(all) Add support for postfix increment and decrement

[perlun]

This is not ready yet, will complete another day. Reasonably complete now.

Things currently missing:

• Support for decrement also, not just increment

• Ensure that we put this at a reasonable level of precedence. (The way I did it now was just a slightly lame approximation)

  Maybe we can live with this as it is now after all. Right now I put it between "assignment" and "logical or". I would love to detect invalid usages such as foo(c++) which I consider very bad practice, but we can perhaps live without that at this stage. Note: even though these are postfix operators at the moment, they work like prefix operators do in other languages. I.e. var c = 0; some_function(c++); will use 1 as the value of the the parameter passed to the function.

  I tried to document this now in the syntax grammar, which is btw again heavily inspired by the Lox specification: (docs) Add syntax-grammar.md document #18.

• Properly handle cases when the user tries to do things like var a = "foo"; a++; and similar folly. This is clearly an area where a static type checker would make things immensely much better; then interpreter/compiler in the future will then have to take much less precautions in ensuring that things are of the expected types all the time.

[perlun]

Making process, but there are still some things that are broken - the failing tests indicate this. There is something causing the resolution of local variables to fail, which is quite odd given that prefix operators work flawlessly.

I tried moving the unary postfix operators into a separate method now (instead of just tucking it into the Assignment() method), but it doesn't change a thing. It feels like there some oddness in the variable resolver going on; I guess I just have to continue digging...

[perlun]

Making process, but there are still some things that are broken - the failing tests indicate this. There is something causing the resolution of local variables to fail, which is quite odd given that prefix operators work flawlessly.

The thing missing was the ResolveLocal() call in VisitUnaryPostfixExpr(); this is necessary to be able to resolve variables and other locals. Isn't it interesting how things sometimes are much easier if you leave it for a while and revisit the problem the next day? 🙂

[perlun]

• Note: even though these are postfix operators at the moment, they work like prefix operators do in other languages. I.e. var c = 0; some_function(c++); will use 1 as the value of the the parameter passed to the function.

This was subsequently revisited in #145. Using different semantics than other languages here was simply discovered to be a pretty bad idea.