egg is a free software package that provides the libraries and utilities needed to produce parsers from EGG input files. EGG (loosely an acronym for Expression Grammar, Grammar) files are yet another BNF adaptation. The intent of the EGG style BNF is to include the ability to use ANY input coding scheme. While the ASCII character set is the default, a grammar using any coding can be described in EGG. Also, the structure of an EGG file is such that producing parsers from it is both encouraged and simplified to a great extent.
Many of the ideas and syntax presented in the original BNF, and subsequently the formal ISO EBNF, and the many variants of ABNF have been incorporated into the EGG syntax. EGG is designed to limit the number of syntax options provided by other extended forms of BNF to those that only produce a more readable form and a clearer understanding of the presented grammar. At the same time, EGG forces the author to abstract almost every element of a grammar into individual phrases, so that both the grammar definition itself is non-ambiguous, and is easily translated into a parser.
egg is designed to produce pure recursive descent parsers. embryo, the utility included in the egg package is an example of this type of parser generator.
embryo is a generic parser generator that uses the EGG grammar format for input. This parser is intended to be able to parse, verify and generate a grammar parser framework from any grammar defined in EGG.
The resultant parser design uses many of the concepts developed and documented in tutorial form by Jack W. Crenshaw PHD. embryo produces pure recursive descent parsers.
Other code modules are also added to provide utility functionality, such as a separate grammar framework generation function library, an input source function library, an input location tracking mechanism, and a callback mechanism for parsers that need to work "on-the-fly" such as interpreters.
Please note that the embryo parser generator folds the traditional tokenizer and grammar parser components into one single parser. While this approach works well for recursive descent parsers, it is recognized that other types of parsers such as LR, LALR, etc. will not work this way. The author of any grammar in EGG can certainly adapt the tokenizer library portion of EGG to a different parser style. This could be accomplished by writing two distinct grammars. One for token recognition and the other for grammar syntax, or by writing one grammar, containing two distinct top-level phrase types, and making synchronized calls to each parser when needed. Many possibilities exist for such needs.
While the embryo parser currently produces correct grammar parsing code and said code is hopefully quite "unfragile", it is recognized by the creator of this package that the generated code is far from being the most efficient code possible.
From a code style viewpoint, the generated code is quite consistent, but includes a lot of possibly unnecessary code bits that are included either out of convenience, or to silence a lot of compiler errors and warnings. Both of these situations can be remedied by doing more look ahead work to determine when a parsing function is producing only literal values, only single sequence phrases, or alternations. All of these cases can be optimized, sometimes dramatically, from a code style standpoint. Doing so should also provide minor improvements in the run time performance of a parser.
The run time efficiency of the generated code can stand a lot of improvement. The author has several ideas for providing optimizations, some by user choice at parser generation time, and others as permanent fixes. These improvements will come along in time, as new versions of this package are released.
For more information on installing the egg package, including embryo please refer to the INSTALL file and other documentation in doc/.
Also, a complete set of user guide, tutorial, manpages, and reference manual can be built and installed from these sources.