1 | /* |
2 | ** Lexical analyzer. |
3 | ** Copyright (C) 2005-2021 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_buf.h" |
16 | #include "lj_str.h" |
17 | #if LJ_HASFFI |
18 | #include "lj_tab.h" |
19 | #include "lj_ctype.h" |
20 | #include "lj_cdata.h" |
21 | #include "lualib.h" |
22 | #endif |
23 | #include "lj_state.h" |
24 | #include "lj_lex.h" |
25 | #include "lj_parse.h" |
26 | #include "lj_char.h" |
27 | #include "lj_strscan.h" |
28 | #include "lj_strfmt.h" |
29 | |
30 | /* Lua lexer token names. */ |
31 | static const char *const tokennames[] = { |
32 | #define TKSTR1(name) #name, |
33 | #define TKSTR2(name, sym) #sym, |
34 | TKDEF(TKSTR1, TKSTR2) |
35 | #undef TKSTR1 |
36 | #undef TKSTR2 |
37 | NULL |
38 | }; |
39 | |
40 | /* -- Buffer handling ----------------------------------------------------- */ |
41 | |
42 | #define LEX_EOF (-1) |
43 | #define lex_iseol(ls) (ls->c == '\n' || ls->c == '\r') |
44 | |
45 | /* Get more input from reader. */ |
46 | static LJ_NOINLINE LexChar lex_more(LexState *ls) |
47 | { |
48 | size_t sz; |
49 | const char *p = ls->rfunc(ls->L, ls->rdata, &sz); |
50 | if (p == NULL || sz == 0) return LEX_EOF; |
51 | if (sz >= LJ_MAX_BUF) { |
52 | if (sz != ~(size_t)0) lj_err_mem(ls->L); |
53 | sz = ~(uintptr_t)0 - (uintptr_t)p; |
54 | if (sz >= LJ_MAX_BUF) sz = LJ_MAX_BUF-1; |
55 | ls->endmark = 1; |
56 | } |
57 | ls->pe = p + sz; |
58 | ls->p = p + 1; |
59 | return (LexChar)(uint8_t)p[0]; |
60 | } |
61 | |
62 | /* Get next character. */ |
63 | static LJ_AINLINE LexChar lex_next(LexState *ls) |
64 | { |
65 | return (ls->c = ls->p < ls->pe ? (LexChar)(uint8_t)*ls->p++ : lex_more(ls)); |
66 | } |
67 | |
68 | /* Save character. */ |
69 | static LJ_AINLINE void lex_save(LexState *ls, LexChar c) |
70 | { |
71 | lj_buf_putb(&ls->sb, c); |
72 | } |
73 | |
74 | /* Save previous character and get next character. */ |
75 | static LJ_AINLINE LexChar lex_savenext(LexState *ls) |
76 | { |
77 | lex_save(ls, ls->c); |
78 | return lex_next(ls); |
79 | } |
80 | |
81 | /* Skip line break. Handles "\n", "\r", "\r\n" or "\n\r". */ |
82 | static void lex_newline(LexState *ls) |
83 | { |
84 | LexChar old = ls->c; |
85 | lj_assertLS(lex_iseol(ls), "bad usage" ); |
86 | lex_next(ls); /* Skip "\n" or "\r". */ |
87 | if (lex_iseol(ls) && ls->c != old) lex_next(ls); /* Skip "\n\r" or "\r\n". */ |
88 | if (++ls->linenumber >= LJ_MAX_LINE) |
89 | lj_lex_error(ls, ls->tok, LJ_ERR_XLINES); |
90 | } |
91 | |
92 | /* -- Scanner for terminals ----------------------------------------------- */ |
93 | |
94 | /* Parse a number literal. */ |
95 | static void lex_number(LexState *ls, TValue *tv) |
96 | { |
97 | StrScanFmt fmt; |
98 | LexChar c, xp = 'e'; |
99 | lj_assertLS(lj_char_isdigit(ls->c), "bad usage" ); |
100 | if ((c = ls->c) == '0' && (lex_savenext(ls) | 0x20) == 'x') |
101 | xp = 'p'; |
102 | while (lj_char_isident(ls->c) || ls->c == '.' || |
103 | ((ls->c == '-' || ls->c == '+') && (c | 0x20) == xp)) { |
104 | c = ls->c; |
105 | lex_savenext(ls); |
106 | } |
107 | lex_save(ls, '\0'); |
108 | fmt = lj_strscan_scan((const uint8_t *)sbufB(&ls->sb), sbuflen(&ls->sb)-1, tv, |
109 | (LJ_DUALNUM ? STRSCAN_OPT_TOINT : STRSCAN_OPT_TONUM) | |
110 | (LJ_HASFFI ? (STRSCAN_OPT_LL|STRSCAN_OPT_IMAG) : 0)); |
111 | if (LJ_DUALNUM && fmt == STRSCAN_INT) { |
112 | setitype(tv, LJ_TISNUM); |
113 | } else if (fmt == STRSCAN_NUM) { |
114 | /* Already in correct format. */ |
115 | #if LJ_HASFFI |
116 | } else if (fmt != STRSCAN_ERROR) { |
117 | lua_State *L = ls->L; |
118 | GCcdata *cd; |
119 | lj_assertLS(fmt == STRSCAN_I64 || fmt == STRSCAN_U64 || fmt == STRSCAN_IMAG, |
120 | "unexpected number format %d" , fmt); |
121 | if (!ctype_ctsG(G(L))) { |
122 | ptrdiff_t oldtop = savestack(L, L->top); |
123 | luaopen_ffi(L); /* Load FFI library on-demand. */ |
124 | L->top = restorestack(L, oldtop); |
125 | } |
126 | if (fmt == STRSCAN_IMAG) { |
127 | cd = lj_cdata_new_(L, CTID_COMPLEX_DOUBLE, 2*sizeof(double)); |
128 | ((double *)cdataptr(cd))[0] = 0; |
129 | ((double *)cdataptr(cd))[1] = numV(tv); |
130 | } else { |
131 | cd = lj_cdata_new_(L, fmt==STRSCAN_I64 ? CTID_INT64 : CTID_UINT64, 8); |
132 | *(uint64_t *)cdataptr(cd) = tv->u64; |
133 | } |
134 | lj_parse_keepcdata(ls, tv, cd); |
135 | #endif |
136 | } else { |
137 | lj_assertLS(fmt == STRSCAN_ERROR, |
138 | "unexpected number format %d" , fmt); |
139 | lj_lex_error(ls, TK_number, LJ_ERR_XNUMBER); |
140 | } |
141 | } |
142 | |
143 | /* Skip equal signs for "[=...=[" and "]=...=]" and return their count. */ |
144 | static int lex_skipeq(LexState *ls) |
145 | { |
146 | int count = 0; |
147 | LexChar s = ls->c; |
148 | lj_assertLS(s == '[' || s == ']', "bad usage" ); |
149 | while (lex_savenext(ls) == '=' && count < 0x20000000) |
150 | count++; |
151 | return (ls->c == s) ? count : (-count) - 1; |
152 | } |
153 | |
154 | /* Parse a long string or long comment (tv set to NULL). */ |
155 | static void lex_longstring(LexState *ls, TValue *tv, int sep) |
156 | { |
157 | lex_savenext(ls); /* Skip second '['. */ |
158 | if (lex_iseol(ls)) /* Skip initial newline. */ |
159 | lex_newline(ls); |
160 | for (;;) { |
161 | switch (ls->c) { |
162 | case LEX_EOF: |
163 | lj_lex_error(ls, TK_eof, tv ? LJ_ERR_XLSTR : LJ_ERR_XLCOM); |
164 | break; |
165 | case ']': |
166 | if (lex_skipeq(ls) == sep) { |
167 | lex_savenext(ls); /* Skip second ']'. */ |
168 | goto endloop; |
169 | } |
170 | break; |
171 | case '\n': |
172 | case '\r': |
173 | lex_save(ls, '\n'); |
174 | lex_newline(ls); |
175 | if (!tv) lj_buf_reset(&ls->sb); /* Don't waste space for comments. */ |
176 | break; |
177 | default: |
178 | lex_savenext(ls); |
179 | break; |
180 | } |
181 | } endloop: |
182 | if (tv) { |
183 | GCstr *str = lj_parse_keepstr(ls, sbufB(&ls->sb) + (2 + (MSize)sep), |
184 | sbuflen(&ls->sb) - 2*(2 + (MSize)sep)); |
185 | setstrV(ls->L, tv, str); |
186 | } |
187 | } |
188 | |
189 | /* Parse a string. */ |
190 | static void lex_string(LexState *ls, TValue *tv) |
191 | { |
192 | LexChar delim = ls->c; /* Delimiter is '\'' or '"'. */ |
193 | lex_savenext(ls); |
194 | while (ls->c != delim) { |
195 | switch (ls->c) { |
196 | case LEX_EOF: |
197 | lj_lex_error(ls, TK_eof, LJ_ERR_XSTR); |
198 | continue; |
199 | case '\n': |
200 | case '\r': |
201 | lj_lex_error(ls, TK_string, LJ_ERR_XSTR); |
202 | continue; |
203 | case '\\': { |
204 | LexChar c = lex_next(ls); /* Skip the '\\'. */ |
205 | switch (c) { |
206 | case 'a': c = '\a'; break; |
207 | case 'b': c = '\b'; break; |
208 | case 'f': c = '\f'; break; |
209 | case 'n': c = '\n'; break; |
210 | case 'r': c = '\r'; break; |
211 | case 't': c = '\t'; break; |
212 | case 'v': c = '\v'; break; |
213 | case 'x': /* Hexadecimal escape '\xXX'. */ |
214 | c = (lex_next(ls) & 15u) << 4; |
215 | if (!lj_char_isdigit(ls->c)) { |
216 | if (!lj_char_isxdigit(ls->c)) goto err_xesc; |
217 | c += 9 << 4; |
218 | } |
219 | c += (lex_next(ls) & 15u); |
220 | if (!lj_char_isdigit(ls->c)) { |
221 | if (!lj_char_isxdigit(ls->c)) goto err_xesc; |
222 | c += 9; |
223 | } |
224 | break; |
225 | case 'u': /* Unicode escape '\u{XX...}'. */ |
226 | if (lex_next(ls) != '{') goto err_xesc; |
227 | lex_next(ls); |
228 | c = 0; |
229 | do { |
230 | c = (c << 4) | (ls->c & 15u); |
231 | if (!lj_char_isdigit(ls->c)) { |
232 | if (!lj_char_isxdigit(ls->c)) goto err_xesc; |
233 | c += 9; |
234 | } |
235 | if (c >= 0x110000) goto err_xesc; /* Out of Unicode range. */ |
236 | } while (lex_next(ls) != '}'); |
237 | if (c < 0x800) { |
238 | if (c < 0x80) break; |
239 | lex_save(ls, 0xc0 | (c >> 6)); |
240 | } else { |
241 | if (c >= 0x10000) { |
242 | lex_save(ls, 0xf0 | (c >> 18)); |
243 | lex_save(ls, 0x80 | ((c >> 12) & 0x3f)); |
244 | } else { |
245 | if (c >= 0xd800 && c < 0xe000) goto err_xesc; /* No surrogates. */ |
246 | lex_save(ls, 0xe0 | (c >> 12)); |
247 | } |
248 | lex_save(ls, 0x80 | ((c >> 6) & 0x3f)); |
249 | } |
250 | c = 0x80 | (c & 0x3f); |
251 | break; |
252 | case 'z': /* Skip whitespace. */ |
253 | lex_next(ls); |
254 | while (lj_char_isspace(ls->c)) |
255 | if (lex_iseol(ls)) lex_newline(ls); else lex_next(ls); |
256 | continue; |
257 | case '\n': case '\r': lex_save(ls, '\n'); lex_newline(ls); continue; |
258 | case '\\': case '\"': case '\'': break; |
259 | case LEX_EOF: continue; |
260 | default: |
261 | if (!lj_char_isdigit(c)) |
262 | goto err_xesc; |
263 | c -= '0'; /* Decimal escape '\ddd'. */ |
264 | if (lj_char_isdigit(lex_next(ls))) { |
265 | c = c*10 + (ls->c - '0'); |
266 | if (lj_char_isdigit(lex_next(ls))) { |
267 | c = c*10 + (ls->c - '0'); |
268 | if (c > 255) { |
269 | err_xesc: |
270 | lj_lex_error(ls, TK_string, LJ_ERR_XESC); |
271 | } |
272 | lex_next(ls); |
273 | } |
274 | } |
275 | lex_save(ls, c); |
276 | continue; |
277 | } |
278 | lex_save(ls, c); |
279 | lex_next(ls); |
280 | continue; |
281 | } |
282 | default: |
283 | lex_savenext(ls); |
284 | break; |
285 | } |
286 | } |
287 | lex_savenext(ls); /* Skip trailing delimiter. */ |
288 | setstrV(ls->L, tv, |
289 | lj_parse_keepstr(ls, sbufB(&ls->sb)+1, sbuflen(&ls->sb)-2)); |
290 | } |
291 | |
292 | /* -- Main lexical scanner ------------------------------------------------ */ |
293 | |
294 | /* Get next lexical token. */ |
295 | static LexToken lex_scan(LexState *ls, TValue *tv) |
296 | { |
297 | lj_buf_reset(&ls->sb); |
298 | for (;;) { |
299 | if (lj_char_isident(ls->c)) { |
300 | GCstr *s; |
301 | if (lj_char_isdigit(ls->c)) { /* Numeric literal. */ |
302 | lex_number(ls, tv); |
303 | return TK_number; |
304 | } |
305 | /* Identifier or reserved word. */ |
306 | do { |
307 | lex_savenext(ls); |
308 | } while (lj_char_isident(ls->c)); |
309 | s = lj_parse_keepstr(ls, sbufB(&ls->sb), sbuflen(&ls->sb)); |
310 | setstrV(ls->L, tv, s); |
311 | if (s->reserved > 0) /* Reserved word? */ |
312 | return TK_OFS + s->reserved; |
313 | return TK_name; |
314 | } |
315 | switch (ls->c) { |
316 | case '\n': |
317 | case '\r': |
318 | lex_newline(ls); |
319 | continue; |
320 | case ' ': |
321 | case '\t': |
322 | case '\v': |
323 | case '\f': |
324 | lex_next(ls); |
325 | continue; |
326 | case '-': |
327 | lex_next(ls); |
328 | if (ls->c != '-') return '-'; |
329 | lex_next(ls); |
330 | if (ls->c == '[') { /* Long comment "--[=*[...]=*]". */ |
331 | int sep = lex_skipeq(ls); |
332 | lj_buf_reset(&ls->sb); /* `lex_skipeq' may dirty the buffer */ |
333 | if (sep >= 0) { |
334 | lex_longstring(ls, NULL, sep); |
335 | lj_buf_reset(&ls->sb); |
336 | continue; |
337 | } |
338 | } |
339 | /* Short comment "--.*\n". */ |
340 | while (!lex_iseol(ls) && ls->c != LEX_EOF) |
341 | lex_next(ls); |
342 | continue; |
343 | case '[': { |
344 | int sep = lex_skipeq(ls); |
345 | if (sep >= 0) { |
346 | lex_longstring(ls, tv, sep); |
347 | return TK_string; |
348 | } else if (sep == -1) { |
349 | return '['; |
350 | } else { |
351 | lj_lex_error(ls, TK_string, LJ_ERR_XLDELIM); |
352 | continue; |
353 | } |
354 | } |
355 | case '=': |
356 | lex_next(ls); |
357 | if (ls->c != '=') return '='; else { lex_next(ls); return TK_eq; } |
358 | case '<': |
359 | lex_next(ls); |
360 | if (ls->c != '=') return '<'; else { lex_next(ls); return TK_le; } |
361 | case '>': |
362 | lex_next(ls); |
363 | if (ls->c != '=') return '>'; else { lex_next(ls); return TK_ge; } |
364 | case '~': |
365 | lex_next(ls); |
366 | if (ls->c != '=') return '~'; else { lex_next(ls); return TK_ne; } |
367 | case ':': |
368 | lex_next(ls); |
369 | if (ls->c != ':') return ':'; else { lex_next(ls); return TK_label; } |
370 | case '"': |
371 | case '\'': |
372 | lex_string(ls, tv); |
373 | return TK_string; |
374 | case '.': |
375 | if (lex_savenext(ls) == '.') { |
376 | lex_next(ls); |
377 | if (ls->c == '.') { |
378 | lex_next(ls); |
379 | return TK_dots; /* ... */ |
380 | } |
381 | return TK_concat; /* .. */ |
382 | } else if (!lj_char_isdigit(ls->c)) { |
383 | return '.'; |
384 | } else { |
385 | lex_number(ls, tv); |
386 | return TK_number; |
387 | } |
388 | case LEX_EOF: |
389 | return TK_eof; |
390 | default: { |
391 | LexChar c = ls->c; |
392 | lex_next(ls); |
393 | return c; /* Single-char tokens (+ - / ...). */ |
394 | } |
395 | } |
396 | } |
397 | } |
398 | |
399 | /* -- Lexer API ----------------------------------------------------------- */ |
400 | |
401 | /* Setup lexer state. */ |
402 | int lj_lex_setup(lua_State *L, LexState *ls) |
403 | { |
404 | int = 0; |
405 | ls->L = L; |
406 | ls->fs = NULL; |
407 | ls->pe = ls->p = NULL; |
408 | ls->vstack = NULL; |
409 | ls->sizevstack = 0; |
410 | ls->vtop = 0; |
411 | ls->bcstack = NULL; |
412 | ls->sizebcstack = 0; |
413 | ls->tok = 0; |
414 | ls->lookahead = TK_eof; /* No look-ahead token. */ |
415 | ls->linenumber = 1; |
416 | ls->lastline = 1; |
417 | ls->endmark = 0; |
418 | lex_next(ls); /* Read-ahead first char. */ |
419 | if (ls->c == 0xef && ls->p + 2 <= ls->pe && (uint8_t)ls->p[0] == 0xbb && |
420 | (uint8_t)ls->p[1] == 0xbf) { /* Skip UTF-8 BOM (if buffered). */ |
421 | ls->p += 2; |
422 | lex_next(ls); |
423 | header = 1; |
424 | } |
425 | if (ls->c == '#') { /* Skip POSIX #! header line. */ |
426 | do { |
427 | lex_next(ls); |
428 | if (ls->c == LEX_EOF) return 0; |
429 | } while (!lex_iseol(ls)); |
430 | lex_newline(ls); |
431 | header = 1; |
432 | } |
433 | if (ls->c == LUA_SIGNATURE[0]) { /* Bytecode dump. */ |
434 | if (header) { |
435 | /* |
436 | ** Loading bytecode with an extra header is disabled for security |
437 | ** reasons. This may circumvent the usual check for bytecode vs. |
438 | ** Lua code by looking at the first char. Since this is a potential |
439 | ** security violation no attempt is made to echo the chunkname either. |
440 | */ |
441 | setstrV(L, L->top++, lj_err_str(L, LJ_ERR_BCBAD)); |
442 | lj_err_throw(L, LUA_ERRSYNTAX); |
443 | } |
444 | return 1; |
445 | } |
446 | return 0; |
447 | } |
448 | |
449 | /* Cleanup lexer state. */ |
450 | void lj_lex_cleanup(lua_State *L, LexState *ls) |
451 | { |
452 | global_State *g = G(L); |
453 | lj_mem_freevec(g, ls->bcstack, ls->sizebcstack, BCInsLine); |
454 | lj_mem_freevec(g, ls->vstack, ls->sizevstack, VarInfo); |
455 | lj_buf_free(g, &ls->sb); |
456 | } |
457 | |
458 | /* Return next lexical token. */ |
459 | void lj_lex_next(LexState *ls) |
460 | { |
461 | ls->lastline = ls->linenumber; |
462 | if (LJ_LIKELY(ls->lookahead == TK_eof)) { /* No lookahead token? */ |
463 | ls->tok = lex_scan(ls, &ls->tokval); /* Get next token. */ |
464 | } else { /* Otherwise return lookahead token. */ |
465 | ls->tok = ls->lookahead; |
466 | ls->lookahead = TK_eof; |
467 | ls->tokval = ls->lookaheadval; |
468 | } |
469 | } |
470 | |
471 | /* Look ahead for the next token. */ |
472 | LexToken lj_lex_lookahead(LexState *ls) |
473 | { |
474 | lj_assertLS(ls->lookahead == TK_eof, "double lookahead" ); |
475 | ls->lookahead = lex_scan(ls, &ls->lookaheadval); |
476 | return ls->lookahead; |
477 | } |
478 | |
479 | /* Convert token to string. */ |
480 | const char *lj_lex_token2str(LexState *ls, LexToken tok) |
481 | { |
482 | if (tok > TK_OFS) |
483 | return tokennames[tok-TK_OFS-1]; |
484 | else if (!lj_char_iscntrl(tok)) |
485 | return lj_strfmt_pushf(ls->L, "%c" , tok); |
486 | else |
487 | return lj_strfmt_pushf(ls->L, "char(%d)" , tok); |
488 | } |
489 | |
490 | /* Lexer error. */ |
491 | void lj_lex_error(LexState *ls, LexToken tok, ErrMsg em, ...) |
492 | { |
493 | const char *tokstr; |
494 | va_list argp; |
495 | if (tok == 0) { |
496 | tokstr = NULL; |
497 | } else if (tok == TK_name || tok == TK_string || tok == TK_number) { |
498 | lex_save(ls, '\0'); |
499 | tokstr = sbufB(&ls->sb); |
500 | } else { |
501 | tokstr = lj_lex_token2str(ls, tok); |
502 | } |
503 | va_start(argp, em); |
504 | lj_err_lex(ls->L, ls->chunkname, tokstr, ls->linenumber, em, argp); |
505 | va_end(argp); |
506 | } |
507 | |
508 | /* Initialize strings for reserved words. */ |
509 | void lj_lex_init(lua_State *L) |
510 | { |
511 | uint32_t i; |
512 | for (i = 0; i < TK_RESERVED; i++) { |
513 | GCstr *s = lj_str_newz(L, tokennames[i]); |
514 | fixstring(s); /* Reserved words are never collected. */ |
515 | s->reserved = (uint8_t)(i+1); |
516 | } |
517 | } |
518 | |
519 | |