Crosscut

Let's move on to the next two problems of the previous prototype. I'm going to tackle syntax and evaluation model together, since they are closely related.

From the start of the project, I was going for a syntax based on postfix operators, and a stack-based evaluation model. I wrote about this decision before. Going through these old thoughts (from 8 months ago!), I still largely agree with the points I made then. (Although I have new thoughts, which I hope to get into at a later date.)

But despite all attractive qualities, this approach caused two major problems. First, what I was trying to build was very different from anything that exists (as far as I know). This turned every hurdle into a major design challenge. Second, this model, despite its simplicity, is quite powerful. Which is nice. But that power came back to bite me later, when working on the type system.