• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 
2 /* Tokenizer implementation */
3 
4 #include "Python.h"
5 #include "pgenheaders.h"
6 
7 #include <ctype.h>
8 #include <assert.h>
9 
10 #include "tokenizer.h"
11 #include "errcode.h"
12 
13 #ifndef PGEN
14 #include "unicodeobject.h"
15 #include "bytesobject.h"
16 #include "fileobject.h"
17 #include "codecs.h"
18 #include "abstract.h"
19 #endif /* PGEN */
20 
21 /* Alternate tab spacing */
22 #define ALTTABSIZE 1
23 
24 #define is_potential_identifier_start(c) (\
25               (c >= 'a' && c <= 'z')\
26                || (c >= 'A' && c <= 'Z')\
27                || c == '_'\
28                || (c >= 128))
29 
30 #define is_potential_identifier_char(c) (\
31               (c >= 'a' && c <= 'z')\
32                || (c >= 'A' && c <= 'Z')\
33                || (c >= '0' && c <= '9')\
34                || c == '_'\
35                || (c >= 128))
36 
37 extern char *PyOS_Readline(FILE *, FILE *, const char *);
38 /* Return malloc'ed string including trailing \n;
39    empty malloc'ed string for EOF;
40    NULL if interrupted */
41 
42 /* Don't ever change this -- it would break the portability of Python code */
43 #define TABSIZE 8
44 
45 /* Forward */
46 static struct tok_state *tok_new(void);
47 static int tok_nextc(struct tok_state *tok);
48 static void tok_backup(struct tok_state *tok, int c);
49 
50 
51 /* Token names */
52 
53 const char *_PyParser_TokenNames[] = {
54     "ENDMARKER",
55     "NAME",
56     "NUMBER",
57     "STRING",
58     "NEWLINE",
59     "INDENT",
60     "DEDENT",
61     "LPAR",
62     "RPAR",
63     "LSQB",
64     "RSQB",
65     "COLON",
66     "COMMA",
67     "SEMI",
68     "PLUS",
69     "MINUS",
70     "STAR",
71     "SLASH",
72     "VBAR",
73     "AMPER",
74     "LESS",
75     "GREATER",
76     "EQUAL",
77     "DOT",
78     "PERCENT",
79     "LBRACE",
80     "RBRACE",
81     "EQEQUAL",
82     "NOTEQUAL",
83     "LESSEQUAL",
84     "GREATEREQUAL",
85     "TILDE",
86     "CIRCUMFLEX",
87     "LEFTSHIFT",
88     "RIGHTSHIFT",
89     "DOUBLESTAR",
90     "PLUSEQUAL",
91     "MINEQUAL",
92     "STAREQUAL",
93     "SLASHEQUAL",
94     "PERCENTEQUAL",
95     "AMPEREQUAL",
96     "VBAREQUAL",
97     "CIRCUMFLEXEQUAL",
98     "LEFTSHIFTEQUAL",
99     "RIGHTSHIFTEQUAL",
100     "DOUBLESTAREQUAL",
101     "DOUBLESLASH",
102     "DOUBLESLASHEQUAL",
103     "AT",
104     "ATEQUAL",
105     "RARROW",
106     "ELLIPSIS",
107     /* This table must match the #defines in token.h! */
108     "OP",
109     "<ERRORTOKEN>",
110     "COMMENT",
111     "NL",
112     "ENCODING",
113     "<N_TOKENS>"
114 };
115 
116 
117 /* Create and initialize a new tok_state structure */
118 
119 static struct tok_state *
tok_new(void)120 tok_new(void)
121 {
122     struct tok_state *tok = (struct tok_state *)PyMem_MALLOC(
123                                             sizeof(struct tok_state));
124     if (tok == NULL)
125         return NULL;
126     tok->buf = tok->cur = tok->end = tok->inp = tok->start = NULL;
127     tok->done = E_OK;
128     tok->fp = NULL;
129     tok->input = NULL;
130     tok->tabsize = TABSIZE;
131     tok->indent = 0;
132     tok->indstack[0] = 0;
133 
134     tok->atbol = 1;
135     tok->pendin = 0;
136     tok->prompt = tok->nextprompt = NULL;
137     tok->lineno = 0;
138     tok->level = 0;
139     tok->altindstack[0] = 0;
140     tok->decoding_state = STATE_INIT;
141     tok->decoding_erred = 0;
142     tok->read_coding_spec = 0;
143     tok->enc = NULL;
144     tok->encoding = NULL;
145     tok->cont_line = 0;
146 #ifndef PGEN
147     tok->filename = NULL;
148     tok->decoding_readline = NULL;
149     tok->decoding_buffer = NULL;
150 #endif
151 
152     return tok;
153 }
154 
155 static char *
new_string(const char * s,Py_ssize_t len,struct tok_state * tok)156 new_string(const char *s, Py_ssize_t len, struct tok_state *tok)
157 {
158     char* result = (char *)PyMem_MALLOC(len + 1);
159     if (!result) {
160         tok->done = E_NOMEM;
161         return NULL;
162     }
163     memcpy(result, s, len);
164     result[len] = '\0';
165     return result;
166 }
167 
168 #ifdef PGEN
169 
170 static char *
decoding_fgets(char * s,int size,struct tok_state * tok)171 decoding_fgets(char *s, int size, struct tok_state *tok)
172 {
173     return fgets(s, size, tok->fp);
174 }
175 
176 static int
decoding_feof(struct tok_state * tok)177 decoding_feof(struct tok_state *tok)
178 {
179     return feof(tok->fp);
180 }
181 
182 static char *
decode_str(const char * str,int exec_input,struct tok_state * tok)183 decode_str(const char *str, int exec_input, struct tok_state *tok)
184 {
185     return new_string(str, strlen(str), tok);
186 }
187 
188 #else /* PGEN */
189 
190 static char *
error_ret(struct tok_state * tok)191 error_ret(struct tok_state *tok) /* XXX */
192 {
193     tok->decoding_erred = 1;
194     if (tok->fp != NULL && tok->buf != NULL) /* see PyTokenizer_Free */
195         PyMem_FREE(tok->buf);
196     tok->buf = tok->cur = tok->end = tok->inp = tok->start = NULL;
197     tok->done = E_DECODE;
198     return NULL;                /* as if it were EOF */
199 }
200 
201 
202 static const char *
get_normal_name(const char * s)203 get_normal_name(const char *s)  /* for utf-8 and latin-1 */
204 {
205     char buf[13];
206     int i;
207     for (i = 0; i < 12; i++) {
208         int c = s[i];
209         if (c == '\0')
210             break;
211         else if (c == '_')
212             buf[i] = '-';
213         else
214             buf[i] = tolower(c);
215     }
216     buf[i] = '\0';
217     if (strcmp(buf, "utf-8") == 0 ||
218         strncmp(buf, "utf-8-", 6) == 0)
219         return "utf-8";
220     else if (strcmp(buf, "latin-1") == 0 ||
221              strcmp(buf, "iso-8859-1") == 0 ||
222              strcmp(buf, "iso-latin-1") == 0 ||
223              strncmp(buf, "latin-1-", 8) == 0 ||
224              strncmp(buf, "iso-8859-1-", 11) == 0 ||
225              strncmp(buf, "iso-latin-1-", 12) == 0)
226         return "iso-8859-1";
227     else
228         return s;
229 }
230 
231 /* Return the coding spec in S, or NULL if none is found.  */
232 
233 static int
get_coding_spec(const char * s,char ** spec,Py_ssize_t size,struct tok_state * tok)234 get_coding_spec(const char *s, char **spec, Py_ssize_t size, struct tok_state *tok)
235 {
236     Py_ssize_t i;
237     *spec = NULL;
238     /* Coding spec must be in a comment, and that comment must be
239      * the only statement on the source code line. */
240     for (i = 0; i < size - 6; i++) {
241         if (s[i] == '#')
242             break;
243         if (s[i] != ' ' && s[i] != '\t' && s[i] != '\014')
244             return 1;
245     }
246     for (; i < size - 6; i++) { /* XXX inefficient search */
247         const char* t = s + i;
248         if (strncmp(t, "coding", 6) == 0) {
249             const char* begin = NULL;
250             t += 6;
251             if (t[0] != ':' && t[0] != '=')
252                 continue;
253             do {
254                 t++;
255             } while (t[0] == '\x20' || t[0] == '\t');
256 
257             begin = t;
258             while (Py_ISALNUM(t[0]) ||
259                    t[0] == '-' || t[0] == '_' || t[0] == '.')
260                 t++;
261 
262             if (begin < t) {
263                 char* r = new_string(begin, t - begin, tok);
264                 const char* q;
265                 if (!r)
266                     return 0;
267                 q = get_normal_name(r);
268                 if (r != q) {
269                     PyMem_FREE(r);
270                     r = new_string(q, strlen(q), tok);
271                     if (!r)
272                         return 0;
273                 }
274                 *spec = r;
275                 break;
276             }
277         }
278     }
279     return 1;
280 }
281 
282 /* Check whether the line contains a coding spec. If it does,
283    invoke the set_readline function for the new encoding.
284    This function receives the tok_state and the new encoding.
285    Return 1 on success, 0 on failure.  */
286 
287 static int
check_coding_spec(const char * line,Py_ssize_t size,struct tok_state * tok,int set_readline (struct tok_state *,const char *))288 check_coding_spec(const char* line, Py_ssize_t size, struct tok_state *tok,
289                   int set_readline(struct tok_state *, const char *))
290 {
291     char *cs;
292     int r = 1;
293 
294     if (tok->cont_line) {
295         /* It's a continuation line, so it can't be a coding spec. */
296         tok->read_coding_spec = 1;
297         return 1;
298     }
299     if (!get_coding_spec(line, &cs, size, tok))
300         return 0;
301     if (!cs) {
302         Py_ssize_t i;
303         for (i = 0; i < size; i++) {
304             if (line[i] == '#' || line[i] == '\n' || line[i] == '\r')
305                 break;
306             if (line[i] != ' ' && line[i] != '\t' && line[i] != '\014') {
307                 /* Stop checking coding spec after a line containing
308                  * anything except a comment. */
309                 tok->read_coding_spec = 1;
310                 break;
311             }
312         }
313         return 1;
314     }
315     tok->read_coding_spec = 1;
316     if (tok->encoding == NULL) {
317         assert(tok->decoding_state == STATE_RAW);
318         if (strcmp(cs, "utf-8") == 0) {
319             tok->encoding = cs;
320         } else {
321             r = set_readline(tok, cs);
322             if (r) {
323                 tok->encoding = cs;
324                 tok->decoding_state = STATE_NORMAL;
325             }
326             else {
327                 PyErr_Format(PyExc_SyntaxError,
328                              "encoding problem: %s", cs);
329                 PyMem_FREE(cs);
330             }
331         }
332     } else {                /* then, compare cs with BOM */
333         r = (strcmp(tok->encoding, cs) == 0);
334         if (!r)
335             PyErr_Format(PyExc_SyntaxError,
336                          "encoding problem: %s with BOM", cs);
337         PyMem_FREE(cs);
338     }
339     return r;
340 }
341 
342 /* See whether the file starts with a BOM. If it does,
343    invoke the set_readline function with the new encoding.
344    Return 1 on success, 0 on failure.  */
345 
346 static int
check_bom(int get_char (struct tok_state *),void unget_char (int,struct tok_state *),int set_readline (struct tok_state *,const char *),struct tok_state * tok)347 check_bom(int get_char(struct tok_state *),
348           void unget_char(int, struct tok_state *),
349           int set_readline(struct tok_state *, const char *),
350           struct tok_state *tok)
351 {
352     int ch1, ch2, ch3;
353     ch1 = get_char(tok);
354     tok->decoding_state = STATE_RAW;
355     if (ch1 == EOF) {
356         return 1;
357     } else if (ch1 == 0xEF) {
358         ch2 = get_char(tok);
359         if (ch2 != 0xBB) {
360             unget_char(ch2, tok);
361             unget_char(ch1, tok);
362             return 1;
363         }
364         ch3 = get_char(tok);
365         if (ch3 != 0xBF) {
366             unget_char(ch3, tok);
367             unget_char(ch2, tok);
368             unget_char(ch1, tok);
369             return 1;
370         }
371 #if 0
372     /* Disable support for UTF-16 BOMs until a decision
373        is made whether this needs to be supported.  */
374     } else if (ch1 == 0xFE) {
375         ch2 = get_char(tok);
376         if (ch2 != 0xFF) {
377             unget_char(ch2, tok);
378             unget_char(ch1, tok);
379             return 1;
380         }
381         if (!set_readline(tok, "utf-16-be"))
382             return 0;
383         tok->decoding_state = STATE_NORMAL;
384     } else if (ch1 == 0xFF) {
385         ch2 = get_char(tok);
386         if (ch2 != 0xFE) {
387             unget_char(ch2, tok);
388             unget_char(ch1, tok);
389             return 1;
390         }
391         if (!set_readline(tok, "utf-16-le"))
392             return 0;
393         tok->decoding_state = STATE_NORMAL;
394 #endif
395     } else {
396         unget_char(ch1, tok);
397         return 1;
398     }
399     if (tok->encoding != NULL)
400         PyMem_FREE(tok->encoding);
401     tok->encoding = new_string("utf-8", 5, tok);
402     if (!tok->encoding)
403         return 0;
404     /* No need to set_readline: input is already utf-8 */
405     return 1;
406 }
407 
408 /* Read a line of text from TOK into S, using the stream in TOK.
409    Return NULL on failure, else S.
410 
411    On entry, tok->decoding_buffer will be one of:
412      1) NULL: need to call tok->decoding_readline to get a new line
413      2) PyUnicodeObject *: decoding_feof has called tok->decoding_readline and
414        stored the result in tok->decoding_buffer
415      3) PyByteArrayObject *: previous call to fp_readl did not have enough room
416        (in the s buffer) to copy entire contents of the line read
417        by tok->decoding_readline.  tok->decoding_buffer has the overflow.
418        In this case, fp_readl is called in a loop (with an expanded buffer)
419        until the buffer ends with a '\n' (or until the end of the file is
420        reached): see tok_nextc and its calls to decoding_fgets.
421 */
422 
423 static char *
fp_readl(char * s,int size,struct tok_state * tok)424 fp_readl(char *s, int size, struct tok_state *tok)
425 {
426     PyObject* bufobj;
427     const char *buf;
428     Py_ssize_t buflen;
429 
430     /* Ask for one less byte so we can terminate it */
431     assert(size > 0);
432     size--;
433 
434     if (tok->decoding_buffer) {
435         bufobj = tok->decoding_buffer;
436         Py_INCREF(bufobj);
437     }
438     else
439     {
440         bufobj = _PyObject_CallNoArg(tok->decoding_readline);
441         if (bufobj == NULL)
442             goto error;
443     }
444     if (PyUnicode_CheckExact(bufobj))
445     {
446         buf = PyUnicode_AsUTF8AndSize(bufobj, &buflen);
447         if (buf == NULL) {
448             goto error;
449         }
450     }
451     else
452     {
453         buf = PyByteArray_AsString(bufobj);
454         if (buf == NULL) {
455             goto error;
456         }
457         buflen = PyByteArray_GET_SIZE(bufobj);
458     }
459 
460     Py_XDECREF(tok->decoding_buffer);
461     if (buflen > size) {
462         /* Too many chars, the rest goes into tok->decoding_buffer */
463         tok->decoding_buffer = PyByteArray_FromStringAndSize(buf+size,
464                                                          buflen-size);
465         if (tok->decoding_buffer == NULL)
466             goto error;
467         buflen = size;
468     }
469     else
470         tok->decoding_buffer = NULL;
471 
472     memcpy(s, buf, buflen);
473     s[buflen] = '\0';
474     if (buflen == 0) /* EOF */
475         s = NULL;
476     Py_DECREF(bufobj);
477     return s;
478 
479 error:
480     Py_XDECREF(bufobj);
481     return error_ret(tok);
482 }
483 
484 /* Set the readline function for TOK to a StreamReader's
485    readline function. The StreamReader is named ENC.
486 
487    This function is called from check_bom and check_coding_spec.
488 
489    ENC is usually identical to the future value of tok->encoding,
490    except for the (currently unsupported) case of UTF-16.
491 
492    Return 1 on success, 0 on failure. */
493 
494 static int
fp_setreadl(struct tok_state * tok,const char * enc)495 fp_setreadl(struct tok_state *tok, const char* enc)
496 {
497     PyObject *readline, *io, *stream;
498     _Py_IDENTIFIER(open);
499     _Py_IDENTIFIER(readline);
500     int fd;
501     long pos;
502 
503     fd = fileno(tok->fp);
504     /* Due to buffering the file offset for fd can be different from the file
505      * position of tok->fp.  If tok->fp was opened in text mode on Windows,
506      * its file position counts CRLF as one char and can't be directly mapped
507      * to the file offset for fd.  Instead we step back one byte and read to
508      * the end of line.*/
509     pos = ftell(tok->fp);
510     if (pos == -1 ||
511         lseek(fd, (off_t)(pos > 0 ? pos - 1 : pos), SEEK_SET) == (off_t)-1) {
512         PyErr_SetFromErrnoWithFilename(PyExc_OSError, NULL);
513         return 0;
514     }
515 
516     io = PyImport_ImportModuleNoBlock("io");
517     if (io == NULL)
518         return 0;
519 
520     stream = _PyObject_CallMethodId(io, &PyId_open, "isisOOO",
521                     fd, "r", -1, enc, Py_None, Py_None, Py_False);
522     Py_DECREF(io);
523     if (stream == NULL)
524         return 0;
525 
526     readline = _PyObject_GetAttrId(stream, &PyId_readline);
527     Py_DECREF(stream);
528     if (readline == NULL)
529         return 0;
530     Py_XSETREF(tok->decoding_readline, readline);
531 
532     if (pos > 0) {
533         PyObject *bufobj = _PyObject_CallNoArg(readline);
534         if (bufobj == NULL)
535             return 0;
536         Py_DECREF(bufobj);
537     }
538 
539     return 1;
540 }
541 
542 /* Fetch the next byte from TOK. */
543 
fp_getc(struct tok_state * tok)544 static int fp_getc(struct tok_state *tok) {
545     return getc(tok->fp);
546 }
547 
548 /* Unfetch the last byte back into TOK.  */
549 
fp_ungetc(int c,struct tok_state * tok)550 static void fp_ungetc(int c, struct tok_state *tok) {
551     ungetc(c, tok->fp);
552 }
553 
554 /* Check whether the characters at s start a valid
555    UTF-8 sequence. Return the number of characters forming
556    the sequence if yes, 0 if not.  */
valid_utf8(const unsigned char * s)557 static int valid_utf8(const unsigned char* s)
558 {
559     int expected = 0;
560     int length;
561     if (*s < 0x80)
562         /* single-byte code */
563         return 1;
564     if (*s < 0xc0)
565         /* following byte */
566         return 0;
567     if (*s < 0xE0)
568         expected = 1;
569     else if (*s < 0xF0)
570         expected = 2;
571     else if (*s < 0xF8)
572         expected = 3;
573     else
574         return 0;
575     length = expected + 1;
576     for (; expected; expected--)
577         if (s[expected] < 0x80 || s[expected] >= 0xC0)
578             return 0;
579     return length;
580 }
581 
582 /* Read a line of input from TOK. Determine encoding
583    if necessary.  */
584 
585 static char *
decoding_fgets(char * s,int size,struct tok_state * tok)586 decoding_fgets(char *s, int size, struct tok_state *tok)
587 {
588     char *line = NULL;
589     int badchar = 0;
590     for (;;) {
591         if (tok->decoding_state == STATE_NORMAL) {
592             /* We already have a codec associated with
593                this input. */
594             line = fp_readl(s, size, tok);
595             break;
596         } else if (tok->decoding_state == STATE_RAW) {
597             /* We want a 'raw' read. */
598             line = Py_UniversalNewlineFgets(s, size,
599                                             tok->fp, NULL);
600             break;
601         } else {
602             /* We have not yet determined the encoding.
603                If an encoding is found, use the file-pointer
604                reader functions from now on. */
605             if (!check_bom(fp_getc, fp_ungetc, fp_setreadl, tok))
606                 return error_ret(tok);
607             assert(tok->decoding_state != STATE_INIT);
608         }
609     }
610     if (line != NULL && tok->lineno < 2 && !tok->read_coding_spec) {
611         if (!check_coding_spec(line, strlen(line), tok, fp_setreadl)) {
612             return error_ret(tok);
613         }
614     }
615 #ifndef PGEN
616     /* The default encoding is UTF-8, so make sure we don't have any
617        non-UTF-8 sequences in it. */
618     if (line && !tok->encoding) {
619         unsigned char *c;
620         int length;
621         for (c = (unsigned char *)line; *c; c += length)
622             if (!(length = valid_utf8(c))) {
623                 badchar = *c;
624                 break;
625             }
626     }
627     if (badchar) {
628         /* Need to add 1 to the line number, since this line
629            has not been counted, yet.  */
630         PyErr_Format(PyExc_SyntaxError,
631                 "Non-UTF-8 code starting with '\\x%.2x' "
632                 "in file %U on line %i, "
633                 "but no encoding declared; "
634                 "see http://python.org/dev/peps/pep-0263/ for details",
635                 badchar, tok->filename, tok->lineno + 1);
636         return error_ret(tok);
637     }
638 #endif
639     return line;
640 }
641 
642 static int
decoding_feof(struct tok_state * tok)643 decoding_feof(struct tok_state *tok)
644 {
645     if (tok->decoding_state != STATE_NORMAL) {
646         return feof(tok->fp);
647     } else {
648         PyObject* buf = tok->decoding_buffer;
649         if (buf == NULL) {
650             buf = _PyObject_CallNoArg(tok->decoding_readline);
651             if (buf == NULL) {
652                 error_ret(tok);
653                 return 1;
654             } else {
655                 tok->decoding_buffer = buf;
656             }
657         }
658         return PyObject_Length(buf) == 0;
659     }
660 }
661 
662 /* Fetch a byte from TOK, using the string buffer. */
663 
664 static int
buf_getc(struct tok_state * tok)665 buf_getc(struct tok_state *tok) {
666     return Py_CHARMASK(*tok->str++);
667 }
668 
669 /* Unfetch a byte from TOK, using the string buffer. */
670 
671 static void
buf_ungetc(int c,struct tok_state * tok)672 buf_ungetc(int c, struct tok_state *tok) {
673     tok->str--;
674     assert(Py_CHARMASK(*tok->str) == c);        /* tok->cur may point to read-only segment */
675 }
676 
677 /* Set the readline function for TOK to ENC. For the string-based
678    tokenizer, this means to just record the encoding. */
679 
680 static int
buf_setreadl(struct tok_state * tok,const char * enc)681 buf_setreadl(struct tok_state *tok, const char* enc) {
682     tok->enc = enc;
683     return 1;
684 }
685 
686 /* Return a UTF-8 encoding Python string object from the
687    C byte string STR, which is encoded with ENC. */
688 
689 static PyObject *
translate_into_utf8(const char * str,const char * enc)690 translate_into_utf8(const char* str, const char* enc) {
691     PyObject *utf8;
692     PyObject* buf = PyUnicode_Decode(str, strlen(str), enc, NULL);
693     if (buf == NULL)
694         return NULL;
695     utf8 = PyUnicode_AsUTF8String(buf);
696     Py_DECREF(buf);
697     return utf8;
698 }
699 
700 
701 static char *
translate_newlines(const char * s,int exec_input,struct tok_state * tok)702 translate_newlines(const char *s, int exec_input, struct tok_state *tok) {
703     int skip_next_lf = 0;
704     size_t needed_length = strlen(s) + 2, final_length;
705     char *buf, *current;
706     char c = '\0';
707     buf = PyMem_MALLOC(needed_length);
708     if (buf == NULL) {
709         tok->done = E_NOMEM;
710         return NULL;
711     }
712     for (current = buf; *s; s++, current++) {
713         c = *s;
714         if (skip_next_lf) {
715             skip_next_lf = 0;
716             if (c == '\n') {
717                 c = *++s;
718                 if (!c)
719                     break;
720             }
721         }
722         if (c == '\r') {
723             skip_next_lf = 1;
724             c = '\n';
725         }
726         *current = c;
727     }
728     /* If this is exec input, add a newline to the end of the string if
729        there isn't one already. */
730     if (exec_input && c != '\n') {
731         *current = '\n';
732         current++;
733     }
734     *current = '\0';
735     final_length = current - buf + 1;
736     if (final_length < needed_length && final_length)
737         /* should never fail */
738         buf = PyMem_REALLOC(buf, final_length);
739     return buf;
740 }
741 
742 /* Decode a byte string STR for use as the buffer of TOK.
743    Look for encoding declarations inside STR, and record them
744    inside TOK.  */
745 
746 static const char *
decode_str(const char * input,int single,struct tok_state * tok)747 decode_str(const char *input, int single, struct tok_state *tok)
748 {
749     PyObject* utf8 = NULL;
750     const char *str;
751     const char *s;
752     const char *newl[2] = {NULL, NULL};
753     int lineno = 0;
754     tok->input = str = translate_newlines(input, single, tok);
755     if (str == NULL)
756         return NULL;
757     tok->enc = NULL;
758     tok->str = str;
759     if (!check_bom(buf_getc, buf_ungetc, buf_setreadl, tok))
760         return error_ret(tok);
761     str = tok->str;             /* string after BOM if any */
762     assert(str);
763     if (tok->enc != NULL) {
764         utf8 = translate_into_utf8(str, tok->enc);
765         if (utf8 == NULL)
766             return error_ret(tok);
767         str = PyBytes_AsString(utf8);
768     }
769     for (s = str;; s++) {
770         if (*s == '\0') break;
771         else if (*s == '\n') {
772             assert(lineno < 2);
773             newl[lineno] = s;
774             lineno++;
775             if (lineno == 2) break;
776         }
777     }
778     tok->enc = NULL;
779     /* need to check line 1 and 2 separately since check_coding_spec
780        assumes a single line as input */
781     if (newl[0]) {
782         if (!check_coding_spec(str, newl[0] - str, tok, buf_setreadl))
783             return error_ret(tok);
784         if (tok->enc == NULL && !tok->read_coding_spec && newl[1]) {
785             if (!check_coding_spec(newl[0]+1, newl[1] - newl[0],
786                                    tok, buf_setreadl))
787                 return error_ret(tok);
788         }
789     }
790     if (tok->enc != NULL) {
791         assert(utf8 == NULL);
792         utf8 = translate_into_utf8(str, tok->enc);
793         if (utf8 == NULL)
794             return error_ret(tok);
795         str = PyBytes_AS_STRING(utf8);
796     }
797     assert(tok->decoding_buffer == NULL);
798     tok->decoding_buffer = utf8; /* CAUTION */
799     return str;
800 }
801 
802 #endif /* PGEN */
803 
804 /* Set up tokenizer for string */
805 
806 struct tok_state *
PyTokenizer_FromString(const char * str,int exec_input)807 PyTokenizer_FromString(const char *str, int exec_input)
808 {
809     struct tok_state *tok = tok_new();
810     if (tok == NULL)
811         return NULL;
812     str = decode_str(str, exec_input, tok);
813     if (str == NULL) {
814         PyTokenizer_Free(tok);
815         return NULL;
816     }
817 
818     /* XXX: constify members. */
819     tok->buf = tok->cur = tok->end = tok->inp = (char*)str;
820     return tok;
821 }
822 
823 struct tok_state *
PyTokenizer_FromUTF8(const char * str,int exec_input)824 PyTokenizer_FromUTF8(const char *str, int exec_input)
825 {
826     struct tok_state *tok = tok_new();
827     if (tok == NULL)
828         return NULL;
829 #ifndef PGEN
830     tok->input = str = translate_newlines(str, exec_input, tok);
831 #endif
832     if (str == NULL) {
833         PyTokenizer_Free(tok);
834         return NULL;
835     }
836     tok->decoding_state = STATE_RAW;
837     tok->read_coding_spec = 1;
838     tok->enc = NULL;
839     tok->str = str;
840     tok->encoding = (char *)PyMem_MALLOC(6);
841     if (!tok->encoding) {
842         PyTokenizer_Free(tok);
843         return NULL;
844     }
845     strcpy(tok->encoding, "utf-8");
846 
847     /* XXX: constify members. */
848     tok->buf = tok->cur = tok->end = tok->inp = (char*)str;
849     return tok;
850 }
851 
852 /* Set up tokenizer for file */
853 
854 struct tok_state *
PyTokenizer_FromFile(FILE * fp,const char * enc,const char * ps1,const char * ps2)855 PyTokenizer_FromFile(FILE *fp, const char* enc,
856                      const char *ps1, const char *ps2)
857 {
858     struct tok_state *tok = tok_new();
859     if (tok == NULL)
860         return NULL;
861     if ((tok->buf = (char *)PyMem_MALLOC(BUFSIZ)) == NULL) {
862         PyTokenizer_Free(tok);
863         return NULL;
864     }
865     tok->cur = tok->inp = tok->buf;
866     tok->end = tok->buf + BUFSIZ;
867     tok->fp = fp;
868     tok->prompt = ps1;
869     tok->nextprompt = ps2;
870     if (enc != NULL) {
871         /* Must copy encoding declaration since it
872            gets copied into the parse tree. */
873         tok->encoding = PyMem_MALLOC(strlen(enc)+1);
874         if (!tok->encoding) {
875             PyTokenizer_Free(tok);
876             return NULL;
877         }
878         strcpy(tok->encoding, enc);
879         tok->decoding_state = STATE_NORMAL;
880     }
881     return tok;
882 }
883 
884 
885 /* Free a tok_state structure */
886 
887 void
PyTokenizer_Free(struct tok_state * tok)888 PyTokenizer_Free(struct tok_state *tok)
889 {
890     if (tok->encoding != NULL)
891         PyMem_FREE(tok->encoding);
892 #ifndef PGEN
893     Py_XDECREF(tok->decoding_readline);
894     Py_XDECREF(tok->decoding_buffer);
895     Py_XDECREF(tok->filename);
896 #endif
897     if (tok->fp != NULL && tok->buf != NULL)
898         PyMem_FREE(tok->buf);
899     if (tok->input)
900         PyMem_FREE((char *)tok->input);
901     PyMem_FREE(tok);
902 }
903 
904 /* Get next char, updating state; error code goes into tok->done */
905 
906 static int
tok_nextc(struct tok_state * tok)907 tok_nextc(struct tok_state *tok)
908 {
909     for (;;) {
910         if (tok->cur != tok->inp) {
911             return Py_CHARMASK(*tok->cur++); /* Fast path */
912         }
913         if (tok->done != E_OK)
914             return EOF;
915         if (tok->fp == NULL) {
916             char *end = strchr(tok->inp, '\n');
917             if (end != NULL)
918                 end++;
919             else {
920                 end = strchr(tok->inp, '\0');
921                 if (end == tok->inp) {
922                     tok->done = E_EOF;
923                     return EOF;
924                 }
925             }
926             if (tok->start == NULL)
927                 tok->buf = tok->cur;
928             tok->line_start = tok->cur;
929             tok->lineno++;
930             tok->inp = end;
931             return Py_CHARMASK(*tok->cur++);
932         }
933         if (tok->prompt != NULL) {
934             char *newtok = PyOS_Readline(stdin, stdout, tok->prompt);
935 #ifndef PGEN
936             if (newtok != NULL) {
937                 char *translated = translate_newlines(newtok, 0, tok);
938                 PyMem_FREE(newtok);
939                 if (translated == NULL)
940                     return EOF;
941                 newtok = translated;
942             }
943             if (tok->encoding && newtok && *newtok) {
944                 /* Recode to UTF-8 */
945                 Py_ssize_t buflen;
946                 const char* buf;
947                 PyObject *u = translate_into_utf8(newtok, tok->encoding);
948                 PyMem_FREE(newtok);
949                 if (!u) {
950                     tok->done = E_DECODE;
951                     return EOF;
952                 }
953                 buflen = PyBytes_GET_SIZE(u);
954                 buf = PyBytes_AS_STRING(u);
955                 newtok = PyMem_MALLOC(buflen+1);
956                 if (newtok == NULL) {
957                     Py_DECREF(u);
958                     tok->done = E_NOMEM;
959                     return EOF;
960                 }
961                 strcpy(newtok, buf);
962                 Py_DECREF(u);
963             }
964 #endif
965             if (tok->nextprompt != NULL)
966                 tok->prompt = tok->nextprompt;
967             if (newtok == NULL)
968                 tok->done = E_INTR;
969             else if (*newtok == '\0') {
970                 PyMem_FREE(newtok);
971                 tok->done = E_EOF;
972             }
973             else if (tok->start != NULL) {
974                 size_t start = tok->start - tok->buf;
975                 size_t oldlen = tok->cur - tok->buf;
976                 size_t newlen = oldlen + strlen(newtok);
977                 char *buf = tok->buf;
978                 buf = (char *)PyMem_REALLOC(buf, newlen+1);
979                 tok->lineno++;
980                 if (buf == NULL) {
981                     PyMem_FREE(tok->buf);
982                     tok->buf = NULL;
983                     PyMem_FREE(newtok);
984                     tok->done = E_NOMEM;
985                     return EOF;
986                 }
987                 tok->buf = buf;
988                 tok->cur = tok->buf + oldlen;
989                 tok->line_start = tok->cur;
990                 strcpy(tok->buf + oldlen, newtok);
991                 PyMem_FREE(newtok);
992                 tok->inp = tok->buf + newlen;
993                 tok->end = tok->inp + 1;
994                 tok->start = tok->buf + start;
995             }
996             else {
997                 tok->lineno++;
998                 if (tok->buf != NULL)
999                     PyMem_FREE(tok->buf);
1000                 tok->buf = newtok;
1001                 tok->cur = tok->buf;
1002                 tok->line_start = tok->buf;
1003                 tok->inp = strchr(tok->buf, '\0');
1004                 tok->end = tok->inp + 1;
1005             }
1006         }
1007         else {
1008             int done = 0;
1009             Py_ssize_t cur = 0;
1010             char *pt;
1011             if (tok->start == NULL) {
1012                 if (tok->buf == NULL) {
1013                     tok->buf = (char *)
1014                         PyMem_MALLOC(BUFSIZ);
1015                     if (tok->buf == NULL) {
1016                         tok->done = E_NOMEM;
1017                         return EOF;
1018                     }
1019                     tok->end = tok->buf + BUFSIZ;
1020                 }
1021                 if (decoding_fgets(tok->buf, (int)(tok->end - tok->buf),
1022                           tok) == NULL) {
1023                     if (!tok->decoding_erred)
1024                         tok->done = E_EOF;
1025                     done = 1;
1026                 }
1027                 else {
1028                     tok->done = E_OK;
1029                     tok->inp = strchr(tok->buf, '\0');
1030                     done = tok->inp == tok->buf || tok->inp[-1] == '\n';
1031                 }
1032             }
1033             else {
1034                 cur = tok->cur - tok->buf;
1035                 if (decoding_feof(tok)) {
1036                     tok->done = E_EOF;
1037                     done = 1;
1038                 }
1039                 else
1040                     tok->done = E_OK;
1041             }
1042             tok->lineno++;
1043             /* Read until '\n' or EOF */
1044             while (!done) {
1045                 Py_ssize_t curstart = tok->start == NULL ? -1 :
1046                           tok->start - tok->buf;
1047                 Py_ssize_t curvalid = tok->inp - tok->buf;
1048                 Py_ssize_t newsize = curvalid + BUFSIZ;
1049                 char *newbuf = tok->buf;
1050                 newbuf = (char *)PyMem_REALLOC(newbuf,
1051                                                newsize);
1052                 if (newbuf == NULL) {
1053                     tok->done = E_NOMEM;
1054                     tok->cur = tok->inp;
1055                     return EOF;
1056                 }
1057                 tok->buf = newbuf;
1058                 tok->cur = tok->buf + cur;
1059                 tok->line_start = tok->cur;
1060                 tok->inp = tok->buf + curvalid;
1061                 tok->end = tok->buf + newsize;
1062                 tok->start = curstart < 0 ? NULL :
1063                          tok->buf + curstart;
1064                 if (decoding_fgets(tok->inp,
1065                                (int)(tok->end - tok->inp),
1066                                tok) == NULL) {
1067                     /* Break out early on decoding
1068                        errors, as tok->buf will be NULL
1069                      */
1070                     if (tok->decoding_erred)
1071                         return EOF;
1072                     /* Last line does not end in \n,
1073                        fake one */
1074                     strcpy(tok->inp, "\n");
1075                 }
1076                 tok->inp = strchr(tok->inp, '\0');
1077                 done = tok->inp[-1] == '\n';
1078             }
1079             if (tok->buf != NULL) {
1080                 tok->cur = tok->buf + cur;
1081                 tok->line_start = tok->cur;
1082                 /* replace "\r\n" with "\n" */
1083                 /* For Mac leave the \r, giving a syntax error */
1084                 pt = tok->inp - 2;
1085                 if (pt >= tok->buf && *pt == '\r') {
1086                     *pt++ = '\n';
1087                     *pt = '\0';
1088                     tok->inp = pt;
1089                 }
1090             }
1091         }
1092         if (tok->done != E_OK) {
1093             if (tok->prompt != NULL)
1094                 PySys_WriteStderr("\n");
1095             tok->cur = tok->inp;
1096             return EOF;
1097         }
1098     }
1099     /*NOTREACHED*/
1100 }
1101 
1102 
1103 /* Back-up one character */
1104 
1105 static void
tok_backup(struct tok_state * tok,int c)1106 tok_backup(struct tok_state *tok, int c)
1107 {
1108     if (c != EOF) {
1109         if (--tok->cur < tok->buf)
1110             Py_FatalError("tok_backup: beginning of buffer");
1111         if (*tok->cur != c)
1112             *tok->cur = c;
1113     }
1114 }
1115 
1116 
1117 /* Return the token corresponding to a single character */
1118 
1119 int
PyToken_OneChar(int c)1120 PyToken_OneChar(int c)
1121 {
1122     switch (c) {
1123     case '(':           return LPAR;
1124     case ')':           return RPAR;
1125     case '[':           return LSQB;
1126     case ']':           return RSQB;
1127     case ':':           return COLON;
1128     case ',':           return COMMA;
1129     case ';':           return SEMI;
1130     case '+':           return PLUS;
1131     case '-':           return MINUS;
1132     case '*':           return STAR;
1133     case '/':           return SLASH;
1134     case '|':           return VBAR;
1135     case '&':           return AMPER;
1136     case '<':           return LESS;
1137     case '>':           return GREATER;
1138     case '=':           return EQUAL;
1139     case '.':           return DOT;
1140     case '%':           return PERCENT;
1141     case '{':           return LBRACE;
1142     case '}':           return RBRACE;
1143     case '^':           return CIRCUMFLEX;
1144     case '~':           return TILDE;
1145     case '@':           return AT;
1146     default:            return OP;
1147     }
1148 }
1149 
1150 
1151 int
PyToken_TwoChars(int c1,int c2)1152 PyToken_TwoChars(int c1, int c2)
1153 {
1154     switch (c1) {
1155     case '=':
1156         switch (c2) {
1157         case '=':               return EQEQUAL;
1158         }
1159         break;
1160     case '!':
1161         switch (c2) {
1162         case '=':               return NOTEQUAL;
1163         }
1164         break;
1165     case '<':
1166         switch (c2) {
1167         case '>':               return NOTEQUAL;
1168         case '=':               return LESSEQUAL;
1169         case '<':               return LEFTSHIFT;
1170         }
1171         break;
1172     case '>':
1173         switch (c2) {
1174         case '=':               return GREATEREQUAL;
1175         case '>':               return RIGHTSHIFT;
1176         }
1177         break;
1178     case '+':
1179         switch (c2) {
1180         case '=':               return PLUSEQUAL;
1181         }
1182         break;
1183     case '-':
1184         switch (c2) {
1185         case '=':               return MINEQUAL;
1186         case '>':               return RARROW;
1187         }
1188         break;
1189     case '*':
1190         switch (c2) {
1191         case '*':               return DOUBLESTAR;
1192         case '=':               return STAREQUAL;
1193         }
1194         break;
1195     case '/':
1196         switch (c2) {
1197         case '/':               return DOUBLESLASH;
1198         case '=':               return SLASHEQUAL;
1199         }
1200         break;
1201     case '|':
1202         switch (c2) {
1203         case '=':               return VBAREQUAL;
1204         }
1205         break;
1206     case '%':
1207         switch (c2) {
1208         case '=':               return PERCENTEQUAL;
1209         }
1210         break;
1211     case '&':
1212         switch (c2) {
1213         case '=':               return AMPEREQUAL;
1214         }
1215         break;
1216     case '^':
1217         switch (c2) {
1218         case '=':               return CIRCUMFLEXEQUAL;
1219         }
1220         break;
1221     case '@':
1222         switch (c2) {
1223         case '=':               return ATEQUAL;
1224         }
1225         break;
1226     }
1227     return OP;
1228 }
1229 
1230 int
PyToken_ThreeChars(int c1,int c2,int c3)1231 PyToken_ThreeChars(int c1, int c2, int c3)
1232 {
1233     switch (c1) {
1234     case '<':
1235         switch (c2) {
1236         case '<':
1237             switch (c3) {
1238             case '=':
1239                 return LEFTSHIFTEQUAL;
1240             }
1241             break;
1242         }
1243         break;
1244     case '>':
1245         switch (c2) {
1246         case '>':
1247             switch (c3) {
1248             case '=':
1249                 return RIGHTSHIFTEQUAL;
1250             }
1251             break;
1252         }
1253         break;
1254     case '*':
1255         switch (c2) {
1256         case '*':
1257             switch (c3) {
1258             case '=':
1259                 return DOUBLESTAREQUAL;
1260             }
1261             break;
1262         }
1263         break;
1264     case '/':
1265         switch (c2) {
1266         case '/':
1267             switch (c3) {
1268             case '=':
1269                 return DOUBLESLASHEQUAL;
1270             }
1271             break;
1272         }
1273         break;
1274     case '.':
1275         switch (c2) {
1276         case '.':
1277             switch (c3) {
1278             case '.':
1279                 return ELLIPSIS;
1280             }
1281             break;
1282         }
1283         break;
1284     }
1285     return OP;
1286 }
1287 
1288 static int
indenterror(struct tok_state * tok)1289 indenterror(struct tok_state *tok)
1290 {
1291     tok->done = E_TABSPACE;
1292     tok->cur = tok->inp;
1293     return ERRORTOKEN;
1294 }
1295 
1296 #ifdef PGEN
1297 #define verify_identifier(tok) 1
1298 #else
1299 /* Verify that the identifier follows PEP 3131.
1300    All identifier strings are guaranteed to be "ready" unicode objects.
1301  */
1302 static int
verify_identifier(struct tok_state * tok)1303 verify_identifier(struct tok_state *tok)
1304 {
1305     PyObject *s;
1306     int result;
1307     if (tok->decoding_erred)
1308         return 0;
1309     s = PyUnicode_DecodeUTF8(tok->start, tok->cur - tok->start, NULL);
1310     if (s == NULL || PyUnicode_READY(s) == -1) {
1311         if (PyErr_ExceptionMatches(PyExc_UnicodeDecodeError)) {
1312             PyErr_Clear();
1313             tok->done = E_IDENTIFIER;
1314         } else {
1315             tok->done = E_ERROR;
1316         }
1317         return 0;
1318     }
1319     result = PyUnicode_IsIdentifier(s);
1320     Py_DECREF(s);
1321     if (result == 0)
1322         tok->done = E_IDENTIFIER;
1323     return result;
1324 }
1325 #endif
1326 
1327 static int
tok_decimal_tail(struct tok_state * tok)1328 tok_decimal_tail(struct tok_state *tok)
1329 {
1330     int c;
1331 
1332     while (1) {
1333         do {
1334             c = tok_nextc(tok);
1335         } while (isdigit(c));
1336         if (c != '_') {
1337             break;
1338         }
1339         c = tok_nextc(tok);
1340         if (!isdigit(c)) {
1341             tok->done = E_TOKEN;
1342             tok_backup(tok, c);
1343             return 0;
1344         }
1345     }
1346     return c;
1347 }
1348 
1349 /* Get next token, after space stripping etc. */
1350 
1351 static int
tok_get(struct tok_state * tok,char ** p_start,char ** p_end)1352 tok_get(struct tok_state *tok, char **p_start, char **p_end)
1353 {
1354     int c;
1355     int blankline, nonascii;
1356 
1357     *p_start = *p_end = NULL;
1358   nextline:
1359     tok->start = NULL;
1360     blankline = 0;
1361 
1362     /* Get indentation level */
1363     if (tok->atbol) {
1364         int col = 0;
1365         int altcol = 0;
1366         tok->atbol = 0;
1367         for (;;) {
1368             c = tok_nextc(tok);
1369             if (c == ' ') {
1370                 col++, altcol++;
1371             }
1372             else if (c == '\t') {
1373                 col = (col / tok->tabsize + 1) * tok->tabsize;
1374                 altcol = (altcol / ALTTABSIZE + 1) * ALTTABSIZE;
1375             }
1376             else if (c == '\014')  {/* Control-L (formfeed) */
1377                 col = altcol = 0; /* For Emacs users */
1378             }
1379             else {
1380                 break;
1381             }
1382         }
1383         tok_backup(tok, c);
1384         if (c == '#' || c == '\n') {
1385             /* Lines with only whitespace and/or comments
1386                shouldn't affect the indentation and are
1387                not passed to the parser as NEWLINE tokens,
1388                except *totally* empty lines in interactive
1389                mode, which signal the end of a command group. */
1390             if (col == 0 && c == '\n' && tok->prompt != NULL) {
1391                 blankline = 0; /* Let it through */
1392             }
1393             else {
1394                 blankline = 1; /* Ignore completely */
1395             }
1396             /* We can't jump back right here since we still
1397                may need to skip to the end of a comment */
1398         }
1399         if (!blankline && tok->level == 0) {
1400             if (col == tok->indstack[tok->indent]) {
1401                 /* No change */
1402                 if (altcol != tok->altindstack[tok->indent]) {
1403                     return indenterror(tok);
1404                 }
1405             }
1406             else if (col > tok->indstack[tok->indent]) {
1407                 /* Indent -- always one */
1408                 if (tok->indent+1 >= MAXINDENT) {
1409                     tok->done = E_TOODEEP;
1410                     tok->cur = tok->inp;
1411                     return ERRORTOKEN;
1412                 }
1413                 if (altcol <= tok->altindstack[tok->indent]) {
1414                     return indenterror(tok);
1415                 }
1416                 tok->pendin++;
1417                 tok->indstack[++tok->indent] = col;
1418                 tok->altindstack[tok->indent] = altcol;
1419             }
1420             else /* col < tok->indstack[tok->indent] */ {
1421                 /* Dedent -- any number, must be consistent */
1422                 while (tok->indent > 0 &&
1423                     col < tok->indstack[tok->indent]) {
1424                     tok->pendin--;
1425                     tok->indent--;
1426                 }
1427                 if (col != tok->indstack[tok->indent]) {
1428                     tok->done = E_DEDENT;
1429                     tok->cur = tok->inp;
1430                     return ERRORTOKEN;
1431                 }
1432                 if (altcol != tok->altindstack[tok->indent]) {
1433                     return indenterror(tok);
1434                 }
1435             }
1436         }
1437     }
1438 
1439     tok->start = tok->cur;
1440 
1441     /* Return pending indents/dedents */
1442     if (tok->pendin != 0) {
1443         if (tok->pendin < 0) {
1444             tok->pendin++;
1445             return DEDENT;
1446         }
1447         else {
1448             tok->pendin--;
1449             return INDENT;
1450         }
1451     }
1452 
1453  again:
1454     tok->start = NULL;
1455     /* Skip spaces */
1456     do {
1457         c = tok_nextc(tok);
1458     } while (c == ' ' || c == '\t' || c == '\014');
1459 
1460     /* Set start of current token */
1461     tok->start = tok->cur - 1;
1462 
1463     /* Skip comment */
1464     if (c == '#') {
1465         while (c != EOF && c != '\n') {
1466             c = tok_nextc(tok);
1467         }
1468     }
1469 
1470     /* Check for EOF and errors now */
1471     if (c == EOF) {
1472         return tok->done == E_EOF ? ENDMARKER : ERRORTOKEN;
1473     }
1474 
1475     /* Identifier (most frequent token!) */
1476     nonascii = 0;
1477     if (is_potential_identifier_start(c)) {
1478         /* Process the various legal combinations of b"", r"", u"", and f"". */
1479         int saw_b = 0, saw_r = 0, saw_u = 0, saw_f = 0;
1480         while (1) {
1481             if (!(saw_b || saw_u || saw_f) && (c == 'b' || c == 'B'))
1482                 saw_b = 1;
1483             /* Since this is a backwards compatibility support literal we don't
1484                want to support it in arbitrary order like byte literals. */
1485             else if (!(saw_b || saw_u || saw_r || saw_f)
1486                      && (c == 'u'|| c == 'U')) {
1487                 saw_u = 1;
1488             }
1489             /* ur"" and ru"" are not supported */
1490             else if (!(saw_r || saw_u) && (c == 'r' || c == 'R')) {
1491                 saw_r = 1;
1492             }
1493             else if (!(saw_f || saw_b || saw_u) && (c == 'f' || c == 'F')) {
1494                 saw_f = 1;
1495             }
1496             else {
1497                 break;
1498             }
1499             c = tok_nextc(tok);
1500             if (c == '"' || c == '\'') {
1501                 goto letter_quote;
1502             }
1503         }
1504         while (is_potential_identifier_char(c)) {
1505             if (c >= 128) {
1506                 nonascii = 1;
1507             }
1508             c = tok_nextc(tok);
1509         }
1510         tok_backup(tok, c);
1511         if (nonascii && !verify_identifier(tok)) {
1512             return ERRORTOKEN;
1513         }
1514         *p_start = tok->start;
1515         *p_end = tok->cur;
1516 
1517         return NAME;
1518     }
1519 
1520     /* Newline */
1521     if (c == '\n') {
1522         tok->atbol = 1;
1523         if (blankline || tok->level > 0) {
1524             goto nextline;
1525         }
1526         *p_start = tok->start;
1527         *p_end = tok->cur - 1; /* Leave '\n' out of the string */
1528         tok->cont_line = 0;
1529         return NEWLINE;
1530     }
1531 
1532     /* Period or number starting with period? */
1533     if (c == '.') {
1534         c = tok_nextc(tok);
1535         if (isdigit(c)) {
1536             goto fraction;
1537         } else if (c == '.') {
1538             c = tok_nextc(tok);
1539             if (c == '.') {
1540                 *p_start = tok->start;
1541                 *p_end = tok->cur;
1542                 return ELLIPSIS;
1543             }
1544             else {
1545                 tok_backup(tok, c);
1546             }
1547             tok_backup(tok, '.');
1548         }
1549         else {
1550             tok_backup(tok, c);
1551         }
1552         *p_start = tok->start;
1553         *p_end = tok->cur;
1554         return DOT;
1555     }
1556 
1557     /* Number */
1558     if (isdigit(c)) {
1559         if (c == '0') {
1560             /* Hex, octal or binary -- maybe. */
1561             c = tok_nextc(tok);
1562             if (c == 'x' || c == 'X') {
1563                 /* Hex */
1564                 c = tok_nextc(tok);
1565                 do {
1566                     if (c == '_') {
1567                         c = tok_nextc(tok);
1568                     }
1569                     if (!isxdigit(c)) {
1570                         tok->done = E_TOKEN;
1571                         tok_backup(tok, c);
1572                         return ERRORTOKEN;
1573                     }
1574                     do {
1575                         c = tok_nextc(tok);
1576                     } while (isxdigit(c));
1577                 } while (c == '_');
1578             }
1579             else if (c == 'o' || c == 'O') {
1580                 /* Octal */
1581                 c = tok_nextc(tok);
1582                 do {
1583                     if (c == '_') {
1584                         c = tok_nextc(tok);
1585                     }
1586                     if (c < '0' || c >= '8') {
1587                         tok->done = E_TOKEN;
1588                         tok_backup(tok, c);
1589                         return ERRORTOKEN;
1590                     }
1591                     do {
1592                         c = tok_nextc(tok);
1593                     } while ('0' <= c && c < '8');
1594                 } while (c == '_');
1595             }
1596             else if (c == 'b' || c == 'B') {
1597                 /* Binary */
1598                 c = tok_nextc(tok);
1599                 do {
1600                     if (c == '_') {
1601                         c = tok_nextc(tok);
1602                     }
1603                     if (c != '0' && c != '1') {
1604                         tok->done = E_TOKEN;
1605                         tok_backup(tok, c);
1606                         return ERRORTOKEN;
1607                     }
1608                     do {
1609                         c = tok_nextc(tok);
1610                     } while (c == '0' || c == '1');
1611                 } while (c == '_');
1612             }
1613             else {
1614                 int nonzero = 0;
1615                 /* maybe old-style octal; c is first char of it */
1616                 /* in any case, allow '0' as a literal */
1617                 while (1) {
1618                     if (c == '_') {
1619                         c = tok_nextc(tok);
1620                         if (!isdigit(c)) {
1621                             tok->done = E_TOKEN;
1622                             tok_backup(tok, c);
1623                             return ERRORTOKEN;
1624                         }
1625                     }
1626                     if (c != '0') {
1627                         break;
1628                     }
1629                     c = tok_nextc(tok);
1630                 }
1631                 if (isdigit(c)) {
1632                     nonzero = 1;
1633                     c = tok_decimal_tail(tok);
1634                     if (c == 0) {
1635                         return ERRORTOKEN;
1636                     }
1637                 }
1638                 if (c == '.') {
1639                     c = tok_nextc(tok);
1640                     goto fraction;
1641                 }
1642                 else if (c == 'e' || c == 'E') {
1643                     goto exponent;
1644                 }
1645                 else if (c == 'j' || c == 'J') {
1646                     goto imaginary;
1647                 }
1648                 else if (nonzero) {
1649                     /* Old-style octal: now disallowed. */
1650                     tok->done = E_TOKEN;
1651                     tok_backup(tok, c);
1652                     return ERRORTOKEN;
1653                 }
1654             }
1655         }
1656         else {
1657             /* Decimal */
1658             c = tok_decimal_tail(tok);
1659             if (c == 0) {
1660                 return ERRORTOKEN;
1661             }
1662             {
1663                 /* Accept floating point numbers. */
1664                 if (c == '.') {
1665                     c = tok_nextc(tok);
1666         fraction:
1667                     /* Fraction */
1668                     if (isdigit(c)) {
1669                         c = tok_decimal_tail(tok);
1670                         if (c == 0) {
1671                             return ERRORTOKEN;
1672                         }
1673                     }
1674                 }
1675                 if (c == 'e' || c == 'E') {
1676                     int e;
1677                   exponent:
1678                     e = c;
1679                     /* Exponent part */
1680                     c = tok_nextc(tok);
1681                     if (c == '+' || c == '-') {
1682                         c = tok_nextc(tok);
1683                         if (!isdigit(c)) {
1684                             tok->done = E_TOKEN;
1685                             tok_backup(tok, c);
1686                             return ERRORTOKEN;
1687                         }
1688                     } else if (!isdigit(c)) {
1689                         tok_backup(tok, c);
1690                         tok_backup(tok, e);
1691                         *p_start = tok->start;
1692                         *p_end = tok->cur;
1693                         return NUMBER;
1694                     }
1695                     c = tok_decimal_tail(tok);
1696                     if (c == 0) {
1697                         return ERRORTOKEN;
1698                     }
1699                 }
1700                 if (c == 'j' || c == 'J') {
1701                     /* Imaginary part */
1702         imaginary:
1703                     c = tok_nextc(tok);
1704                 }
1705             }
1706         }
1707         tok_backup(tok, c);
1708         *p_start = tok->start;
1709         *p_end = tok->cur;
1710         return NUMBER;
1711     }
1712 
1713   letter_quote:
1714     /* String */
1715     if (c == '\'' || c == '"') {
1716         int quote = c;
1717         int quote_size = 1;             /* 1 or 3 */
1718         int end_quote_size = 0;
1719 
1720         /* Find the quote size and start of string */
1721         c = tok_nextc(tok);
1722         if (c == quote) {
1723             c = tok_nextc(tok);
1724             if (c == quote) {
1725                 quote_size = 3;
1726             }
1727             else {
1728                 end_quote_size = 1;     /* empty string found */
1729             }
1730         }
1731         if (c != quote) {
1732             tok_backup(tok, c);
1733         }
1734 
1735         /* Get rest of string */
1736         while (end_quote_size != quote_size) {
1737             c = tok_nextc(tok);
1738             if (c == EOF) {
1739                 if (quote_size == 3) {
1740                     tok->done = E_EOFS;
1741                 }
1742                 else {
1743                     tok->done = E_EOLS;
1744                 }
1745                 tok->cur = tok->inp;
1746                 return ERRORTOKEN;
1747             }
1748             if (quote_size == 1 && c == '\n') {
1749                 tok->done = E_EOLS;
1750                 tok->cur = tok->inp;
1751                 return ERRORTOKEN;
1752             }
1753             if (c == quote) {
1754                 end_quote_size += 1;
1755             }
1756             else {
1757                 end_quote_size = 0;
1758                 if (c == '\\') {
1759                     tok_nextc(tok);  /* skip escaped char */
1760                 }
1761             }
1762         }
1763 
1764         *p_start = tok->start;
1765         *p_end = tok->cur;
1766         return STRING;
1767     }
1768 
1769     /* Line continuation */
1770     if (c == '\\') {
1771         c = tok_nextc(tok);
1772         if (c != '\n') {
1773             tok->done = E_LINECONT;
1774             tok->cur = tok->inp;
1775             return ERRORTOKEN;
1776         }
1777         tok->cont_line = 1;
1778         goto again; /* Read next line */
1779     }
1780 
1781     /* Check for two-character token */
1782     {
1783         int c2 = tok_nextc(tok);
1784         int token = PyToken_TwoChars(c, c2);
1785         if (token != OP) {
1786             int c3 = tok_nextc(tok);
1787             int token3 = PyToken_ThreeChars(c, c2, c3);
1788             if (token3 != OP) {
1789                 token = token3;
1790             }
1791             else {
1792                 tok_backup(tok, c3);
1793             }
1794             *p_start = tok->start;
1795             *p_end = tok->cur;
1796             return token;
1797         }
1798         tok_backup(tok, c2);
1799     }
1800 
1801     /* Keep track of parentheses nesting level */
1802     switch (c) {
1803     case '(':
1804     case '[':
1805     case '{':
1806         tok->level++;
1807         break;
1808     case ')':
1809     case ']':
1810     case '}':
1811         tok->level--;
1812         break;
1813     }
1814 
1815     /* Punctuation character */
1816     *p_start = tok->start;
1817     *p_end = tok->cur;
1818     return PyToken_OneChar(c);
1819 }
1820 
1821 int
PyTokenizer_Get(struct tok_state * tok,char ** p_start,char ** p_end)1822 PyTokenizer_Get(struct tok_state *tok, char **p_start, char **p_end)
1823 {
1824     int result = tok_get(tok, p_start, p_end);
1825     if (tok->decoding_erred) {
1826         result = ERRORTOKEN;
1827         tok->done = E_DECODE;
1828     }
1829     return result;
1830 }
1831 
1832 /* Get the encoding of a Python file. Check for the coding cookie and check if
1833    the file starts with a BOM.
1834 
1835    PyTokenizer_FindEncodingFilename() returns NULL when it can't find the
1836    encoding in the first or second line of the file (in which case the encoding
1837    should be assumed to be UTF-8).
1838 
1839    The char* returned is malloc'ed via PyMem_MALLOC() and thus must be freed
1840    by the caller. */
1841 
1842 char *
PyTokenizer_FindEncodingFilename(int fd,PyObject * filename)1843 PyTokenizer_FindEncodingFilename(int fd, PyObject *filename)
1844 {
1845     struct tok_state *tok;
1846     FILE *fp;
1847     char *p_start =NULL , *p_end =NULL , *encoding = NULL;
1848 
1849 #ifndef PGEN
1850     fd = _Py_dup(fd);
1851 #else
1852     fd = dup(fd);
1853 #endif
1854     if (fd < 0) {
1855         return NULL;
1856     }
1857 
1858     fp = fdopen(fd, "r");
1859     if (fp == NULL) {
1860         return NULL;
1861     }
1862     tok = PyTokenizer_FromFile(fp, NULL, NULL, NULL);
1863     if (tok == NULL) {
1864         fclose(fp);
1865         return NULL;
1866     }
1867 #ifndef PGEN
1868     if (filename != NULL) {
1869         Py_INCREF(filename);
1870         tok->filename = filename;
1871     }
1872     else {
1873         tok->filename = PyUnicode_FromString("<string>");
1874         if (tok->filename == NULL) {
1875             fclose(fp);
1876             PyTokenizer_Free(tok);
1877             return encoding;
1878         }
1879     }
1880 #endif
1881     while (tok->lineno < 2 && tok->done == E_OK) {
1882         PyTokenizer_Get(tok, &p_start, &p_end);
1883     }
1884     fclose(fp);
1885     if (tok->encoding) {
1886         encoding = (char *)PyMem_MALLOC(strlen(tok->encoding) + 1);
1887         if (encoding)
1888         strcpy(encoding, tok->encoding);
1889     }
1890     PyTokenizer_Free(tok);
1891     return encoding;
1892 }
1893 
1894 char *
PyTokenizer_FindEncoding(int fd)1895 PyTokenizer_FindEncoding(int fd)
1896 {
1897     return PyTokenizer_FindEncodingFilename(fd, NULL);
1898 }
1899 
1900 #ifdef Py_DEBUG
1901 
1902 void
tok_dump(int type,char * start,char * end)1903 tok_dump(int type, char *start, char *end)
1904 {
1905     printf("%s", _PyParser_TokenNames[type]);
1906     if (type == NAME || type == NUMBER || type == STRING || type == OP)
1907         printf("(%.*s)", (int)(end - start), start);
1908 }
1909 
1910 #endif
1911