2 This is a list of things I definitely want to do. I don't go into too many
3 specifics, because the specifics change often as I think about the problem
6 Compiler / Grammar Analysis (all changes should have test-cases):
7 * properly support EOF in lookahead (like s -> "A" | "A" "B"; should predict 1).
8 * properly support lookahead that predicts a return from a final RTN state
9 like s -> a "X"; a -> "Y"*;
10 * try to detect when we are in an infinite loop (requires much thought).
11 * detect cases where some RTN alternatives have no GLA final state.
12 (shouldn't happen, modulo bugs).
13 * deal with lexer-level ambiguity. longest match will do for now, but
14 we're not currently detecting s -> "A" | "AB";
17 * Richer callback specifiers.
18 * Bring back slotbufs: a cheap (stack only, no heap) way of saving parse_vals for
19 the currently-open nodes of the parse tree.
20 * Provide a buffering layer.
21 * As the runtime starts to mature: language bindings.
23 Tests! Everything should have tests, as much as possible:
24 * all aspects of compilation
25 * bitcode format (both reading and writing)
26 * JSON output from gzlparse. Both well-formedness and accuracy.
28 Major design areas that exist only in my head:
29 * Character sets other than ASCII.
30 * Embedding Lua to do things only an imperative language can do.
31 * Operator-precedence parsing using the shunting yard algorithm.
32 * Parallel parsers (for both embedded languages and things like whitespace/comments)
33 * Error recovery (basically: just yield to an imperative function).
34 * Generate imperative bytecode / JIT compile.