Describe the PR

• For the symbol table, only classes that were clearly internal API are deprecated. Other classes, like scope implementations, are still the only way to get a name declaration. So I don't think we should deprecate them immediately on master
• The replacement of TypeHelper is in the package, that [java] New type resolution #2689 uses, so that it is forward-compatible
  • Many methods are not replaced. I only kept isA, and isExactlyA. Stuff like isExactlyNone, isEither, are very specific and barely used in the codebase. I'd rather have 4 public methods with a solid contract than 10 loosely consistent ones.
  • The new isA/isExactlyA have changed a bit:
    • They take the node as the second parameter. This is because usually, the longer expression is the one that produces the node (the type is usually a class literal). I think isA(String.class, someNode.getTypeNode()) reads better, even when the second expression grows longer
    • They return false if the node is null. This simplifies calling code, which usually needed a null check, and declared a temporary variable to do so. The previous behavior was mostly to throw if the node was null, though that wasn't specified, and was probably not consistent (I didn't check)
    • They throw if the class argument is null. This isn't really needed, just that I think it's also the better default, and it simplifies the implementation. The previous behavior was to return false, or throw in some cases, inconsistently. This wasn't specified.

The old methods of TypeHelper keep the same behavior but are deprecated, and implemented using TypeTestUtil

I think this is necessary before #2689 is merged

Related issues

• [java] Java-grammar progress document #2701

Ready?

• Added unit tests for fixed bug/feature
• Passing all unit tests
• Complete build ./mvnw clean verify passes (checked automatically by travis)
• Added (in-code) documentation (if needed)