1 | /* |
2 | ** Lexical analyzer. |
3 | ** Copyright (C) 2005-2014 Mike Pall. See Copyright Notice in luajit.h |
4 | ** |
5 | ** Major portions taken verbatim or adapted from the Lua interpreter. |
6 | ** Copyright (C) 1994-2008 Lua.org, PUC-Rio. See Copyright Notice in lua.h |
7 | */ |
8 | |
9 | #define lj_lex_c |
10 | #define LUA_CORE |
11 | |
12 | #include "lj_obj.h" |
13 | #include "lj_gc.h" |
14 | #include "lj_err.h" |
15 | #include "lj_str.h" |
16 | #if LJ_HASFFI |
17 | #include "lj_tab.h" |
18 | #include "lj_ctype.h" |
19 | #include "lj_cdata.h" |
20 | #include "lualib.h" |
21 | #endif |
22 | #include "lj_state.h" |
23 | #include "lj_lex.h" |
24 | #include "lj_parse.h" |
25 | #include "lj_char.h" |
26 | #include "lj_strscan.h" |
27 | |
28 | /* Lua lexer token names. */ |
29 | static const char *const tokennames[] = { |
30 | #define TKSTR1(name) #name, |
31 | #define TKSTR2(name, sym) #sym, |
32 | TKDEF(TKSTR1, TKSTR2) |
33 | #undef TKSTR1 |
34 | #undef TKSTR2 |
35 | NULL |
36 | }; |
37 | |
38 | /* -- Buffer handling ----------------------------------------------------- */ |
39 | |
40 | #define char2int(c) ((int)(uint8_t)(c)) |
41 | #define next(ls) \ |
42 | (ls->current = (ls->n--) > 0 ? char2int(*ls->p++) : fillbuf(ls)) |
43 | #define save_and_next(ls) (save(ls, ls->current), next(ls)) |
44 | #define currIsNewline(ls) (ls->current == '\n' || ls->current == '\r') |
45 | #define END_OF_STREAM (-1) |
46 | |
47 | static int fillbuf(LexState *ls) |
48 | { |
49 | size_t sz; |
50 | const char *buf = ls->rfunc(ls->L, ls->rdata, &sz); |
51 | if (buf == NULL || sz == 0) return END_OF_STREAM; |
52 | ls->n = (MSize)sz - 1; |
53 | ls->p = buf; |
54 | return char2int(*(ls->p++)); |
55 | } |
56 | |
57 | static LJ_NOINLINE void save_grow(LexState *ls, int c) |
58 | { |
59 | MSize newsize; |
60 | if (ls->sb.sz >= LJ_MAX_STR/2) |
61 | lj_lex_error(ls, 0, LJ_ERR_XELEM); |
62 | newsize = ls->sb.sz * 2; |
63 | lj_str_resizebuf(ls->L, &ls->sb, newsize); |
64 | ls->sb.buf[ls->sb.n++] = (char)c; |
65 | } |
66 | |
67 | static LJ_AINLINE void save(LexState *ls, int c) |
68 | { |
69 | if (LJ_UNLIKELY(ls->sb.n + 1 > ls->sb.sz)) |
70 | save_grow(ls, c); |
71 | else |
72 | ls->sb.buf[ls->sb.n++] = (char)c; |
73 | } |
74 | |
75 | static void inclinenumber(LexState *ls) |
76 | { |
77 | int old = ls->current; |
78 | lua_assert(currIsNewline(ls)); |
79 | next(ls); /* skip `\n' or `\r' */ |
80 | if (currIsNewline(ls) && ls->current != old) |
81 | next(ls); /* skip `\n\r' or `\r\n' */ |
82 | if (++ls->linenumber >= LJ_MAX_LINE) |
83 | lj_lex_error(ls, ls->token, LJ_ERR_XLINES); |
84 | } |
85 | |
86 | /* -- Scanner for terminals ----------------------------------------------- */ |
87 | |
88 | /* Parse a number literal. */ |
89 | static void lex_number(LexState *ls, TValue *tv) |
90 | { |
91 | StrScanFmt fmt; |
92 | int c, xp = 'e'; |
93 | lua_assert(lj_char_isdigit(ls->current)); |
94 | if ((c = ls->current) == '0') { |
95 | save_and_next(ls); |
96 | if ((ls->current | 0x20) == 'x') xp = 'p'; |
97 | } |
98 | while (lj_char_isident(ls->current) || ls->current == '.' || |
99 | ((ls->current == '-' || ls->current == '+') && (c | 0x20) == xp)) { |
100 | c = ls->current; |
101 | save_and_next(ls); |
102 | } |
103 | save(ls, '\0'); |
104 | fmt = lj_strscan_scan((const uint8_t *)ls->sb.buf, tv, |
105 | (LJ_DUALNUM ? STRSCAN_OPT_TOINT : STRSCAN_OPT_TONUM) | |
106 | (LJ_HASFFI ? (STRSCAN_OPT_LL|STRSCAN_OPT_IMAG) : 0)); |
107 | if (LJ_DUALNUM && fmt == STRSCAN_INT) { |
108 | setitype(tv, LJ_TISNUM); |
109 | } else if (fmt == STRSCAN_NUM) { |
110 | /* Already in correct format. */ |
111 | #if LJ_HASFFI |
112 | } else if (fmt != STRSCAN_ERROR) { |
113 | lua_State *L = ls->L; |
114 | GCcdata *cd; |
115 | lua_assert(fmt == STRSCAN_I64 || fmt == STRSCAN_U64 || fmt == STRSCAN_IMAG); |
116 | if (!ctype_ctsG(G(L))) { |
117 | ptrdiff_t oldtop = savestack(L, L->top); |
118 | luaopen_ffi(L); /* Load FFI library on-demand. */ |
119 | L->top = restorestack(L, oldtop); |
120 | } |
121 | if (fmt == STRSCAN_IMAG) { |
122 | cd = lj_cdata_new_(L, CTID_COMPLEX_DOUBLE, 2*sizeof(double)); |
123 | ((double *)cdataptr(cd))[0] = 0; |
124 | ((double *)cdataptr(cd))[1] = numV(tv); |
125 | } else { |
126 | cd = lj_cdata_new_(L, fmt==STRSCAN_I64 ? CTID_INT64 : CTID_UINT64, 8); |
127 | *(uint64_t *)cdataptr(cd) = tv->u64; |
128 | } |
129 | lj_parse_keepcdata(ls, tv, cd); |
130 | #endif |
131 | } else { |
132 | lua_assert(fmt == STRSCAN_ERROR); |
133 | lj_lex_error(ls, TK_number, LJ_ERR_XNUMBER); |
134 | } |
135 | } |
136 | |
137 | static int skip_sep(LexState *ls) |
138 | { |
139 | int count = 0; |
140 | int s = ls->current; |
141 | lua_assert(s == '[' || s == ']'); |
142 | save_and_next(ls); |
143 | while (ls->current == '=') { |
144 | save_and_next(ls); |
145 | count++; |
146 | } |
147 | return (ls->current == s) ? count : (-count) - 1; |
148 | } |
149 | |
150 | static void read_long_string(LexState *ls, TValue *tv, int sep) |
151 | { |
152 | save_and_next(ls); /* skip 2nd `[' */ |
153 | if (currIsNewline(ls)) /* string starts with a newline? */ |
154 | inclinenumber(ls); /* skip it */ |
155 | for (;;) { |
156 | switch (ls->current) { |
157 | case END_OF_STREAM: |
158 | lj_lex_error(ls, TK_eof, tv ? LJ_ERR_XLSTR : LJ_ERR_XLCOM); |
159 | break; |
160 | case ']': |
161 | if (skip_sep(ls) == sep) { |
162 | save_and_next(ls); /* skip 2nd `]' */ |
163 | goto endloop; |
164 | } |
165 | break; |
166 | case '\n': |
167 | case '\r': |
168 | save(ls, '\n'); |
169 | inclinenumber(ls); |
170 | if (!tv) lj_str_resetbuf(&ls->sb); /* avoid wasting space */ |
171 | break; |
172 | default: |
173 | if (tv) save_and_next(ls); |
174 | else next(ls); |
175 | break; |
176 | } |
177 | } endloop: |
178 | if (tv) { |
179 | GCstr *str = lj_parse_keepstr(ls, ls->sb.buf + (2 + (MSize)sep), |
180 | ls->sb.n - 2*(2 + (MSize)sep)); |
181 | setstrV(ls->L, tv, str); |
182 | } |
183 | } |
184 | |
185 | static void read_string(LexState *ls, int delim, TValue *tv) |
186 | { |
187 | save_and_next(ls); |
188 | while (ls->current != delim) { |
189 | switch (ls->current) { |
190 | case END_OF_STREAM: |
191 | lj_lex_error(ls, TK_eof, LJ_ERR_XSTR); |
192 | continue; |
193 | case '\n': |
194 | case '\r': |
195 | lj_lex_error(ls, TK_string, LJ_ERR_XSTR); |
196 | continue; |
197 | case '\\': { |
198 | int c = next(ls); /* Skip the '\\'. */ |
199 | switch (c) { |
200 | case 'a': c = '\a'; break; |
201 | case 'b': c = '\b'; break; |
202 | case 'f': c = '\f'; break; |
203 | case 'n': c = '\n'; break; |
204 | case 'r': c = '\r'; break; |
205 | case 't': c = '\t'; break; |
206 | case 'v': c = '\v'; break; |
207 | case 'x': /* Hexadecimal escape '\xXX'. */ |
208 | c = (next(ls) & 15u) << 4; |
209 | if (!lj_char_isdigit(ls->current)) { |
210 | if (!lj_char_isxdigit(ls->current)) goto err_xesc; |
211 | c += 9 << 4; |
212 | } |
213 | c += (next(ls) & 15u); |
214 | if (!lj_char_isdigit(ls->current)) { |
215 | if (!lj_char_isxdigit(ls->current)) goto err_xesc; |
216 | c += 9; |
217 | } |
218 | break; |
219 | case 'z': /* Skip whitespace. */ |
220 | next(ls); |
221 | while (lj_char_isspace(ls->current)) |
222 | if (currIsNewline(ls)) inclinenumber(ls); else next(ls); |
223 | continue; |
224 | case '\n': case '\r': save(ls, '\n'); inclinenumber(ls); continue; |
225 | case '\\': case '\"': case '\'': break; |
226 | case END_OF_STREAM: continue; |
227 | default: |
228 | if (!lj_char_isdigit(c)) |
229 | goto err_xesc; |
230 | c -= '0'; /* Decimal escape '\ddd'. */ |
231 | if (lj_char_isdigit(next(ls))) { |
232 | c = c*10 + (ls->current - '0'); |
233 | if (lj_char_isdigit(next(ls))) { |
234 | c = c*10 + (ls->current - '0'); |
235 | if (c > 255) { |
236 | err_xesc: |
237 | lj_lex_error(ls, TK_string, LJ_ERR_XESC); |
238 | } |
239 | next(ls); |
240 | } |
241 | } |
242 | save(ls, c); |
243 | continue; |
244 | } |
245 | save(ls, c); |
246 | next(ls); |
247 | continue; |
248 | } |
249 | default: |
250 | save_and_next(ls); |
251 | break; |
252 | } |
253 | } |
254 | save_and_next(ls); /* skip delimiter */ |
255 | setstrV(ls->L, tv, lj_parse_keepstr(ls, ls->sb.buf + 1, ls->sb.n - 2)); |
256 | } |
257 | |
258 | /* -- Main lexical scanner ------------------------------------------------ */ |
259 | |
260 | static int llex(LexState *ls, TValue *tv) |
261 | { |
262 | lj_str_resetbuf(&ls->sb); |
263 | for (;;) { |
264 | if (lj_char_isident(ls->current)) { |
265 | GCstr *s; |
266 | if (lj_char_isdigit(ls->current)) { /* Numeric literal. */ |
267 | lex_number(ls, tv); |
268 | return TK_number; |
269 | } |
270 | /* Identifier or reserved word. */ |
271 | do { |
272 | save_and_next(ls); |
273 | } while (lj_char_isident(ls->current)); |
274 | s = lj_parse_keepstr(ls, ls->sb.buf, ls->sb.n); |
275 | setstrV(ls->L, tv, s); |
276 | if (s->reserved > 0) /* Reserved word? */ |
277 | return TK_OFS + s->reserved; |
278 | return TK_name; |
279 | } |
280 | switch (ls->current) { |
281 | case '\n': |
282 | case '\r': |
283 | inclinenumber(ls); |
284 | continue; |
285 | case ' ': |
286 | case '\t': |
287 | case '\v': |
288 | case '\f': |
289 | next(ls); |
290 | continue; |
291 | case '-': |
292 | next(ls); |
293 | if (ls->current != '-') return '-'; |
294 | /* else is a comment */ |
295 | next(ls); |
296 | if (ls->current == '[') { |
297 | int sep = skip_sep(ls); |
298 | lj_str_resetbuf(&ls->sb); /* `skip_sep' may dirty the buffer */ |
299 | if (sep >= 0) { |
300 | read_long_string(ls, NULL, sep); /* long comment */ |
301 | lj_str_resetbuf(&ls->sb); |
302 | continue; |
303 | } |
304 | } |
305 | /* else short comment */ |
306 | while (!currIsNewline(ls) && ls->current != END_OF_STREAM) |
307 | next(ls); |
308 | continue; |
309 | case '[': { |
310 | int sep = skip_sep(ls); |
311 | if (sep >= 0) { |
312 | read_long_string(ls, tv, sep); |
313 | return TK_string; |
314 | } else if (sep == -1) { |
315 | return '['; |
316 | } else { |
317 | lj_lex_error(ls, TK_string, LJ_ERR_XLDELIM); |
318 | continue; |
319 | } |
320 | } |
321 | case '=': |
322 | next(ls); |
323 | if (ls->current != '=') return '='; else { next(ls); return TK_eq; } |
324 | case '<': |
325 | next(ls); |
326 | if (ls->current != '=') return '<'; else { next(ls); return TK_le; } |
327 | case '>': |
328 | next(ls); |
329 | if (ls->current != '=') return '>'; else { next(ls); return TK_ge; } |
330 | case '~': |
331 | next(ls); |
332 | if (ls->current != '=') return '~'; else { next(ls); return TK_ne; } |
333 | case ':': |
334 | next(ls); |
335 | if (ls->current != ':') return ':'; else { next(ls); return TK_label; } |
336 | case '"': |
337 | case '\'': |
338 | read_string(ls, ls->current, tv); |
339 | return TK_string; |
340 | case '.': |
341 | save_and_next(ls); |
342 | if (ls->current == '.') { |
343 | next(ls); |
344 | if (ls->current == '.') { |
345 | next(ls); |
346 | return TK_dots; /* ... */ |
347 | } |
348 | return TK_concat; /* .. */ |
349 | } else if (!lj_char_isdigit(ls->current)) { |
350 | return '.'; |
351 | } else { |
352 | lex_number(ls, tv); |
353 | return TK_number; |
354 | } |
355 | case END_OF_STREAM: |
356 | return TK_eof; |
357 | default: { |
358 | int c = ls->current; |
359 | next(ls); |
360 | return c; /* Single-char tokens (+ - / ...). */ |
361 | } |
362 | } |
363 | } |
364 | } |
365 | |
366 | /* -- Lexer API ----------------------------------------------------------- */ |
367 | |
368 | /* Setup lexer state. */ |
369 | int lj_lex_setup(lua_State *L, LexState *ls) |
370 | { |
371 | int = 0; |
372 | ls->L = L; |
373 | ls->fs = NULL; |
374 | ls->n = 0; |
375 | ls->p = NULL; |
376 | ls->vstack = NULL; |
377 | ls->sizevstack = 0; |
378 | ls->vtop = 0; |
379 | ls->bcstack = NULL; |
380 | ls->sizebcstack = 0; |
381 | ls->lookahead = TK_eof; /* No look-ahead token. */ |
382 | ls->linenumber = 1; |
383 | ls->lastline = 1; |
384 | lj_str_resizebuf(ls->L, &ls->sb, LJ_MIN_SBUF); |
385 | next(ls); /* Read-ahead first char. */ |
386 | if (ls->current == 0xef && ls->n >= 2 && char2int(ls->p[0]) == 0xbb && |
387 | char2int(ls->p[1]) == 0xbf) { /* Skip UTF-8 BOM (if buffered). */ |
388 | ls->n -= 2; |
389 | ls->p += 2; |
390 | next(ls); |
391 | header = 1; |
392 | } |
393 | if (ls->current == '#') { /* Skip POSIX #! header line. */ |
394 | do { |
395 | next(ls); |
396 | if (ls->current == END_OF_STREAM) return 0; |
397 | } while (!currIsNewline(ls)); |
398 | inclinenumber(ls); |
399 | header = 1; |
400 | } |
401 | if (ls->current == LUA_SIGNATURE[0]) { /* Bytecode dump. */ |
402 | if (header) { |
403 | /* |
404 | ** Loading bytecode with an extra header is disabled for security |
405 | ** reasons. This may circumvent the usual check for bytecode vs. |
406 | ** Lua code by looking at the first char. Since this is a potential |
407 | ** security violation no attempt is made to echo the chunkname either. |
408 | */ |
409 | setstrV(L, L->top++, lj_err_str(L, LJ_ERR_BCBAD)); |
410 | lj_err_throw(L, LUA_ERRSYNTAX); |
411 | } |
412 | return 1; |
413 | } |
414 | return 0; |
415 | } |
416 | |
417 | /* Cleanup lexer state. */ |
418 | void lj_lex_cleanup(lua_State *L, LexState *ls) |
419 | { |
420 | global_State *g = G(L); |
421 | lj_mem_freevec(g, ls->bcstack, ls->sizebcstack, BCInsLine); |
422 | lj_mem_freevec(g, ls->vstack, ls->sizevstack, VarInfo); |
423 | lj_str_freebuf(g, &ls->sb); |
424 | } |
425 | |
426 | void lj_lex_next(LexState *ls) |
427 | { |
428 | ls->lastline = ls->linenumber; |
429 | if (LJ_LIKELY(ls->lookahead == TK_eof)) { /* No lookahead token? */ |
430 | ls->token = llex(ls, &ls->tokenval); /* Get next token. */ |
431 | } else { /* Otherwise return lookahead token. */ |
432 | ls->token = ls->lookahead; |
433 | ls->lookahead = TK_eof; |
434 | ls->tokenval = ls->lookaheadval; |
435 | } |
436 | } |
437 | |
438 | LexToken lj_lex_lookahead(LexState *ls) |
439 | { |
440 | lua_assert(ls->lookahead == TK_eof); |
441 | ls->lookahead = llex(ls, &ls->lookaheadval); |
442 | return ls->lookahead; |
443 | } |
444 | |
445 | const char *lj_lex_token2str(LexState *ls, LexToken token) |
446 | { |
447 | if (token > TK_OFS) |
448 | return tokennames[token-TK_OFS-1]; |
449 | else if (!lj_char_iscntrl(token)) |
450 | return lj_str_pushf(ls->L, "%c" , token); |
451 | else |
452 | return lj_str_pushf(ls->L, "char(%d)" , token); |
453 | } |
454 | |
455 | void lj_lex_error(LexState *ls, LexToken token, ErrMsg em, ...) |
456 | { |
457 | const char *tok; |
458 | va_list argp; |
459 | if (token == 0) { |
460 | tok = NULL; |
461 | } else if (token == TK_name || token == TK_string || token == TK_number) { |
462 | save(ls, '\0'); |
463 | tok = ls->sb.buf; |
464 | } else { |
465 | tok = lj_lex_token2str(ls, token); |
466 | } |
467 | va_start(argp, em); |
468 | lj_err_lex(ls->L, ls->chunkname, tok, ls->linenumber, em, argp); |
469 | va_end(argp); |
470 | } |
471 | |
472 | void lj_lex_init(lua_State *L) |
473 | { |
474 | uint32_t i; |
475 | for (i = 0; i < TK_RESERVED; i++) { |
476 | GCstr *s = lj_str_newz(L, tokennames[i]); |
477 | fixstring(s); /* Reserved words are never collected. */ |
478 | s->reserved = (uint8_t)(i+1); |
479 | } |
480 | } |
481 | |
482 | |