would :“a()”, for example, even though it might evoke the method-creating side of the attr\_\* family). David