feat(compiler): access modifiers

[yoav-steinberg]

See #108
This PR adds public and protected support to class members and while doing so makes all other members private by default. It also verifies interface method implementations must be public. It also includes support for JSII access modifiers. To get this working I did the following refactors:

• Cleaned up fields in the SymbolEnv and instead added a kind field distinguishing between 3 types of environments: a simple code block (Scope) a function/method/closure environemt (Function) and an environment used to store members of a type (Type). This was primarily needed so I can use the result of a symbol env lookup to figure out in what ancestor class a given member was defined. The added benefit was code clarity and sanity checks which, btw, found a non-related bug in super constructor type checking (see below). One piece of such clarity was the long awaited removal of return_type from SymbolEnv.
• Cleaned up the ContextVisitor which I needed to add to the TypeChecker. Here I removed the requirement to store the init_env in push_class which didn't make sense, and is probably a leftover of some iteration on our lifting code. I also cleaned up the push_method code to make more sense and now it's a more general push_function stack with a helper for when we're just interested in methods.
• I added some more stuff to VisitorWithContext trait, for easy "push -> do stuff or even return -> pop" pattern. Specifically I needed the TypeChecker to push function definitions and current statement safely.
• Use a VisitorContext in TypeChecker to track context so I know in what class I am for access modifier checking. This approach also provides a better way of tracking the current return_type so I could remove it from the env. This also removed all sorts of dirt from the TypeChecker struct (like in_json, etc.).
• The type_check_statement method was getting way to big, so I moved the contents of its match statements into separate methods.
• The sanity checks when creating new SymbolEnvs detected that we're creating a weird env in type_check_super_constructor_against_parent_initializer. I removed a large bunch of code there that didn't seem correct to me.
• The actual access modifier verification takes place inside get_property_from_class_like (<= this is where you want to look). This way I avoid cascading errors and "unhydrated" classes. To make this work for all cases I updated resolve_referece for static type references to also use this method instead of doing its own thing. I think this is cleaner.

Still left and possible candidate for followup PRs:

• Tests for JSII imports with access modifiers
• Handle method overriding rules
• Support public for types (export types from module) -> out of scope, will be handled added by our module system.
• Decide if we allow public fields: tracking issue consider removing support for public class fields #4180
• Decide about internal: tracking issue internal access modifier #4156

Checklist

• Title matches Winglang's style guide
• Description explains motivation and solution
• Tests added (always)
• Docs updated (only required for features)
• Added pr/e2e-full label if this feature requires end-to-end testing

By submitting this pull request, I confirm that my contribution is made under the terms of the Wing Cloud Contribution License.

[monadabot]

Console preview environment is available at https://wing-console-pr-4133.fly.dev 🚀

Updated (UTC): 2023-09-15 18:12

[mergify]

Thanks for contributing, @yoav-steinberg! This PR will now be added to the merge queue, or immediately merged if yoav/access_modifiers is up-to-date with main and the queue is empty.

[monadabot]

Congrats! 🚀 This was released in Wing 0.30.9.