grammar2parser
Goal: parser generator, to generate a (Parsica-)parser from a grammar. The grammar-input is in EBNF. The grammar defines the parser, which can be automatically built from it.
Parsica is a toolbox to build parsers, written in PHP, using principles from Functional Programming, like referential transparency, immutability and above all: composability. In this project we want to keep that same style of programming.
Grammar is the language of languages; a grammar defines the well-formed strings in a language. A grammar is a set of production-rules. In Extended Backus-Naur form (EBNF) we can express a Context Free Grammar (CFG). Symbols produce expressions, containing other symbols and/or terminal expressions. Here is an incomplete part of a grammar that describes a JSON-array, with terminals in single quotes (and the symbol "value" not resolved):
array = '[' whitespace ']' | '[' elements ']'
elements = element ',' elements | element
element = whitespace value whitespace
whitespace = ' ' | '\n' | '\r' | '\t' | ''
| indicates OR.
We use a PEG (Parsing Expression Grammar) = a grammar where the choices are ordered. So when we have a rule:
protocol = 'https' | 'http'
We first try to match 'https' and only if that would not succeed, we try 'http'.
See: https://parsica.verraes.net/docs/tutorial/order_matters
Parts:
- Grammar in PHP: an array of rules, each rule consisting of a symbol and an array of expressions
- Parser for EBNF, so we can read a grammar into our PHP-format
- The generator, that can build the parser source-code from the grammar.
Investigate
- can we (automatically) convert any CFG into a Parsica-parser? Are there limitations to such a LL-parser?
- how to define indentation, used to indicate begin and end of blocks, in a grammar?
- can we do anything with context sensitivity? For instance; if we want to check uniqueness of keys in JSON-objects?