1 files changed, 123 insertions, 0 deletions
diff --git a/parser.moon b/parser.moon
new file mode 100644
index 0000000..cac126c
--- /dev/null
+++ b/parser.moon
@@ -0,0 +1,123 @@
+lpeg = require 'lpeg'
+re = require 're'
+lpeg.setmaxstack 10000
+{:P,:R,:V,:S,:Cg,:C,:Cp,:B,:Cmt,:Carg} = lpeg
+utils = require 'utils'
+{:match, :sub, :rep, :gsub, :format, :byte, :match, :find} = string
+{:NomsuCode, :LuaCode, :Source} = require "code_obj"
+AST = require "nomsu_tree"
+
+NOMSU_DEFS = with {}
+    -- Newline supports either windows-style CR+LF or unix-style LF
+    .nl = P("\r")^-1 * P("\n")
+    .ws = S(" \t")
+    .tonumber = tonumber
+    string_escapes = n:"\n", t:"\t", b:"\b", a:"\a", v:"\v", f:"\f", r:"\r"
+    digit, hex = R('09'), R('09','af','AF')
+    .escaped_char = (P("\\")*S("xX")*C(hex*hex)) / => string.char(tonumber(@, 16))
+    .escaped_char += (P("\\")*C(digit*(digit^-2))) / => string.char(tonumber @)
+    .escaped_char += (P("\\")*C(S("ntbavfr"))) / string_escapes
+    .operator_char = S("'`~!@$^&*-+=|<>?/")
+    .utf8_char = (
+        R("\194\223")*R("\128\191") +
+        R("\224\239")*R("\128\191")*R("\128\191") +
+        R("\240\244")*R("\128\191")*R("\128\191")*R("\128\191"))
+    .ident_char = R("az","AZ","09") + P("_") + .utf8_char
+
+    -- If the line begins with #indent+4 spaces, the pattern matches *those* spaces
+    -- and adds them to the current indent (not any more).
+    .indent = Cmt Carg(1), (start, userdata)=>
+        indented = userdata.indent..'    '
+        if sub(@, start, start+#indented-1) == indented
+            userdata.indent = indented
+            return start + #indented
+    -- If the number of leading space characters is <= the number of spaces in the current
+    -- indent minus 4, this pattern matches and decrements the current indent exactly once.
+    .dedent = Cmt Carg(1), (start, userdata)=>
+        dedented = sub(userdata.indent, 1, -5)
+        if #match(@, "^[ ]*", start) <= #dedented
+            userdata.indent = dedented
+            return start
+    -- If the number of leading space characters is >= the number of spaces in the current
+    -- indent, this pattern matches and does not modify the indent.
+    .nodent = Cmt Carg(1), (start, userdata)=>
+        if sub(@, start, start+#userdata.indent-1) == userdata.indent
+            return start + #userdata.indent
+
+    .userdata = Carg(1)
+
+    .error = (src,end_pos,start_pos,err_msg,userdata)->
+        seen_errors = userdata.errors
+        if seen_errors[start_pos]
+            return true
+        if utils.size(seen_errors) >= 10
+            seen_errors[start_pos+1] = colored.bright colored.yellow colored.onred "Too many errors, canceling parsing..."
+            return #src+1
+        err_pos = start_pos
+        line_no = pos_to_line(src, err_pos)
+        src = FILE_CACHE[userdata.source.filename]
+        line_starts = LINE_STARTS[src]
+        prev_line = line_no == 1 and "" or src\sub(line_starts[line_no-1] or 1, line_starts[line_no]-2)
+        err_line = src\sub(line_starts[line_no], (line_starts[line_no+1] or 0)-2)
+        next_line = src\sub(line_starts[line_no+1] or -1, (line_starts[line_no+2] or 0)-2)
+        i = err_pos-line_starts[line_no]
+        pointer = ("-")\rep(i) .. "^"
+        err_msg = colored.bright colored.yellow colored.onred (err_msg or "Parse error").." at #{userdata.source.filename}:#{line_no}:"
+        if #prev_line > 0 then err_msg ..= "\n"..colored.dim(prev_line)
+        err_line = colored.white(err_line\sub(1, i))..colored.bright(colored.red(err_line\sub(i+1,i+1)))..colored.dim(err_line\sub(i+2,-1))
+        err_msg ..= "\n#{err_line}\n#{colored.red pointer}"
+        if #next_line > 0 then err_msg ..= "\n"..colored.dim(next_line)
+        --error(err_msg)
+        seen_errors[start_pos] = err_msg
+        return true
+
+setmetatable(NOMSU_DEFS, {__index:(key)=>
+    make_node = (start, value, stop, userdata)->
+        if userdata.source
+            with userdata.source
+                value.source = Source(.filename, .start + start-1, .start + stop-1)
+        setmetatable(value, AST[key])
+        if value.__init then value\__init!
+        for i=1,#value do assert(value[i])
+        return value
+
+    self[key] = make_node
+    return make_node
+})
+
+NOMSU_PATTERN = do
+    -- Just for cleanliness, I put the language spec in its own file using a slightly modified
+    -- version of the lpeg.re syntax.
+    peg_tidier = re.compile [[
+    file <- {~ %nl* (def/comment) (%nl+ (def/comment))* %nl* ~}
+    def <- anon_def / captured_def
+    anon_def <- ({ident} (" "*) ":"
+        {((%nl " "+ [^%nl]*)+) / ([^%nl]*)}) -> "%1 <- %2"
+    captured_def <- ({ident} (" "*) "(" {ident} ")" (" "*) ":"
+        {((%nl " "+ [^%nl]*)+) / ([^%nl]*)}) -> "%1 <- (({} %3 {} %%userdata) -> %2)"
+    ident <- [a-zA-Z_][a-zA-Z0-9_]*
+    comment <- "--" [^%nl]*
+    ]]
+    nomsu_peg = peg_tidier\match(FILE_CACHE["nomsu.peg"])
+    re.compile(nomsu_peg, NOMSU_DEFS)
+
+parse = (nomsu_code, source=nil)->
+    nomsu_code = tostring(nomsu_code)
+    userdata = {
+        indent: "", errors: {}, :source
+    }
+    tree = NOMSU_PATTERN\match(nomsu_code, nil, userdata)
+    unless tree
+        error "In file #{colored.blue tostring(source or "<unknown>")} failed to parse:\n#{colored.onyellow colored.black nomsu_code}"
+    if type(tree) == 'number'
+        tree = nil
+
+    if next(userdata.errors)
+        keys = utils.keys(userdata.errors)
+        table.sort(keys)
+        errors = [userdata.errors[k] for k in *keys]
+        error(table.concat(errors, "\n\n"), 0)
+    
+    return tree
+
+return parse