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 (backslashes, digits)
70 if #backslashes%2==0 then
71 -- Even number of backslashes, they escape each other, not the digits.
72 -- Return them so that unesc_letter() can treaat them
73 return backslashes..digits
75 -- Remove the odd backslash, which escapes the number sequence.
76 -- The rest will be returned and parsed by unesc_letter()
77 backslashes = backslashes :sub (1,-2)
79 local k, j, i = digits:reverse():byte(1, 3)
80 local z = _G.string.byte "0"
81 local code = (k or z) + 10*(j or z) + 100*(i or z) - 111*z
83 error ("Illegal escape sequence '\\"..digits.."' in string: ASCII codes must be in [0..255]")
85 return backslashes .. string.char (code)
88 -- Take a letter [x], and returns the character represented by the
89 -- sequence ['\\'..x], e.g. [unesc_letter "n" == "\n"].
90 local function unesc_letter(x)
92 a = "\a", b = "\b", f = "\f",
93 n = "\n", r = "\r", t = "\t", v = "\v",
94 ["\\"] = "\\", ["'"] = "'", ['"'] = '"', ["\n"] = "\n" }
95 return t[x] or error([[Unknown escape sequence '\]]..x..[[']])
99 :gsub ("(\\+)([0-9][0-9]?[0-9]?)", unesc_digits)
100 :gsub ("\\(%D)",unesc_letter)
104 "skip_whitespaces_and_comments",
105 "extract_short_string", "extract_word", "extract_number",
106 "extract_long_string", "extract_symbol" }
108 lexer.token_metatable = {
109 -- __tostring = function(a)
110 -- return string.format ("`%s{'%s'}",a.tag, a[1])
114 lexer.lineinfo_metatable = { }
116 -- The presence of this function prevents serialization by Pluto,
117 -- I can't figure out why :(
118 function lexer.lineinfo_metatable:__tostring()
119 local txt = string.format("%s:%i(%i,%i)", self[4], self[3], self[1], self[2])
120 if self.comments then
122 for comment in ivalues(self.comments) do
123 local content, loc1, loc2, kind = unpack(comment)
124 table.insert (acc, string.format ("%s@%i..%i:%q", kind, loc1, loc2, content))
126 txt = txt.."["..table.concat(acc,"; ").."]"
132 ----------------------------------------------------------------------
133 -- Really extract next token fron the raw string
134 -- (and update the index).
135 -- loc: offset of the position just after spaces and comments
136 -- previous_i: offset in src before extraction began
137 ----------------------------------------------------------------------
138 function lexer:extract ()
139 local previous_i = self.i
143 -- Put line info, comments and metatable arount the tag and content
144 -- provided by extractors, thus returning a complete lexer token.
145 -- first_line: line # at the beginning of token
146 -- first_column_offset: char # of the last '\n' before beginning of token
147 -- i: scans from beginning of prefix spaces/comments to end of token.
148 local function build_token (tag, content)
149 assert (tag and content)
150 local i, first_line, first_column_offset, previous_line_length =
151 previous_i, self.line, self.column_offset, nil
153 -- update self.line and first_line. i := indexes of '\n' chars
155 i = self.src :find ("\n", i+1, true)
156 if not i or i>self.i then break end -- no more '\n' until end of token
157 previous_line_length = i - self.column_offset
158 if loc and i <= loc then -- '\n' before beginning of token
159 first_column_offset = i
160 first_line = first_line+1
162 self.line = self.line+1
163 self.column_offset = i
166 -- lineinfo entries: [1]=line, [2]=column, [3]=char, [4]=filename
167 local fli = { first_line, loc-first_column_offset, loc, self.src_name }
168 local lli = { self.line, self.i-self.column_offset-1, self.i-1, self.src_name }
169 --Pluto barfes when the metatable is set:(
170 setmetatable(fli, lexer.lineinfo_metatable)
171 setmetatable(lli, lexer.lineinfo_metatable)
172 local a = { tag = tag, lineinfo = { first=fli, last=lli }, content }
173 if lli[2]==-1 then lli[1], lli[2] = lli[1]-1, previous_line_length-1 end
174 if #self.attached_comments > 0 then
175 a.lineinfo.comments = self.attached_comments
176 fli.comments = self.attached_comments
177 if self.lineinfo_last then
178 self.lineinfo_last.comments = self.attached_comments
181 self.attached_comments = { }
182 return setmetatable (a, self.token_metatable)
183 end --</function build_token>
185 for ext_idx, extractor in ipairs(self.extractors) do
186 -- printf("method = %s", method)
187 local tag, content = self [extractor] (self)
188 -- [loc] is placed just after the leading whitespaces and comments;
189 -- for this to work, the whitespace extractor *must be* at index 1.
190 if ext_idx==1 then loc = self.i end
193 --printf("`%s{ %q }\t%i", tag, content, loc);
194 return build_token (tag, content)
198 error "None of the lexer extractors returned anything!"
201 ----------------------------------------------------------------------
202 -- skip whites and comments
203 -- FIXME: doesn't take into account:
204 -- - unterminated long comments
205 -- - short comments at last line without a final \n
206 ----------------------------------------------------------------------
207 function lexer:skip_whitespaces_and_comments()
208 local table_insert = _G.table.insert
209 repeat -- loop as long as a space or comment chunk is found
212 local last_comment_content = nil
214 self.i = self.src:match (self.patterns.spaces, self.i)
215 -- skip a long comment if any
216 _, last_comment_content, j =
217 self.src :match (self.patterns.long_comment, self.i)
219 table_insert(self.attached_comments,
220 {last_comment_content, self.i, j, "long"})
223 -- skip a short comment if any
224 last_comment_content, j = self.src:match (self.patterns.short_comment, self.i)
226 table_insert(self.attached_comments,
227 {last_comment_content, self.i, j, "short"})
230 if self.i>#self.src then return "Eof", "eof" end
233 if self.src:match (self.patterns.final_short_comment, self.i) then
234 return "Eof", "eof" end
235 --assert (not self.src:match(self.patterns.short_comment, self.i))
236 --assert (not self.src:match(self.patterns.long_comment, self.i))
237 -- --assert (not self.src:match(self.patterns.spaces, self.i))
241 ----------------------------------------------------------------------
242 -- extract a '...' or "..." short string
243 ----------------------------------------------------------------------
244 function lexer:extract_short_string()
245 -- [k] is the first unread char, [self.i] points to [k] in [self.src]
246 local j, k = self.i, self.src :sub (self.i,self.i)
247 if k~="'" and k~='"' then return end
251 -- k = opening char: either simple-quote or double-quote
252 -- i = index of beginning-of-string
253 -- x = next "interesting" character
254 -- j = position after interesting char
255 -- y = char just after x
257 x, j, y = self.src :match ("([\\\r\n"..k.."])()(.?)", j)
258 if x == '\\' then j=j+1 -- don't parse escaped char
259 elseif x == k then break -- unescaped end of string
260 else -- eof or '\r' or '\n' reached before end of string
261 assert (not x or x=="\r" or x=="\n")
262 error "Unterminated string"
267 return "String", unescape_string (self.src:sub (i,j-2))
270 ----------------------------------------------------------------------
272 ----------------------------------------------------------------------
273 function lexer:extract_word()
275 local word, j = self.src:match (self.patterns.word, self.i)
278 if self.alpha [word] then return "Keyword", word
279 else return "Id", word end
283 ----------------------------------------------------------------------
285 ----------------------------------------------------------------------
286 function lexer:extract_number()
288 local j = self.src:match(self.patterns.number_hex, self.i)
290 j = self.src:match (self.patterns.number_mantissa[1], self.i) or
291 self.src:match (self.patterns.number_mantissa[2], self.i)
293 j = self.src:match (self.patterns.number_exponant, j) or j;
296 if not j then return end
297 -- Number found, interpret with tonumber() and return it
298 local n = tonumber (self.src:sub (self.i, j-1))
303 ----------------------------------------------------------------------
305 ----------------------------------------------------------------------
306 function lexer:extract_long_string()
308 local _, content, j = self.src:match (self.patterns.long_string, self.i)
309 if j then self.i = j; return "String", content end
312 ----------------------------------------------------------------------
314 ----------------------------------------------------------------------
315 function lexer:extract_symbol()
317 local k = self.src:sub (self.i,self.i)
318 local symk = self.sym [k]
323 for _, sym in pairs (symk) do
324 if sym == self.src:sub (self.i, self.i + #sym - 1) then
325 self.i = self.i + #sym;
326 return "Keyword", sym
329 -- single char symbol
334 ----------------------------------------------------------------------
335 -- Add a keyword to the list of keywords recognized by the lexer.
336 ----------------------------------------------------------------------
337 function lexer:add (w, ...)
338 assert(not ..., "lexer:add() takes only one arg, although possibly a table")
339 if type (w) == "table" then
340 for _, x in ipairs (w) do self:add (x) end
342 if w:match (self.patterns.word .. "$") then self.alpha [w] = true
343 elseif w:match "^%p%p+$" then
345 local list = self.sym [k]
346 if not list then list = { }; self.sym [k] = list end
347 _G.table.insert (list, w)
348 elseif w:match "^%p$" then return
349 else error "Invalid keyword" end
353 ----------------------------------------------------------------------
354 -- Return the [n]th next token, without consumming it.
355 -- [n] defaults to 1. If it goes pass the end of the stream, an EOF
356 -- token is returned.
357 ----------------------------------------------------------------------
358 function lexer:peek (n)
359 if not n then n=1 end
360 if n > #self.peeked then
361 for i = #self.peeked+1, n do
362 self.peeked [i] = self:extract()
365 return self.peeked [n]
368 ----------------------------------------------------------------------
369 -- Return the [n]th next token, removing it as well as the 0..n-1
370 -- previous tokens. [n] defaults to 1. If it goes pass the end of the
371 -- stream, an EOF token is returned.
372 ----------------------------------------------------------------------
373 function lexer:next (n)
378 a = _G.table.remove (self.peeked, 1)
380 --debugf ("lexer:next() ==> %s %s",
381 -- table.tostring(a), tostring(a))
383 self.lastline = a.lineinfo.last[1]
385 self.lineinfo_last = a.lineinfo.last
386 return a or eof_token
389 ----------------------------------------------------------------------
390 -- Returns an object which saves the stream's current state.
391 ----------------------------------------------------------------------
392 -- FIXME there are more fields than that to save
393 function lexer:save () return { self.i; _G.table.cat(self.peeked) } end
395 ----------------------------------------------------------------------
396 -- Restore the stream's state, as saved by method [save].
397 ----------------------------------------------------------------------
398 -- FIXME there are more fields than that to restore
399 function lexer:restore (s) self.i=s[1]; self.peeked=s[2] end
401 ----------------------------------------------------------------------
402 -- Resynchronize: cancel any token in self.peeked, by emptying the
403 -- list and resetting the indexes
404 ----------------------------------------------------------------------
405 function lexer:sync()
406 local p1 = self.peeked[1]
408 li = p1.lineinfo.first
409 self.line, self.i = li[1], li[3]
410 self.column_offset = self.i - li[2]
412 self.attached_comments = p1.lineinfo.first.comments or { }
416 ----------------------------------------------------------------------
417 -- Take the source and offset of an old lexer.
418 ----------------------------------------------------------------------
419 function lexer:takeover(old)
421 self.line, self.column_offset, self.i, self.src, self.attached_comments =
422 old.line, old.column_offset, old.i, old.src, old.attached_comments
426 -- function lexer:lineinfo()
427 -- if self.peeked[1] then return self.peeked[1].lineinfo.first
428 -- else return { self.line, self.i-self.column_offset, self.i } end
432 ----------------------------------------------------------------------
433 -- Return the current position in the sources. This position is between
434 -- two tokens, and can be within a space / comment area, and therefore
435 -- have a non-null width. :lineinfo_left() returns the beginning of the
436 -- separation area, :lineinfo_right() returns the end of that area.
438 -- ____ last consummed token ____ first unconsummed token
440 -- XXXXX <spaces and comments> YYYYY
442 -- :lineinfo_left() :lineinfo_right()
443 ----------------------------------------------------------------------
444 function lexer:lineinfo_right()
445 return self:peek(1).lineinfo.first
448 function lexer:lineinfo_left()
449 return self.lineinfo_last
452 ----------------------------------------------------------------------
453 -- Create a new lexstream.
454 ----------------------------------------------------------------------
455 function lexer:newstream (src_or_stream, name)
457 if type(src_or_stream)=='table' then -- it's a stream
458 return setmetatable ({ }, self) :takeover (src_or_stream)
459 elseif type(src_or_stream)=='string' then -- it's a source string
460 local src = src_or_stream
462 src_name = name; -- Name of the file
463 src = src; -- The source, as a single string
464 peeked = { }; -- Already peeked, but not discarded yet, tokens
465 i = 1; -- Character offset in src
466 line = 1; -- Current line number
467 column_offset = 0; -- distance from beginning of file to last '\n'
468 attached_comments = { },-- comments accumulator
469 lineinfo_last = { 1, 1, 1, name }
471 setmetatable (stream, self)
473 -- skip initial sharp-bang for unix scripts
474 -- FIXME: redundant with mlp.chunk()
475 if src and src :match "^#" then stream.i = src :find "\n" + 1 end
478 assert(false, ":newstream() takes a source string or a stream, not a "..
483 ----------------------------------------------------------------------
484 -- if there's no ... args, return the token a (whose truth value is
485 -- true) if it's a `Keyword{ }, or nil. If there are ... args, they
486 -- have to be strings. if the token a is a keyword, and it's content
487 -- is one of the ... args, then returns it (it's truth value is
488 -- true). If no a keyword or not in ..., return nil.
489 ----------------------------------------------------------------------
490 function lexer:is_keyword (a, ...)
491 if not a or a.tag ~= "Keyword" then return false end
493 if #words == 0 then return a[1] end
494 for _, w in ipairs (words) do
495 if w == a[1] then return w end
500 ----------------------------------------------------------------------
501 -- Cause an error if the next token isn't a keyword whose content
502 -- is listed among ... args (which have to be strings).
503 ----------------------------------------------------------------------
504 function lexer:check (...)
506 local a = self:next()
507 local function err ()
508 error ("Got " .. tostring (a) ..
509 ", expected one of these keywords : '" ..
510 _G.table.concat (words,"', '") .. "'") end
512 if not a or a.tag ~= "Keyword" then err () end
513 if #words == 0 then return a[1] end
514 for _, w in ipairs (words) do
515 if w == a[1] then return w end
520 ----------------------------------------------------------------------
522 ----------------------------------------------------------------------
523 function lexer:clone()
525 alpha = table.deep_copy(self.alpha),
526 sym = table.deep_copy(self.sym) }
527 setmetatable(clone, self)
528 clone.__index = clone