source: python/vendor/current/Objects/fileobject.c

Last change on this file was 388, checked in by dmik, 11 years ago

python: Update vendor to 2.7.6.

  • Property svn:eol-style set to native
File size: 86.2 KB
Line 
1/* File object implementation */
2
3#define PY_SSIZE_T_CLEAN
4#include "Python.h"
5#include "structmember.h"
6
7#ifdef HAVE_SYS_TYPES_H
8#include <sys/types.h>
9#endif /* HAVE_SYS_TYPES_H */
10
11#ifdef MS_WINDOWS
12#define fileno _fileno
13/* can simulate truncate with Win32 API functions; see file_truncate */
14#define HAVE_FTRUNCATE
15#define WIN32_LEAN_AND_MEAN
16#include <windows.h>
17#endif
18
19#if defined(PYOS_OS2) && defined(PYCC_GCC)
20#include <io.h>
21#endif
22
23#define BUF(v) PyString_AS_STRING((PyStringObject *)v)
24
25#ifdef HAVE_ERRNO_H
26#include <errno.h>
27#endif
28
29#ifdef HAVE_GETC_UNLOCKED
30#define GETC(f) getc_unlocked(f)
31#define FLOCKFILE(f) flockfile(f)
32#define FUNLOCKFILE(f) funlockfile(f)
33#else
34#define GETC(f) getc(f)
35#define FLOCKFILE(f)
36#define FUNLOCKFILE(f)
37#endif
38
39/* Bits in f_newlinetypes */
40#define NEWLINE_UNKNOWN 0 /* No newline seen, yet */
41#define NEWLINE_CR 1 /* \r newline seen */
42#define NEWLINE_LF 2 /* \n newline seen */
43#define NEWLINE_CRLF 4 /* \r\n newline seen */
44
45/*
46 * These macros release the GIL while preventing the f_close() function being
47 * called in the interval between them. For that purpose, a running total of
48 * the number of currently running unlocked code sections is kept in
49 * the unlocked_count field of the PyFileObject. The close() method raises
50 * an IOError if that field is non-zero. See issue #815646, #595601.
51 */
52
53#define FILE_BEGIN_ALLOW_THREADS(fobj) \
54{ \
55 fobj->unlocked_count++; \
56 Py_BEGIN_ALLOW_THREADS
57
58#define FILE_END_ALLOW_THREADS(fobj) \
59 Py_END_ALLOW_THREADS \
60 fobj->unlocked_count--; \
61 assert(fobj->unlocked_count >= 0); \
62}
63
64#define FILE_ABORT_ALLOW_THREADS(fobj) \
65 Py_BLOCK_THREADS \
66 fobj->unlocked_count--; \
67 assert(fobj->unlocked_count >= 0);
68
69#ifdef __cplusplus
70extern "C" {
71#endif
72
73FILE *
74PyFile_AsFile(PyObject *f)
75{
76 if (f == NULL || !PyFile_Check(f))
77 return NULL;
78 else
79 return ((PyFileObject *)f)->f_fp;
80}
81
82void PyFile_IncUseCount(PyFileObject *fobj)
83{
84 fobj->unlocked_count++;
85}
86
87void PyFile_DecUseCount(PyFileObject *fobj)
88{
89 fobj->unlocked_count--;
90 assert(fobj->unlocked_count >= 0);
91}
92
93PyObject *
94PyFile_Name(PyObject *f)
95{
96 if (f == NULL || !PyFile_Check(f))
97 return NULL;
98 else
99 return ((PyFileObject *)f)->f_name;
100}
101
102/* This is a safe wrapper around PyObject_Print to print to the FILE
103 of a PyFileObject. PyObject_Print releases the GIL but knows nothing
104 about PyFileObject. */
105static int
106file_PyObject_Print(PyObject *op, PyFileObject *f, int flags)
107{
108 int result;
109 PyFile_IncUseCount(f);
110 result = PyObject_Print(op, f->f_fp, flags);
111 PyFile_DecUseCount(f);
112 return result;
113}
114
115/* On Unix, fopen will succeed for directories.
116 In Python, there should be no file objects referring to
117 directories, so we need a check. */
118
119static PyFileObject*
120dircheck(PyFileObject* f)
121{
122#if defined(HAVE_FSTAT) && defined(S_IFDIR) && defined(EISDIR)
123 struct stat buf;
124 if (f->f_fp == NULL)
125 return f;
126 if (fstat(fileno(f->f_fp), &buf) == 0 &&
127 S_ISDIR(buf.st_mode)) {
128 char *msg = strerror(EISDIR);
129 PyObject *exc = PyObject_CallFunction(PyExc_IOError, "(isO)",
130 EISDIR, msg, f->f_name);
131 PyErr_SetObject(PyExc_IOError, exc);
132 Py_XDECREF(exc);
133 return NULL;
134 }
135#endif
136 return f;
137}
138
139
140static PyObject *
141fill_file_fields(PyFileObject *f, FILE *fp, PyObject *name, char *mode,
142 int (*close)(FILE *))
143{
144 assert(name != NULL);
145 assert(f != NULL);
146 assert(PyFile_Check(f));
147 assert(f->f_fp == NULL);
148
149 Py_DECREF(f->f_name);
150 Py_DECREF(f->f_mode);
151 Py_DECREF(f->f_encoding);
152 Py_DECREF(f->f_errors);
153
154 Py_INCREF(name);
155 f->f_name = name;
156
157 f->f_mode = PyString_FromString(mode);
158
159 f->f_close = close;
160 f->f_softspace = 0;
161 f->f_binary = strchr(mode,'b') != NULL;
162 f->f_buf = NULL;
163 f->f_univ_newline = (strchr(mode, 'U') != NULL);
164 f->f_newlinetypes = NEWLINE_UNKNOWN;
165 f->f_skipnextlf = 0;
166 Py_INCREF(Py_None);
167 f->f_encoding = Py_None;
168 Py_INCREF(Py_None);
169 f->f_errors = Py_None;
170 f->readable = f->writable = 0;
171 if (strchr(mode, 'r') != NULL || f->f_univ_newline)
172 f->readable = 1;
173 if (strchr(mode, 'w') != NULL || strchr(mode, 'a') != NULL)
174 f->writable = 1;
175 if (strchr(mode, '+') != NULL)
176 f->readable = f->writable = 1;
177
178 if (f->f_mode == NULL)
179 return NULL;
180 f->f_fp = fp;
181 f = dircheck(f);
182 return (PyObject *) f;
183}
184
185#if defined _MSC_VER && _MSC_VER >= 1400 && defined(__STDC_SECURE_LIB__)
186#define Py_VERIFY_WINNT
187/* The CRT on windows compiled with Visual Studio 2005 and higher may
188 * assert if given invalid mode strings. This is all fine and well
189 * in static languages like C where the mode string is typcially hard
190 * coded. But in Python, were we pass in the mode string from the user,
191 * we need to verify it first manually
192 */
193static int _PyVerify_Mode_WINNT(const char *mode)
194{
195 /* See if mode string is valid on Windows to avoid hard assertions */
196 /* remove leading spacese */
197 int singles = 0;
198 int pairs = 0;
199 int encoding = 0;
200 const char *s, *c;
201
202 while(*mode == ' ') /* strip initial spaces */
203 ++mode;
204 if (!strchr("rwa", *mode)) /* must start with one of these */
205 return 0;
206 while (*++mode) {
207 if (*mode == ' ' || *mode == 'N') /* ignore spaces and N */
208 continue;
209 s = "+TD"; /* each of this can appear only once */
210 c = strchr(s, *mode);
211 if (c) {
212 ptrdiff_t idx = s-c;
213 if (singles & (1<<idx))
214 return 0;
215 singles |= (1<<idx);
216 continue;
217 }
218 s = "btcnSR"; /* only one of each letter in the pairs allowed */
219 c = strchr(s, *mode);
220 if (c) {
221 ptrdiff_t idx = (s-c)/2;
222 if (pairs & (1<<idx))
223 return 0;
224 pairs |= (1<<idx);
225 continue;
226 }
227 if (*mode == ',') {
228 encoding = 1;
229 break;
230 }
231 return 0; /* found an invalid char */
232 }
233
234 if (encoding) {
235 char *e[] = {"UTF-8", "UTF-16LE", "UNICODE"};
236 while (*mode == ' ')
237 ++mode;
238 /* find 'ccs =' */
239 if (strncmp(mode, "ccs", 3))
240 return 0;
241 mode += 3;
242 while (*mode == ' ')
243 ++mode;
244 if (*mode != '=')
245 return 0;
246 while (*mode == ' ')
247 ++mode;
248 for(encoding = 0; encoding<_countof(e); ++encoding) {
249 size_t l = strlen(e[encoding]);
250 if (!strncmp(mode, e[encoding], l)) {
251 mode += l; /* found a valid encoding */
252 break;
253 }
254 }
255 if (encoding == _countof(e))
256 return 0;
257 }
258 /* skip trailing spaces */
259 while (*mode == ' ')
260 ++mode;
261
262 return *mode == '\0'; /* must be at the end of the string */
263}
264#endif
265
266/* check for known incorrect mode strings - problem is, platforms are
267 free to accept any mode characters they like and are supposed to
268 ignore stuff they don't understand... write or append mode with
269 universal newline support is expressly forbidden by PEP 278.
270 Additionally, remove the 'U' from the mode string as platforms
271 won't know what it is. Non-zero return signals an exception */
272int
273_PyFile_SanitizeMode(char *mode)
274{
275 char *upos;
276 size_t len = strlen(mode);
277
278 if (!len) {
279 PyErr_SetString(PyExc_ValueError, "empty mode string");
280 return -1;
281 }
282
283 upos = strchr(mode, 'U');
284 if (upos) {
285 memmove(upos, upos+1, len-(upos-mode)); /* incl null char */
286
287 if (mode[0] == 'w' || mode[0] == 'a') {
288 PyErr_Format(PyExc_ValueError, "universal newline "
289 "mode can only be used with modes "
290 "starting with 'r'");
291 return -1;
292 }
293
294 if (mode[0] != 'r') {
295 memmove(mode+1, mode, strlen(mode)+1);
296 mode[0] = 'r';
297 }
298
299 if (!strchr(mode, 'b')) {
300 memmove(mode+2, mode+1, strlen(mode));
301 mode[1] = 'b';
302 }
303 } else if (mode[0] != 'r' && mode[0] != 'w' && mode[0] != 'a') {
304 PyErr_Format(PyExc_ValueError, "mode string must begin with "
305 "one of 'r', 'w', 'a' or 'U', not '%.200s'", mode);
306 return -1;
307 }
308#ifdef Py_VERIFY_WINNT
309 /* additional checks on NT with visual studio 2005 and higher */
310 if (!_PyVerify_Mode_WINNT(mode)) {
311 PyErr_Format(PyExc_ValueError, "Invalid mode ('%.50s')", mode);
312 return -1;
313 }
314#endif
315 return 0;
316}
317
318static PyObject *
319open_the_file(PyFileObject *f, char *name, char *mode)
320{
321 char *newmode;
322 assert(f != NULL);
323 assert(PyFile_Check(f));
324#ifdef MS_WINDOWS
325 /* windows ignores the passed name in order to support Unicode */
326 assert(f->f_name != NULL);
327#else
328 assert(name != NULL);
329#endif
330 assert(mode != NULL);
331 assert(f->f_fp == NULL);
332
333 /* probably need to replace 'U' by 'rb' */
334 newmode = PyMem_MALLOC(strlen(mode) + 3);
335 if (!newmode) {
336 PyErr_NoMemory();
337 return NULL;
338 }
339 strcpy(newmode, mode);
340
341 if (_PyFile_SanitizeMode(newmode)) {
342 f = NULL;
343 goto cleanup;
344 }
345
346 /* rexec.py can't stop a user from getting the file() constructor --
347 all they have to do is get *any* file object f, and then do
348 type(f). Here we prevent them from doing damage with it. */
349 if (PyEval_GetRestricted()) {
350 PyErr_SetString(PyExc_IOError,
351 "file() constructor not accessible in restricted mode");
352 f = NULL;
353 goto cleanup;
354 }
355 errno = 0;
356
357#ifdef MS_WINDOWS
358 if (PyUnicode_Check(f->f_name)) {
359 PyObject *wmode;
360 wmode = PyUnicode_DecodeASCII(newmode, strlen(newmode), NULL);
361 if (f->f_name && wmode) {
362 FILE_BEGIN_ALLOW_THREADS(f)
363 /* PyUnicode_AS_UNICODE OK without thread
364 lock as it is a simple dereference. */
365 f->f_fp = _wfopen(PyUnicode_AS_UNICODE(f->f_name),
366 PyUnicode_AS_UNICODE(wmode));
367 FILE_END_ALLOW_THREADS(f)
368 }
369 Py_XDECREF(wmode);
370 }
371#endif
372 if (NULL == f->f_fp && NULL != name) {
373 FILE_BEGIN_ALLOW_THREADS(f)
374 f->f_fp = fopen(name, newmode);
375 FILE_END_ALLOW_THREADS(f)
376 }
377
378 if (f->f_fp == NULL) {
379#if defined _MSC_VER && (_MSC_VER < 1400 || !defined(__STDC_SECURE_LIB__))
380 /* MSVC 6 (Microsoft) leaves errno at 0 for bad mode strings,
381 * across all Windows flavors. When it sets EINVAL varies
382 * across Windows flavors, the exact conditions aren't
383 * documented, and the answer lies in the OS's implementation
384 * of Win32's CreateFile function (whose source is secret).
385 * Seems the best we can do is map EINVAL to ENOENT.
386 * Starting with Visual Studio .NET 2005, EINVAL is correctly
387 * set by our CRT error handler (set in exceptions.c.)
388 */
389 if (errno == 0) /* bad mode string */
390 errno = EINVAL;
391 else if (errno == EINVAL) /* unknown, but not a mode string */
392 errno = ENOENT;
393#endif
394 /* EINVAL is returned when an invalid filename or
395 * an invalid mode is supplied. */
396 if (errno == EINVAL) {
397 PyObject *v;
398 char message[100];
399 PyOS_snprintf(message, 100,
400 "invalid mode ('%.50s') or filename", mode);
401 v = Py_BuildValue("(isO)", errno, message, f->f_name);
402 if (v != NULL) {
403 PyErr_SetObject(PyExc_IOError, v);
404 Py_DECREF(v);
405 }
406 }
407 else
408 PyErr_SetFromErrnoWithFilenameObject(PyExc_IOError, f->f_name);
409 f = NULL;
410 }
411 if (f != NULL)
412 f = dircheck(f);
413
414cleanup:
415 PyMem_FREE(newmode);
416
417 return (PyObject *)f;
418}
419
420static PyObject *
421close_the_file(PyFileObject *f)
422{
423 int sts = 0;
424 int (*local_close)(FILE *);
425 FILE *local_fp = f->f_fp;
426 char *local_setbuf = f->f_setbuf;
427 if (local_fp != NULL) {
428 local_close = f->f_close;
429 if (local_close != NULL && f->unlocked_count > 0) {
430 if (f->ob_refcnt > 0) {
431 PyErr_SetString(PyExc_IOError,
432 "close() called during concurrent "
433 "operation on the same file object.");
434 } else {
435 /* This should not happen unless someone is
436 * carelessly playing with the PyFileObject
437 * struct fields and/or its associated FILE
438 * pointer. */
439 PyErr_SetString(PyExc_SystemError,
440 "PyFileObject locking error in "
441 "destructor (refcnt <= 0 at close).");
442 }
443 return NULL;
444 }
445 /* NULL out the FILE pointer before releasing the GIL, because
446 * it will not be valid anymore after the close() function is
447 * called. */
448 f->f_fp = NULL;
449 if (local_close != NULL) {
450 /* Issue #9295: must temporarily reset f_setbuf so that another
451 thread doesn't free it when running file_close() concurrently.
452 Otherwise this close() will crash when flushing the buffer. */
453 f->f_setbuf = NULL;
454 Py_BEGIN_ALLOW_THREADS
455 errno = 0;
456 sts = (*local_close)(local_fp);
457 Py_END_ALLOW_THREADS
458 f->f_setbuf = local_setbuf;
459 if (sts == EOF)
460 return PyErr_SetFromErrno(PyExc_IOError);
461 if (sts != 0)
462 return PyInt_FromLong((long)sts);
463 }
464 }
465 Py_RETURN_NONE;
466}
467
468PyObject *
469PyFile_FromFile(FILE *fp, char *name, char *mode, int (*close)(FILE *))
470{
471 PyFileObject *f;
472 PyObject *o_name;
473
474 f = (PyFileObject *)PyFile_Type.tp_new(&PyFile_Type, NULL, NULL);
475 if (f == NULL)
476 return NULL;
477 o_name = PyString_FromString(name);
478 if (o_name == NULL) {
479 if (close != NULL && fp != NULL)
480 close(fp);
481 Py_DECREF(f);
482 return NULL;
483 }
484 if (fill_file_fields(f, fp, o_name, mode, close) == NULL) {
485 Py_DECREF(f);
486 Py_DECREF(o_name);
487 return NULL;
488 }
489 Py_DECREF(o_name);
490 return (PyObject *)f;
491}
492
493PyObject *
494PyFile_FromString(char *name, char *mode)
495{
496 extern int fclose(FILE *);
497 PyFileObject *f;
498
499 f = (PyFileObject *)PyFile_FromFile((FILE *)NULL, name, mode, fclose);
500 if (f != NULL) {
501 if (open_the_file(f, name, mode) == NULL) {
502 Py_DECREF(f);
503 f = NULL;
504 }
505 }
506 return (PyObject *)f;
507}
508
509void
510PyFile_SetBufSize(PyObject *f, int bufsize)
511{
512 PyFileObject *file = (PyFileObject *)f;
513 if (bufsize >= 0) {
514 int type;
515 switch (bufsize) {
516 case 0:
517 type = _IONBF;
518 break;
519#ifdef HAVE_SETVBUF
520 case 1:
521 type = _IOLBF;
522 bufsize = BUFSIZ;
523 break;
524#endif
525 default:
526 type = _IOFBF;
527#ifndef HAVE_SETVBUF
528 bufsize = BUFSIZ;
529#endif
530 break;
531 }
532 fflush(file->f_fp);
533 if (type == _IONBF) {
534 PyMem_Free(file->f_setbuf);
535 file->f_setbuf = NULL;
536 } else {
537 file->f_setbuf = (char *)PyMem_Realloc(file->f_setbuf,
538 bufsize);
539 }
540#ifdef HAVE_SETVBUF
541 setvbuf(file->f_fp, file->f_setbuf, type, bufsize);
542#else /* !HAVE_SETVBUF */
543 setbuf(file->f_fp, file->f_setbuf);
544#endif /* !HAVE_SETVBUF */
545 }
546}
547
548/* Set the encoding used to output Unicode strings.
549 Return 1 on success, 0 on failure. */
550
551int
552PyFile_SetEncoding(PyObject *f, const char *enc)
553{
554 return PyFile_SetEncodingAndErrors(f, enc, NULL);
555}
556
557int
558PyFile_SetEncodingAndErrors(PyObject *f, const char *enc, char* errors)
559{
560 PyFileObject *file = (PyFileObject*)f;
561 PyObject *str, *oerrors;
562
563 assert(PyFile_Check(f));
564 str = PyString_FromString(enc);
565 if (!str)
566 return 0;
567 if (errors) {
568 oerrors = PyString_FromString(errors);
569 if (!oerrors) {
570 Py_DECREF(str);
571 return 0;
572 }
573 } else {
574 oerrors = Py_None;
575 Py_INCREF(Py_None);
576 }
577 Py_DECREF(file->f_encoding);
578 file->f_encoding = str;
579 Py_DECREF(file->f_errors);
580 file->f_errors = oerrors;
581 return 1;
582}
583
584static PyObject *
585err_closed(void)
586{
587 PyErr_SetString(PyExc_ValueError, "I/O operation on closed file");
588 return NULL;
589}
590
591static PyObject *
592err_mode(char *action)
593{
594 PyErr_Format(PyExc_IOError, "File not open for %s", action);
595 return NULL;
596}
597
598/* Refuse regular file I/O if there's data in the iteration-buffer.
599 * Mixing them would cause data to arrive out of order, as the read*
600 * methods don't use the iteration buffer. */
601static PyObject *
602err_iterbuffered(void)
603{
604 PyErr_SetString(PyExc_ValueError,
605 "Mixing iteration and read methods would lose data");
606 return NULL;
607}
608
609static void drop_readahead(PyFileObject *);
610
611/* Methods */
612
613static void
614file_dealloc(PyFileObject *f)
615{
616 PyObject *ret;
617 if (f->weakreflist != NULL)
618 PyObject_ClearWeakRefs((PyObject *) f);
619 ret = close_the_file(f);
620 if (!ret) {
621 PySys_WriteStderr("close failed in file object destructor:\n");
622 PyErr_Print();
623 }
624 else {
625 Py_DECREF(ret);
626 }
627 PyMem_Free(f->f_setbuf);
628 Py_XDECREF(f->f_name);
629 Py_XDECREF(f->f_mode);
630 Py_XDECREF(f->f_encoding);
631 Py_XDECREF(f->f_errors);
632 drop_readahead(f);
633 Py_TYPE(f)->tp_free((PyObject *)f);
634}
635
636static PyObject *
637file_repr(PyFileObject *f)
638{
639 PyObject *ret = NULL;
640 PyObject *name = NULL;
641 if (PyUnicode_Check(f->f_name)) {
642#ifdef Py_USING_UNICODE
643 const char *name_str;
644 name = PyUnicode_AsUnicodeEscapeString(f->f_name);
645 name_str = name ? PyString_AsString(name) : "?";
646 ret = PyString_FromFormat("<%s file u'%s', mode '%s' at %p>",
647 f->f_fp == NULL ? "closed" : "open",
648 name_str,
649 PyString_AsString(f->f_mode),
650 f);
651 Py_XDECREF(name);
652 return ret;
653#endif
654 } else {
655 name = PyObject_Repr(f->f_name);
656 if (name == NULL)
657 return NULL;
658 ret = PyString_FromFormat("<%s file %s, mode '%s' at %p>",
659 f->f_fp == NULL ? "closed" : "open",
660 PyString_AsString(name),
661 PyString_AsString(f->f_mode),
662 f);
663 Py_XDECREF(name);
664 return ret;
665 }
666}
667
668static PyObject *
669file_close(PyFileObject *f)
670{
671 PyObject *sts = close_the_file(f);
672 if (sts) {
673 PyMem_Free(f->f_setbuf);
674 f->f_setbuf = NULL;
675 }
676 return sts;
677}
678
679
680/* Our very own off_t-like type, 64-bit if possible */
681#if !defined(HAVE_LARGEFILE_SUPPORT)
682typedef off_t Py_off_t;
683#elif SIZEOF_OFF_T >= 8
684typedef off_t Py_off_t;
685#elif SIZEOF_FPOS_T >= 8
686typedef fpos_t Py_off_t;
687#else
688#error "Large file support, but neither off_t nor fpos_t is large enough."
689#endif
690
691
692/* a portable fseek() function
693 return 0 on success, non-zero on failure (with errno set) */
694static int
695_portable_fseek(FILE *fp, Py_off_t offset, int whence)
696{
697#if !defined(HAVE_LARGEFILE_SUPPORT)
698 return fseek(fp, offset, whence);
699#elif defined(HAVE_FSEEKO) && SIZEOF_OFF_T >= 8
700 return fseeko(fp, offset, whence);
701#elif defined(HAVE_FSEEK64)
702 return fseek64(fp, offset, whence);
703#elif defined(__BEOS__)
704 return _fseek(fp, offset, whence);
705#elif SIZEOF_FPOS_T >= 8
706 /* lacking a 64-bit capable fseek(), use a 64-bit capable fsetpos()
707 and fgetpos() to implement fseek()*/
708 fpos_t pos;
709 switch (whence) {
710 case SEEK_END:
711#ifdef MS_WINDOWS
712 fflush(fp);
713 if (_lseeki64(fileno(fp), 0, 2) == -1)
714 return -1;
715#else
716 if (fseek(fp, 0, SEEK_END) != 0)
717 return -1;
718#endif
719 /* fall through */
720 case SEEK_CUR:
721 if (fgetpos(fp, &pos) != 0)
722 return -1;
723 offset += pos;
724 break;
725 /* case SEEK_SET: break; */
726 }
727 return fsetpos(fp, &offset);
728#else
729#error "Large file support, but no way to fseek."
730#endif
731}
732
733
734/* a portable ftell() function
735 Return -1 on failure with errno set appropriately, current file
736 position on success */
737static Py_off_t
738_portable_ftell(FILE* fp)
739{
740#if !defined(HAVE_LARGEFILE_SUPPORT)
741 return ftell(fp);
742#elif defined(HAVE_FTELLO) && SIZEOF_OFF_T >= 8
743 return ftello(fp);
744#elif defined(HAVE_FTELL64)
745 return ftell64(fp);
746#elif SIZEOF_FPOS_T >= 8
747 fpos_t pos;
748 if (fgetpos(fp, &pos) != 0)
749 return -1;
750 return pos;
751#else
752#error "Large file support, but no way to ftell."
753#endif
754}
755
756
757static PyObject *
758file_seek(PyFileObject *f, PyObject *args)
759{
760 int whence;
761 int ret;
762 Py_off_t offset;
763 PyObject *offobj, *off_index;
764
765 if (f->f_fp == NULL)
766 return err_closed();
767 drop_readahead(f);
768 whence = 0;
769 if (!PyArg_ParseTuple(args, "O|i:seek", &offobj, &whence))
770 return NULL;
771 off_index = PyNumber_Index(offobj);
772 if (!off_index) {
773 if (!PyFloat_Check(offobj))
774 return NULL;
775 /* Deprecated in 2.6 */
776 PyErr_Clear();
777 if (PyErr_WarnEx(PyExc_DeprecationWarning,
778 "integer argument expected, got float",
779 1) < 0)
780 return NULL;
781 off_index = offobj;
782 Py_INCREF(offobj);
783 }
784#if !defined(HAVE_LARGEFILE_SUPPORT)
785 offset = PyInt_AsLong(off_index);
786#else
787 offset = PyLong_Check(off_index) ?
788 PyLong_AsLongLong(off_index) : PyInt_AsLong(off_index);
789#endif
790 Py_DECREF(off_index);
791 if (PyErr_Occurred())
792 return NULL;
793
794 FILE_BEGIN_ALLOW_THREADS(f)
795 errno = 0;
796 ret = _portable_fseek(f->f_fp, offset, whence);
797 FILE_END_ALLOW_THREADS(f)
798
799 if (ret != 0) {
800 PyErr_SetFromErrno(PyExc_IOError);
801 clearerr(f->f_fp);
802 return NULL;
803 }
804 f->f_skipnextlf = 0;
805 Py_INCREF(Py_None);
806 return Py_None;
807}
808
809
810#ifdef HAVE_FTRUNCATE
811static PyObject *
812file_truncate(PyFileObject *f, PyObject *args)
813{
814 Py_off_t newsize;
815 PyObject *newsizeobj = NULL;
816 Py_off_t initialpos;
817 int ret;
818
819 if (f->f_fp == NULL)
820 return err_closed();
821 if (!f->writable)
822 return err_mode("writing");
823 if (!PyArg_UnpackTuple(args, "truncate", 0, 1, &newsizeobj))
824 return NULL;
825
826 /* Get current file position. If the file happens to be open for
827 * update and the last operation was an input operation, C doesn't
828 * define what the later fflush() will do, but we promise truncate()
829 * won't change the current position (and fflush() *does* change it
830 * then at least on Windows). The easiest thing is to capture
831 * current pos now and seek back to it at the end.
832 */
833 FILE_BEGIN_ALLOW_THREADS(f)
834 errno = 0;
835 initialpos = _portable_ftell(f->f_fp);
836 FILE_END_ALLOW_THREADS(f)
837 if (initialpos == -1)
838 goto onioerror;
839
840 /* Set newsize to current postion if newsizeobj NULL, else to the
841 * specified value.
842 */
843 if (newsizeobj != NULL) {
844#if !defined(HAVE_LARGEFILE_SUPPORT)
845 newsize = PyInt_AsLong(newsizeobj);
846#else
847 newsize = PyLong_Check(newsizeobj) ?
848 PyLong_AsLongLong(newsizeobj) :
849 PyInt_AsLong(newsizeobj);
850#endif
851 if (PyErr_Occurred())
852 return NULL;
853 }
854 else /* default to current position */
855 newsize = initialpos;
856
857 /* Flush the stream. We're mixing stream-level I/O with lower-level
858 * I/O, and a flush may be necessary to synch both platform views
859 * of the current file state.
860 */
861 FILE_BEGIN_ALLOW_THREADS(f)
862 errno = 0;
863 ret = fflush(f->f_fp);
864 FILE_END_ALLOW_THREADS(f)
865 if (ret != 0)
866 goto onioerror;
867
868#ifdef MS_WINDOWS
869 /* MS _chsize doesn't work if newsize doesn't fit in 32 bits,
870 so don't even try using it. */
871 {
872 HANDLE hFile;
873
874 /* Have to move current pos to desired endpoint on Windows. */
875 FILE_BEGIN_ALLOW_THREADS(f)
876 errno = 0;
877 ret = _portable_fseek(f->f_fp, newsize, SEEK_SET) != 0;
878 FILE_END_ALLOW_THREADS(f)
879 if (ret)
880 goto onioerror;
881
882 /* Truncate. Note that this may grow the file! */
883 FILE_BEGIN_ALLOW_THREADS(f)
884 errno = 0;
885 hFile = (HANDLE)_get_osfhandle(fileno(f->f_fp));
886 ret = hFile == (HANDLE)-1;
887 if (ret == 0) {
888 ret = SetEndOfFile(hFile) == 0;
889 if (ret)
890 errno = EACCES;
891 }
892 FILE_END_ALLOW_THREADS(f)
893 if (ret)
894 goto onioerror;
895 }
896#else
897 FILE_BEGIN_ALLOW_THREADS(f)
898 errno = 0;
899 ret = ftruncate(fileno(f->f_fp), newsize);
900 FILE_END_ALLOW_THREADS(f)
901 if (ret != 0)
902 goto onioerror;
903#endif /* !MS_WINDOWS */
904
905 /* Restore original file position. */
906 FILE_BEGIN_ALLOW_THREADS(f)
907 errno = 0;
908 ret = _portable_fseek(f->f_fp, initialpos, SEEK_SET) != 0;
909 FILE_END_ALLOW_THREADS(f)
910 if (ret)
911 goto onioerror;
912
913 Py_INCREF(Py_None);
914 return Py_None;
915
916onioerror:
917 PyErr_SetFromErrno(PyExc_IOError);
918 clearerr(f->f_fp);
919 return NULL;
920}
921#endif /* HAVE_FTRUNCATE */
922
923static PyObject *
924file_tell(PyFileObject *f)
925{
926 Py_off_t pos;
927
928 if (f->f_fp == NULL)
929 return err_closed();
930 FILE_BEGIN_ALLOW_THREADS(f)
931 errno = 0;
932 pos = _portable_ftell(f->f_fp);
933 FILE_END_ALLOW_THREADS(f)
934
935 if (pos == -1) {
936 PyErr_SetFromErrno(PyExc_IOError);
937 clearerr(f->f_fp);
938 return NULL;
939 }
940 if (f->f_skipnextlf) {
941 int c;
942 c = GETC(f->f_fp);
943 if (c == '\n') {
944 f->f_newlinetypes |= NEWLINE_CRLF;
945 pos++;
946 f->f_skipnextlf = 0;
947 } else if (c != EOF) ungetc(c, f->f_fp);
948 }
949#if !defined(HAVE_LARGEFILE_SUPPORT)
950 return PyInt_FromLong(pos);
951#else
952 return PyLong_FromLongLong(pos);
953#endif
954}
955
956static PyObject *
957file_fileno(PyFileObject *f)
958{
959 if (f->f_fp == NULL)
960 return err_closed();
961 return PyInt_FromLong((long) fileno(f->f_fp));
962}
963
964static PyObject *
965file_flush(PyFileObject *f)
966{
967 int res;
968
969 if (f->f_fp == NULL)
970 return err_closed();
971 FILE_BEGIN_ALLOW_THREADS(f)
972 errno = 0;
973 res = fflush(f->f_fp);
974 FILE_END_ALLOW_THREADS(f)
975 if (res != 0) {
976 PyErr_SetFromErrno(PyExc_IOError);
977 clearerr(f->f_fp);
978 return NULL;
979 }
980 Py_INCREF(Py_None);
981 return Py_None;
982}
983
984static PyObject *
985file_isatty(PyFileObject *f)
986{
987 long res;
988 if (f->f_fp == NULL)
989 return err_closed();
990 FILE_BEGIN_ALLOW_THREADS(f)
991 res = isatty((int)fileno(f->f_fp));
992 FILE_END_ALLOW_THREADS(f)
993 return PyBool_FromLong(res);
994}
995
996
997#if BUFSIZ < 8192
998#define SMALLCHUNK 8192
999#else
1000#define SMALLCHUNK BUFSIZ
1001#endif
1002
1003static size_t
1004new_buffersize(PyFileObject *f, size_t currentsize)
1005{
1006#ifdef HAVE_FSTAT
1007 off_t pos, end;
1008 struct stat st;
1009 if (fstat(fileno(f->f_fp), &st) == 0) {
1010 end = st.st_size;
1011 /* The following is not a bug: we really need to call lseek()
1012 *and* ftell(). The reason is that some stdio libraries
1013 mistakenly flush their buffer when ftell() is called and
1014 the lseek() call it makes fails, thereby throwing away
1015 data that cannot be recovered in any way. To avoid this,
1016 we first test lseek(), and only call ftell() if lseek()
1017 works. We can't use the lseek() value either, because we
1018 need to take the amount of buffered data into account.
1019 (Yet another reason why stdio stinks. :-) */
1020 pos = lseek(fileno(f->f_fp), 0L, SEEK_CUR);
1021 if (pos >= 0) {
1022 pos = ftell(f->f_fp);
1023 }
1024 if (pos < 0)
1025 clearerr(f->f_fp);
1026 if (end > pos && pos >= 0)
1027 return currentsize + end - pos + 1;
1028 /* Add 1 so if the file were to grow we'd notice. */
1029 }
1030#endif
1031 /* Expand the buffer by an amount proportional to the current size,
1032 giving us amortized linear-time behavior. Use a less-than-double
1033 growth factor to avoid excessive allocation. */
1034 return currentsize + (currentsize >> 3) + 6;
1035}
1036
1037#if defined(EWOULDBLOCK) && defined(EAGAIN) && EWOULDBLOCK != EAGAIN
1038#define BLOCKED_ERRNO(x) ((x) == EWOULDBLOCK || (x) == EAGAIN)
1039#else
1040#ifdef EWOULDBLOCK
1041#define BLOCKED_ERRNO(x) ((x) == EWOULDBLOCK)
1042#else
1043#ifdef EAGAIN
1044#define BLOCKED_ERRNO(x) ((x) == EAGAIN)
1045#else
1046#define BLOCKED_ERRNO(x) 0
1047#endif
1048#endif
1049#endif
1050
1051static PyObject *
1052file_read(PyFileObject *f, PyObject *args)
1053{
1054 long bytesrequested = -1;
1055 size_t bytesread, buffersize, chunksize;
1056 PyObject *v;
1057
1058 if (f->f_fp == NULL)
1059 return err_closed();
1060 if (!f->readable)
1061 return err_mode("reading");
1062 /* refuse to mix with f.next() */
1063 if (f->f_buf != NULL &&
1064 (f->f_bufend - f->f_bufptr) > 0 &&
1065 f->f_buf[0] != '\0')
1066 return err_iterbuffered();
1067 if (!PyArg_ParseTuple(args, "|l:read", &bytesrequested))
1068 return NULL;
1069 if (bytesrequested < 0)
1070 buffersize = new_buffersize(f, (size_t)0);
1071 else
1072 buffersize = bytesrequested;
1073 if (buffersize > PY_SSIZE_T_MAX) {
1074 PyErr_SetString(PyExc_OverflowError,
1075 "requested number of bytes is more than a Python string can hold");
1076 return NULL;
1077 }
1078 v = PyString_FromStringAndSize((char *)NULL, buffersize);
1079 if (v == NULL)
1080 return NULL;
1081 bytesread = 0;
1082 for (;;) {
1083 int interrupted;
1084 FILE_BEGIN_ALLOW_THREADS(f)
1085 errno = 0;
1086 chunksize = Py_UniversalNewlineFread(BUF(v) + bytesread,
1087 buffersize - bytesread, f->f_fp, (PyObject *)f);
1088 interrupted = ferror(f->f_fp) && errno == EINTR;
1089 FILE_END_ALLOW_THREADS(f)
1090 if (interrupted) {
1091 clearerr(f->f_fp);
1092 if (PyErr_CheckSignals()) {
1093 Py_DECREF(v);
1094 return NULL;
1095 }
1096 }
1097 if (chunksize == 0) {
1098 if (interrupted)
1099 continue;
1100 if (!ferror(f->f_fp))
1101 break;
1102 clearerr(f->f_fp);
1103 /* When in non-blocking mode, data shouldn't
1104 * be discarded if a blocking signal was
1105 * received. That will also happen if
1106 * chunksize != 0, but bytesread < buffersize. */
1107 if (bytesread > 0 && BLOCKED_ERRNO(errno))
1108 break;
1109 PyErr_SetFromErrno(PyExc_IOError);
1110 Py_DECREF(v);
1111 return NULL;
1112 }
1113 bytesread += chunksize;
1114 if (bytesread < buffersize && !interrupted) {
1115 clearerr(f->f_fp);
1116 break;
1117 }
1118 if (bytesrequested < 0) {
1119 buffersize = new_buffersize(f, buffersize);
1120 if (_PyString_Resize(&v, buffersize) < 0)
1121 return NULL;
1122 } else {
1123 /* Got what was requested. */
1124 break;
1125 }
1126 }
1127 if (bytesread != buffersize && _PyString_Resize(&v, bytesread))
1128 return NULL;
1129 return v;
1130}
1131
1132static PyObject *
1133file_readinto(PyFileObject *f, PyObject *args)
1134{
1135 char *ptr;
1136 Py_ssize_t ntodo;
1137 Py_ssize_t ndone, nnow;
1138 Py_buffer pbuf;
1139
1140 if (f->f_fp == NULL)
1141 return err_closed();
1142 if (!f->readable)
1143 return err_mode("reading");
1144 /* refuse to mix with f.next() */
1145 if (f->f_buf != NULL &&
1146 (f->f_bufend - f->f_bufptr) > 0 &&
1147 f->f_buf[0] != '\0')
1148 return err_iterbuffered();
1149 if (!PyArg_ParseTuple(args, "w*", &pbuf))
1150 return NULL;
1151 ptr = pbuf.buf;
1152 ntodo = pbuf.len;
1153 ndone = 0;
1154 while (ntodo > 0) {
1155 int interrupted;
1156 FILE_BEGIN_ALLOW_THREADS(f)
1157 errno = 0;
1158 nnow = Py_UniversalNewlineFread(ptr+ndone, ntodo, f->f_fp,
1159 (PyObject *)f);
1160 interrupted = ferror(f->f_fp) && errno == EINTR;
1161 FILE_END_ALLOW_THREADS(f)
1162 if (interrupted) {
1163 clearerr(f->f_fp);
1164 if (PyErr_CheckSignals()) {
1165 PyBuffer_Release(&pbuf);
1166 return NULL;
1167 }
1168 }
1169 if (nnow == 0) {
1170 if (interrupted)
1171 continue;
1172 if (!ferror(f->f_fp))
1173 break;
1174 PyErr_SetFromErrno(PyExc_IOError);
1175 clearerr(f->f_fp);
1176 PyBuffer_Release(&pbuf);
1177 return NULL;
1178 }
1179 ndone += nnow;
1180 ntodo -= nnow;
1181 }
1182 PyBuffer_Release(&pbuf);
1183 return PyInt_FromSsize_t(ndone);
1184}
1185
1186/**************************************************************************
1187Routine to get next line using platform fgets().
1188
1189Under MSVC 6:
1190
1191+ MS threadsafe getc is very slow (multiple layers of function calls before+
1192 after each character, to lock+unlock the stream).
1193+ The stream-locking functions are MS-internal -- can't access them from user
1194 code.
1195+ There's nothing Tim could find in the MS C or platform SDK libraries that
1196 can worm around this.
1197+ MS fgets locks/unlocks only once per line; it's the only hook we have.
1198
1199So we use fgets for speed(!), despite that it's painful.
1200
1201MS realloc is also slow.
1202
1203Reports from other platforms on this method vs getc_unlocked (which MS doesn't
1204have):
1205 Linux a wash
1206 Solaris a wash
1207 Tru64 Unix getline_via_fgets significantly faster
1208
1209CAUTION: The C std isn't clear about this: in those cases where fgets
1210writes something into the buffer, can it write into any position beyond the
1211required trailing null byte? MSVC 6 fgets does not, and no platform is (yet)
1212known on which it does; and it would be a strange way to code fgets. Still,
1213getline_via_fgets may not work correctly if it does. The std test
1214test_bufio.py should fail if platform fgets() routinely writes beyond the
1215trailing null byte. #define DONT_USE_FGETS_IN_GETLINE to disable this code.
1216**************************************************************************/
1217
1218/* Use this routine if told to, or by default on non-get_unlocked()
1219 * platforms unless told not to. Yikes! Let's spell that out:
1220 * On a platform with getc_unlocked():
1221 * By default, use getc_unlocked().
1222 * If you want to use fgets() instead, #define USE_FGETS_IN_GETLINE.
1223 * On a platform without getc_unlocked():
1224 * By default, use fgets().
1225 * If you don't want to use fgets(), #define DONT_USE_FGETS_IN_GETLINE.
1226 */
1227#if !defined(USE_FGETS_IN_GETLINE) && !defined(HAVE_GETC_UNLOCKED)
1228#define USE_FGETS_IN_GETLINE
1229#endif
1230
1231#if defined(DONT_USE_FGETS_IN_GETLINE) && defined(USE_FGETS_IN_GETLINE)
1232#undef USE_FGETS_IN_GETLINE
1233#endif
1234
1235#ifdef USE_FGETS_IN_GETLINE
1236static PyObject*
1237getline_via_fgets(PyFileObject *f, FILE *fp)
1238{
1239/* INITBUFSIZE is the maximum line length that lets us get away with the fast
1240 * no-realloc, one-fgets()-call path. Boosting it isn't free, because we have
1241 * to fill this much of the buffer with a known value in order to figure out
1242 * how much of the buffer fgets() overwrites. So if INITBUFSIZE is larger
1243 * than "most" lines, we waste time filling unused buffer slots. 100 is
1244 * surely adequate for most peoples' email archives, chewing over source code,
1245 * etc -- "regular old text files".
1246 * MAXBUFSIZE is the maximum line length that lets us get away with the less
1247 * fast (but still zippy) no-realloc, two-fgets()-call path. See above for
1248 * cautions about boosting that. 300 was chosen because the worst real-life
1249 * text-crunching job reported on Python-Dev was a mail-log crawler where over
1250 * half the lines were 254 chars.
1251 */
1252#define INITBUFSIZE 100
1253#define MAXBUFSIZE 300
1254 char* p; /* temp */
1255 char buf[MAXBUFSIZE];
1256 PyObject* v; /* the string object result */
1257 char* pvfree; /* address of next free slot */
1258 char* pvend; /* address one beyond last free slot */
1259 size_t nfree; /* # of free buffer slots; pvend-pvfree */
1260 size_t total_v_size; /* total # of slots in buffer */
1261 size_t increment; /* amount to increment the buffer */
1262 size_t prev_v_size;
1263
1264 /* Optimize for normal case: avoid _PyString_Resize if at all
1265 * possible via first reading into stack buffer "buf".
1266 */
1267 total_v_size = INITBUFSIZE; /* start small and pray */
1268 pvfree = buf;
1269 for (;;) {
1270 FILE_BEGIN_ALLOW_THREADS(f)
1271 pvend = buf + total_v_size;
1272 nfree = pvend - pvfree;
1273 memset(pvfree, '\n', nfree);
1274 assert(nfree < INT_MAX); /* Should be atmost MAXBUFSIZE */
1275 p = fgets(pvfree, (int)nfree, fp);
1276 FILE_END_ALLOW_THREADS(f)
1277
1278 if (p == NULL) {
1279 clearerr(fp);
1280 if (PyErr_CheckSignals())
1281 return NULL;
1282 v = PyString_FromStringAndSize(buf, pvfree - buf);
1283 return v;
1284 }
1285 /* fgets read *something* */
1286 p = memchr(pvfree, '\n', nfree);
1287 if (p != NULL) {
1288 /* Did the \n come from fgets or from us?
1289 * Since fgets stops at the first \n, and then writes
1290 * \0, if it's from fgets a \0 must be next. But if
1291 * that's so, it could not have come from us, since
1292 * the \n's we filled the buffer with have only more
1293 * \n's to the right.
1294 */
1295 if (p+1 < pvend && *(p+1) == '\0') {
1296 /* It's from fgets: we win! In particular,
1297 * we haven't done any mallocs yet, and can
1298 * build the final result on the first try.
1299 */
1300 ++p; /* include \n from fgets */
1301 }
1302 else {
1303 /* Must be from us: fgets didn't fill the
1304 * buffer and didn't find a newline, so it
1305 * must be the last and newline-free line of
1306 * the file.
1307 */
1308 assert(p > pvfree && *(p-1) == '\0');
1309 --p; /* don't include \0 from fgets */
1310 }
1311 v = PyString_FromStringAndSize(buf, p - buf);
1312 return v;
1313 }
1314 /* yuck: fgets overwrote all the newlines, i.e. the entire
1315 * buffer. So this line isn't over yet, or maybe it is but
1316 * we're exactly at EOF. If we haven't already, try using the
1317 * rest of the stack buffer.
1318 */
1319 assert(*(pvend-1) == '\0');
1320 if (pvfree == buf) {
1321 pvfree = pvend - 1; /* overwrite trailing null */
1322 total_v_size = MAXBUFSIZE;
1323 }
1324 else
1325 break;
1326 }
1327
1328 /* The stack buffer isn't big enough; malloc a string object and read
1329 * into its buffer.
1330 */
1331 total_v_size = MAXBUFSIZE << 1;
1332 v = PyString_FromStringAndSize((char*)NULL, (int)total_v_size);
1333 if (v == NULL)
1334 return v;
1335 /* copy over everything except the last null byte */
1336 memcpy(BUF(v), buf, MAXBUFSIZE-1);
1337 pvfree = BUF(v) + MAXBUFSIZE - 1;
1338
1339 /* Keep reading stuff into v; if it ever ends successfully, break
1340 * after setting p one beyond the end of the line. The code here is
1341 * very much like the code above, except reads into v's buffer; see
1342 * the code above for detailed comments about the logic.
1343 */
1344 for (;;) {
1345 FILE_BEGIN_ALLOW_THREADS(f)
1346 pvend = BUF(v) + total_v_size;
1347 nfree = pvend - pvfree;
1348 memset(pvfree, '\n', nfree);
1349 assert(nfree < INT_MAX);
1350 p = fgets(pvfree, (int)nfree, fp);
1351 FILE_END_ALLOW_THREADS(f)
1352
1353 if (p == NULL) {
1354 clearerr(fp);
1355 if (PyErr_CheckSignals()) {
1356 Py_DECREF(v);
1357 return NULL;
1358 }
1359 p = pvfree;
1360 break;
1361 }
1362 p = memchr(pvfree, '\n', nfree);
1363 if (p != NULL) {
1364 if (p+1 < pvend && *(p+1) == '\0') {
1365 /* \n came from fgets */
1366 ++p;
1367 break;
1368 }
1369 /* \n came from us; last line of file, no newline */
1370 assert(p > pvfree && *(p-1) == '\0');
1371 --p;
1372 break;
1373 }
1374 /* expand buffer and try again */
1375 assert(*(pvend-1) == '\0');
1376 increment = total_v_size >> 2; /* mild exponential growth */
1377 prev_v_size = total_v_size;
1378 total_v_size += increment;
1379 /* check for overflow */
1380 if (total_v_size <= prev_v_size ||
1381 total_v_size > PY_SSIZE_T_MAX) {
1382 PyErr_SetString(PyExc_OverflowError,
1383 "line is longer than a Python string can hold");
1384 Py_DECREF(v);
1385 return NULL;
1386 }
1387 if (_PyString_Resize(&v, (int)total_v_size) < 0)
1388 return NULL;
1389 /* overwrite the trailing null byte */
1390 pvfree = BUF(v) + (prev_v_size - 1);
1391 }
1392 if (BUF(v) + total_v_size != p && _PyString_Resize(&v, p - BUF(v)))
1393 return NULL;
1394 return v;
1395#undef INITBUFSIZE
1396#undef MAXBUFSIZE
1397}
1398#endif /* ifdef USE_FGETS_IN_GETLINE */
1399
1400/* Internal routine to get a line.
1401 Size argument interpretation:
1402 > 0: max length;
1403 <= 0: read arbitrary line
1404*/
1405
1406static PyObject *
1407get_line(PyFileObject *f, int n)
1408{
1409 FILE *fp = f->f_fp;
1410 int c;
1411 char *buf, *end;
1412 size_t total_v_size; /* total # of slots in buffer */
1413 size_t used_v_size; /* # used slots in buffer */
1414 size_t increment; /* amount to increment the buffer */
1415 PyObject *v;
1416 int newlinetypes = f->f_newlinetypes;
1417 int skipnextlf = f->f_skipnextlf;
1418 int univ_newline = f->f_univ_newline;
1419
1420#if defined(USE_FGETS_IN_GETLINE)
1421 if (n <= 0 && !univ_newline )
1422 return getline_via_fgets(f, fp);
1423#endif
1424 total_v_size = n > 0 ? n : 100;
1425 v = PyString_FromStringAndSize((char *)NULL, total_v_size);
1426 if (v == NULL)
1427 return NULL;
1428 buf = BUF(v);
1429 end = buf + total_v_size;
1430
1431 for (;;) {
1432 FILE_BEGIN_ALLOW_THREADS(f)
1433 FLOCKFILE(fp);
1434 if (univ_newline) {
1435 c = 'x'; /* Shut up gcc warning */
1436 while ( buf != end && (c = GETC(fp)) != EOF ) {
1437 if (skipnextlf ) {
1438 skipnextlf = 0;
1439 if (c == '\n') {
1440 /* Seeing a \n here with
1441 * skipnextlf true means we
1442 * saw a \r before.
1443 */
1444 newlinetypes |= NEWLINE_CRLF;
1445 c = GETC(fp);
1446 if (c == EOF) break;
1447 } else {
1448 newlinetypes |= NEWLINE_CR;
1449 }
1450 }
1451 if (c == '\r') {
1452 skipnextlf = 1;
1453 c = '\n';
1454 } else if ( c == '\n')
1455 newlinetypes |= NEWLINE_LF;
1456 *buf++ = c;
1457 if (c == '\n') break;
1458 }
1459 if (c == EOF) {
1460 if (ferror(fp) && errno == EINTR) {
1461 FUNLOCKFILE(fp);
1462 FILE_ABORT_ALLOW_THREADS(f)
1463 f->f_newlinetypes = newlinetypes;
1464 f->f_skipnextlf = skipnextlf;
1465
1466 if (PyErr_CheckSignals()) {
1467 Py_DECREF(v);
1468 return NULL;
1469 }
1470 /* We executed Python signal handlers and got no exception.
1471 * Now back to reading the line where we left off. */
1472 clearerr(fp);
1473 continue;
1474 }
1475 if (skipnextlf)
1476 newlinetypes |= NEWLINE_CR;
1477 }
1478 } else /* If not universal newlines use the normal loop */
1479 while ((c = GETC(fp)) != EOF &&
1480 (*buf++ = c) != '\n' &&
1481 buf != end)
1482 ;
1483 FUNLOCKFILE(fp);
1484 FILE_END_ALLOW_THREADS(f)
1485 f->f_newlinetypes = newlinetypes;
1486 f->f_skipnextlf = skipnextlf;
1487 if (c == '\n')
1488 break;
1489 if (c == EOF) {
1490 if (ferror(fp)) {
1491 if (errno == EINTR) {
1492 if (PyErr_CheckSignals()) {
1493 Py_DECREF(v);
1494 return NULL;
1495 }
1496 /* We executed Python signal handlers and got no exception.
1497 * Now back to reading the line where we left off. */
1498 clearerr(fp);
1499 continue;
1500 }
1501 PyErr_SetFromErrno(PyExc_IOError);
1502 clearerr(fp);
1503 Py_DECREF(v);
1504 return NULL;
1505 }
1506 clearerr(fp);
1507 if (PyErr_CheckSignals()) {
1508 Py_DECREF(v);
1509 return NULL;
1510 }
1511 break;
1512 }
1513 /* Must be because buf == end */
1514 if (n > 0)
1515 break;
1516 used_v_size = total_v_size;
1517 increment = total_v_size >> 2; /* mild exponential growth */
1518 total_v_size += increment;
1519 if (total_v_size > PY_SSIZE_T_MAX) {
1520 PyErr_SetString(PyExc_OverflowError,
1521 "line is longer than a Python string can hold");
1522 Py_DECREF(v);
1523 return NULL;
1524 }
1525 if (_PyString_Resize(&v, total_v_size) < 0)
1526 return NULL;
1527 buf = BUF(v) + used_v_size;
1528 end = BUF(v) + total_v_size;
1529 }
1530
1531 used_v_size = buf - BUF(v);
1532 if (used_v_size != total_v_size && _PyString_Resize(&v, used_v_size))
1533 return NULL;
1534 return v;
1535}
1536
1537/* External C interface */
1538
1539PyObject *
1540PyFile_GetLine(PyObject *f, int n)
1541{
1542 PyObject *result;
1543
1544 if (f == NULL) {
1545 PyErr_BadInternalCall();
1546 return NULL;
1547 }
1548
1549 if (PyFile_Check(f)) {
1550 PyFileObject *fo = (PyFileObject *)f;
1551 if (fo->f_fp == NULL)
1552 return err_closed();
1553 if (!fo->readable)
1554 return err_mode("reading");
1555 /* refuse to mix with f.next() */
1556 if (fo->f_buf != NULL &&
1557 (fo->f_bufend - fo->f_bufptr) > 0 &&
1558 fo->f_buf[0] != '\0')
1559 return err_iterbuffered();
1560 result = get_line(fo, n);
1561 }
1562 else {
1563 PyObject *reader;
1564 PyObject *args;
1565
1566 reader = PyObject_GetAttrString(f, "readline");
1567 if (reader == NULL)
1568 return NULL;
1569 if (n <= 0)
1570 args = PyTuple_New(0);
1571 else
1572 args = Py_BuildValue("(i)", n);
1573 if (args == NULL) {
1574 Py_DECREF(reader);
1575 return NULL;
1576 }
1577 result = PyEval_CallObject(reader, args);
1578 Py_DECREF(reader);
1579 Py_DECREF(args);
1580 if (result != NULL && !PyString_Check(result) &&
1581 !PyUnicode_Check(result)) {
1582 Py_DECREF(result);
1583 result = NULL;
1584 PyErr_SetString(PyExc_TypeError,
1585 "object.readline() returned non-string");
1586 }
1587 }
1588
1589 if (n < 0 && result != NULL && PyString_Check(result)) {
1590 char *s = PyString_AS_STRING(result);
1591 Py_ssize_t len = PyString_GET_SIZE(result);
1592 if (len == 0) {
1593 Py_DECREF(result);
1594 result = NULL;
1595 PyErr_SetString(PyExc_EOFError,
1596 "EOF when reading a line");
1597 }
1598 else if (s[len-1] == '\n') {
1599 if (result->ob_refcnt == 1) {
1600 if (_PyString_Resize(&result, len-1))
1601 return NULL;
1602 }
1603 else {
1604 PyObject *v;
1605 v = PyString_FromStringAndSize(s, len-1);
1606 Py_DECREF(result);
1607 result = v;
1608 }
1609 }
1610 }
1611#ifdef Py_USING_UNICODE
1612 if (n < 0 && result != NULL && PyUnicode_Check(result)) {
1613 Py_UNICODE *s = PyUnicode_AS_UNICODE(result);
1614 Py_ssize_t len = PyUnicode_GET_SIZE(result);
1615 if (len == 0) {
1616 Py_DECREF(result);
1617 result = NULL;
1618 PyErr_SetString(PyExc_EOFError,
1619 "EOF when reading a line");
1620 }
1621 else if (s[len-1] == '\n') {
1622 if (result->ob_refcnt == 1)
1623 PyUnicode_Resize(&result, len-1);
1624 else {
1625 PyObject *v;
1626 v = PyUnicode_FromUnicode(s, len-1);
1627 Py_DECREF(result);
1628 result = v;
1629 }
1630 }
1631 }
1632#endif
1633 return result;
1634}
1635
1636/* Python method */
1637
1638static PyObject *
1639file_readline(PyFileObject *f, PyObject *args)
1640{
1641 int n = -1;
1642
1643 if (f->f_fp == NULL)
1644 return err_closed();
1645 if (!f->readable)
1646 return err_mode("reading");
1647 /* refuse to mix with f.next() */
1648 if (f->f_buf != NULL &&
1649 (f->f_bufend - f->f_bufptr) > 0 &&
1650 f->f_buf[0] != '\0')
1651 return err_iterbuffered();
1652 if (!PyArg_ParseTuple(args, "|i:readline", &n))
1653 return NULL;
1654 if (n == 0)
1655 return PyString_FromString("");
1656 if (n < 0)
1657 n = 0;
1658 return get_line(f, n);
1659}
1660
1661static PyObject *
1662file_readlines(PyFileObject *f, PyObject *args)
1663{
1664 long sizehint = 0;
1665 PyObject *list = NULL;
1666 PyObject *line;
1667 char small_buffer[SMALLCHUNK];
1668 char *buffer = small_buffer;
1669 size_t buffersize = SMALLCHUNK;
1670 PyObject *big_buffer = NULL;
1671 size_t nfilled = 0;
1672 size_t nread;
1673 size_t totalread = 0;
1674 char *p, *q, *end;
1675 int err;
1676 int shortread = 0; /* bool, did the previous read come up short? */
1677
1678 if (f->f_fp == NULL)
1679 return err_closed();
1680 if (!f->readable)
1681 return err_mode("reading");
1682 /* refuse to mix with f.next() */
1683 if (f->f_buf != NULL &&
1684 (f->f_bufend - f->f_bufptr) > 0 &&
1685 f->f_buf[0] != '\0')
1686 return err_iterbuffered();
1687 if (!PyArg_ParseTuple(args, "|l:readlines", &sizehint))
1688 return NULL;
1689 if ((list = PyList_New(0)) == NULL)
1690 return NULL;
1691 for (;;) {
1692 if (shortread)
1693 nread = 0;
1694 else {
1695 FILE_BEGIN_ALLOW_THREADS(f)
1696 errno = 0;
1697 nread = Py_UniversalNewlineFread(buffer+nfilled,
1698 buffersize-nfilled, f->f_fp, (PyObject *)f);
1699 FILE_END_ALLOW_THREADS(f)
1700 shortread = (nread < buffersize-nfilled);
1701 }
1702 if (nread == 0) {
1703 sizehint = 0;
1704 if (!ferror(f->f_fp))
1705 break;
1706 if (errno == EINTR) {
1707 if (PyErr_CheckSignals()) {
1708 goto error;
1709 }
1710 clearerr(f->f_fp);
1711 shortread = 0;
1712 continue;
1713 }
1714 PyErr_SetFromErrno(PyExc_IOError);
1715 clearerr(f->f_fp);
1716 goto error;
1717 }
1718 totalread += nread;
1719 p = (char *)memchr(buffer+nfilled, '\n', nread);
1720 if (p == NULL) {
1721 /* Need a larger buffer to fit this line */
1722 nfilled += nread;
1723 buffersize *= 2;
1724 if (buffersize > PY_SSIZE_T_MAX) {
1725 PyErr_SetString(PyExc_OverflowError,
1726 "line is longer than a Python string can hold");
1727 goto error;
1728 }
1729 if (big_buffer == NULL) {
1730 /* Create the big buffer */
1731 big_buffer = PyString_FromStringAndSize(
1732 NULL, buffersize);
1733 if (big_buffer == NULL)
1734 goto error;
1735 buffer = PyString_AS_STRING(big_buffer);
1736 memcpy(buffer, small_buffer, nfilled);
1737 }
1738 else {
1739 /* Grow the big buffer */
1740 if ( _PyString_Resize(&big_buffer, buffersize) < 0 )
1741 goto error;
1742 buffer = PyString_AS_STRING(big_buffer);
1743 }
1744 continue;
1745 }
1746 end = buffer+nfilled+nread;
1747 q = buffer;
1748 do {
1749 /* Process complete lines */
1750 p++;
1751 line = PyString_FromStringAndSize(q, p-q);
1752 if (line == NULL)
1753 goto error;
1754 err = PyList_Append(list, line);
1755 Py_DECREF(line);
1756 if (err != 0)
1757 goto error;
1758 q = p;
1759 p = (char *)memchr(q, '\n', end-q);
1760 } while (p != NULL);
1761 /* Move the remaining incomplete line to the start */
1762 nfilled = end-q;
1763 memmove(buffer, q, nfilled);
1764 if (sizehint > 0)
1765 if (totalread >= (size_t)sizehint)
1766 break;
1767 }
1768 if (nfilled != 0) {
1769 /* Partial last line */
1770 line = PyString_FromStringAndSize(buffer, nfilled);
1771 if (line == NULL)
1772 goto error;
1773 if (sizehint > 0) {
1774 /* Need to complete the last line */
1775 PyObject *rest = get_line(f, 0);
1776 if (rest == NULL) {
1777 Py_DECREF(line);
1778 goto error;
1779 }
1780 PyString_Concat(&line, rest);
1781 Py_DECREF(rest);
1782 if (line == NULL)
1783 goto error;
1784 }
1785 err = PyList_Append(list, line);
1786 Py_DECREF(line);
1787 if (err != 0)
1788 goto error;
1789 }
1790
1791cleanup:
1792 Py_XDECREF(big_buffer);
1793 return list;
1794
1795error:
1796 Py_CLEAR(list);
1797 goto cleanup;
1798}
1799
1800static PyObject *
1801file_write(PyFileObject *f, PyObject *args)
1802{
1803 Py_buffer pbuf;
1804 const char *s;
1805 Py_ssize_t n, n2;
1806 PyObject *encoded = NULL;
1807
1808 if (f->f_fp == NULL)
1809 return err_closed();
1810 if (!f->writable)
1811 return err_mode("writing");
1812 if (f->f_binary) {
1813 if (!PyArg_ParseTuple(args, "s*", &pbuf))
1814 return NULL;
1815 s = pbuf.buf;
1816 n = pbuf.len;
1817 }
1818 else {
1819 PyObject *text;
1820 if (!PyArg_ParseTuple(args, "O", &text))
1821 return NULL;
1822
1823 if (PyString_Check(text)) {
1824 s = PyString_AS_STRING(text);
1825 n = PyString_GET_SIZE(text);
1826#ifdef Py_USING_UNICODE
1827 } else if (PyUnicode_Check(text)) {
1828 const char *encoding, *errors;
1829 if (f->f_encoding != Py_None)
1830 encoding = PyString_AS_STRING(f->f_encoding);
1831 else
1832 encoding = PyUnicode_GetDefaultEncoding();
1833 if (f->f_errors != Py_None)
1834 errors = PyString_AS_STRING(f->f_errors);
1835 else
1836 errors = "strict";
1837 encoded = PyUnicode_AsEncodedString(text, encoding, errors);
1838 if (encoded == NULL)
1839 return NULL;
1840 s = PyString_AS_STRING(encoded);
1841 n = PyString_GET_SIZE(encoded);
1842#endif
1843 } else {
1844 if (PyObject_AsCharBuffer(text, &s, &n))
1845 return NULL;
1846 }
1847 }
1848 f->f_softspace = 0;
1849 FILE_BEGIN_ALLOW_THREADS(f)
1850 errno = 0;
1851 n2 = fwrite(s, 1, n, f->f_fp);
1852 FILE_END_ALLOW_THREADS(f)
1853 Py_XDECREF(encoded);
1854 if (f->f_binary)
1855 PyBuffer_Release(&pbuf);
1856 if (n2 != n) {
1857 PyErr_SetFromErrno(PyExc_IOError);
1858 clearerr(f->f_fp);
1859 return NULL;
1860 }
1861 Py_INCREF(Py_None);
1862 return Py_None;
1863}
1864
1865static PyObject *
1866file_writelines(PyFileObject *f, PyObject *seq)
1867{
1868#define CHUNKSIZE 1000
1869 PyObject *list, *line;
1870 PyObject *it; /* iter(seq) */
1871 PyObject *result;
1872 int index, islist;
1873 Py_ssize_t i, j, nwritten, len;
1874
1875 assert(seq != NULL);
1876 if (f->f_fp == NULL)
1877 return err_closed();
1878 if (!f->writable)
1879 return err_mode("writing");
1880
1881 result = NULL;
1882 list = NULL;
1883 islist = PyList_Check(seq);
1884 if (islist)
1885 it = NULL;
1886 else {
1887 it = PyObject_GetIter(seq);
1888 if (it == NULL) {
1889 PyErr_SetString(PyExc_TypeError,
1890 "writelines() requires an iterable argument");
1891 return NULL;
1892 }
1893 /* From here on, fail by going to error, to reclaim "it". */
1894 list = PyList_New(CHUNKSIZE);
1895 if (list == NULL)
1896 goto error;
1897 }
1898
1899 /* Strategy: slurp CHUNKSIZE lines into a private list,
1900 checking that they are all strings, then write that list
1901 without holding the interpreter lock, then come back for more. */
1902 for (index = 0; ; index += CHUNKSIZE) {
1903 if (islist) {
1904 Py_XDECREF(list);
1905 list = PyList_GetSlice(seq, index, index+CHUNKSIZE);
1906 if (list == NULL)
1907 goto error;
1908 j = PyList_GET_SIZE(list);
1909 }
1910 else {
1911 for (j = 0; j < CHUNKSIZE; j++) {
1912 line = PyIter_Next(it);
1913 if (line == NULL) {
1914 if (PyErr_Occurred())
1915 goto error;
1916 break;
1917 }
1918 PyList_SetItem(list, j, line);
1919 }
1920 /* The iterator might have closed the file on us. */
1921 if (f->f_fp == NULL) {
1922 err_closed();
1923 goto error;
1924 }
1925 }
1926 if (j == 0)
1927 break;
1928
1929 /* Check that all entries are indeed strings. If not,
1930 apply the same rules as for file.write() and
1931 convert the results to strings. This is slow, but
1932 seems to be the only way since all conversion APIs
1933 could potentially execute Python code. */
1934 for (i = 0; i < j; i++) {
1935 PyObject *v = PyList_GET_ITEM(list, i);
1936 if (!PyString_Check(v)) {
1937 const char *buffer;
1938 if (((f->f_binary &&
1939 PyObject_AsReadBuffer(v,
1940 (const void**)&buffer,
1941 &len)) ||
1942 PyObject_AsCharBuffer(v,
1943 &buffer,
1944 &len))) {
1945 PyErr_SetString(PyExc_TypeError,
1946 "writelines() argument must be a sequence of strings");
1947 goto error;
1948 }
1949 line = PyString_FromStringAndSize(buffer,
1950 len);
1951 if (line == NULL)
1952 goto error;
1953 Py_DECREF(v);
1954 PyList_SET_ITEM(list, i, line);
1955 }
1956 }
1957
1958 /* Since we are releasing the global lock, the
1959 following code may *not* execute Python code. */
1960 f->f_softspace = 0;
1961 FILE_BEGIN_ALLOW_THREADS(f)
1962 errno = 0;
1963 for (i = 0; i < j; i++) {
1964 line = PyList_GET_ITEM(list, i);
1965 len = PyString_GET_SIZE(line);
1966 nwritten = fwrite(PyString_AS_STRING(line),
1967 1, len, f->f_fp);
1968 if (nwritten != len) {
1969 FILE_ABORT_ALLOW_THREADS(f)
1970 PyErr_SetFromErrno(PyExc_IOError);
1971 clearerr(f->f_fp);
1972 goto error;
1973 }
1974 }
1975 FILE_END_ALLOW_THREADS(f)
1976
1977 if (j < CHUNKSIZE)
1978 break;
1979 }
1980
1981 Py_INCREF(Py_None);
1982 result = Py_None;
1983 error:
1984 Py_XDECREF(list);
1985 Py_XDECREF(it);
1986 return result;
1987#undef CHUNKSIZE
1988}
1989
1990static PyObject *
1991file_self(PyFileObject *f)
1992{
1993 if (f->f_fp == NULL)
1994 return err_closed();
1995 Py_INCREF(f);
1996 return (PyObject *)f;
1997}
1998
1999static PyObject *
2000file_xreadlines(PyFileObject *f)
2001{
2002 if (PyErr_WarnPy3k("f.xreadlines() not supported in 3.x, "
2003 "try 'for line in f' instead", 1) < 0)
2004 return NULL;
2005 return file_self(f);
2006}
2007
2008static PyObject *
2009file_exit(PyObject *f, PyObject *args)
2010{
2011 PyObject *ret = PyObject_CallMethod(f, "close", NULL);
2012 if (!ret)
2013 /* If error occurred, pass through */
2014 return NULL;
2015 Py_DECREF(ret);
2016 /* We cannot return the result of close since a true
2017 * value will be interpreted as "yes, swallow the
2018 * exception if one was raised inside the with block". */
2019 Py_RETURN_NONE;
2020}
2021
2022PyDoc_STRVAR(readline_doc,
2023"readline([size]) -> next line from the file, as a string.\n"
2024"\n"
2025"Retain newline. A non-negative size argument limits the maximum\n"
2026"number of bytes to return (an incomplete line may be returned then).\n"
2027"Return an empty string at EOF.");
2028
2029PyDoc_STRVAR(read_doc,
2030"read([size]) -> read at most size bytes, returned as a string.\n"
2031"\n"
2032"If the size argument is negative or omitted, read until EOF is reached.\n"
2033"Notice that when in non-blocking mode, less data than what was requested\n"
2034"may be returned, even if no size parameter was given.");
2035
2036PyDoc_STRVAR(write_doc,
2037"write(str) -> None. Write string str to file.\n"
2038"\n"
2039"Note that due to buffering, flush() or close() may be needed before\n"
2040"the file on disk reflects the data written.");
2041
2042PyDoc_STRVAR(fileno_doc,
2043"fileno() -> integer \"file descriptor\".\n"
2044"\n"
2045"This is needed for lower-level file interfaces, such os.read().");
2046
2047PyDoc_STRVAR(seek_doc,
2048"seek(offset[, whence]) -> None. Move to new file position.\n"
2049"\n"
2050"Argument offset is a byte count. Optional argument whence defaults to\n"
2051"0 (offset from start of file, offset should be >= 0); other values are 1\n"
2052"(move relative to current position, positive or negative), and 2 (move\n"
2053"relative to end of file, usually negative, although many platforms allow\n"
2054"seeking beyond the end of a file). If the file is opened in text mode,\n"
2055"only offsets returned by tell() are legal. Use of other offsets causes\n"
2056"undefined behavior."
2057"\n"
2058"Note that not all file objects are seekable.");
2059
2060#ifdef HAVE_FTRUNCATE
2061PyDoc_STRVAR(truncate_doc,
2062"truncate([size]) -> None. Truncate the file to at most size bytes.\n"
2063"\n"
2064"Size defaults to the current file position, as returned by tell().");
2065#endif
2066
2067PyDoc_STRVAR(tell_doc,
2068"tell() -> current file position, an integer (may be a long integer).");
2069
2070PyDoc_STRVAR(readinto_doc,
2071"readinto() -> Undocumented. Don't use this; it may go away.");
2072
2073PyDoc_STRVAR(readlines_doc,
2074"readlines([size]) -> list of strings, each a line from the file.\n"
2075"\n"
2076"Call readline() repeatedly and return a list of the lines so read.\n"
2077"The optional size argument, if given, is an approximate bound on the\n"
2078"total number of bytes in the lines returned.");
2079
2080PyDoc_STRVAR(xreadlines_doc,
2081"xreadlines() -> returns self.\n"
2082"\n"
2083"For backward compatibility. File objects now include the performance\n"
2084"optimizations previously implemented in the xreadlines module.");
2085
2086PyDoc_STRVAR(writelines_doc,
2087"writelines(sequence_of_strings) -> None. Write the strings to the file.\n"
2088"\n"
2089"Note that newlines are not added. The sequence can be any iterable object\n"
2090"producing strings. This is equivalent to calling write() for each string.");
2091
2092PyDoc_STRVAR(flush_doc,
2093"flush() -> None. Flush the internal I/O buffer.");
2094
2095PyDoc_STRVAR(close_doc,
2096"close() -> None or (perhaps) an integer. Close the file.\n"
2097"\n"
2098"Sets data attribute .closed to True. A closed file cannot be used for\n"
2099"further I/O operations. close() may be called more than once without\n"
2100"error. Some kinds of file objects (for example, opened by popen())\n"
2101"may return an exit status upon closing.");
2102
2103PyDoc_STRVAR(isatty_doc,
2104"isatty() -> true or false. True if the file is connected to a tty device.");
2105
2106PyDoc_STRVAR(enter_doc,
2107 "__enter__() -> self.");
2108
2109PyDoc_STRVAR(exit_doc,
2110 "__exit__(*excinfo) -> None. Closes the file.");
2111
2112static PyMethodDef file_methods[] = {
2113 {"readline", (PyCFunction)file_readline, METH_VARARGS, readline_doc},
2114 {"read", (PyCFunction)file_read, METH_VARARGS, read_doc},
2115 {"write", (PyCFunction)file_write, METH_VARARGS, write_doc},
2116 {"fileno", (PyCFunction)file_fileno, METH_NOARGS, fileno_doc},
2117 {"seek", (PyCFunction)file_seek, METH_VARARGS, seek_doc},
2118#ifdef HAVE_FTRUNCATE
2119 {"truncate", (PyCFunction)file_truncate, METH_VARARGS, truncate_doc},
2120#endif
2121 {"tell", (PyCFunction)file_tell, METH_NOARGS, tell_doc},
2122 {"readinto", (PyCFunction)file_readinto, METH_VARARGS, readinto_doc},
2123 {"readlines", (PyCFunction)file_readlines, METH_VARARGS, readlines_doc},
2124 {"xreadlines",(PyCFunction)file_xreadlines, METH_NOARGS, xreadlines_doc},
2125 {"writelines",(PyCFunction)file_writelines, METH_O, writelines_doc},
2126 {"flush", (PyCFunction)file_flush, METH_NOARGS, flush_doc},
2127 {"close", (PyCFunction)file_close, METH_NOARGS, close_doc},
2128 {"isatty", (PyCFunction)file_isatty, METH_NOARGS, isatty_doc},
2129 {"__enter__", (PyCFunction)file_self, METH_NOARGS, enter_doc},
2130 {"__exit__", (PyCFunction)file_exit, METH_VARARGS, exit_doc},
2131 {NULL, NULL} /* sentinel */
2132};
2133
2134#define OFF(x) offsetof(PyFileObject, x)
2135
2136static PyMemberDef file_memberlist[] = {
2137 {"mode", T_OBJECT, OFF(f_mode), RO,
2138 "file mode ('r', 'U', 'w', 'a', possibly with 'b' or '+' added)"},
2139 {"name", T_OBJECT, OFF(f_name), RO,
2140 "file name"},
2141 {"encoding", T_OBJECT, OFF(f_encoding), RO,
2142 "file encoding"},
2143 {"errors", T_OBJECT, OFF(f_errors), RO,
2144 "Unicode error handler"},
2145 /* getattr(f, "closed") is implemented without this table */
2146 {NULL} /* Sentinel */
2147};
2148
2149static PyObject *
2150get_closed(PyFileObject *f, void *closure)
2151{
2152 return PyBool_FromLong((long)(f->f_fp == 0));
2153}
2154static PyObject *
2155get_newlines(PyFileObject *f, void *closure)
2156{
2157 switch (f->f_newlinetypes) {
2158 case NEWLINE_UNKNOWN:
2159 Py_INCREF(Py_None);
2160 return Py_None;
2161 case NEWLINE_CR:
2162 return PyString_FromString("\r");
2163 case NEWLINE_LF:
2164 return PyString_FromString("\n");
2165 case NEWLINE_CR|NEWLINE_LF:
2166 return Py_BuildValue("(ss)", "\r", "\n");
2167 case NEWLINE_CRLF:
2168 return PyString_FromString("\r\n");
2169 case NEWLINE_CR|NEWLINE_CRLF:
2170 return Py_BuildValue("(ss)", "\r", "\r\n");
2171 case NEWLINE_LF|NEWLINE_CRLF:
2172 return Py_BuildValue("(ss)", "\n", "\r\n");
2173 case NEWLINE_CR|NEWLINE_LF|NEWLINE_CRLF:
2174 return Py_BuildValue("(sss)", "\r", "\n", "\r\n");
2175 default:
2176 PyErr_Format(PyExc_SystemError,
2177 "Unknown newlines value 0x%x\n",
2178 f->f_newlinetypes);
2179 return NULL;
2180 }
2181}
2182
2183static PyObject *
2184get_softspace(PyFileObject *f, void *closure)
2185{
2186 if (PyErr_WarnPy3k("file.softspace not supported in 3.x", 1) < 0)
2187 return NULL;
2188 return PyInt_FromLong(f->f_softspace);
2189}
2190
2191static int
2192set_softspace(PyFileObject *f, PyObject *value)
2193{
2194 int new;
2195 if (PyErr_WarnPy3k("file.softspace not supported in 3.x", 1) < 0)
2196 return -1;
2197
2198 if (value == NULL) {
2199 PyErr_SetString(PyExc_TypeError,
2200 "can't delete softspace attribute");
2201 return -1;
2202 }
2203
2204 new = PyInt_AsLong(value);
2205 if (new == -1 && PyErr_Occurred())
2206 return -1;
2207 f->f_softspace = new;
2208 return 0;
2209}
2210
2211static PyGetSetDef file_getsetlist[] = {
2212 {"closed", (getter)get_closed, NULL, "True if the file is closed"},
2213 {"newlines", (getter)get_newlines, NULL,
2214 "end-of-line convention used in this file"},
2215 {"softspace", (getter)get_softspace, (setter)set_softspace,
2216 "flag indicating that a space needs to be printed; used by print"},
2217 {0},
2218};
2219
2220static void
2221drop_readahead(PyFileObject *f)
2222{
2223 if (f->f_buf != NULL) {
2224 PyMem_Free(f->f_buf);
2225 f->f_buf = NULL;
2226 }
2227}
2228
2229/* Make sure that file has a readahead buffer with at least one byte
2230 (unless at EOF) and no more than bufsize. Returns negative value on
2231 error, will set MemoryError if bufsize bytes cannot be allocated. */
2232static int
2233readahead(PyFileObject *f, int bufsize)
2234{
2235 Py_ssize_t chunksize;
2236
2237 if (f->f_buf != NULL) {
2238 if( (f->f_bufend - f->f_bufptr) >= 1)
2239 return 0;
2240 else
2241 drop_readahead(f);
2242 }
2243 if ((f->f_buf = (char *)PyMem_Malloc(bufsize)) == NULL) {
2244 PyErr_NoMemory();
2245 return -1;
2246 }
2247 FILE_BEGIN_ALLOW_THREADS(f)
2248 errno = 0;
2249 chunksize = Py_UniversalNewlineFread(
2250 f->f_buf, bufsize, f->f_fp, (PyObject *)f);
2251 FILE_END_ALLOW_THREADS(f)
2252 if (chunksize == 0) {
2253 if (ferror(f->f_fp)) {
2254 PyErr_SetFromErrno(PyExc_IOError);
2255 clearerr(f->f_fp);
2256 drop_readahead(f);
2257 return -1;
2258 }
2259 }
2260 f->f_bufptr = f->f_buf;
2261 f->f_bufend = f->f_buf + chunksize;
2262 return 0;
2263}
2264
2265/* Used by file_iternext. The returned string will start with 'skip'
2266 uninitialized bytes followed by the remainder of the line. Don't be
2267 horrified by the recursive call: maximum recursion depth is limited by
2268 logarithmic buffer growth to about 50 even when reading a 1gb line. */
2269
2270static PyStringObject *
2271readahead_get_line_skip(PyFileObject *f, int skip, int bufsize)
2272{
2273 PyStringObject* s;
2274 char *bufptr;
2275 char *buf;
2276 Py_ssize_t len;
2277
2278 if (f->f_buf == NULL)
2279 if (readahead(f, bufsize) < 0)
2280 return NULL;
2281
2282 len = f->f_bufend - f->f_bufptr;
2283 if (len == 0)
2284 return (PyStringObject *)
2285 PyString_FromStringAndSize(NULL, skip);
2286 bufptr = (char *)memchr(f->f_bufptr, '\n', len);
2287 if (bufptr != NULL) {
2288 bufptr++; /* Count the '\n' */
2289 len = bufptr - f->f_bufptr;
2290 s = (PyStringObject *)
2291 PyString_FromStringAndSize(NULL, skip+len);
2292 if (s == NULL)
2293 return NULL;
2294 memcpy(PyString_AS_STRING(s)+skip, f->f_bufptr, len);
2295 f->f_bufptr = bufptr;
2296 if (bufptr == f->f_bufend)
2297 drop_readahead(f);
2298 } else {
2299 bufptr = f->f_bufptr;
2300 buf = f->f_buf;
2301 f->f_buf = NULL; /* Force new readahead buffer */
2302 assert(skip+len < INT_MAX);
2303 s = readahead_get_line_skip(
2304 f, (int)(skip+len), bufsize + (bufsize>>2) );
2305 if (s == NULL) {
2306 PyMem_Free(buf);
2307 return NULL;
2308 }
2309 memcpy(PyString_AS_STRING(s)+skip, bufptr, len);
2310 PyMem_Free(buf);
2311 }
2312 return s;
2313}
2314
2315/* A larger buffer size may actually decrease performance. */
2316#define READAHEAD_BUFSIZE 8192
2317
2318static PyObject *
2319file_iternext(PyFileObject *f)
2320{
2321 PyStringObject* l;
2322
2323 if (f->f_fp == NULL)
2324 return err_closed();
2325 if (!f->readable)
2326 return err_mode("reading");
2327
2328 l = readahead_get_line_skip(f, 0, READAHEAD_BUFSIZE);
2329 if (l == NULL || PyString_GET_SIZE(l) == 0) {
2330 Py_XDECREF(l);
2331 return NULL;
2332 }
2333 return (PyObject *)l;
2334}
2335
2336
2337static PyObject *
2338file_new(PyTypeObject *type, PyObject *args, PyObject *kwds)
2339{
2340 PyObject *self;
2341 static PyObject *not_yet_string;
2342
2343 assert(type != NULL && type->tp_alloc != NULL);
2344
2345 if (not_yet_string == NULL) {
2346 not_yet_string = PyString_InternFromString("<uninitialized file>");
2347 if (not_yet_string == NULL)
2348 return NULL;
2349 }
2350
2351 self = type->tp_alloc(type, 0);
2352 if (self != NULL) {
2353 /* Always fill in the name and mode, so that nobody else
2354 needs to special-case NULLs there. */
2355 Py_INCREF(not_yet_string);
2356 ((PyFileObject *)self)->f_name = not_yet_string;
2357 Py_INCREF(not_yet_string);
2358 ((PyFileObject *)self)->f_mode = not_yet_string;
2359 Py_INCREF(Py_None);
2360 ((PyFileObject *)self)->f_encoding = Py_None;
2361 Py_INCREF(Py_None);
2362 ((PyFileObject *)self)->f_errors = Py_None;
2363 ((PyFileObject *)self)->weakreflist = NULL;
2364 ((PyFileObject *)self)->unlocked_count = 0;
2365 }
2366 return self;
2367}
2368
2369static int
2370file_init(PyObject *self, PyObject *args, PyObject *kwds)
2371{
2372 PyFileObject *foself = (PyFileObject *)self;
2373 int ret = 0;
2374 static char *kwlist[] = {"name", "mode", "buffering", 0};
2375 char *name = NULL;
2376 char *mode = "r";
2377 int bufsize = -1;
2378 int wideargument = 0;
2379#ifdef MS_WINDOWS
2380 PyObject *po;
2381#endif
2382
2383 assert(PyFile_Check(self));
2384 if (foself->f_fp != NULL) {
2385 /* Have to close the existing file first. */
2386 PyObject *closeresult = file_close(foself);
2387 if (closeresult == NULL)
2388 return -1;
2389 Py_DECREF(closeresult);
2390 }
2391
2392#ifdef MS_WINDOWS
2393 if (PyArg_ParseTupleAndKeywords(args, kwds, "U|si:file",
2394 kwlist, &po, &mode, &bufsize)) {
2395 wideargument = 1;
2396 if (fill_file_fields(foself, NULL, po, mode,
2397 fclose) == NULL)
2398 goto Error;
2399 } else {
2400 /* Drop the argument parsing error as narrow
2401 strings are also valid. */
2402 PyErr_Clear();
2403 }
2404#endif
2405
2406 if (!wideargument) {
2407 PyObject *o_name;
2408
2409 if (!PyArg_ParseTupleAndKeywords(args, kwds, "et|si:file", kwlist,
2410 Py_FileSystemDefaultEncoding,
2411 &name,
2412 &mode, &bufsize))
2413 return -1;
2414
2415 /* We parse again to get the name as a PyObject */
2416 if (!PyArg_ParseTupleAndKeywords(args, kwds, "O|si:file",
2417 kwlist, &o_name, &mode,
2418 &bufsize))
2419 goto Error;
2420
2421 if (fill_file_fields(foself, NULL, o_name, mode,
2422 fclose) == NULL)
2423 goto Error;
2424 }
2425 if (open_the_file(foself, name, mode) == NULL)
2426 goto Error;
2427 foself->f_setbuf = NULL;
2428 PyFile_SetBufSize(self, bufsize);
2429 goto Done;
2430
2431Error:
2432 ret = -1;
2433 /* fall through */
2434Done:
2435 PyMem_Free(name); /* free the encoded string */
2436 return ret;
2437}
2438
2439PyDoc_VAR(file_doc) =
2440PyDoc_STR(
2441"file(name[, mode[, buffering]]) -> file object\n"
2442"\n"
2443"Open a file. The mode can be 'r', 'w' or 'a' for reading (default),\n"
2444"writing or appending. The file will be created if it doesn't exist\n"
2445"when opened for writing or appending; it will be truncated when\n"
2446"opened for writing. Add a 'b' to the mode for binary files.\n"
2447"Add a '+' to the mode to allow simultaneous reading and writing.\n"
2448"If the buffering argument is given, 0 means unbuffered, 1 means line\n"
2449"buffered, and larger numbers specify the buffer size. The preferred way\n"
2450"to open a file is with the builtin open() function.\n"
2451)
2452PyDoc_STR(
2453"Add a 'U' to mode to open the file for input with universal newline\n"
2454"support. Any line ending in the input file will be seen as a '\\n'\n"
2455"in Python. Also, a file so opened gains the attribute 'newlines';\n"
2456"the value for this attribute is one of None (no newline read yet),\n"
2457"'\\r', '\\n', '\\r\\n' or a tuple containing all the newline types seen.\n"
2458"\n"
2459"'U' cannot be combined with 'w' or '+' mode.\n"
2460);
2461
2462PyTypeObject PyFile_Type = {
2463 PyVarObject_HEAD_INIT(&PyType_Type, 0)
2464 "file",
2465 sizeof(PyFileObject),
2466 0,
2467 (destructor)file_dealloc, /* tp_dealloc */
2468 0, /* tp_print */
2469 0, /* tp_getattr */
2470 0, /* tp_setattr */
2471 0, /* tp_compare */
2472 (reprfunc)file_repr, /* tp_repr */
2473 0, /* tp_as_number */
2474 0, /* tp_as_sequence */
2475 0, /* tp_as_mapping */
2476 0, /* tp_hash */
2477 0, /* tp_call */
2478 0, /* tp_str */
2479 PyObject_GenericGetAttr, /* tp_getattro */
2480 /* softspace is writable: we must supply tp_setattro */
2481 PyObject_GenericSetAttr, /* tp_setattro */
2482 0, /* tp_as_buffer */
2483 Py_TPFLAGS_DEFAULT | Py_TPFLAGS_BASETYPE | Py_TPFLAGS_HAVE_WEAKREFS, /* tp_flags */
2484 file_doc, /* tp_doc */
2485 0, /* tp_traverse */
2486 0, /* tp_clear */
2487 0, /* tp_richcompare */
2488 offsetof(PyFileObject, weakreflist), /* tp_weaklistoffset */
2489 (getiterfunc)file_self, /* tp_iter */
2490 (iternextfunc)file_iternext, /* tp_iternext */
2491 file_methods, /* tp_methods */
2492 file_memberlist, /* tp_members */
2493 file_getsetlist, /* tp_getset */
2494 0, /* tp_base */
2495 0, /* tp_dict */
2496 0, /* tp_descr_get */
2497 0, /* tp_descr_set */
2498 0, /* tp_dictoffset */
2499 file_init, /* tp_init */
2500 PyType_GenericAlloc, /* tp_alloc */
2501 file_new, /* tp_new */
2502 PyObject_Del, /* tp_free */
2503};
2504
2505/* Interface for the 'soft space' between print items. */
2506
2507int
2508PyFile_SoftSpace(PyObject *f, int newflag)
2509{
2510 long oldflag = 0;
2511 if (f == NULL) {
2512 /* Do nothing */
2513 }
2514 else if (PyFile_Check(f)) {
2515 oldflag = ((PyFileObject *)f)->f_softspace;
2516 ((PyFileObject *)f)->f_softspace = newflag;
2517 }
2518 else {
2519 PyObject *v;
2520 v = PyObject_GetAttrString(f, "softspace");
2521 if (v == NULL)
2522 PyErr_Clear();
2523 else {
2524 if (PyInt_Check(v))
2525 oldflag = PyInt_AsLong(v);
2526 assert(oldflag < INT_MAX);
2527 Py_DECREF(v);
2528 }
2529 v = PyInt_FromLong((long)newflag);
2530 if (v == NULL)
2531 PyErr_Clear();
2532 else {
2533 if (PyObject_SetAttrString(f, "softspace", v) != 0)
2534 PyErr_Clear();
2535 Py_DECREF(v);
2536 }
2537 }
2538 return (int)oldflag;
2539}
2540
2541/* Interfaces to write objects/strings to file-like objects */
2542
2543int
2544PyFile_WriteObject(PyObject *v, PyObject *f, int flags)
2545{
2546 PyObject *writer, *value, *args, *result;
2547 if (f == NULL) {
2548 PyErr_SetString(PyExc_TypeError, "writeobject with NULL file");
2549 return -1;
2550 }
2551 else if (PyFile_Check(f)) {
2552 PyFileObject *fobj = (PyFileObject *) f;
2553#ifdef Py_USING_UNICODE
2554 PyObject *enc = fobj->f_encoding;
2555 int result;
2556#endif
2557 if (fobj->f_fp == NULL) {
2558 err_closed();
2559 return -1;
2560 }
2561#ifdef Py_USING_UNICODE
2562 if ((flags & Py_PRINT_RAW) &&
2563 PyUnicode_Check(v) && enc != Py_None) {
2564 char *cenc = PyString_AS_STRING(enc);
2565 char *errors = fobj->f_errors == Py_None ?
2566 "strict" : PyString_AS_STRING(fobj->f_errors);
2567 value = PyUnicode_AsEncodedString(v, cenc, errors);
2568 if (value == NULL)
2569 return -1;
2570 } else {
2571 value = v;
2572 Py_INCREF(value);
2573 }
2574 result = file_PyObject_Print(value, fobj, flags);
2575 Py_DECREF(value);
2576 return result;
2577#else
2578 return file_PyObject_Print(v, fobj, flags);
2579#endif
2580 }
2581 writer = PyObject_GetAttrString(f, "write");
2582 if (writer == NULL)
2583 return -1;
2584 if (flags & Py_PRINT_RAW) {
2585 if (PyUnicode_Check(v)) {
2586 value = v;
2587 Py_INCREF(value);
2588 } else
2589 value = PyObject_Str(v);
2590 }
2591 else
2592 value = PyObject_Repr(v);
2593 if (value == NULL) {
2594 Py_DECREF(writer);
2595 return -1;
2596 }
2597 args = PyTuple_Pack(1, value);
2598 if (args == NULL) {
2599 Py_DECREF(value);
2600 Py_DECREF(writer);
2601 return -1;
2602 }
2603 result = PyEval_CallObject(writer, args);
2604 Py_DECREF(args);
2605 Py_DECREF(value);
2606 Py_DECREF(writer);
2607 if (result == NULL)
2608 return -1;
2609 Py_DECREF(result);
2610 return 0;
2611}
2612
2613int
2614PyFile_WriteString(const char *s, PyObject *f)
2615{
2616
2617 if (f == NULL) {
2618 /* Should be caused by a pre-existing error */
2619 if (!PyErr_Occurred())
2620 PyErr_SetString(PyExc_SystemError,
2621 "null file for PyFile_WriteString");
2622 return -1;
2623 }
2624 else if (PyFile_Check(f)) {
2625 PyFileObject *fobj = (PyFileObject *) f;
2626 FILE *fp = PyFile_AsFile(f);
2627 if (fp == NULL) {
2628 err_closed();
2629 return -1;
2630 }
2631 FILE_BEGIN_ALLOW_THREADS(fobj)
2632 fputs(s, fp);
2633 FILE_END_ALLOW_THREADS(fobj)
2634 return 0;
2635 }
2636 else if (!PyErr_Occurred()) {
2637 PyObject *v = PyString_FromString(s);
2638 int err;
2639 if (v == NULL)
2640 return -1;
2641 err = PyFile_WriteObject(v, f, Py_PRINT_RAW);
2642 Py_DECREF(v);
2643 return err;
2644 }
2645 else
2646 return -1;
2647}
2648
2649/* Try to get a file-descriptor from a Python object. If the object
2650 is an integer or long integer, its value is returned. If not, the
2651 object's fileno() method is called if it exists; the method must return
2652 an integer or long integer, which is returned as the file descriptor value.
2653 -1 is returned on failure.
2654*/
2655
2656int PyObject_AsFileDescriptor(PyObject *o)
2657{
2658 int fd;
2659 PyObject *meth;
2660
2661 if (PyInt_Check(o)) {
2662 fd = _PyInt_AsInt(o);
2663 }
2664 else if (PyLong_Check(o)) {
2665 fd = _PyLong_AsInt(o);
2666 }
2667 else if ((meth = PyObject_GetAttrString(o, "fileno")) != NULL)
2668 {
2669 PyObject *fno = PyEval_CallObject(meth, NULL);
2670 Py_DECREF(meth);
2671 if (fno == NULL)
2672 return -1;
2673
2674 if (PyInt_Check(fno)) {
2675 fd = _PyInt_AsInt(fno);
2676 Py_DECREF(fno);
2677 }
2678 else if (PyLong_Check(fno)) {
2679 fd = _PyLong_AsInt(fno);
2680 Py_DECREF(fno);
2681 }
2682 else {
2683 PyErr_SetString(PyExc_TypeError,
2684 "fileno() returned a non-integer");
2685 Py_DECREF(fno);
2686 return -1;
2687 }
2688 }
2689 else {
2690 PyErr_SetString(PyExc_TypeError,
2691 "argument must be an int, or have a fileno() method.");
2692 return -1;
2693 }
2694
2695 if (fd < 0) {
2696 PyErr_Format(PyExc_ValueError,
2697 "file descriptor cannot be a negative integer (%i)",
2698 fd);
2699 return -1;
2700 }
2701 return fd;
2702}
2703
2704/* From here on we need access to the real fgets and fread */
2705#undef fgets
2706#undef fread
2707
2708/*
2709** Py_UniversalNewlineFgets is an fgets variation that understands
2710** all of \r, \n and \r\n conventions.
2711** The stream should be opened in binary mode.
2712** If fobj is NULL the routine always does newline conversion, and
2713** it may peek one char ahead to gobble the second char in \r\n.
2714** If fobj is non-NULL it must be a PyFileObject. In this case there
2715** is no readahead but in stead a flag is used to skip a following
2716** \n on the next read. Also, if the file is open in binary mode
2717** the whole conversion is skipped. Finally, the routine keeps track of
2718** the different types of newlines seen.
2719** Note that we need no error handling: fgets() treats error and eof
2720** identically.
2721*/
2722char *
2723Py_UniversalNewlineFgets(char *buf, int n, FILE *stream, PyObject *fobj)
2724{
2725 char *p = buf;
2726 int c;
2727 int newlinetypes = 0;
2728 int skipnextlf = 0;
2729 int univ_newline = 1;
2730
2731 if (fobj) {
2732 if (!PyFile_Check(fobj)) {
2733 errno = ENXIO; /* What can you do... */
2734 return NULL;
2735 }
2736 univ_newline = ((PyFileObject *)fobj)->f_univ_newline;
2737 if ( !univ_newline )
2738 return fgets(buf, n, stream);
2739 newlinetypes = ((PyFileObject *)fobj)->f_newlinetypes;
2740 skipnextlf = ((PyFileObject *)fobj)->f_skipnextlf;
2741 }
2742 FLOCKFILE(stream);
2743 c = 'x'; /* Shut up gcc warning */
2744 while (--n > 0 && (c = GETC(stream)) != EOF ) {
2745 if (skipnextlf ) {
2746 skipnextlf = 0;
2747 if (c == '\n') {
2748 /* Seeing a \n here with skipnextlf true
2749 ** means we saw a \r before.
2750 */
2751 newlinetypes |= NEWLINE_CRLF;
2752 c = GETC(stream);
2753 if (c == EOF) break;
2754 } else {
2755 /*
2756 ** Note that c == EOF also brings us here,
2757 ** so we're okay if the last char in the file
2758 ** is a CR.
2759 */
2760 newlinetypes |= NEWLINE_CR;
2761 }
2762 }
2763 if (c == '\r') {
2764 /* A \r is translated into a \n, and we skip
2765 ** an adjacent \n, if any. We don't set the
2766 ** newlinetypes flag until we've seen the next char.
2767 */
2768 skipnextlf = 1;
2769 c = '\n';
2770 } else if ( c == '\n') {
2771 newlinetypes |= NEWLINE_LF;
2772 }
2773 *p++ = c;
2774 if (c == '\n') break;
2775 }
2776 if ( c == EOF && skipnextlf )
2777 newlinetypes |= NEWLINE_CR;
2778 FUNLOCKFILE(stream);
2779 *p = '\0';
2780 if (fobj) {
2781 ((PyFileObject *)fobj)->f_newlinetypes = newlinetypes;
2782 ((PyFileObject *)fobj)->f_skipnextlf = skipnextlf;
2783 } else if ( skipnextlf ) {
2784 /* If we have no file object we cannot save the
2785 ** skipnextlf flag. We have to readahead, which
2786 ** will cause a pause if we're reading from an
2787 ** interactive stream, but that is very unlikely
2788 ** unless we're doing something silly like
2789 ** execfile("/dev/tty").
2790 */
2791 c = GETC(stream);
2792 if ( c != '\n' )
2793 ungetc(c, stream);
2794 }
2795 if (p == buf)
2796 return NULL;
2797 return buf;
2798}
2799
2800/*
2801** Py_UniversalNewlineFread is an fread variation that understands
2802** all of \r, \n and \r\n conventions.
2803** The stream should be opened in binary mode.
2804** fobj must be a PyFileObject. In this case there
2805** is no readahead but in stead a flag is used to skip a following
2806** \n on the next read. Also, if the file is open in binary mode
2807** the whole conversion is skipped. Finally, the routine keeps track of
2808** the different types of newlines seen.
2809*/
2810size_t
2811Py_UniversalNewlineFread(char *buf, size_t n,
2812 FILE *stream, PyObject *fobj)
2813{
2814 char *dst = buf;
2815 PyFileObject *f = (PyFileObject *)fobj;
2816 int newlinetypes, skipnextlf;
2817
2818 assert(buf != NULL);
2819 assert(stream != NULL);
2820
2821 if (!fobj || !PyFile_Check(fobj)) {
2822 errno = ENXIO; /* What can you do... */
2823 return 0;
2824 }
2825 if (!f->f_univ_newline)
2826 return fread(buf, 1, n, stream);
2827 newlinetypes = f->f_newlinetypes;
2828 skipnextlf = f->f_skipnextlf;
2829 /* Invariant: n is the number of bytes remaining to be filled
2830 * in the buffer.
2831 */
2832 while (n) {
2833 size_t nread;
2834 int shortread;
2835 char *src = dst;
2836
2837 nread = fread(dst, 1, n, stream);
2838 assert(nread <= n);
2839 if (nread == 0)
2840 break;
2841
2842 n -= nread; /* assuming 1 byte out for each in; will adjust */
2843 shortread = n != 0; /* true iff EOF or error */
2844 while (nread--) {
2845 char c = *src++;
2846 if (c == '\r') {
2847 /* Save as LF and set flag to skip next LF. */
2848 *dst++ = '\n';
2849 skipnextlf = 1;
2850 }
2851 else if (skipnextlf && c == '\n') {
2852 /* Skip LF, and remember we saw CR LF. */
2853 skipnextlf = 0;
2854 newlinetypes |= NEWLINE_CRLF;
2855 ++n;
2856 }
2857 else {
2858 /* Normal char to be stored in buffer. Also
2859 * update the newlinetypes flag if either this
2860 * is an LF or the previous char was a CR.
2861 */
2862 if (c == '\n')
2863 newlinetypes |= NEWLINE_LF;
2864 else if (skipnextlf)
2865 newlinetypes |= NEWLINE_CR;
2866 *dst++ = c;
2867 skipnextlf = 0;
2868 }
2869 }
2870 if (shortread) {
2871 /* If this is EOF, update type flags. */
2872 if (skipnextlf && feof(stream))
2873 newlinetypes |= NEWLINE_CR;
2874 break;
2875 }
2876 }
2877 f->f_newlinetypes = newlinetypes;
2878 f->f_skipnextlf = skipnextlf;
2879 return dst - buf;
2880}
2881
2882#ifdef __cplusplus
2883}
2884#endif
Note: See TracBrowser for help on using the repository browser.