code / nomsu

   1 -- This file contains the parser, which converts text into abstract syntax trees
   2 lpeg = require 'lpeg'
   3 re = require 're'
   4 lpeg.setmaxstack 20000
   5 {:P,:R,:S,:C,:Cmt,:Carg,:Cc} = lpeg
   6 
   7 -- foldr {A{a1,a2,...},B{b1,b2,...},C{c1,c2,...}} -> C{B{A{a1,a2,...},b1,b2...},c1,c2...}
   8 foldr = (...)->
   9     inner = select(1,...)
  10     for i=2,select('#',...) do
  11         assert inner.type
  12         outer = select(i,...)
  13         table.insert(outer, 1, inner)
  14         outer.source.start = inner.source.start
  15         inner = outer
  16     assert inner.type
  17     return inner
  18 
  19 DEFS = with {}
  20     -- Newline supports either windows-style CR+LF or unix-style LF
  21     .nl = P("\r")^-1 * P("\n")
  22     .tab = P("\t")
  23     .at_break = lpeg.B(lpeg.S(";,. \r\n\t({[")) + -lpeg.B(1)
  24     .tonumber = tonumber
  25     .tochar = string.char
  26     .unpack = unpack or table.unpack
  27     .nil = Cc(nil)
  28     .userdata = Carg(1)
  29     -- Always match and capture the indentation (spaces only) of the current line
  30     -- i.e. the leading space of the chunk of non-newline characters leading up to s[i]
  31     .indentation = lpeg.Cmt P(0),
  32         (s, i)->
  33             sub = string.sub
  34             while i > 1 and sub(s,i-1,i-1) != '\n' do i -= 1
  35             return true, (s\match("^ *", i))
  36     .utf8_char = (
  37         R("\194\223")*R("\128\191") +
  38         R("\224\239")*R("\128\191")*R("\128\191") +
  39         R("\240\244")*R("\128\191")*R("\128\191")*R("\128\191"))
  40     .operator_char = S("#'`~@^&*+=<>?/%!|\\-") + (P("\xE2") * (R("\x88\x8B")+R("\xA8\xAB")) * R("\128\191"))
  41     .Tree = (t, userdata)-> userdata.make_tree(t, userdata)
  42     .foldr = foldr
  43 
  44 setmetatable(DEFS, {__index:(key)=>
  45     if i = key\match("^ascii_(%d+)$")
  46         c = string.char(tonumber(i))
  47         self[key] = c
  48         return c
  49     elseif i = key\match("^number_(%d+)$")
  50         p = Cc(tonumber(i))
  51         self[key] = p
  52         return p
  53 })
  54 
  55 -- Just for cleanliness, I put the language spec in its own file using a slightly
  56 -- extended version of the lpeg.re syntax.
  57 peg_tidier = re.compile [[
  58     file <- %nl* {~ (captured_def/line) (%nl+ (captured_def/line))* %nl* ~}
  59     ident <- [a-zA-Z_][a-zA-Z0-9_]*
  60     line <- [^%nl]*
  61     captured_def <-
  62         ({ident} (" "*) "(" {ident} ")" (" "*) "<-" {[^%nl]* (%nl+ " "+ [^%nl]*)*}) ->
  63 "%1 <- ({| {:type:''->'%2':} {:start:{}:}
  64     (%3)
  65     {:stop:{}:} |} %%userdata) -> Tree"
  66 ]]
  67 
  68 make_parser = (peg, make_tree=nil)->
  69     peg = assert(peg_tidier\match(peg))
  70     peg = assert(re.compile(peg, DEFS))
  71     return (input, filename='???')->
  72         input = tostring(input)
  73         tree_mt = {__index: {source:input, filename:filename}}
  74         userdata = {
  75             make_tree: make_tree or ((t)->setmetatable(t, tree_mt))
  76             :filename, file:input
  77         }
  78         tree = peg\match(input, nil, userdata)
  79         if not tree or type(tree) == 'number'
  80             error "File #{filename} failed to parse:\n#{input}"
  81         return tree
  82 
  83 return make_parser