Fix quote & region scope problems

[NeilKleistGao]

It seems that I'm not authorized to push to the branch #366 directly.

In this PR:

• Collect symbols in quasiquote and region (in BBML/InvalML) for Scoped
• Have correct error message for cross-stage references in diff test

[LPTK]

Thanks!

[LPTK]

Let's change these to `===, then.

But it's strange that we can't use a global thing like Predef – that's not truly a cross-stage reference.

[NeilKleistGao]

It is because we need to compute which files should be included in the next stage. So, we need to use import.meta to get the relative path. In this example, == in Predef shadows the default one. That's why we get the error.

[LPTK]

I still don't get it (see my other message).

[LPTK]

Why would import.meta be needed? All we need is to make the genearted code import Predef, not the current file...

[NeilKleistGao]

Assume that we have a file A.mls containing a quoted symbol that is defined in ./B.mls (a relative path to A.mls). If we generate the quoted code fragment into a file ./C.mls (** a relative path to the pwd, which is the root dir for now **), we must know the relative path of B.mls to C.mls, which requires knowing where A.mls is. In the ES Module, __filename is not defined, and we need to use import.meta to retrieve the full file path of A.mls. However, it is not available in the REPL mode. That's why I did this check. Of course, it is possible to always generate an import statement for Predef. But for now, we are still using a relative path to import Predef. If we can write import "Predef", and the compiler can locate the file by the given paths, then the reference to symbols in Predef will no longer rely on import.meta.

[LPTK]

Clearly, we should be able to know the path toPredef here, because it is being imported in the code! We know it at compile time, so we should be able to generate appropriate code.