1 ----------------------------------------------------------------------
2 -- Metalua: $Id: mll.lua,v 1.3 2006/11/15 09:07:50 fab13n Exp $
4 -- Summary: generic Lua-style lexer definition. You need this plus
5 -- some keyword additions to create the complete Lua lexer,
6 -- as is done in mlp_lexer.lua.
10 -- * Make it possible to change lexer on the fly. This implies the
11 -- ability to easily undo any pre-extracted tokens;
13 -- * Make it easy to define new flavors of strings. Replacing the
14 -- lexer.patterns.long_string regexp by an extensible list, with
15 -- customizable token tag, would probably be enough. Maybe add:
16 -- + an index of capture for the regexp, that would specify
17 -- which capture holds the content of the string-like token
19 -- + or a string->string transformer function.
21 -- * There are some _G.table to prevent a namespace clash which has
22 -- now disappered. remove them.
23 ----------------------------------------------------------------------
25 -- Copyright (c) 2006, Fabien Fleutot <metalua@gmail.com>.
27 -- This software is released under the MIT Licence, see licence.txt
30 ----------------------------------------------------------------------
32 module ("lexer", package.seeall)
34 require 'metalua.runtime'
37 lexer = { alpha={ }, sym={ } }
40 local debugf = function() end
43 ----------------------------------------------------------------------
44 -- Patterns used by [lexer:extract] to decompose the raw string into
45 -- correctly tagged tokens.
46 ----------------------------------------------------------------------
48 spaces = "^[ \r\n\t]*()",
49 short_comment = "^%-%-([^\n]*)()\n",
50 final_short_comment = "^%-%-([^\n]*)()$",
51 long_comment = "^%-%-%[(=*)%[\n?(.-)%]%1%]()",
52 long_string = "^%[(=*)%[\n?(.-)%]%1%]()",
56 number_exponant = "^[eE][%+%-]?%d+()",
57 number_hex = "^0[xX]%x+()",
58 word = "^([%a_][%w_]*)()"
61 ----------------------------------------------------------------------
62 -- unescape a whole string, applying [unesc_digits] and
63 -- [unesc_letter] as many times as required.
64 ----------------------------------------------------------------------
65 local function unescape_string (s)
67 -- Turn the digits of an escape sequence into the corresponding
68 -- character, e.g. [unesc_digits("123") == string.char(123)].
69 local function unesc_digits (x)
70 if x:sub(1,1)=="\\" then return x end -- Hack to parse correctly "\\123"
71 local k, j, i = x:reverse():byte(1, 3)
72 local z = _G.string.byte "0"
73 local code = (k or z) + 10*(j or z) + 100*(i or z) - 111*z
75 error ("Illegal escape sequence '\\"..x.."' in string: ASCII codes must be in [0..255]")
77 return _G.string.char ((k or z) + 10*(j or z) + 100*(i or z) - 111*z)
80 -- Take a letter [x], and returns the character represented by the
81 -- sequence ['\\'..x], e.g. [unesc_letter "n" == "\n"].
82 local function unesc_letter(x)
84 a = "\a", b = "\b", f = "\f",
85 n = "\n", r = "\r", t = "\t", v = "\v",
86 ["\\"] = "\\", ["'"] = "'", ['"'] = '"', ["\n"] = "\n" }
87 return t[x] or error([[Unknown escape sequence '\]]..x..[[']])
91 :gsub ("\\(%D)",unesc_letter)
92 :gsub ("\\(\\?[0-9][0-9]?[0-9]?)", unesc_digits)
96 "skip_whitespaces_and_comments",
97 "extract_short_string", "extract_word", "extract_number",
98 "extract_long_string", "extract_symbol" }
100 lexer.token_metatable = {
101 -- __tostring = function(a)
102 -- return string.format ("`%s{'%s'}",a.tag, a[1])
106 lexer.lineinfo_metatable = { }
108 -- The presence of this function prevents serialization by Pluto,
109 -- I can't figure out why :(
110 function lexer.lineinfo_metatable:__tostring()
111 local txt = string.format("%s:%i(%i,%i)", self[4], self[3], self[1], self[2])
112 if self.comments then
114 for comment in ivalues(self.comments) do
115 local content, loc1, loc2, kind = unpack(comment)
116 table.insert (acc, string.format ("%s@%i..%i:%q", kind, loc1, loc2, content))
118 txt = txt.."["..table.concat(acc,"; ").."]"
124 ----------------------------------------------------------------------
125 -- Really extract next token fron the raw string
126 -- (and update the index).
127 -- loc: offset of the position just after spaces and comments
128 -- previous_i: offset in src before extraction began
129 ----------------------------------------------------------------------
130 function lexer:extract ()
131 local previous_i = self.i
135 -- Put line info, comments and metatable arount the tag and content
136 -- provided by extractors, thus returning a complete lexer token.
137 -- first_line: line # at the beginning of token
138 -- first_column_offset: char # of the last '\n' before beginning of token
139 -- i: scans from beginning of prefix spaces/comments to end of token.
140 local function build_token (tag, content)
141 assert (tag and content)
142 local i, first_line, first_column_offset, previous_line_length =
143 previous_i, self.line, self.column_offset, nil
145 -- update self.line and first_line. i := indexes of '\n' chars
147 i = self.src :find ("\n", i+1, true)
148 if not i or i>self.i then break end -- no more '\n' until end of token
149 previous_line_length = i - self.column_offset
150 if loc and i <= loc then -- '\n' before beginning of token
151 first_column_offset = i
152 first_line = first_line+1
154 self.line = self.line+1
155 self.column_offset = i
158 -- lineinfo entries: [1]=line, [2]=column, [3]=char, [4]=filename
159 local fli = { first_line, loc-first_column_offset, loc, self.src_name }
160 local lli = { self.line, self.i-self.column_offset-1, self.i-1, self.src_name }
161 --Pluto barfes when the metatable is set:(
162 setmetatable(fli, lexer.lineinfo_metatable)
163 setmetatable(lli, lexer.lineinfo_metatable)
164 local a = { tag = tag, lineinfo = { first=fli, last=lli }, content }
165 if lli[2]==-1 then lli[1], lli[2] = lli[1]-1, previous_line_length-1 end
166 if #self.attached_comments > 0 then
167 a.lineinfo.comments = self.attached_comments
168 fli.comments = self.attached_comments
169 if self.lineinfo_last then
170 self.lineinfo_last.comments = self.attached_comments
173 self.attached_comments = { }
174 return setmetatable (a, self.token_metatable)
175 end --</function build_token>
177 for ext_idx, extractor in ipairs(self.extractors) do
178 -- printf("method = %s", method)
179 local tag, content = self [extractor] (self)
180 -- [loc] is placed just after the leading whitespaces and comments;
181 -- for this to work, the whitespace extractor *must be* at index 1.
182 if ext_idx==1 then loc = self.i end
185 --printf("`%s{ %q }\t%i", tag, content, loc);
186 return build_token (tag, content)
190 error "None of the lexer extractors returned anything!"
193 ----------------------------------------------------------------------
194 -- skip whites and comments
195 -- FIXME: doesn't take into account:
196 -- - unterminated long comments
197 -- - short comments at last line without a final \n
198 ----------------------------------------------------------------------
199 function lexer:skip_whitespaces_and_comments()
200 local table_insert = _G.table.insert
201 repeat -- loop as long as a space or comment chunk is found
204 local last_comment_content = nil
206 self.i = self.src:match (self.patterns.spaces, self.i)
207 -- skip a long comment if any
208 _, last_comment_content, j =
209 self.src :match (self.patterns.long_comment, self.i)
211 table_insert(self.attached_comments,
212 {last_comment_content, self.i, j, "long"})
215 -- skip a short comment if any
216 last_comment_content, j = self.src:match (self.patterns.short_comment, self.i)
218 table_insert(self.attached_comments,
219 {last_comment_content, self.i, j, "short"})
222 if self.i>#self.src then return "Eof", "eof" end
225 if self.src:match (self.patterns.final_short_comment, self.i) then
226 return "Eof", "eof" end
227 --assert (not self.src:match(self.patterns.short_comment, self.i))
228 --assert (not self.src:match(self.patterns.long_comment, self.i))
229 -- --assert (not self.src:match(self.patterns.spaces, self.i))
233 ----------------------------------------------------------------------
234 -- extract a '...' or "..." short string
235 ----------------------------------------------------------------------
236 function lexer:extract_short_string()
237 -- [k] is the first unread char, [self.i] points to [k] in [self.src]
238 local j, k = self.i, self.src :sub (self.i,self.i)
239 if k~="'" and k~='"' then return end
243 -- k = opening char: either simple-quote or double-quote
244 -- i = index of beginning-of-string
245 -- x = next "interesting" character
246 -- j = position after interesting char
247 -- y = char just after x
249 x, j, y = self.src :match ("([\\\r\n"..k.."])()(.?)", j)
250 if x == '\\' then j=j+1 -- don't parse escaped char
251 elseif x == k then break -- unescaped end of string
252 else -- eof or '\r' or '\n' reached before end of string
253 assert (not x or x=="\r" or x=="\n")
254 error "Unterminated string"
259 return "String", unescape_string (self.src:sub (i,j-2))
262 ----------------------------------------------------------------------
264 ----------------------------------------------------------------------
265 function lexer:extract_word()
267 local word, j = self.src:match (self.patterns.word, self.i)
270 if self.alpha [word] then return "Keyword", word
271 else return "Id", word end
275 ----------------------------------------------------------------------
277 ----------------------------------------------------------------------
278 function lexer:extract_number()
280 local j = self.src:match(self.patterns.number_hex, self.i)
282 j = self.src:match (self.patterns.number_mantissa[1], self.i) or
283 self.src:match (self.patterns.number_mantissa[2], self.i)
285 j = self.src:match (self.patterns.number_exponant, j) or j;
288 if not j then return end
289 -- Number found, interpret with tonumber() and return it
290 local n = tonumber (self.src:sub (self.i, j-1))
295 ----------------------------------------------------------------------
297 ----------------------------------------------------------------------
298 function lexer:extract_long_string()
300 local _, content, j = self.src:match (self.patterns.long_string, self.i)
301 if j then self.i = j; return "String", content end
304 ----------------------------------------------------------------------
306 ----------------------------------------------------------------------
307 function lexer:extract_symbol()
309 local k = self.src:sub (self.i,self.i)
310 local symk = self.sym [k]
315 for _, sym in pairs (symk) do
316 if sym == self.src:sub (self.i, self.i + #sym - 1) then
317 self.i = self.i + #sym;
318 return "Keyword", sym
321 -- single char symbol
326 ----------------------------------------------------------------------
327 -- Add a keyword to the list of keywords recognized by the lexer.
328 ----------------------------------------------------------------------
329 function lexer:add (w, ...)
330 assert(not ..., "lexer:add() takes only one arg, although possibly a table")
331 if type (w) == "table" then
332 for _, x in ipairs (w) do self:add (x) end
334 if w:match (self.patterns.word .. "$") then self.alpha [w] = true
335 elseif w:match "^%p%p+$" then
337 local list = self.sym [k]
338 if not list then list = { }; self.sym [k] = list end
339 _G.table.insert (list, w)
340 elseif w:match "^%p$" then return
341 else error "Invalid keyword" end
345 ----------------------------------------------------------------------
346 -- Return the [n]th next token, without consumming it.
347 -- [n] defaults to 1. If it goes pass the end of the stream, an EOF
348 -- token is returned.
349 ----------------------------------------------------------------------
350 function lexer:peek (n)
351 if not n then n=1 end
352 if n > #self.peeked then
353 for i = #self.peeked+1, n do
354 self.peeked [i] = self:extract()
357 return self.peeked [n]
360 ----------------------------------------------------------------------
361 -- Return the [n]th next token, removing it as well as the 0..n-1
362 -- previous tokens. [n] defaults to 1. If it goes pass the end of the
363 -- stream, an EOF token is returned.
364 ----------------------------------------------------------------------
365 function lexer:next (n)
370 a = _G.table.remove (self.peeked, 1)
372 --debugf ("lexer:next() ==> %s %s",
373 -- table.tostring(a), tostring(a))
375 self.lastline = a.lineinfo.last[1]
377 self.lineinfo_last = a.lineinfo.last
378 return a or eof_token
381 ----------------------------------------------------------------------
382 -- Returns an object which saves the stream's current state.
383 ----------------------------------------------------------------------
384 -- FIXME there are more fields than that to save
385 function lexer:save () return { self.i; _G.table.cat(self.peeked) } end
387 ----------------------------------------------------------------------
388 -- Restore the stream's state, as saved by method [save].
389 ----------------------------------------------------------------------
390 -- FIXME there are more fields than that to restore
391 function lexer:restore (s) self.i=s[1]; self.peeked=s[2] end
393 ----------------------------------------------------------------------
394 -- Resynchronize: cancel any token in self.peeked, by emptying the
395 -- list and resetting the indexes
396 ----------------------------------------------------------------------
397 function lexer:sync()
398 local p1 = self.peeked[1]
400 li = p1.lineinfo.first
401 self.line, self.i = li[1], li[3]
402 self.column_offset = self.i - li[2]
404 self.attached_comments = p1.lineinfo.first.comments or { }
408 ----------------------------------------------------------------------
409 -- Take the source and offset of an old lexer.
410 ----------------------------------------------------------------------
411 function lexer:takeover(old)
413 self.line, self.column_offset, self.i, self.src, self.attached_comments =
414 old.line, old.column_offset, old.i, old.src, old.attached_comments
418 -- function lexer:lineinfo()
419 -- if self.peeked[1] then return self.peeked[1].lineinfo.first
420 -- else return { self.line, self.i-self.column_offset, self.i } end
424 ----------------------------------------------------------------------
425 -- Return the current position in the sources. This position is between
426 -- two tokens, and can be within a space / comment area, and therefore
427 -- have a non-null width. :lineinfo_left() returns the beginning of the
428 -- separation area, :lineinfo_right() returns the end of that area.
430 -- ____ last consummed token ____ first unconsummed token
432 -- XXXXX <spaces and comments> YYYYY
434 -- :lineinfo_left() :lineinfo_right()
435 ----------------------------------------------------------------------
436 function lexer:lineinfo_right()
437 return self:peek(1).lineinfo.first
440 function lexer:lineinfo_left()
441 return self.lineinfo_last
444 ----------------------------------------------------------------------
445 -- Create a new lexstream.
446 ----------------------------------------------------------------------
447 function lexer:newstream (src_or_stream, name)
449 if type(src_or_stream)=='table' then -- it's a stream
450 return setmetatable ({ }, self) :takeover (src_or_stream)
451 elseif type(src_or_stream)=='string' then -- it's a source string
452 local src = src_or_stream
454 src_name = name; -- Name of the file
455 src = src; -- The source, as a single string
456 peeked = { }; -- Already peeked, but not discarded yet, tokens
457 i = 1; -- Character offset in src
458 line = 1; -- Current line number
459 column_offset = 0; -- distance from beginning of file to last '\n'
460 attached_comments = { },-- comments accumulator
461 lineinfo_last = { 1, 1, 1, name }
463 setmetatable (stream, self)
465 -- skip initial sharp-bang for unix scripts
466 -- FIXME: redundant with mlp.chunk()
467 if src and src :match "^#" then stream.i = src :find "\n" + 1 end
470 assert(false, ":newstream() takes a source string or a stream, not a "..
475 ----------------------------------------------------------------------
476 -- if there's no ... args, return the token a (whose truth value is
477 -- true) if it's a `Keyword{ }, or nil. If there are ... args, they
478 -- have to be strings. if the token a is a keyword, and it's content
479 -- is one of the ... args, then returns it (it's truth value is
480 -- true). If no a keyword or not in ..., return nil.
481 ----------------------------------------------------------------------
482 function lexer:is_keyword (a, ...)
483 if not a or a.tag ~= "Keyword" then return false end
485 if #words == 0 then return a[1] end
486 for _, w in ipairs (words) do
487 if w == a[1] then return w end
492 ----------------------------------------------------------------------
493 -- Cause an error if the next token isn't a keyword whose content
494 -- is listed among ... args (which have to be strings).
495 ----------------------------------------------------------------------
496 function lexer:check (...)
498 local a = self:next()
499 local function err ()
500 error ("Got " .. tostring (a) ..
501 ", expected one of these keywords : '" ..
502 _G.table.concat (words,"', '") .. "'") end
504 if not a or a.tag ~= "Keyword" then err () end
505 if #words == 0 then return a[1] end
506 for _, w in ipairs (words) do
507 if w == a[1] then return w end
512 ----------------------------------------------------------------------
514 ----------------------------------------------------------------------
515 function lexer:clone()
517 alpha = table.deep_copy(self.alpha),
518 sym = table.deep_copy(self.sym) }
519 setmetatable(clone, self)
520 clone.__index = clone