Current Problems / Tasks #1

TODOs

• ability to import a parser into the REPL, just can't construct anything in the repl
  • so class A(Rule): ... or A = B|C|D, or any other expression over a Rule should raise an exception, but importing and using should be fine
• add support for building grammars across multiple files
• error messages using the prettyerr library
  • consider user defined error hooks into parsing process. TBD exactly what interface would work
• improve the README. have 1 good example, mention grammars must be defined in files, better explanation of GLL backend, etc.
• make it so optional elements are None, not empty array. e.g. missing exponent or missing fractional part of json
• [-] pull json imports out of test json imports and put in test hydration
• mixins
  • probably better to just support user defined def convert(self) -> Whatever: ...
• add more grammar examples + corresponding tests
  • test CSV grammar
  • improve the CSV grammar
• add support for longest match disambiguation rule
• make disambiguation rules filter during parsing rather than after parsing so that the algorithm is more efficient on ambiguous grammars
• support whitespace policy
  • default Explicit
  • SkipUnexpectedWhitespace
    • blacklist or whitelist to specify which characters are skipable
    • basically before trying to eat the next part of a rule, skip all whitespace (assuming rule doesn't contain whitespace)
    • this could also be useful for e.g. phone numbers where various delimiters are unimportant (parenthesis, dash, space, etc.)
• ability to skip junk before and or after a parse. in principle we could have a single meta rule that you parameterize with your rule? we would need a new disambiguation which is essentially eat until next rule could start. Alternatively a dumb approach would be to try to eat, progressing one token at a time when failing, until we find the spot we can start from
• filter for rejecting parses. e.g. say we want to parse a number 0-255, should be able to easily say if > 255 fail this branch
• helper for wrapping and optional wrapping. E.g. wrap[A, ['(',')'], None] -> (A)|A. Experiment to find best ergonomics