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 local k, j, i = x:reverse():byte(1, 3)
71 local z = _G.string.byte "0"
72 return _G.string.char ((k or z) + 10*(j or z) + 100*(i or z) - 111*z)
75 -- Take a letter [x], and returns the character represented by the
76 -- sequence ['\\'..x], e.g. [unesc_letter "n" == "\n"].
77 local function unesc_letter(x)
79 a = "\a", b = "\b", f = "\f",
80 n = "\n", r = "\r", t = "\t", v = "\v",
81 ["\\"] = "\\", ["'"] = "'", ['"'] = '"', ["\n"] = "\n" }
82 return t[x] or error([[Unknown escape sequence '\]]..x..[[']])
86 :gsub ("\\(%D)",unesc_letter)
87 :gsub ("\\([0-9][0-9]?[0-9]?)", unesc_digits)
91 "skip_whitespaces_and_comments",
92 "extract_short_string", "extract_word", "extract_number",
93 "extract_long_string", "extract_symbol" }
95 lexer.token_metatable = {
96 -- __tostring = function(a)
97 -- return string.format ("`%s{'%s'}",a.tag, a[1])
101 lexer.lineinfo_metatable = { }
103 -- The presence of this function prevents serialization by Pluto,
104 -- I can't figure out why :(
105 function lexer.lineinfo_metatable:__tostring()
106 local txt = string.format("%s:%i(%i,%i)", self[4], self[3], self[1], self[2])
107 if self.comments then
109 for comment in ivalues(self.comments) do
110 local content, loc1, loc2, kind = unpack(comment)
111 table.insert (acc, string.format ("%s@%i..%i:%q", kind, loc1, loc2, content))
113 txt = txt.."["..table.concat(acc,"; ").."]"
119 ----------------------------------------------------------------------
120 -- Really extract next token fron the raw string
121 -- (and update the index).
122 -- loc: offset of the position just after spaces and comments
123 -- previous_i: offset in src before extraction began
124 ----------------------------------------------------------------------
125 function lexer:extract ()
126 local previous_i = self.i
130 -- Put line info, comments and metatable arount the tag and content
131 -- provided by extractors, thus returning a complete lexer token.
132 -- first_line: line # at the beginning of token
133 -- first_column_offset: char # of the last '\n' before beginning of token
134 -- i: scans from beginning of prefix spaces/comments to end of token.
135 local function build_token (tag, content)
136 assert (tag and content)
137 local i, first_line, first_column_offset, previous_line_length =
138 previous_i, self.line, self.column_offset, nil
140 -- update self.line and first_line. i := indexes of '\n' chars
142 i = self.src :find ("\n", i+1, true)
143 if not i or i>self.i then break end -- no more '\n' until end of token
144 previous_line_length = i - self.column_offset
145 if loc and i <= loc then -- '\n' before beginning of token
146 first_column_offset = i
147 first_line = first_line+1
149 self.line = self.line+1
150 self.column_offset = i
153 -- lineinfo entries: [1]=line, [2]=column, [3]=char, [4]=filename
154 local fli = { first_line, loc-first_column_offset, loc, self.src_name }
155 local lli = { self.line, self.i-self.column_offset-1, self.i-1, self.src_name }
156 --Pluto barfes when the metatable is set:(
157 setmetatable(fli, lexer.lineinfo_metatable)
158 setmetatable(lli, lexer.lineinfo_metatable)
159 local a = { tag = tag, lineinfo = { first=fli, last=lli }, content }
160 if lli[2]==-1 then lli[1], lli[2] = lli[1]-1, previous_line_length-1 end
161 if #self.attached_comments > 0 then
162 a.lineinfo.comments = self.attached_comments
163 fli.comments = self.attached_comments
164 if self.lineinfo_last then
165 self.lineinfo_last.comments = self.attached_comments
168 self.attached_comments = { }
169 return setmetatable (a, self.token_metatable)
170 end --</function build_token>
172 for ext_idx, extractor in ipairs(self.extractors) do
173 -- printf("method = %s", method)
174 local tag, content = self [extractor] (self)
175 -- [loc] is placed just after the leading whitespaces and comments;
176 -- for this to work, the whitespace extractor *must be* at index 1.
177 if ext_idx==1 then loc = self.i end
180 --printf("`%s{ %q }\t%i", tag, content, loc);
181 return build_token (tag, content)
185 error "None of the lexer extractors returned anything!"
188 ----------------------------------------------------------------------
189 -- skip whites and comments
190 -- FIXME: doesn't take into account:
191 -- - unterminated long comments
192 -- - short comments at last line without a final \n
193 ----------------------------------------------------------------------
194 function lexer:skip_whitespaces_and_comments()
195 local table_insert = _G.table.insert
196 repeat -- loop as long as a space or comment chunk is found
199 local last_comment_content = nil
201 self.i = self.src:match (self.patterns.spaces, self.i)
202 -- skip a long comment if any
203 _, last_comment_content, j =
204 self.src :match (self.patterns.long_comment, self.i)
206 table_insert(self.attached_comments,
207 {last_comment_content, self.i, j, "long"})
210 -- skip a short comment if any
211 last_comment_content, j = self.src:match (self.patterns.short_comment, self.i)
213 table_insert(self.attached_comments,
214 {last_comment_content, self.i, j, "short"})
217 if self.i>#self.src then return "Eof", "eof" end
220 if self.src:match (self.patterns.final_short_comment, self.i) then
221 return "Eof", "eof" end
222 --assert (not self.src:match(self.patterns.short_comment, self.i))
223 --assert (not self.src:match(self.patterns.long_comment, self.i))
224 -- --assert (not self.src:match(self.patterns.spaces, self.i))
228 ----------------------------------------------------------------------
229 -- extract a '...' or "..." short string
230 ----------------------------------------------------------------------
231 function lexer:extract_short_string()
232 -- [k] is the first unread char, [self.i] points to [k] in [self.src]
233 local j, k = self.i, self.src :sub (self.i,self.i)
234 if k~="'" and k~='"' then return end
238 -- k = opening char: either simple-quote or double-quote
239 -- i = index of beginning-of-string
240 -- x = next "interesting" character
241 -- j = position after interesting char
242 -- y = char just after x
244 x, j, y = self.src :match ("([\\\r\n"..k.."])()(.?)", j)
245 if x == '\\' then j=j+1 -- don't parse escaped char
246 elseif x == k then break -- unescaped end of string
247 else -- eof or '\r' or '\n' reached before end of string
248 assert (not x or x=="\r" or x=="\n")
249 error "Unterminated string"
254 return "String", unescape_string (self.src:sub (i,j-2))
257 ----------------------------------------------------------------------
259 ----------------------------------------------------------------------
260 function lexer:extract_word()
262 local word, j = self.src:match (self.patterns.word, self.i)
265 if self.alpha [word] then return "Keyword", word
266 else return "Id", word end
270 ----------------------------------------------------------------------
272 ----------------------------------------------------------------------
273 function lexer:extract_number()
275 local j = self.src:match(self.patterns.number_hex, self.i)
277 j = self.src:match (self.patterns.number_mantissa[1], self.i) or
278 self.src:match (self.patterns.number_mantissa[2], self.i)
280 j = self.src:match (self.patterns.number_exponant, j) or j;
283 if not j then return end
284 -- Number found, interpret with tonumber() and return it
285 local n = tonumber (self.src:sub (self.i, j-1))
290 ----------------------------------------------------------------------
292 ----------------------------------------------------------------------
293 function lexer:extract_long_string()
295 local _, content, j = self.src:match (self.patterns.long_string, self.i)
296 if j then self.i = j; return "String", content end
299 ----------------------------------------------------------------------
301 ----------------------------------------------------------------------
302 function lexer:extract_symbol()
304 local k = self.src:sub (self.i,self.i)
305 local symk = self.sym [k]
310 for _, sym in pairs (symk) do
311 if sym == self.src:sub (self.i, self.i + #sym - 1) then
312 self.i = self.i + #sym;
313 return "Keyword", sym
316 -- single char symbol
321 ----------------------------------------------------------------------
322 -- Add a keyword to the list of keywords recognized by the lexer.
323 ----------------------------------------------------------------------
324 function lexer:add (w, ...)
325 assert(not ..., "lexer:add() takes only one arg, although possibly a table")
326 if type (w) == "table" then
327 for _, x in ipairs (w) do self:add (x) end
329 if w:match (self.patterns.word .. "$") then self.alpha [w] = true
330 elseif w:match "^%p%p+$" then
332 local list = self.sym [k]
333 if not list then list = { }; self.sym [k] = list end
334 _G.table.insert (list, w)
335 elseif w:match "^%p$" then return
336 else error "Invalid keyword" end
340 ----------------------------------------------------------------------
341 -- Return the [n]th next token, without consumming it.
342 -- [n] defaults to 1. If it goes pass the end of the stream, an EOF
343 -- token is returned.
344 ----------------------------------------------------------------------
345 function lexer:peek (n)
346 if not n then n=1 end
347 if n > #self.peeked then
348 for i = #self.peeked+1, n do
349 self.peeked [i] = self:extract()
352 return self.peeked [n]
355 ----------------------------------------------------------------------
356 -- Return the [n]th next token, removing it as well as the 0..n-1
357 -- previous tokens. [n] defaults to 1. If it goes pass the end of the
358 -- stream, an EOF token is returned.
359 ----------------------------------------------------------------------
360 function lexer:next (n)
365 a = _G.table.remove (self.peeked, 1)
367 --debugf ("lexer:next() ==> %s %s",
368 -- table.tostring(a), tostring(a))
370 self.lastline = a.lineinfo.last[1]
372 self.lineinfo_last = a.lineinfo.last
373 return a or eof_token
376 ----------------------------------------------------------------------
377 -- Returns an object which saves the stream's current state.
378 ----------------------------------------------------------------------
379 -- FIXME there are more fields than that to save
380 function lexer:save () return { self.i; _G.table.cat(self.peeked) } end
382 ----------------------------------------------------------------------
383 -- Restore the stream's state, as saved by method [save].
384 ----------------------------------------------------------------------
385 -- FIXME there are more fields than that to restore
386 function lexer:restore (s) self.i=s[1]; self.peeked=s[2] end
388 ----------------------------------------------------------------------
389 -- Resynchronize: cancel any token in self.peeked, by emptying the
390 -- list and resetting the indexes
391 ----------------------------------------------------------------------
392 function lexer:sync()
393 local p1 = self.peeked[1]
395 li = p1.lineinfo.first
396 self.line, self.i = li[1], li[3]
397 self.column_offset = self.i - li[2]
399 self.attached_comments = p1.lineinfo.first.comments or { }
403 ----------------------------------------------------------------------
404 -- Take the source and offset of an old lexer.
405 ----------------------------------------------------------------------
406 function lexer:takeover(old)
408 self.line, self.column_offset, self.i, self.src, self.attached_comments =
409 old.line, old.column_offset, old.i, old.src, old.attached_comments
413 -- function lexer:lineinfo()
414 -- if self.peeked[1] then return self.peeked[1].lineinfo.first
415 -- else return { self.line, self.i-self.column_offset, self.i } end
419 ----------------------------------------------------------------------
420 -- Return the current position in the sources. This position is between
421 -- two tokens, and can be within a space / comment area, and therefore
422 -- have a non-null width. :lineinfo_left() returns the beginning of the
423 -- separation area, :lineinfo_right() returns the end of that area.
425 -- ____ last consummed token ____ first unconsummed token
427 -- XXXXX <spaces and comments> YYYYY
429 -- :lineinfo_left() :lineinfo_right()
430 ----------------------------------------------------------------------
431 function lexer:lineinfo_right()
432 return self:peek(1).lineinfo.first
435 function lexer:lineinfo_left()
436 return self.lineinfo_last
439 ----------------------------------------------------------------------
440 -- Create a new lexstream.
441 ----------------------------------------------------------------------
442 function lexer:newstream (src_or_stream, name)
444 if type(src_or_stream)=='table' then -- it's a stream
445 return setmetatable ({ }, self) :takeover (src_or_stream)
446 elseif type(src_or_stream)=='string' then -- it's a source string
447 local src = src_or_stream
449 src_name = name; -- Name of the file
450 src = src; -- The source, as a single string
451 peeked = { }; -- Already peeked, but not discarded yet, tokens
452 i = 1; -- Character offset in src
453 line = 1; -- Current line number
454 column_offset = 0; -- distance from beginning of file to last '\n'
455 attached_comments = { },-- comments accumulator
456 lineinfo_last = { 1, 1, 1, name }
458 setmetatable (stream, self)
460 -- skip initial sharp-bang for unix scripts
461 -- FIXME: redundant with mlp.chunk()
462 if src and src :match "^#" then stream.i = src :find "\n" + 1 end
465 assert(false, ":newstream() takes a source string or a stream, not a "..
470 ----------------------------------------------------------------------
471 -- if there's no ... args, return the token a (whose truth value is
472 -- true) if it's a `Keyword{ }, or nil. If there are ... args, they
473 -- have to be strings. if the token a is a keyword, and it's content
474 -- is one of the ... args, then returns it (it's truth value is
475 -- true). If no a keyword or not in ..., return nil.
476 ----------------------------------------------------------------------
477 function lexer:is_keyword (a, ...)
478 if not a or a.tag ~= "Keyword" then return false end
480 if #words == 0 then return a[1] end
481 for _, w in ipairs (words) do
482 if w == a[1] then return w end
487 ----------------------------------------------------------------------
488 -- Cause an error if the next token isn't a keyword whose content
489 -- is listed among ... args (which have to be strings).
490 ----------------------------------------------------------------------
491 function lexer:check (...)
493 local a = self:next()
494 local function err ()
495 error ("Got " .. tostring (a) ..
496 ", expected one of these keywords : '" ..
497 _G.table.concat (words,"', '") .. "'") end
499 if not a or a.tag ~= "Keyword" then err () end
500 if #words == 0 then return a[1] end
501 for _, w in ipairs (words) do
502 if w == a[1] then return w end
507 ----------------------------------------------------------------------
509 ----------------------------------------------------------------------
510 function lexer:clone()
512 alpha = table.deep_copy(self.alpha),
513 sym = table.deep_copy(self.sym) }
514 setmetatable(clone, self)
515 clone.__index = clone