Why I love Smalltalk
This post was extracted from a small talk I gave at Simplificator, where I work, titled “Why I love Smalltalk and Lisp”. There should be another post, “Why I love Lisp” following this one.
After I learned my basic coding skill in more or less traditional languages, like C, C++, Python, there were four languages that really taught me something new. Those languages changed my way of thinking and even if I never use them, they were worth learning. They are:
- Smalltalk
- Lisp
- Erlang
- Haskell
You can probably add Prolog to that list, but I never learned Prolog. This post is about Smalltalk.
My goal is not to teach Smalltalk but to show things that you can do with Smalltalk that you can’t do with any other language (disclaimer: surely other languages can do it, and we’ll call them Smalltalk dialects). Nevertheless I need to show you some basics of the language to be able to show you the good stuff, so here we go, a first program:
1 + 1
That of course, evaluates to 2. If we want to store it in a variable:
m := 1 + 1
Statements are finished by a period, like this:
m := 1. m := m + 1
In Squeak, a Smalltalk implementation, there’s an object called Transcript and you can send messages to it to be displayed on the screen. It’s more or less like a log window. It works like this:
Transcript show: 'Hello world'
and it looks like this:
The syntax is quite unique to Smalltalk. The message, otherwise known as “method call” in other languages, is called show: (including the colon) and it takes an argument. We can run it 10 times in a row with the following snippet:
10 timesRepeat: [ Transcript show: 'Hello world' ]
There you can start to see how Smalltalk is special. I’m sending the message timesRepeat: to the object 10, an Integer. Doing something N times repeatedly is handled by the Integer class, which if you think about it, makes sense.
The second interesting part, is the block. The part inside squared brackets. You might thing that’s the equivalent of other language’s block syntax, like in this Java example:
for(int i=1; i<11; i++) {
System.out.println("Hello world");
}
but Smalltalk version’s is a bit more powerful. It’s a real closure. Look at this:
t := [ Transcript show: 'Hello world' ]
Now I have a variable named t, of type BlockClosure, and I can do anything I want with that variable. If I send it the class message it’ll return its class:
t class
and if I sed it the value message, it’ll execute and leave a “Hello World” in the transcript:
t value
Let’s see some more code. A message without any arguments:
10 printString
a message with one argument:
10 printStringBase: 2
and a message with two arguments:
10 printStringBase: 2 nDigits: 10
Isn’t it cute? That method is called printStringBase:nDigits:. I never seen that syntax anywhere else; well, except in Objective-C, which copied it from Smalltalk.
Enough toying around, let’s start building serious stuff. Let’s create a class:
Object subclass: #MyClass
instanceVariableNames: ''
classVariableNames: ''
poolDictionaries: ''
category: 'Pupeno'
Notice that a class is created by sending a message to another class telling it to subclass itself with the name and a few other arguments. It’s a message, a method call like any other. Object is a class, classes are objects. The object model of Smalltalk is a beauty but that’s a subject for another post.
Now that we have a class, let’s create a method called greet: in that class.
greet: name "Greets the user named name" | message | message := 'Hello ', name. Transcript show: message.
In that method definition first we have a comment for the method, then the list of local variables within pipes (“|”), and then the implementation, which sets the variable message to contain “Hello ” and the comma concatenates name to it. Then we just send it to the transcript.
It looks like this:
Ok, let’s use it:
m := MyClass new. m greet: 'Pupeno'
To create an object of class MyClass, we send the new message to that class. There’s no new keyword like in Java. new is just a method. You can read its code, override it, etc. Don’t mess with it unless you really know what you are doing.
Actually, if you think about it, we haven’t seen a single keyword. Look all the code we wrote without having to memorize any keywords! What’s even more important is that by now you essentially know Smalltalk. That’s all there is, but like LEGO bricks, this simple and small building blocks allow you to build whatever you want.
Yes, that’s it, that’s all there is to it. We already saw that Smalltalk doesn’t need loops, it has integers and that class implements the timesRepeat: message which allows you to do something N times. There are many other looping methods here and there.
What about the if keyword you ask? Surely Smalltalk has an if? Well, no, it doesn’t. What you can recognize as an if is actually implemented in Smalltalk using the same mechanism of classes and message passing you saw already. Just for fun let’s re-implement it.
We starte by creating the class PBoolean and then two classes inheriting from it, PTrue and PFalse.
Object subclass: #PBoolean
instanceVariableNames: ''
classVariableNames: ''
poolDictionaries: ''
category: 'Pupeno'
PBoolean subclass: #PTrue
instanceVariableNames: ''
classVariableNames: ''
poolDictionaries: ''
category: 'Pupeno'
PBoolean subclass: #PFalse
instanceVariableNames: ''
classVariableNames: ''
poolDictionaries: ''
category: 'Pupeno'
For the class we created before, MyClass, we define a equals: method that will return either true or false, or rather, PTrue or PFalse.
equals: other ^ PTrue new
The little hat, ^, means return. For now, just a hardcoded true. Now we can do this in the workspace:
m1 := MyClass new. m2 := MyClass new. m1 equals: m2
and get true, that is PTrue, as a result. We are getting close but no if yet. How should if look like? It’ll look like something like this:
m1 := MyClass new. m2 := MyClass new. (m1 equals: m2) ifTrue: [ Transcript show: 'They are equal'; cr ] else: [ Transcript show: 'They are false'; cr ]
and you can start to imagine how to implement it. In PTrue we add the method:
ifTrue: do else: notdo ^ do value
That method basically takes two parameters, evaluates the first one and ignores the second one. For PFalse we create the oposite:
ifTrue: notdo else: do ^ do value
and that’s it. A working if! If you ask me, I think this is truly amazing. And if you check Squeak itself, you’ll find the if is actually implemented this way:
If your programming language allows you to create something as basic as the if conditional, then it allows you to create anything you want.
