Awesome

This is a modified version of Go package peg, written by pointlander, for the purpose of supporting LEG grammars, a variant of PEG grammars as defined in peg(1). See README.orig for the original README.
If you do not intend to use a LEG grammar, please take a look at the original package instead.

The subdirectory cmd/leg contains source files for the LEG parser. Using this parser, the peg-markdown package, which contains a LEG definition, has been ported to Go.

To download and install, run

go get github.com/knieriem/peg

Run make or make prepare to bootstrap the peg parser, and to create the leg parser and the example parsers. There should be a binary peg in ./cmd/peg now.

To delete the generated source files and binaries that are not part of the project, run make clean.

The desk calculator example from peg(1) can be built by typing go build in directory ./cmd/legcalc.

The parser generators now take on option -O to turn on various optimizations, with a single argument consisting either of a number of colon-separated flags, or the string "all". For the possible values of these flags, see util.go.

Summary of other modifications:

• AddPackage, AddPeg and AddState methods have been replaced by a new method AddDefine, which stores different values into a map. This way additional strings (like yystype) can easily be specified.

• As markdown_parser.leg makes heavy use of yytext, I replaced the action arguments buffer string, begin, end int by yytext string, which equals to buffer[begin:end].

• Headers %{ ... %} and a Trailer %% ..., which are used in LEG Grammers, are supported by new methods AddHeader, and AddTrailer. (Both probably could be replaced by AddDefine ...)

• Parse() has got an integer argument `ruleId', to allow rules different from rule 0 to be applied, as it is needed by peg-markdown. The output file now also contains a const block containing names and IDs of all rules. Defined but unused rules are not deleted anymore (the warning has been preserved), because they might be called directly.

• Added support for semantic values as described in peg(1). Results of sub-rules can be referred to from within actions, whereas $$ can be used to store the current rule's return value. At the moment this only works without the -inline option.

• Added ResetBuffer closure to parser. The user can set a new buffer to be processed, the remaining part of the old buffer is returned. This way a parser can be reused without calling Init again. See ./leg/calc.leg for an example.

--
Michael Teichgräber