Thursday, February 03, 2005

Why Smalltalk?

Because it's powerful.

1 comment:

Howard Stearns said...

Cute, and of course, I agree.

I'm also interested in ways in which Smalltalk is not powerful, in order to see what can be done to improve things. As with your comments on profiling, I like to be able to measure stuff. One way I like to measure the power of a programming language is to measure the cost (in lines of code) of making a change. This raises two issues in my mind:

1. In the Crqouet code that I've seen, a lot of code is duplicated between classes. Initialize methods do a lot. Operational methods in analogous but non-inheriting classes end up repeating the same (or nearly the same) code. So making a simple change often requires making the same kind of change in multiple places. The bigger the code base, the more often I have to do this. It's just cut and paste, but it's a lot of cust and paste. I hate cut and paste. It offends me. The way this is avoided in my former favorite language is with multiple inheritance. (Including multiple overridable inheritance of initialization methods.) I understand that there are people who have issues with multiple-inheritance -- or ANY form of class inheritance. OK. Maybe that's not the answer. But I would like to address the problem in some way. Any thoughts? I find myself drawn to delegation techniques, but I don't have much experience with them.

2. There's a fair amount of setup and boilerplate that have to do with getting related objects to cooperate and initialize themselves properly. In my formerly favorite language, this was addressed with macros. Instead of writing application-specific stuff directly in the language, you typically create a domain-specific language using a few macros that you write once, and then use that language to create the application code. When issues arise for performance or interfaces to external systems, we change the macros rather than the boilerplate that is distributed throughout the code. Again, any thoughts?

I like to see the cost of making a change be proportional to the the application-domain scope of the change. (I don't have a good way of measuring the application-domain scope of the change. All I have is a subjective feel. But the general idea is that "easy" things should be easy.) In both of the situations I describe, the cost of making a change ends up being proportional to the size of the code base.