Neal Ford made a blog post on meta-programming, Ruby, and Smalltalk recently. It caused some discussion from
long-time Smalltalker, James Robertson, and
Rubyist, Glenn Vandenburg. Even
Avi Bryant got in the discussion. There was a lot of talking past one another on the issues of code generation in Ruby and Smalltalk.
I'll start with a quote from Neal Ford:
the Smalltalk version is a great example of accidental complexity, not essential complexity.
There is no example given of what the "accidental complexity" was in Smalltalk. But, let's look at an example that Rubyists are familiar with and then I'll go forward:
class Person
attr_accessor :name
end
This will generate the accessors :name and :name= at run-time. It's a nice way to describe a public field. There's even read and write only variations. It allows you to describe your intent of the field. Now, in Smalltalk you could do the same thing with class-side methods. Like so:
attributeAccessors
^#(name)
This could add the instance variable (if it didn't exist), and generate the getters and setters when the class is initialized. It is true in this case, we would use tools to do the generation and generate the code before hand. I think this is what Neal was talking about accidental complexity because we are adding methods that we later have to browse through. But, this is where they are mistaken.
Smalltalk has categories and usually, these code generated methods are placed in them. Accessors are generally placed in their own category so do not add to the cognitive friction. In fact, when I use code generation, I group the methods in a category with "AUTO" in it. This is so that I can later delete those methods and start over. Also, it allows me to not have to look at them when looking at the more important methods of the class.
There is one last way to do things like this in Smalltalk, but is rare. Classes are created in Smalltalk by sending messages. It's not some hard-coded construct in the language. It's a message. We can create our own messages to create our own classes. Here's what the example could look like in Squeak:
Object subclass: #Person
attributeAccessors: 'names'
classVariables: ''
poolDictionaries: ''
category: 'MetaExample'
And lastly, there is a initialize method for all Smalltalk classes that would give a more Ruby feel:
initialize
self attributeAccessors: 'name'
Now, with all of that said, these are just examples to show what is possible. It's all a matter of taste and what you are trying to accomplish to which method you choose.
finally, my mouth dropped when I saw this in the same blog post:
Smalltalk had (and has) an awesome environment, including incredible tool support. Because the tool is pervasive (literally part of the project itself), Smalltalkers generally shied away from the kind of meta-programming described above because you have to build tool support along with the meta-programmed code. This is a conscious trade off. One of the best things about the Ruby world (up until now) is the lack of tool support, meaning that the tools never constrained (either literally or influentially) what you can do. Fortunately, this level of power in Ruby is pretty ingrained (look at Rails for lots of examples), so even when the tools finally come out, they need to support the incredible things you can do in Ruby.
I will answer this as bluntly and respectively as I can. Smalltalkers have NEVER shied away from meta-programming because of having to build a tool. Most Smalltalkers have a bag of tools that they use and add to their environment. Tools are so easy to write in Smalltalk that most programmers in it have at some point written one or two to help them. It's trivial to add functionality to the browser and inspectors. Generally, Smalltalkers only stay away from things that are hard to debug (ala method_missing, doesNotUnderstand:), but do them when necessary. Readability is always the utmost importance to Smalltalkers. Abbreviations are generally frowned upon even.
The last thing I want to remark on is the quote in bold above. I'll repeat it here because it makes me both shocked and sad:
One of the best things about the Ruby world (up until now) is the lack of tool support
I hardly call that a strength. Really. Why would I want to go back to bear skin and stone development? I have finally become productive in Java because of Eclipse (code browsing, auto-format, code completion, refactoring, etc). Why would I want to go back to command line brute force methods? I keep hoping for a great Ruby IDE (some are close like Ruby's Eclipse plug-in, Arachno, etc). Besides, I think while tools like ri and rdoc are nice, I want to look at source code and it's difficult to find the definitions of methods (without using grep) when going through new source code. A friend once told me he never trusted a language you couldn't write an IDE for it. There's truth to that. I never got FreeRIDE to last longer than a few minutes of development. Tools and easy to read syntax are necessary.
Ruby has great potential, but the syntax needs to be heavily reafactored (%w while being nice shorthand is unreadable) and made more consistent (blocks do not take the same types of arguments as methods, I can not send a block with & into a block example, lambda {|&block| block.call } gives a compile error). But, those are my gripes and I do still like Ruby. I think boasting that your language has no tools is short sighted at best.
I think both Smalltalk and Ruby developers could learn a lot from one another. So, if any Ruby developer has questions about Smaltallk, please feel free to email me. I will even extend the same to Smalltalk developers curious about Ruby. The point of this post was mainly to educate and inspire both Rubyists and Smalltalkers to be better.
Posts
