Han-Wen Nienhuys <hanw...@gmail.com> writes: > We have a dump of the bison grammar in the contributor guide (see > http://lilypond.org/doc/v2.19/Documentation/contributor/lilypond-grammar). > > Is there any value in keeping this? It complicates the generation, as > it is a cross-directory dependency.
Much of LilyPond's language has been offloaded to music functions and the parsing of music function arguments uses synthetic tokens and to a good degree is directed not as much from the rules but the underlying actions. As an end user tool, it reflects far too little of what the input language of LilyPond is about. And it does not contain enough to work with when placed, say, in the CG. While one might want to think about whether the responsible scripts could in any useful manner be contributed to Bison (after all, Texinfo is the official GNU documentation language), for LilyPond itself it does no longer make much sense in my opinion. It allows interpreting the output of -ddebug-parser of a binary corresponding to the version of the NR. But the complexity of LilyPond's grammar is such that I would not expect somebody not working with a full checkout-out source to be likely in a capacity of interpreting the respective traces of Bison. -- David Kastrup
