Frontend-For-Free

A bootstrap of RBNF.hs to generate standalone parsers targeting multiple programming languages.

Standalone: the generated code can run without runtime dependencies other than the language and standard libraries.

Installation

Install from Sources

You can install binary files via: The Haskell Tool Stack.

sh> stack install .

Install from Binaries

Otherwise, binary files for various platforms(Win64, Generic Linux, MAC OSX 10.13-10.15) are released on GitHub.

Download it from Releases, add fff-lex and fff-pgen to your PATH.

Besides, You Need a Python Wrapper

frontend-for-free now provides a wrapper for Python only:

pip install frontend-for-free or install it from GitHub.

Usage

sh> fff <xxx>.rbnf --trace [--lexer_out <xxx>_lex.py] [--parser_out <xxx>_parser.py] 
sh> # note that you should also provide a <xxx>.rlex file
sh> ls | grep <xxx>
<xxx>_parser.py <xxx>_lex.py

See examples at runtest.

What is Frontend-For-Free?

A framework for generating context-free parsers with the following features:

• cross-language
• distributed with a lexer generator, but feel free to use your own lexers.
• LL(k) capability
• efficient left recursions
• standalone No 3rd party library is introduced, while the generator requires Python3.6+ with a few dependencies.
• defined with a most intuitive and expressive BNF derivative

  • action/rewrite:

    pair := a b { ($1, $2) }

  • parameterised polymorphisms for productions:

    nonEmpty[A] := A { [$1] } | hd=A tl=nonEmpty[A] { tl.append(hd); tl }

    where append shall be provided by the user code.

Currently,

• the parser generator support for a programming language is hard coded in src/RBNF/BackEnds/<LanguageName>.hs.
• the lexer generator support for a programming language is hard coded in ffflex.py.

Galleries

• Parsing JSON

  • lexer: json.rlex
  • parser: json.rbnf

• Parser as Interpreter: Implementing a Programming Language within 20 Minutes

  • parser/interpreter: https://github.com/thautwarm/simple-pl/blob/master/easylang.gg
  • lexer: https://github.com/thautwarm/simple-pl/blob/master/easylang.rlex

• Parsing LaTeX

  • lexer: gkdtex.rlex
  • parser: gkdtex.gg

• Parsing LLVM IR(A major subset)

  • lexer: llvmir.rlex
  • parser: llvmir.rbnf

• Parsing nested arithmetic expressions

  • lexer: arith.rlex
  • parser: arith.rbnf

• Parsing the BNF derivative used by FFF(bootstrap)

  • lexer: fffbnf.rlex
  • parser: fffbnf.rbnf

• Parsing ML syntax:

  • lexer: mlfs.rlex
  • parser: mlfs.bnf

• (OLD VER 0)Parsing ML syntax and convert it to DrRacket

  • lexer: yesml.rlex
  • parser: yesml.rbnf

• (OLD VER 1)Muridesu: 以木兰的方式, 三小时做出强比Python，形似GoLang的语言

  • lexer: muridesu.rlex
  • parser: muridesu.exrbnf

• (OLD VER 2)Parsing Python ASDL files

  • lexer: asdl.rlex
  • parser: asdl.exrbnf

OLD VER 2, OLD VER 1 and OLD VER 0 are out-of-date, hence the code generation does not work with the master branch.

However, the generated code is permanent and now still working.

Further, OLD VER 2 can be easily up-to-date by manually performing the following transformations:

1. changing slots $0, $1, $2, ... to $1, $2, $3, ...

2. changing list(rule) to list[rule], and provide the definition of list production:

   list[p] ::= p        { [$1] }
           |  list[p] p { $1.append($2); $1 }
   

3. changing separated_list(sep, rule) to separated_list[sep, rule], and provide the definition of separated_list production:

   separated_list[sep, p] ::= 
            p             { [$1] }
         |  list[p] sep p { $1.append($3); $1 }
   

End-To-End: A Common Pattern for Using the Generated Parser

For most cases, you don't need to understand any parsing components like lexers, token tables, states, etc.

In fact, you can easily access your generated parser simply via the following function parse(source_code, filename="<unknown>"):

from <the generated parser module> import *
from <the generated lexer module> import lexer

__all__ = ["parse"]
_parse = mk_parser()


def parse(text: str, filename: str = "unknown"):
    tokens = lexer(filename, text)
    status, res_or_err = _parse(None, Tokens(tokens))
    if status:
        return res_or_err

    msgs = []
    lineno = None
    colno = None
    filename = None
    offset = 0
    msg = ""
    for each in res_or_err:
        i, msg = each
        token = tokens[i]
        lineno = token.lineno + 1
        colno = token.colno
        offset = token.offset
        filename = token.filename
        break
    e = SyntaxError(msg)
    e.lineno = lineno
    e.colno = colno
    e.filename = filename
    e.text = text[offset - colno:text.find('\n', offset)]
    e.offset = colno
    raise e

Calling parse will get you the expected result, or a considerably readable error message.