TODO.md

Merger of epsilon and gamma

• epsilon is a full generator based on an LL(1) parser
• gamma is just an LALR(1) parser generator
• both need a lexer and parser for EAG specifications
• by now, we already changed gamma to use epsilon' s lexer
• since gamma uses a modern grammar model, we try the following approach:
  • we analyze grammar properties only with gamma.input
  • we finish gamma.input (handling of affix forms)
  • we transform the gamma grammar model into the epsilon grammar model (to reuse the compiler generation)
  • we remove the epsilon analyzer and move the lexer to gamma.input.epsilang

Escape Sequences

• epsilon blindly copies escape sequences to the generated compiler
• but this makes "\"" and '"' two different terminal symbols
• take, for example, the code from the gamma Scanner

compiled compiler

• swap options --info and --verbose
  • --info should give detailed parser error messages
  • --verbose should print debug output

epsilang.eag

• specify the lowering for EBNF and various abbreviations as an EAG

Expand

• use ~ instead of Expand

Code Generation

• create abstract syntax tree instead of writing code fragments

Table Files

• include .Tab and .EvalTab files in executable

Code Generation

• create intermediate language nodes

History Set

• remove history state, tailored for Oberon's command execution

UTF-8

• rewrite decision-tree implementation in generated lexer (current one is a workaround for the change to UTF-8 support)
• add reserved nonterminal "alpha" for Unicode letter?

RISC Assembler

• revive RISC assembler?

Lexer Generation

WhiteSpaces:
    "WHITESPACES" "ARE" char { "," char } ";" .

Comments:
    Comment { ";" Comment } ";" .
Comment:
    "COMMENTS" "FROM" char [ char ] "TO" char [ char ] ["NESTED"] .

Replace Subtyping with Sum Type

• https://github.com/pbackus/sumtype

Expressive Diagnostics

• https://clang.llvm.org/diagnostics.html

Color