prog8/CompilerDevelopment.md
2021-07-03 15:11:34 +02:00

3.8 KiB

Just a few remarks upfront:

  • There is the (gradle/IDEA) module parser: that's the parser generated by ANTLR4, in Java. The only file to be edited here is the grammar, prog8.g4.
  • Then we have the module compilerAst - in Kotlin - which uses parser and adds AST nodes. Here we put our additions to the generated thing, including any tests of the parsing stage.
    • the name is a bit misleading, as this module isn't (or, resp. shouldn't be; see below) about compiling, only the parsing stage
    • also, the tree that comes out isn't much of an abstraction, but rather still more or less a parse tree (this might very well change).
    • However, let's not yet rename the module. We'll find a good name during refactoring.

Problems with compilerAst:

  • ModuleImporter.kt, doing (Prog8-) module resolution. That's not the parser's job.
  • ParsingFailedError (in ModuleParsing.kt): this exception (it is actually not a java.lang.Error...) is thrown in a number of places, where other exceptions would make more sense. For example: not finding a file should just yield a NoSuchFileException, not this one. The other problem with it is that it does not provide any additional information about the source of parsing error, in particular a Position.
  • During parsing, character literals are turned into UBYTEs (since there is no basic type e.g. CHAR). That's bad because it depends on a specific character encoding (IStringEncoding in compilerAst/src/prog8/ast/AstToplevel.kt) of/for some target platform. Note that strings are indeed encoded later, in the compiler module.
  • The same argument applies to IMemSizer, and - not entirely sure about that - IBuiltinFunctions.

Steps to take, in conceptual (!) order:

  1. introduce an abstraction SourceCode that encapsulates the origin and actual loading of Prog8 source code
    • from the local file system (use case: user programs)
    • from resources (prog8lib)
    • from plain strings (for testing)
  2. add subclass ParseError : ParsingFailedError which adds information about the source of parsing error (SourceCode and Position). We cannot just replace ParsingFailedError right away because it is so widely used (even in the compiler module). Therefore we'll just subclass for the time being, add more and more tests requiring the new one to be thrown (or, resp., NOT to be thrown), and gradually transition.
  3. introduce a minimal interface to the outside, input: SourceCode, output: a tree with a Module node as the root
    • this will be the Kotlin singleton Prog8Parser with the main method parseModule
    • plus, optionally, method's for registering/unregistering a listener with the parser
    • the only exception ever thrown / reported to listeners (TBD) will be ParseError
    • anything related to the lexer, error strategies, character/token streams is hidden from the outside
    • to make a clear distinction between the generated parser (and lexer) vs. Prog8Parser, and to discourage directly using the generated stuff, we'll rename the existing prog8Parser/prog8Lexer to Prog8ANTLRParser and Prog8ANTLRLexer and move them to package prog8.parser.generated
  4. introduce AST node CharLiteral and keep them until after identifier resolution and type checking; insert there an AST transformation step that turns them in UBYTE constants (literals)
  5. remove uses of IStringEncoding from module compilerAst - none should be necessary anymore
  6. move IStringEncoding to module compiler
  7. same with ModuleImporter, then rewrite that (addressing #46)
  8. refactor AST nodes and grammar: less generated parse tree nodes (XyzContext), less intermediary stuff (private classes in Antrl2Kotlin.kt), more compact code. Also: nicer names such as simply StringLiteral instead of StringLiteralValue
  9. re-think IStringEncoding to address #38