The columns that are reported by CPD were inconsistent across languages
before. A language like Java (using a JavaCC-based tokenizer) would use
a width of 8 for tabs, whereas a language like C# (using an Antlr-based
tokenizer) would use 1 instead.
This includes unit tests for most languages to ensure a tab character is
counted as 1. The configuration for JavaCC has been adjusted to respect
this as well.
Antlr root nodes need to implement RootNode.
Also, I didn't like the fact that the swift
module didn't use the conventional package
structure (lang.swift.ast) but was rolling
its own convention (lang.swift.antlr4).
I moved base classes for the antlr implementations
into nspmd.lang.ast.impl.antlr4. The fact that a
module is implemented with antlr is an implementation
detail, and it doesn't deserve its own toplevel package
in nspmd.lang.
Andreas Dangel