Parsing Expression grammars. Parsing expression grammars allow for full lookahead and backtracking in time linear to the input size. This makes them more expressive than YACC or BNF, which are limited in the how far they look ahead, and as a consequence how far they can backtrack. They also require more memory than either YACC or BNF to parse an equivalently sized input.
PEG grammars also do not have a separate lexing stage. Lexing and parsing are performed at the same time, using the same language for both.
See wikipedia for additional general information on Parsing Expression Grammars.
4th Baseline Machine Grammar Proposal
.alyn.post. is working on a proposal for a 4th Machine Grammar Baseline, replacing the 3rd baseline's YACC grammar with a PEG grammar. This work is scheduled for inclusion into CLL version 2.0 or CLL version 2.1.
Since PEG does not have a separate lexing stage, any PEG Machine Grammar will also need to express the Lojban Morphology in PEG.
Technical Points of the PEG language
- The '.' characters means any character, in any character set. It is only used after fa'o, which unconditionally consumes the remaining characters.
- '!.' is the way EOF is tested in PEG.
- space is defined as the literal '.' (as opposed to the '.' operator of PEG), whitespace, and all punctuation other than ',' and '''.
The CLL will need to be updated to account for changes resulting from the translation of the grammar to PEG.
Main Article: CLL PEG Errata
Open Discussion Points
- jbogenturfa'i further transforms the PEG grammar into an idealized representation. This parse tree is suitable for programmatic manipulation. Why is this idealized parse tree not the way the PEG is written?
Lojban parsers that use PEG
Of the current official grammar
- camxes, the original PEG parser from which all others are based.
- jbominje, John Leuner's PEG grammar. The lojban_grammar.peg is derived from camxes.
Of proposed grammars
- iocixes written in Haskell by la.iocikun. based on zasni gerna with MEX grammar proposal of la xorxes. (zasni gerna peg)
- ilmentufa: Created by la.ilmen. based on the mhagiwara's camxes.js with plenty of experimental propositions; because the core grammar is camxes.js, it does not include all the debuggings that were done in zasni gerna by la xorxes. (ilmentufa's peg)
- Another frontend of ilmentufa: Created by la.uilym. based on ilmentufa. It is not necessarily based on the latest version of ilmentufa's peg.
- Robin Powell's PEG Grammar Page. This document builds on the work camgusmis and xorxes have done, documented on this page.
- Grammar, for a discussion of Lojban's grammar beyond PEG.
- YACC, the language in which Lojban's official grammar is defined.
- BNF, widely considered easier to read than the YACC grammar.