Hello, As I understand it, in Smalltalk, the instance variables of a class C are "protected" - able to be referenced by methods of C or its subclasses, but not by other objects. This is a useful feature as it clearly points out which pieces of data are not available to other objects, and thereby simplifies code. However, I am often unsure of whether to use this feature or not, as it conflicts in my mind with the practice of using accessor methods. I like accessor methods because they make it easy to change behavior later -- If I have a dozen calls to an accessor method, then I only need to change it in one place. If these were instance variable references, I would have to do more work. The disadvantage of accessor methods is that they obscure the protected status of data -- it becomes unclear how protected an instance variable is meant to be. (Accessors also make it harder for me to find users of the data when browsing, when there are senders from totally unrelated classes, although I suspect I have just not figured out how to browse scoped in the right way for this). It occurs to me that a tool could be (easily?) developed that would solve this problem. It would take existing variable references and turn them into calls to accessor methods. That way, I could have protection when I want it, and easy ability to change code as well. Does something like this exist, or is it feasible to build? How do others think about this issue? -Lyn |
Hi,
> On 30 Sep 2015, at 21:35, Lyn Headley <[hidden email]> wrote: > > Hello, > > As I understand it, in Smalltalk, the instance variables of a class C are "protected" - able to be referenced by methods of C or its subclasses, but not by other objects. This is a useful feature as it clearly points out which pieces of data are not available to other objects, and thereby simplifies code. > > However, I am often unsure of whether to use this feature or not, as it conflicts in my mind with the practice of using accessor methods. I like accessor methods because they make it easy to change behavior later -- If I have a dozen calls to an accessor method, then I only need to change it in one place. If these were instance variable references, I would have to do more work. The disadvantage of accessor methods is that they obscure the protected status of data -- it becomes unclear how protected an instance variable is meant to be. (Accessors also make it harder for me to find users of the data when browsing, when there are senders from totally unrelated classes, although I suspect I have just not figured out how to browse scoped in the right way for this). > > It occurs to me that a tool could be (easily?) developed that would solve this problem. It would take existing variable references and turn them into calls to accessor methods. That way, I could have protection when I want it, and easy ability to change code as well. Does something like this exist, or is it feasible to build? > > How do others think about this issue? > > -Lyn I think that with the current options in the tools (auto generation of accessors based on instance variables, checking references to instance variables, renaming either accessors or instance variables, syntax coloring) the difference between the two is becoming somewhat smaller. I used to consistently use accessors for the reasons you describe, but lately I find it more important to think as object oriented as possible, to maximise encapsulation and minimise the interface to the outside world. Now, I only add accessors if that really makes sense, the lack of accessors should then signal the private nature of the instance variable. Sven |
In reply to this post by laheadle
Hi,
This is a bugbear for me. On 2015/09/30 09:35 PM, Lyn Headley wrote: > Hello, > > As I understand it, in Smalltalk, the instance variables of a class C > are "protected" - able to be referenced by methods of C or its > subclasses, but not by other objects. This is a useful feature as it > clearly points out which pieces of data are not available to other > objects, and thereby simplifies code. > > However, I am often unsure of whether to use this feature or not, as > it conflicts in my mind with the practice of using accessor methods. I > like accessor methods because they make it easy to change behavior > later -- If I have a dozen calls to an accessor method, then I only > need to change it in one place. If these were instance variable > references, I would have to do more work. The disadvantage of accessor > methods is that they obscure the protected status of data -- it > becomes unclear how protected an instance variable is meant to be. > (Accessors also make it harder for me to find users of the data when > browsing, when there are senders from totally unrelated classes, > although I suspect I have just not figured out how to browse scoped in > the right way for this). > > It occurs to me that a tool could be (easily?) developed that would > solve this problem. It would take existing variable references and > turn them into calls to accessor methods. That way, I could have > protection when I want it, and easy ability to change code as well. > Does something like this exist, or is it feasible to build? > > How do others think about this issue? > > -Lyn > encapsulation, they cloud the object interface, and most importantly they allow bugs to be introduced by undocumented modification of the object's internal state. Nobody has shown me a good reason for their existence and I never use them except where I'm forced to use them in Spec and Glorp. Craig PS: I'm forced to use them in C# as well (my day job). Because in C# it's only way to promote a local variable to a property which is required by most frameworks. |
Nobody has shown me a good reason for their existence and I never use them except where I'm forced to use them in Spec How so? Spec requires only read accessor for your (sub)components. |
In reply to this post by Sven Van Caekenberghe-2
Hi lyn
I have the same experience than sven. Stef -Lyn > I think that with the current options in the tools (auto generation of accessors based on instance variables, checking references to instance variables, renaming either accessors or instance variables, syntax coloring) the difference between the two is becoming somewhat smaller. > > I used to consistently use accessors for the reasons you describe, but lately I find it more important to think as object oriented as possible, to maximise encapsulation and minimise the interface to the outside world. Now, I only add accessors if that really makes sense, the lack of accessors should then signal the private nature of the instance variable. > > Sven > > > |
Free forum by Nabble | Edit this page |