Bill Tozier

My heartfelt and carefully-considered frustration comes, to be honest, from trying to imagine how to make simple, small changes that could be incorporated into your working codebase. And failing. Because...

In the meantime I've kicked to wall too many times, and started [this project](https://github.com/Vaguery/push-in-closure). I'm not a good enough programmer to write any interpreter without tests, and definitely am not...

Dude, I'm totally not quitting the idea of separating the interpreter. I'm just not able to _heal_ the current interpreter.

Absolutely. Except with complete coverage with automated tests, and without 729 duplicates of "are there two integers on the :integer stack?" written in sixteen different dialects :). The first thing...

That said, I am still going to implement this weird-ass (no offense) hard precision limit on integers and floats as a "quirks mode" behavior, for backwards-compatibility.

OK, this really should get addressed at some point before I am able to say that the new interpreter works "the same" as the old one. At the moment (and...

This is also a problem with `lein midje`, since [MIdje](https://github.com/marick/Midje) checks everything in `/src` and `/test`, and thus loads all those overlapping global doodads.

Actually I've been thinking and I guess I'm even more confused now. Looking at [this whole sequence](https://github.com/lspector/Clojush/blob/master/src/clojush/pushstate.clj#L36-L50) it seems as if there is a `set` called `registered-instructions` and also a...

Actually, once the interpreter is a sequestered entity in its own right (and not a global pile of everything) this becomes moot, and it would be Really Weird if anybody...

I also note that if I completely delete all the problems, this trouble goes away :) While I'm not suggesting that, I am wondering now about a modular architecture in...