[dan.from.tokyo]

This is why I stopped working on Walnut almost two years ago.

It was failing to inline the overridden op*, or much else for that matter.

Argument structs were being rep movsd'd in and out of the stack even as inout, rather than being accessed like a closure from the stack above like I thought.

D was introduced to std.variant.

It would have been much cleaner written if I was allowed to opAssign_r

While writing the parser, I went from wanting an AST, to bytecode, to native code generation. It was all written for AST I think.

While writing the parser, I realized that most stuff can have the type information and casting discarded IFF the parser maintains type coloring on Identifiers, which means Variable/Variant doesn't need to be used. It also means that alot of the Object Oriented stuff is only relevant during compile time provided we work around eval and with.

ECMA262 Edition 5 was looking imminent, but ended up taking a long time. It was released in December and changes the internals a bit.

The bank I was working at was draining my soul.

At the time I was writing it, there was practically no views or replies in this forum, so I was lonely.

What I'm doing now:

I have recently become unemployed and have taken up writing this up again. This time I am writing it in C99 to ECMA262 Edition 5.

My language choice is disappointing, but I needed to avoid the 80kb of cruft that D attaches to everything including a "hello world".