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 | #ifdef _MSC_VER
|
---|
20 | /* Need GetVersion to see if on NT so safe to use _wfopen */
|
---|
21 | #define WIN32_LEAN_AND_MEAN
|
---|
22 | #include <windows.h>
|
---|
23 | #endif /* _MSC_VER */
|
---|
24 |
|
---|
25 | #if defined(PYOS_OS2) && defined(PYCC_GCC)
|
---|
26 | #include <io.h>
|
---|
27 | #endif
|
---|
28 |
|
---|
29 | #define BUF(v) PyString_AS_STRING((PyStringObject *)v)
|
---|
30 |
|
---|
31 | #ifndef DONT_HAVE_ERRNO_H
|
---|
32 | #include <errno.h>
|
---|
33 | #endif
|
---|
34 |
|
---|
35 | #ifdef HAVE_GETC_UNLOCKED
|
---|
36 | #define GETC(f) getc_unlocked(f)
|
---|
37 | #define FLOCKFILE(f) flockfile(f)
|
---|
38 | #define FUNLOCKFILE(f) funlockfile(f)
|
---|
39 | #else
|
---|
40 | #define GETC(f) getc(f)
|
---|
41 | #define FLOCKFILE(f)
|
---|
42 | #define FUNLOCKFILE(f)
|
---|
43 | #endif
|
---|
44 |
|
---|
45 | /* Bits in f_newlinetypes */
|
---|
46 | #define NEWLINE_UNKNOWN 0 /* No newline seen, yet */
|
---|
47 | #define NEWLINE_CR 1 /* \r newline seen */
|
---|
48 | #define NEWLINE_LF 2 /* \n newline seen */
|
---|
49 | #define NEWLINE_CRLF 4 /* \r\n newline seen */
|
---|
50 |
|
---|
51 | /*
|
---|
52 | * These macros release the GIL while preventing the f_close() function being
|
---|
53 | * called in the interval between them. For that purpose, a running total of
|
---|
54 | * the number of currently running unlocked code sections is kept in
|
---|
55 | * the unlocked_count field of the PyFileObject. The close() method raises
|
---|
56 | * an IOError if that field is non-zero. See issue #815646, #595601.
|
---|
57 | */
|
---|
58 |
|
---|
59 | #define FILE_BEGIN_ALLOW_THREADS(fobj) \
|
---|
60 | { \
|
---|
61 | fobj->unlocked_count++; \
|
---|
62 | Py_BEGIN_ALLOW_THREADS
|
---|
63 |
|
---|
64 | #define FILE_END_ALLOW_THREADS(fobj) \
|
---|
65 | Py_END_ALLOW_THREADS \
|
---|
66 | fobj->unlocked_count--; \
|
---|
67 | assert(fobj->unlocked_count >= 0); \
|
---|
68 | }
|
---|
69 |
|
---|
70 | #define FILE_ABORT_ALLOW_THREADS(fobj) \
|
---|
71 | Py_BLOCK_THREADS \
|
---|
72 | fobj->unlocked_count--; \
|
---|
73 | assert(fobj->unlocked_count >= 0);
|
---|
74 |
|
---|
75 | #ifdef __cplusplus
|
---|
76 | extern "C" {
|
---|
77 | #endif
|
---|
78 |
|
---|
79 | FILE *
|
---|
80 | PyFile_AsFile(PyObject *f)
|
---|
81 | {
|
---|
82 | if (f == NULL || !PyFile_Check(f))
|
---|
83 | return NULL;
|
---|
84 | else
|
---|
85 | return ((PyFileObject *)f)->f_fp;
|
---|
86 | }
|
---|
87 |
|
---|
88 | void PyFile_IncUseCount(PyFileObject *fobj)
|
---|
89 | {
|
---|
90 | fobj->unlocked_count++;
|
---|
91 | }
|
---|
92 |
|
---|
93 | void PyFile_DecUseCount(PyFileObject *fobj)
|
---|
94 | {
|
---|
95 | fobj->unlocked_count--;
|
---|
96 | assert(fobj->unlocked_count >= 0);
|
---|
97 | }
|
---|
98 |
|
---|
99 | PyObject *
|
---|
100 | PyFile_Name(PyObject *f)
|
---|
101 | {
|
---|
102 | if (f == NULL || !PyFile_Check(f))
|
---|
103 | return NULL;
|
---|
104 | else
|
---|
105 | return ((PyFileObject *)f)->f_name;
|
---|
106 | }
|
---|
107 |
|
---|
108 | /* This is a safe wrapper around PyObject_Print to print to the FILE
|
---|
109 | of a PyFileObject. PyObject_Print releases the GIL but knows nothing
|
---|
110 | about PyFileObject. */
|
---|
111 | static int
|
---|
112 | file_PyObject_Print(PyObject *op, PyFileObject *f, int flags)
|
---|
113 | {
|
---|
114 | int result;
|
---|
115 | PyFile_IncUseCount(f);
|
---|
116 | result = PyObject_Print(op, f->f_fp, flags);
|
---|
117 | PyFile_DecUseCount(f);
|
---|
118 | return result;
|
---|
119 | }
|
---|
120 |
|
---|
121 | /* On Unix, fopen will succeed for directories.
|
---|
122 | In Python, there should be no file objects referring to
|
---|
123 | directories, so we need a check. */
|
---|
124 |
|
---|
125 | static PyFileObject*
|
---|
126 | dircheck(PyFileObject* f)
|
---|
127 | {
|
---|
128 | #if defined(HAVE_FSTAT) && defined(S_IFDIR) && defined(EISDIR)
|
---|
129 | struct stat buf;
|
---|
130 | if (f->f_fp == NULL)
|
---|
131 | return f;
|
---|
132 | if (fstat(fileno(f->f_fp), &buf) == 0 &&
|
---|
133 | S_ISDIR(buf.st_mode)) {
|
---|
134 | char *msg = strerror(EISDIR);
|
---|
135 | PyObject *exc = PyObject_CallFunction(PyExc_IOError, "(isO)",
|
---|
136 | EISDIR, msg, f->f_name);
|
---|
137 | PyErr_SetObject(PyExc_IOError, exc);
|
---|
138 | Py_XDECREF(exc);
|
---|
139 | return NULL;
|
---|
140 | }
|
---|
141 | #endif
|
---|
142 | return f;
|
---|
143 | }
|
---|
144 |
|
---|
145 |
|
---|
146 | static PyObject *
|
---|
147 | fill_file_fields(PyFileObject *f, FILE *fp, PyObject *name, char *mode,
|
---|
148 | int (*close)(FILE *))
|
---|
149 | {
|
---|
150 | assert(name != NULL);
|
---|
151 | assert(f != NULL);
|
---|
152 | assert(PyFile_Check(f));
|
---|
153 | assert(f->f_fp == NULL);
|
---|
154 |
|
---|
155 | Py_DECREF(f->f_name);
|
---|
156 | Py_DECREF(f->f_mode);
|
---|
157 | Py_DECREF(f->f_encoding);
|
---|
158 | Py_DECREF(f->f_errors);
|
---|
159 |
|
---|
160 | Py_INCREF(name);
|
---|
161 | f->f_name = name;
|
---|
162 |
|
---|
163 | f->f_mode = PyString_FromString(mode);
|
---|
164 |
|
---|
165 | f->f_close = close;
|
---|
166 | f->f_softspace = 0;
|
---|
167 | f->f_binary = strchr(mode,'b') != NULL;
|
---|
168 | f->f_buf = NULL;
|
---|
169 | f->f_univ_newline = (strchr(mode, 'U') != NULL);
|
---|
170 | f->f_newlinetypes = NEWLINE_UNKNOWN;
|
---|
171 | f->f_skipnextlf = 0;
|
---|
172 | Py_INCREF(Py_None);
|
---|
173 | f->f_encoding = Py_None;
|
---|
174 | Py_INCREF(Py_None);
|
---|
175 | f->f_errors = Py_None;
|
---|
176 | f->readable = f->writable = 0;
|
---|
177 | if (strchr(mode, 'r') != NULL || f->f_univ_newline)
|
---|
178 | f->readable = 1;
|
---|
179 | if (strchr(mode, 'w') != NULL || strchr(mode, 'a') != NULL)
|
---|
180 | f->writable = 1;
|
---|
181 | if (strchr(mode, '+') != NULL)
|
---|
182 | f->readable = f->writable = 1;
|
---|
183 |
|
---|
184 | if (f->f_mode == NULL)
|
---|
185 | return NULL;
|
---|
186 | f->f_fp = fp;
|
---|
187 | f = dircheck(f);
|
---|
188 | return (PyObject *) f;
|
---|
189 | }
|
---|
190 |
|
---|
191 | /* check for known incorrect mode strings - problem is, platforms are
|
---|
192 | free to accept any mode characters they like and are supposed to
|
---|
193 | ignore stuff they don't understand... write or append mode with
|
---|
194 | universal newline support is expressly forbidden by PEP 278.
|
---|
195 | Additionally, remove the 'U' from the mode string as platforms
|
---|
196 | won't know what it is. Non-zero return signals an exception */
|
---|
197 | int
|
---|
198 | _PyFile_SanitizeMode(char *mode)
|
---|
199 | {
|
---|
200 | char *upos;
|
---|
201 | size_t len = strlen(mode);
|
---|
202 |
|
---|
203 | if (!len) {
|
---|
204 | PyErr_SetString(PyExc_ValueError, "empty mode string");
|
---|
205 | return -1;
|
---|
206 | }
|
---|
207 |
|
---|
208 | upos = strchr(mode, 'U');
|
---|
209 | if (upos) {
|
---|
210 | memmove(upos, upos+1, len-(upos-mode)); /* incl null char */
|
---|
211 |
|
---|
212 | if (mode[0] == 'w' || mode[0] == 'a') {
|
---|
213 | PyErr_Format(PyExc_ValueError, "universal newline "
|
---|
214 | "mode can only be used with modes "
|
---|
215 | "starting with 'r'");
|
---|
216 | return -1;
|
---|
217 | }
|
---|
218 |
|
---|
219 | if (mode[0] != 'r') {
|
---|
220 | memmove(mode+1, mode, strlen(mode)+1);
|
---|
221 | mode[0] = 'r';
|
---|
222 | }
|
---|
223 |
|
---|
224 | if (!strchr(mode, 'b')) {
|
---|
225 | memmove(mode+2, mode+1, strlen(mode));
|
---|
226 | mode[1] = 'b';
|
---|
227 | }
|
---|
228 | } else if (mode[0] != 'r' && mode[0] != 'w' && mode[0] != 'a') {
|
---|
229 | PyErr_Format(PyExc_ValueError, "mode string must begin with "
|
---|
230 | "one of 'r', 'w', 'a' or 'U', not '%.200s'", mode);
|
---|
231 | return -1;
|
---|
232 | }
|
---|
233 |
|
---|
234 | return 0;
|
---|
235 | }
|
---|
236 |
|
---|
237 | static PyObject *
|
---|
238 | open_the_file(PyFileObject *f, char *name, char *mode)
|
---|
239 | {
|
---|
240 | char *newmode;
|
---|
241 | assert(f != NULL);
|
---|
242 | assert(PyFile_Check(f));
|
---|
243 | #ifdef MS_WINDOWS
|
---|
244 | /* windows ignores the passed name in order to support Unicode */
|
---|
245 | assert(f->f_name != NULL);
|
---|
246 | #else
|
---|
247 | assert(name != NULL);
|
---|
248 | #endif
|
---|
249 | assert(mode != NULL);
|
---|
250 | assert(f->f_fp == NULL);
|
---|
251 |
|
---|
252 | /* probably need to replace 'U' by 'rb' */
|
---|
253 | newmode = PyMem_MALLOC(strlen(mode) + 3);
|
---|
254 | if (!newmode) {
|
---|
255 | PyErr_NoMemory();
|
---|
256 | return NULL;
|
---|
257 | }
|
---|
258 | strcpy(newmode, mode);
|
---|
259 |
|
---|
260 | if (_PyFile_SanitizeMode(newmode)) {
|
---|
261 | f = NULL;
|
---|
262 | goto cleanup;
|
---|
263 | }
|
---|
264 |
|
---|
265 | /* rexec.py can't stop a user from getting the file() constructor --
|
---|
266 | all they have to do is get *any* file object f, and then do
|
---|
267 | type(f). Here we prevent them from doing damage with it. */
|
---|
268 | if (PyEval_GetRestricted()) {
|
---|
269 | PyErr_SetString(PyExc_IOError,
|
---|
270 | "file() constructor not accessible in restricted mode");
|
---|
271 | f = NULL;
|
---|
272 | goto cleanup;
|
---|
273 | }
|
---|
274 | errno = 0;
|
---|
275 |
|
---|
276 | #ifdef MS_WINDOWS
|
---|
277 | if (PyUnicode_Check(f->f_name)) {
|
---|
278 | PyObject *wmode;
|
---|
279 | wmode = PyUnicode_DecodeASCII(newmode, strlen(newmode), NULL);
|
---|
280 | if (f->f_name && wmode) {
|
---|
281 | FILE_BEGIN_ALLOW_THREADS(f)
|
---|
282 | /* PyUnicode_AS_UNICODE OK without thread
|
---|
283 | lock as it is a simple dereference. */
|
---|
284 | f->f_fp = _wfopen(PyUnicode_AS_UNICODE(f->f_name),
|
---|
285 | PyUnicode_AS_UNICODE(wmode));
|
---|
286 | FILE_END_ALLOW_THREADS(f)
|
---|
287 | }
|
---|
288 | Py_XDECREF(wmode);
|
---|
289 | }
|
---|
290 | #endif
|
---|
291 | if (NULL == f->f_fp && NULL != name) {
|
---|
292 | FILE_BEGIN_ALLOW_THREADS(f)
|
---|
293 | f->f_fp = fopen(name, newmode);
|
---|
294 | FILE_END_ALLOW_THREADS(f)
|
---|
295 | }
|
---|
296 |
|
---|
297 | if (f->f_fp == NULL) {
|
---|
298 | #if defined _MSC_VER && (_MSC_VER < 1400 || !defined(__STDC_SECURE_LIB__))
|
---|
299 | /* MSVC 6 (Microsoft) leaves errno at 0 for bad mode strings,
|
---|
300 | * across all Windows flavors. When it sets EINVAL varies
|
---|
301 | * across Windows flavors, the exact conditions aren't
|
---|
302 | * documented, and the answer lies in the OS's implementation
|
---|
303 | * of Win32's CreateFile function (whose source is secret).
|
---|
304 | * Seems the best we can do is map EINVAL to ENOENT.
|
---|
305 | * Starting with Visual Studio .NET 2005, EINVAL is correctly
|
---|
306 | * set by our CRT error handler (set in exceptions.c.)
|
---|
307 | */
|
---|
308 | if (errno == 0) /* bad mode string */
|
---|
309 | errno = EINVAL;
|
---|
310 | else if (errno == EINVAL) /* unknown, but not a mode string */
|
---|
311 | errno = ENOENT;
|
---|
312 | #endif
|
---|
313 | /* EINVAL is returned when an invalid filename or
|
---|
314 | * an invalid mode is supplied. */
|
---|
315 | if (errno == EINVAL) {
|
---|
316 | PyObject *v;
|
---|
317 | char message[100];
|
---|
318 | PyOS_snprintf(message, 100,
|
---|
319 | "invalid mode ('%.50s') or filename", mode);
|
---|
320 | v = Py_BuildValue("(isO)", errno, message, f->f_name);
|
---|
321 | if (v != NULL) {
|
---|
322 | PyErr_SetObject(PyExc_IOError, v);
|
---|
323 | Py_DECREF(v);
|
---|
324 | }
|
---|
325 | }
|
---|
326 | else
|
---|
327 | PyErr_SetFromErrnoWithFilenameObject(PyExc_IOError, f->f_name);
|
---|
328 | f = NULL;
|
---|
329 | }
|
---|
330 | if (f != NULL)
|
---|
331 | f = dircheck(f);
|
---|
332 |
|
---|
333 | cleanup:
|
---|
334 | PyMem_FREE(newmode);
|
---|
335 |
|
---|
336 | return (PyObject *)f;
|
---|
337 | }
|
---|
338 |
|
---|
339 | static PyObject *
|
---|
340 | close_the_file(PyFileObject *f)
|
---|
341 | {
|
---|
342 | int sts = 0;
|
---|
343 | int (*local_close)(FILE *);
|
---|
344 | FILE *local_fp = f->f_fp;
|
---|
345 | if (local_fp != NULL) {
|
---|
346 | local_close = f->f_close;
|
---|
347 | if (local_close != NULL && f->unlocked_count > 0) {
|
---|
348 | if (f->ob_refcnt > 0) {
|
---|
349 | PyErr_SetString(PyExc_IOError,
|
---|
350 | "close() called during concurrent "
|
---|
351 | "operation on the same file object.");
|
---|
352 | } else {
|
---|
353 | /* This should not happen unless someone is
|
---|
354 | * carelessly playing with the PyFileObject
|
---|
355 | * struct fields and/or its associated FILE
|
---|
356 | * pointer. */
|
---|
357 | PyErr_SetString(PyExc_SystemError,
|
---|
358 | "PyFileObject locking error in "
|
---|
359 | "destructor (refcnt <= 0 at close).");
|
---|
360 | }
|
---|
361 | return NULL;
|
---|
362 | }
|
---|
363 | /* NULL out the FILE pointer before releasing the GIL, because
|
---|
364 | * it will not be valid anymore after the close() function is
|
---|
365 | * called. */
|
---|
366 | f->f_fp = NULL;
|
---|
367 | if (local_close != NULL) {
|
---|
368 | Py_BEGIN_ALLOW_THREADS
|
---|
369 | errno = 0;
|
---|
370 | sts = (*local_close)(local_fp);
|
---|
371 | Py_END_ALLOW_THREADS
|
---|
372 | if (sts == EOF)
|
---|
373 | return PyErr_SetFromErrno(PyExc_IOError);
|
---|
374 | if (sts != 0)
|
---|
375 | return PyInt_FromLong((long)sts);
|
---|
376 | }
|
---|
377 | }
|
---|
378 | Py_RETURN_NONE;
|
---|
379 | }
|
---|
380 |
|
---|
381 | PyObject *
|
---|
382 | PyFile_FromFile(FILE *fp, char *name, char *mode, int (*close)(FILE *))
|
---|
383 | {
|
---|
384 | PyFileObject *f = (PyFileObject *)PyFile_Type.tp_new(&PyFile_Type,
|
---|
385 | NULL, NULL);
|
---|
386 | if (f != NULL) {
|
---|
387 | PyObject *o_name = PyString_FromString(name);
|
---|
388 | if (o_name == NULL)
|
---|
389 | return NULL;
|
---|
390 | if (fill_file_fields(f, fp, o_name, mode, close) == NULL) {
|
---|
391 | Py_DECREF(f);
|
---|
392 | f = NULL;
|
---|
393 | }
|
---|
394 | Py_DECREF(o_name);
|
---|
395 | }
|
---|
396 | return (PyObject *) f;
|
---|
397 | }
|
---|
398 |
|
---|
399 | PyObject *
|
---|
400 | PyFile_FromString(char *name, char *mode)
|
---|
401 | {
|
---|
402 | extern int fclose(FILE *);
|
---|
403 | PyFileObject *f;
|
---|
404 |
|
---|
405 | f = (PyFileObject *)PyFile_FromFile((FILE *)NULL, name, mode, fclose);
|
---|
406 | if (f != NULL) {
|
---|
407 | if (open_the_file(f, name, mode) == NULL) {
|
---|
408 | Py_DECREF(f);
|
---|
409 | f = NULL;
|
---|
410 | }
|
---|
411 | }
|
---|
412 | return (PyObject *)f;
|
---|
413 | }
|
---|
414 |
|
---|
415 | void
|
---|
416 | PyFile_SetBufSize(PyObject *f, int bufsize)
|
---|
417 | {
|
---|
418 | PyFileObject *file = (PyFileObject *)f;
|
---|
419 | if (bufsize >= 0) {
|
---|
420 | int type;
|
---|
421 | switch (bufsize) {
|
---|
422 | case 0:
|
---|
423 | type = _IONBF;
|
---|
424 | break;
|
---|
425 | #ifdef HAVE_SETVBUF
|
---|
426 | case 1:
|
---|
427 | type = _IOLBF;
|
---|
428 | bufsize = BUFSIZ;
|
---|
429 | break;
|
---|
430 | #endif
|
---|
431 | default:
|
---|
432 | type = _IOFBF;
|
---|
433 | #ifndef HAVE_SETVBUF
|
---|
434 | bufsize = BUFSIZ;
|
---|
435 | #endif
|
---|
436 | break;
|
---|
437 | }
|
---|
438 | fflush(file->f_fp);
|
---|
439 | if (type == _IONBF) {
|
---|
440 | PyMem_Free(file->f_setbuf);
|
---|
441 | file->f_setbuf = NULL;
|
---|
442 | } else {
|
---|
443 | file->f_setbuf = (char *)PyMem_Realloc(file->f_setbuf,
|
---|
444 | bufsize);
|
---|
445 | }
|
---|
446 | #ifdef HAVE_SETVBUF
|
---|
447 | setvbuf(file->f_fp, file->f_setbuf, type, bufsize);
|
---|
448 | #else /* !HAVE_SETVBUF */
|
---|
449 | setbuf(file->f_fp, file->f_setbuf);
|
---|
450 | #endif /* !HAVE_SETVBUF */
|
---|
451 | }
|
---|
452 | }
|
---|
453 |
|
---|
454 | /* Set the encoding used to output Unicode strings.
|
---|
455 | Return 1 on success, 0 on failure. */
|
---|
456 |
|
---|
457 | int
|
---|
458 | PyFile_SetEncoding(PyObject *f, const char *enc)
|
---|
459 | {
|
---|
460 | return PyFile_SetEncodingAndErrors(f, enc, NULL);
|
---|
461 | }
|
---|
462 |
|
---|
463 | int
|
---|
464 | PyFile_SetEncodingAndErrors(PyObject *f, const char *enc, char* errors)
|
---|
465 | {
|
---|
466 | PyFileObject *file = (PyFileObject*)f;
|
---|
467 | PyObject *str, *oerrors;
|
---|
468 |
|
---|
469 | assert(PyFile_Check(f));
|
---|
470 | str = PyString_FromString(enc);
|
---|
471 | if (!str)
|
---|
472 | return 0;
|
---|
473 | if (errors) {
|
---|
474 | oerrors = PyString_FromString(errors);
|
---|
475 | if (!oerrors) {
|
---|
476 | Py_DECREF(str);
|
---|
477 | return 0;
|
---|
478 | }
|
---|
479 | } else {
|
---|
480 | oerrors = Py_None;
|
---|
481 | Py_INCREF(Py_None);
|
---|
482 | }
|
---|
483 | Py_DECREF(file->f_encoding);
|
---|
484 | file->f_encoding = str;
|
---|
485 | Py_DECREF(file->f_errors);
|
---|
486 | file->f_errors = oerrors;
|
---|
487 | return 1;
|
---|
488 | }
|
---|
489 |
|
---|
490 | static PyObject *
|
---|
491 | err_closed(void)
|
---|
492 | {
|
---|
493 | PyErr_SetString(PyExc_ValueError, "I/O operation on closed file");
|
---|
494 | return NULL;
|
---|
495 | }
|
---|
496 |
|
---|
497 | static PyObject *
|
---|
498 | err_mode(char *action)
|
---|
499 | {
|
---|
500 | PyErr_Format(PyExc_IOError, "File not open for %s", action);
|
---|
501 | return NULL;
|
---|
502 | }
|
---|
503 |
|
---|
504 | /* Refuse regular file I/O if there's data in the iteration-buffer.
|
---|
505 | * Mixing them would cause data to arrive out of order, as the read*
|
---|
506 | * methods don't use the iteration buffer. */
|
---|
507 | static PyObject *
|
---|
508 | err_iterbuffered(void)
|
---|
509 | {
|
---|
510 | PyErr_SetString(PyExc_ValueError,
|
---|
511 | "Mixing iteration and read methods would lose data");
|
---|
512 | return NULL;
|
---|
513 | }
|
---|
514 |
|
---|
515 | static void drop_readahead(PyFileObject *);
|
---|
516 |
|
---|
517 | /* Methods */
|
---|
518 |
|
---|
519 | static void
|
---|
520 | file_dealloc(PyFileObject *f)
|
---|
521 | {
|
---|
522 | PyObject *ret;
|
---|
523 | if (f->weakreflist != NULL)
|
---|
524 | PyObject_ClearWeakRefs((PyObject *) f);
|
---|
525 | ret = close_the_file(f);
|
---|
526 | if (!ret) {
|
---|
527 | PySys_WriteStderr("close failed in file object destructor:\n");
|
---|
528 | PyErr_Print();
|
---|
529 | }
|
---|
530 | else {
|
---|
531 | Py_DECREF(ret);
|
---|
532 | }
|
---|
533 | PyMem_Free(f->f_setbuf);
|
---|
534 | Py_XDECREF(f->f_name);
|
---|
535 | Py_XDECREF(f->f_mode);
|
---|
536 | Py_XDECREF(f->f_encoding);
|
---|
537 | Py_XDECREF(f->f_errors);
|
---|
538 | drop_readahead(f);
|
---|
539 | Py_TYPE(f)->tp_free((PyObject *)f);
|
---|
540 | }
|
---|
541 |
|
---|
542 | static PyObject *
|
---|
543 | file_repr(PyFileObject *f)
|
---|
544 | {
|
---|
545 | if (PyUnicode_Check(f->f_name)) {
|
---|
546 | #ifdef Py_USING_UNICODE
|
---|
547 | PyObject *ret = NULL;
|
---|
548 | PyObject *name = PyUnicode_AsUnicodeEscapeString(f->f_name);
|
---|
549 | const char *name_str = name ? PyString_AsString(name) : "?";
|
---|
550 | ret = PyString_FromFormat("<%s file u'%s', mode '%s' at %p>",
|
---|
551 | f->f_fp == NULL ? "closed" : "open",
|
---|
552 | name_str,
|
---|
553 | PyString_AsString(f->f_mode),
|
---|
554 | f);
|
---|
555 | Py_XDECREF(name);
|
---|
556 | return ret;
|
---|
557 | #endif
|
---|
558 | } else {
|
---|
559 | return PyString_FromFormat("<%s file '%s', mode '%s' at %p>",
|
---|
560 | f->f_fp == NULL ? "closed" : "open",
|
---|
561 | PyString_AsString(f->f_name),
|
---|
562 | PyString_AsString(f->f_mode),
|
---|
563 | f);
|
---|
564 | }
|
---|
565 | }
|
---|
566 |
|
---|
567 | static PyObject *
|
---|
568 | file_close(PyFileObject *f)
|
---|
569 | {
|
---|
570 | PyObject *sts = close_the_file(f);
|
---|
571 | PyMem_Free(f->f_setbuf);
|
---|
572 | f->f_setbuf = NULL;
|
---|
573 | return sts;
|
---|
574 | }
|
---|
575 |
|
---|
576 |
|
---|
577 | /* Our very own off_t-like type, 64-bit if possible */
|
---|
578 | #if !defined(HAVE_LARGEFILE_SUPPORT)
|
---|
579 | typedef off_t Py_off_t;
|
---|
580 | #elif SIZEOF_OFF_T >= 8
|
---|
581 | typedef off_t Py_off_t;
|
---|
582 | #elif SIZEOF_FPOS_T >= 8
|
---|
583 | typedef fpos_t Py_off_t;
|
---|
584 | #else
|
---|
585 | #error "Large file support, but neither off_t nor fpos_t is large enough."
|
---|
586 | #endif
|
---|
587 |
|
---|
588 |
|
---|
589 | /* a portable fseek() function
|
---|
590 | return 0 on success, non-zero on failure (with errno set) */
|
---|
591 | static int
|
---|
592 | _portable_fseek(FILE *fp, Py_off_t offset, int whence)
|
---|
593 | {
|
---|
594 | #if !defined(HAVE_LARGEFILE_SUPPORT)
|
---|
595 | return fseek(fp, offset, whence);
|
---|
596 | #elif defined(HAVE_FSEEKO) && SIZEOF_OFF_T >= 8
|
---|
597 | return fseeko(fp, offset, whence);
|
---|
598 | #elif defined(HAVE_FSEEK64)
|
---|
599 | return fseek64(fp, offset, whence);
|
---|
600 | #elif defined(__BEOS__)
|
---|
601 | return _fseek(fp, offset, whence);
|
---|
602 | #elif SIZEOF_FPOS_T >= 8
|
---|
603 | /* lacking a 64-bit capable fseek(), use a 64-bit capable fsetpos()
|
---|
604 | and fgetpos() to implement fseek()*/
|
---|
605 | fpos_t pos;
|
---|
606 | switch (whence) {
|
---|
607 | case SEEK_END:
|
---|
608 | #ifdef MS_WINDOWS
|
---|
609 | fflush(fp);
|
---|
610 | if (_lseeki64(fileno(fp), 0, 2) == -1)
|
---|
611 | return -1;
|
---|
612 | #else
|
---|
613 | if (fseek(fp, 0, SEEK_END) != 0)
|
---|
614 | return -1;
|
---|
615 | #endif
|
---|
616 | /* fall through */
|
---|
617 | case SEEK_CUR:
|
---|
618 | if (fgetpos(fp, &pos) != 0)
|
---|
619 | return -1;
|
---|
620 | offset += pos;
|
---|
621 | break;
|
---|
622 | /* case SEEK_SET: break; */
|
---|
623 | }
|
---|
624 | return fsetpos(fp, &offset);
|
---|
625 | #else
|
---|
626 | #error "Large file support, but no way to fseek."
|
---|
627 | #endif
|
---|
628 | }
|
---|
629 |
|
---|
630 |
|
---|
631 | /* a portable ftell() function
|
---|
632 | Return -1 on failure with errno set appropriately, current file
|
---|
633 | position on success */
|
---|
634 | static Py_off_t
|
---|
635 | _portable_ftell(FILE* fp)
|
---|
636 | {
|
---|
637 | #if !defined(HAVE_LARGEFILE_SUPPORT)
|
---|
638 | return ftell(fp);
|
---|
639 | #elif defined(HAVE_FTELLO) && SIZEOF_OFF_T >= 8
|
---|
640 | return ftello(fp);
|
---|
641 | #elif defined(HAVE_FTELL64)
|
---|
642 | return ftell64(fp);
|
---|
643 | #elif SIZEOF_FPOS_T >= 8
|
---|
644 | fpos_t pos;
|
---|
645 | if (fgetpos(fp, &pos) != 0)
|
---|
646 | return -1;
|
---|
647 | return pos;
|
---|
648 | #else
|
---|
649 | #error "Large file support, but no way to ftell."
|
---|
650 | #endif
|
---|
651 | }
|
---|
652 |
|
---|
653 |
|
---|
654 | static PyObject *
|
---|
655 | file_seek(PyFileObject *f, PyObject *args)
|
---|
656 | {
|
---|
657 | int whence;
|
---|
658 | int ret;
|
---|
659 | Py_off_t offset;
|
---|
660 | PyObject *offobj, *off_index;
|
---|
661 |
|
---|
662 | if (f->f_fp == NULL)
|
---|
663 | return err_closed();
|
---|
664 | drop_readahead(f);
|
---|
665 | whence = 0;
|
---|
666 | if (!PyArg_ParseTuple(args, "O|i:seek", &offobj, &whence))
|
---|
667 | return NULL;
|
---|
668 | off_index = PyNumber_Index(offobj);
|
---|
669 | if (!off_index) {
|
---|
670 | if (!PyFloat_Check(offobj))
|
---|
671 | return NULL;
|
---|
672 | /* Deprecated in 2.6 */
|
---|
673 | PyErr_Clear();
|
---|
674 | if (PyErr_WarnEx(PyExc_DeprecationWarning,
|
---|
675 | "integer argument expected, got float",
|
---|
676 | 1) < 0)
|
---|
677 | return NULL;
|
---|
678 | off_index = offobj;
|
---|
679 | Py_INCREF(offobj);
|
---|
680 | }
|
---|
681 | #if !defined(HAVE_LARGEFILE_SUPPORT)
|
---|
682 | offset = PyInt_AsLong(off_index);
|
---|
683 | #else
|
---|
684 | offset = PyLong_Check(off_index) ?
|
---|
685 | PyLong_AsLongLong(off_index) : PyInt_AsLong(off_index);
|
---|
686 | #endif
|
---|
687 | Py_DECREF(off_index);
|
---|
688 | if (PyErr_Occurred())
|
---|
689 | return NULL;
|
---|
690 |
|
---|
691 | FILE_BEGIN_ALLOW_THREADS(f)
|
---|
692 | errno = 0;
|
---|
693 | ret = _portable_fseek(f->f_fp, offset, whence);
|
---|
694 | FILE_END_ALLOW_THREADS(f)
|
---|
695 |
|
---|
696 | if (ret != 0) {
|
---|
697 | PyErr_SetFromErrno(PyExc_IOError);
|
---|
698 | clearerr(f->f_fp);
|
---|
699 | return NULL;
|
---|
700 | }
|
---|
701 | f->f_skipnextlf = 0;
|
---|
702 | Py_INCREF(Py_None);
|
---|
703 | return Py_None;
|
---|
704 | }
|
---|
705 |
|
---|
706 |
|
---|
707 | #ifdef HAVE_FTRUNCATE
|
---|
708 | static PyObject *
|
---|
709 | file_truncate(PyFileObject *f, PyObject *args)
|
---|
710 | {
|
---|
711 | Py_off_t newsize;
|
---|
712 | PyObject *newsizeobj = NULL;
|
---|
713 | Py_off_t initialpos;
|
---|
714 | int ret;
|
---|
715 |
|
---|
716 | if (f->f_fp == NULL)
|
---|
717 | return err_closed();
|
---|
718 | if (!f->writable)
|
---|
719 | return err_mode("writing");
|
---|
720 | if (!PyArg_UnpackTuple(args, "truncate", 0, 1, &newsizeobj))
|
---|
721 | return NULL;
|
---|
722 |
|
---|
723 | /* Get current file position. If the file happens to be open for
|
---|
724 | * update and the last operation was an input operation, C doesn't
|
---|
725 | * define what the later fflush() will do, but we promise truncate()
|
---|
726 | * won't change the current position (and fflush() *does* change it
|
---|
727 | * then at least on Windows). The easiest thing is to capture
|
---|
728 | * current pos now and seek back to it at the end.
|
---|
729 | */
|
---|
730 | FILE_BEGIN_ALLOW_THREADS(f)
|
---|
731 | errno = 0;
|
---|
732 | initialpos = _portable_ftell(f->f_fp);
|
---|
733 | FILE_END_ALLOW_THREADS(f)
|
---|
734 | if (initialpos == -1)
|
---|
735 | goto onioerror;
|
---|
736 |
|
---|
737 | /* Set newsize to current postion if newsizeobj NULL, else to the
|
---|
738 | * specified value.
|
---|
739 | */
|
---|
740 | if (newsizeobj != NULL) {
|
---|
741 | #if !defined(HAVE_LARGEFILE_SUPPORT)
|
---|
742 | newsize = PyInt_AsLong(newsizeobj);
|
---|
743 | #else
|
---|
744 | newsize = PyLong_Check(newsizeobj) ?
|
---|
745 | PyLong_AsLongLong(newsizeobj) :
|
---|
746 | PyInt_AsLong(newsizeobj);
|
---|
747 | #endif
|
---|
748 | if (PyErr_Occurred())
|
---|
749 | return NULL;
|
---|
750 | }
|
---|
751 | else /* default to current position */
|
---|
752 | newsize = initialpos;
|
---|
753 |
|
---|
754 | /* Flush the stream. We're mixing stream-level I/O with lower-level
|
---|
755 | * I/O, and a flush may be necessary to synch both platform views
|
---|
756 | * of the current file state.
|
---|
757 | */
|
---|
758 | FILE_BEGIN_ALLOW_THREADS(f)
|
---|
759 | errno = 0;
|
---|
760 | ret = fflush(f->f_fp);
|
---|
761 | FILE_END_ALLOW_THREADS(f)
|
---|
762 | if (ret != 0)
|
---|
763 | goto onioerror;
|
---|
764 |
|
---|
765 | #ifdef MS_WINDOWS
|
---|
766 | /* MS _chsize doesn't work if newsize doesn't fit in 32 bits,
|
---|
767 | so don't even try using it. */
|
---|
768 | {
|
---|
769 | HANDLE hFile;
|
---|
770 |
|
---|
771 | /* Have to move current pos to desired endpoint on Windows. */
|
---|
772 | FILE_BEGIN_ALLOW_THREADS(f)
|
---|
773 | errno = 0;
|
---|
774 | ret = _portable_fseek(f->f_fp, newsize, SEEK_SET) != 0;
|
---|
775 | FILE_END_ALLOW_THREADS(f)
|
---|
776 | if (ret)
|
---|
777 | goto onioerror;
|
---|
778 |
|
---|
779 | /* Truncate. Note that this may grow the file! */
|
---|
780 | FILE_BEGIN_ALLOW_THREADS(f)
|
---|
781 | errno = 0;
|
---|
782 | hFile = (HANDLE)_get_osfhandle(fileno(f->f_fp));
|
---|
783 | ret = hFile == (HANDLE)-1;
|
---|
784 | if (ret == 0) {
|
---|
785 | ret = SetEndOfFile(hFile) == 0;
|
---|
786 | if (ret)
|
---|
787 | errno = EACCES;
|
---|
788 | }
|
---|
789 | FILE_END_ALLOW_THREADS(f)
|
---|
790 | if (ret)
|
---|
791 | goto onioerror;
|
---|
792 | }
|
---|
793 | #else
|
---|
794 | FILE_BEGIN_ALLOW_THREADS(f)
|
---|
795 | errno = 0;
|
---|
796 | ret = ftruncate(fileno(f->f_fp), newsize);
|
---|
797 | FILE_END_ALLOW_THREADS(f)
|
---|
798 | if (ret != 0)
|
---|
799 | goto onioerror;
|
---|
800 | #endif /* !MS_WINDOWS */
|
---|
801 |
|
---|
802 | /* Restore original file position. */
|
---|
803 | FILE_BEGIN_ALLOW_THREADS(f)
|
---|
804 | errno = 0;
|
---|
805 | ret = _portable_fseek(f->f_fp, initialpos, SEEK_SET) != 0;
|
---|
806 | FILE_END_ALLOW_THREADS(f)
|
---|
807 | if (ret)
|
---|
808 | goto onioerror;
|
---|
809 |
|
---|
810 | Py_INCREF(Py_None);
|
---|
811 | return Py_None;
|
---|
812 |
|
---|
813 | onioerror:
|
---|
814 | PyErr_SetFromErrno(PyExc_IOError);
|
---|
815 | clearerr(f->f_fp);
|
---|
816 | return NULL;
|
---|
817 | }
|
---|
818 | #endif /* HAVE_FTRUNCATE */
|
---|
819 |
|
---|
820 | static PyObject *
|
---|
821 | file_tell(PyFileObject *f)
|
---|
822 | {
|
---|
823 | Py_off_t pos;
|
---|
824 |
|
---|
825 | if (f->f_fp == NULL)
|
---|
826 | return err_closed();
|
---|
827 | FILE_BEGIN_ALLOW_THREADS(f)
|
---|
828 | errno = 0;
|
---|
829 | pos = _portable_ftell(f->f_fp);
|
---|
830 | FILE_END_ALLOW_THREADS(f)
|
---|
831 |
|
---|
832 | if (pos == -1) {
|
---|
833 | PyErr_SetFromErrno(PyExc_IOError);
|
---|
834 | clearerr(f->f_fp);
|
---|
835 | return NULL;
|
---|
836 | }
|
---|
837 | if (f->f_skipnextlf) {
|
---|
838 | int c;
|
---|
839 | c = GETC(f->f_fp);
|
---|
840 | if (c == '\n') {
|
---|
841 | f->f_newlinetypes |= NEWLINE_CRLF;
|
---|
842 | pos++;
|
---|
843 | f->f_skipnextlf = 0;
|
---|
844 | } else if (c != EOF) ungetc(c, f->f_fp);
|
---|
845 | }
|
---|
846 | #if !defined(HAVE_LARGEFILE_SUPPORT)
|
---|
847 | return PyInt_FromLong(pos);
|
---|
848 | #else
|
---|
849 | return PyLong_FromLongLong(pos);
|
---|
850 | #endif
|
---|
851 | }
|
---|
852 |
|
---|
853 | static PyObject *
|
---|
854 | file_fileno(PyFileObject *f)
|
---|
855 | {
|
---|
856 | if (f->f_fp == NULL)
|
---|
857 | return err_closed();
|
---|
858 | return PyInt_FromLong((long) fileno(f->f_fp));
|
---|
859 | }
|
---|
860 |
|
---|
861 | static PyObject *
|
---|
862 | file_flush(PyFileObject *f)
|
---|
863 | {
|
---|
864 | int res;
|
---|
865 |
|
---|
866 | if (f->f_fp == NULL)
|
---|
867 | return err_closed();
|
---|
868 | FILE_BEGIN_ALLOW_THREADS(f)
|
---|
869 | errno = 0;
|
---|
870 | res = fflush(f->f_fp);
|
---|
871 | FILE_END_ALLOW_THREADS(f)
|
---|
872 | if (res != 0) {
|
---|
873 | PyErr_SetFromErrno(PyExc_IOError);
|
---|
874 | clearerr(f->f_fp);
|
---|
875 | return NULL;
|
---|
876 | }
|
---|
877 | Py_INCREF(Py_None);
|
---|
878 | return Py_None;
|
---|
879 | }
|
---|
880 |
|
---|
881 | static PyObject *
|
---|
882 | file_isatty(PyFileObject *f)
|
---|
883 | {
|
---|
884 | long res;
|
---|
885 | if (f->f_fp == NULL)
|
---|
886 | return err_closed();
|
---|
887 | FILE_BEGIN_ALLOW_THREADS(f)
|
---|
888 | res = isatty((int)fileno(f->f_fp));
|
---|
889 | FILE_END_ALLOW_THREADS(f)
|
---|
890 | return PyBool_FromLong(res);
|
---|
891 | }
|
---|
892 |
|
---|
893 |
|
---|
894 | #if BUFSIZ < 8192
|
---|
895 | #define SMALLCHUNK 8192
|
---|
896 | #else
|
---|
897 | #define SMALLCHUNK BUFSIZ
|
---|
898 | #endif
|
---|
899 |
|
---|
900 | #if SIZEOF_INT < 4
|
---|
901 | #define BIGCHUNK (512 * 32)
|
---|
902 | #else
|
---|
903 | #define BIGCHUNK (512 * 1024)
|
---|
904 | #endif
|
---|
905 |
|
---|
906 | static size_t
|
---|
907 | new_buffersize(PyFileObject *f, size_t currentsize)
|
---|
908 | {
|
---|
909 | #ifdef HAVE_FSTAT
|
---|
910 | off_t pos, end;
|
---|
911 | struct stat st;
|
---|
912 | if (fstat(fileno(f->f_fp), &st) == 0) {
|
---|
913 | end = st.st_size;
|
---|
914 | /* The following is not a bug: we really need to call lseek()
|
---|
915 | *and* ftell(). The reason is that some stdio libraries
|
---|
916 | mistakenly flush their buffer when ftell() is called and
|
---|
917 | the lseek() call it makes fails, thereby throwing away
|
---|
918 | data that cannot be recovered in any way. To avoid this,
|
---|
919 | we first test lseek(), and only call ftell() if lseek()
|
---|
920 | works. We can't use the lseek() value either, because we
|
---|
921 | need to take the amount of buffered data into account.
|
---|
922 | (Yet another reason why stdio stinks. :-) */
|
---|
923 | pos = lseek(fileno(f->f_fp), 0L, SEEK_CUR);
|
---|
924 | if (pos >= 0) {
|
---|
925 | pos = ftell(f->f_fp);
|
---|
926 | }
|
---|
927 | if (pos < 0)
|
---|
928 | clearerr(f->f_fp);
|
---|
929 | if (end > pos && pos >= 0)
|
---|
930 | return currentsize + end - pos + 1;
|
---|
931 | /* Add 1 so if the file were to grow we'd notice. */
|
---|
932 | }
|
---|
933 | #endif
|
---|
934 | if (currentsize > SMALLCHUNK) {
|
---|
935 | /* Keep doubling until we reach BIGCHUNK;
|
---|
936 | then keep adding BIGCHUNK. */
|
---|
937 | if (currentsize <= BIGCHUNK)
|
---|
938 | return currentsize + currentsize;
|
---|
939 | else
|
---|
940 | return currentsize + BIGCHUNK;
|
---|
941 | }
|
---|
942 | return currentsize + SMALLCHUNK;
|
---|
943 | }
|
---|
944 |
|
---|
945 | #if defined(EWOULDBLOCK) && defined(EAGAIN) && EWOULDBLOCK != EAGAIN
|
---|
946 | #define BLOCKED_ERRNO(x) ((x) == EWOULDBLOCK || (x) == EAGAIN)
|
---|
947 | #else
|
---|
948 | #ifdef EWOULDBLOCK
|
---|
949 | #define BLOCKED_ERRNO(x) ((x) == EWOULDBLOCK)
|
---|
950 | #else
|
---|
951 | #ifdef EAGAIN
|
---|
952 | #define BLOCKED_ERRNO(x) ((x) == EAGAIN)
|
---|
953 | #else
|
---|
954 | #define BLOCKED_ERRNO(x) 0
|
---|
955 | #endif
|
---|
956 | #endif
|
---|
957 | #endif
|
---|
958 |
|
---|
959 | static PyObject *
|
---|
960 | file_read(PyFileObject *f, PyObject *args)
|
---|
961 | {
|
---|
962 | long bytesrequested = -1;
|
---|
963 | size_t bytesread, buffersize, chunksize;
|
---|
964 | PyObject *v;
|
---|
965 |
|
---|
966 | if (f->f_fp == NULL)
|
---|
967 | return err_closed();
|
---|
968 | if (!f->readable)
|
---|
969 | return err_mode("reading");
|
---|
970 | /* refuse to mix with f.next() */
|
---|
971 | if (f->f_buf != NULL &&
|
---|
972 | (f->f_bufend - f->f_bufptr) > 0 &&
|
---|
973 | f->f_buf[0] != '\0')
|
---|
974 | return err_iterbuffered();
|
---|
975 | if (!PyArg_ParseTuple(args, "|l:read", &bytesrequested))
|
---|
976 | return NULL;
|
---|
977 | if (bytesrequested < 0)
|
---|
978 | buffersize = new_buffersize(f, (size_t)0);
|
---|
979 | else
|
---|
980 | buffersize = bytesrequested;
|
---|
981 | if (buffersize > PY_SSIZE_T_MAX) {
|
---|
982 | PyErr_SetString(PyExc_OverflowError,
|
---|
983 | "requested number of bytes is more than a Python string can hold");
|
---|
984 | return NULL;
|
---|
985 | }
|
---|
986 | v = PyString_FromStringAndSize((char *)NULL, buffersize);
|
---|
987 | if (v == NULL)
|
---|
988 | return NULL;
|
---|
989 | bytesread = 0;
|
---|
990 | for (;;) {
|
---|
991 | FILE_BEGIN_ALLOW_THREADS(f)
|
---|
992 | errno = 0;
|
---|
993 | chunksize = Py_UniversalNewlineFread(BUF(v) + bytesread,
|
---|
994 | buffersize - bytesread, f->f_fp, (PyObject *)f);
|
---|
995 | FILE_END_ALLOW_THREADS(f)
|
---|
996 | if (chunksize == 0) {
|
---|
997 | if (!ferror(f->f_fp))
|
---|
998 | break;
|
---|
999 | clearerr(f->f_fp);
|
---|
1000 | /* When in non-blocking mode, data shouldn't
|
---|
1001 | * be discarded if a blocking signal was
|
---|
1002 | * received. That will also happen if
|
---|
1003 | * chunksize != 0, but bytesread < buffersize. */
|
---|
1004 | if (bytesread > 0 && BLOCKED_ERRNO(errno))
|
---|
1005 | break;
|
---|
1006 | PyErr_SetFromErrno(PyExc_IOError);
|
---|
1007 | Py_DECREF(v);
|
---|
1008 | return NULL;
|
---|
1009 | }
|
---|
1010 | bytesread += chunksize;
|
---|
1011 | if (bytesread < buffersize) {
|
---|
1012 | clearerr(f->f_fp);
|
---|
1013 | break;
|
---|
1014 | }
|
---|
1015 | if (bytesrequested < 0) {
|
---|
1016 | buffersize = new_buffersize(f, buffersize);
|
---|
1017 | if (_PyString_Resize(&v, buffersize) < 0)
|
---|
1018 | return NULL;
|
---|
1019 | } else {
|
---|
1020 | /* Got what was requested. */
|
---|
1021 | break;
|
---|
1022 | }
|
---|
1023 | }
|
---|
1024 | if (bytesread != buffersize)
|
---|
1025 | _PyString_Resize(&v, bytesread);
|
---|
1026 | return v;
|
---|
1027 | }
|
---|
1028 |
|
---|
1029 | static PyObject *
|
---|
1030 | file_readinto(PyFileObject *f, PyObject *args)
|
---|
1031 | {
|
---|
1032 | char *ptr;
|
---|
1033 | Py_ssize_t ntodo;
|
---|
1034 | Py_ssize_t ndone, nnow;
|
---|
1035 | Py_buffer pbuf;
|
---|
1036 |
|
---|
1037 | if (f->f_fp == NULL)
|
---|
1038 | return err_closed();
|
---|
1039 | if (!f->readable)
|
---|
1040 | return err_mode("reading");
|
---|
1041 | /* refuse to mix with f.next() */
|
---|
1042 | if (f->f_buf != NULL &&
|
---|
1043 | (f->f_bufend - f->f_bufptr) > 0 &&
|
---|
1044 | f->f_buf[0] != '\0')
|
---|
1045 | return err_iterbuffered();
|
---|
1046 | if (!PyArg_ParseTuple(args, "w*", &pbuf))
|
---|
1047 | return NULL;
|
---|
1048 | ptr = pbuf.buf;
|
---|
1049 | ntodo = pbuf.len;
|
---|
1050 | ndone = 0;
|
---|
1051 | while (ntodo > 0) {
|
---|
1052 | FILE_BEGIN_ALLOW_THREADS(f)
|
---|
1053 | errno = 0;
|
---|
1054 | nnow = Py_UniversalNewlineFread(ptr+ndone, ntodo, f->f_fp,
|
---|
1055 | (PyObject *)f);
|
---|
1056 | FILE_END_ALLOW_THREADS(f)
|
---|
1057 | if (nnow == 0) {
|
---|
1058 | if (!ferror(f->f_fp))
|
---|
1059 | break;
|
---|
1060 | PyErr_SetFromErrno(PyExc_IOError);
|
---|
1061 | clearerr(f->f_fp);
|
---|
1062 | PyBuffer_Release(&pbuf);
|
---|
1063 | return NULL;
|
---|
1064 | }
|
---|
1065 | ndone += nnow;
|
---|
1066 | ntodo -= nnow;
|
---|
1067 | }
|
---|
1068 | PyBuffer_Release(&pbuf);
|
---|
1069 | return PyInt_FromSsize_t(ndone);
|
---|
1070 | }
|
---|
1071 |
|
---|
1072 | /**************************************************************************
|
---|
1073 | Routine to get next line using platform fgets().
|
---|
1074 |
|
---|
1075 | Under MSVC 6:
|
---|
1076 |
|
---|
1077 | + MS threadsafe getc is very slow (multiple layers of function calls before+
|
---|
1078 | after each character, to lock+unlock the stream).
|
---|
1079 | + The stream-locking functions are MS-internal -- can't access them from user
|
---|
1080 | code.
|
---|
1081 | + There's nothing Tim could find in the MS C or platform SDK libraries that
|
---|
1082 | can worm around this.
|
---|
1083 | + MS fgets locks/unlocks only once per line; it's the only hook we have.
|
---|
1084 |
|
---|
1085 | So we use fgets for speed(!), despite that it's painful.
|
---|
1086 |
|
---|
1087 | MS realloc is also slow.
|
---|
1088 |
|
---|
1089 | Reports from other platforms on this method vs getc_unlocked (which MS doesn't
|
---|
1090 | have):
|
---|
1091 | Linux a wash
|
---|
1092 | Solaris a wash
|
---|
1093 | Tru64 Unix getline_via_fgets significantly faster
|
---|
1094 |
|
---|
1095 | CAUTION: The C std isn't clear about this: in those cases where fgets
|
---|
1096 | writes something into the buffer, can it write into any position beyond the
|
---|
1097 | required trailing null byte? MSVC 6 fgets does not, and no platform is (yet)
|
---|
1098 | known on which it does; and it would be a strange way to code fgets. Still,
|
---|
1099 | getline_via_fgets may not work correctly if it does. The std test
|
---|
1100 | test_bufio.py should fail if platform fgets() routinely writes beyond the
|
---|
1101 | trailing null byte. #define DONT_USE_FGETS_IN_GETLINE to disable this code.
|
---|
1102 | **************************************************************************/
|
---|
1103 |
|
---|
1104 | /* Use this routine if told to, or by default on non-get_unlocked()
|
---|
1105 | * platforms unless told not to. Yikes! Let's spell that out:
|
---|
1106 | * On a platform with getc_unlocked():
|
---|
1107 | * By default, use getc_unlocked().
|
---|
1108 | * If you want to use fgets() instead, #define USE_FGETS_IN_GETLINE.
|
---|
1109 | * On a platform without getc_unlocked():
|
---|
1110 | * By default, use fgets().
|
---|
1111 | * If you don't want to use fgets(), #define DONT_USE_FGETS_IN_GETLINE.
|
---|
1112 | */
|
---|
1113 | #if !defined(USE_FGETS_IN_GETLINE) && !defined(HAVE_GETC_UNLOCKED)
|
---|
1114 | #define USE_FGETS_IN_GETLINE
|
---|
1115 | #endif
|
---|
1116 |
|
---|
1117 | #if defined(DONT_USE_FGETS_IN_GETLINE) && defined(USE_FGETS_IN_GETLINE)
|
---|
1118 | #undef USE_FGETS_IN_GETLINE
|
---|
1119 | #endif
|
---|
1120 |
|
---|
1121 | #ifdef USE_FGETS_IN_GETLINE
|
---|
1122 | static PyObject*
|
---|
1123 | getline_via_fgets(PyFileObject *f, FILE *fp)
|
---|
1124 | {
|
---|
1125 | /* INITBUFSIZE is the maximum line length that lets us get away with the fast
|
---|
1126 | * no-realloc, one-fgets()-call path. Boosting it isn't free, because we have
|
---|
1127 | * to fill this much of the buffer with a known value in order to figure out
|
---|
1128 | * how much of the buffer fgets() overwrites. So if INITBUFSIZE is larger
|
---|
1129 | * than "most" lines, we waste time filling unused buffer slots. 100 is
|
---|
1130 | * surely adequate for most peoples' email archives, chewing over source code,
|
---|
1131 | * etc -- "regular old text files".
|
---|
1132 | * MAXBUFSIZE is the maximum line length that lets us get away with the less
|
---|
1133 | * fast (but still zippy) no-realloc, two-fgets()-call path. See above for
|
---|
1134 | * cautions about boosting that. 300 was chosen because the worst real-life
|
---|
1135 | * text-crunching job reported on Python-Dev was a mail-log crawler where over
|
---|
1136 | * half the lines were 254 chars.
|
---|
1137 | */
|
---|
1138 | #define INITBUFSIZE 100
|
---|
1139 | #define MAXBUFSIZE 300
|
---|
1140 | char* p; /* temp */
|
---|
1141 | char buf[MAXBUFSIZE];
|
---|
1142 | PyObject* v; /* the string object result */
|
---|
1143 | char* pvfree; /* address of next free slot */
|
---|
1144 | char* pvend; /* address one beyond last free slot */
|
---|
1145 | size_t nfree; /* # of free buffer slots; pvend-pvfree */
|
---|
1146 | size_t total_v_size; /* total # of slots in buffer */
|
---|
1147 | size_t increment; /* amount to increment the buffer */
|
---|
1148 | size_t prev_v_size;
|
---|
1149 |
|
---|
1150 | /* Optimize for normal case: avoid _PyString_Resize if at all
|
---|
1151 | * possible via first reading into stack buffer "buf".
|
---|
1152 | */
|
---|
1153 | total_v_size = INITBUFSIZE; /* start small and pray */
|
---|
1154 | pvfree = buf;
|
---|
1155 | for (;;) {
|
---|
1156 | FILE_BEGIN_ALLOW_THREADS(f)
|
---|
1157 | pvend = buf + total_v_size;
|
---|
1158 | nfree = pvend - pvfree;
|
---|
1159 | memset(pvfree, '\n', nfree);
|
---|
1160 | assert(nfree < INT_MAX); /* Should be atmost MAXBUFSIZE */
|
---|
1161 | p = fgets(pvfree, (int)nfree, fp);
|
---|
1162 | FILE_END_ALLOW_THREADS(f)
|
---|
1163 |
|
---|
1164 | if (p == NULL) {
|
---|
1165 | clearerr(fp);
|
---|
1166 | if (PyErr_CheckSignals())
|
---|
1167 | return NULL;
|
---|
1168 | v = PyString_FromStringAndSize(buf, pvfree - buf);
|
---|
1169 | return v;
|
---|
1170 | }
|
---|
1171 | /* fgets read *something* */
|
---|
1172 | p = memchr(pvfree, '\n', nfree);
|
---|
1173 | if (p != NULL) {
|
---|
1174 | /* Did the \n come from fgets or from us?
|
---|
1175 | * Since fgets stops at the first \n, and then writes
|
---|
1176 | * \0, if it's from fgets a \0 must be next. But if
|
---|
1177 | * that's so, it could not have come from us, since
|
---|
1178 | * the \n's we filled the buffer with have only more
|
---|
1179 | * \n's to the right.
|
---|
1180 | */
|
---|
1181 | if (p+1 < pvend && *(p+1) == '\0') {
|
---|
1182 | /* It's from fgets: we win! In particular,
|
---|
1183 | * we haven't done any mallocs yet, and can
|
---|
1184 | * build the final result on the first try.
|
---|
1185 | */
|
---|
1186 | ++p; /* include \n from fgets */
|
---|
1187 | }
|
---|
1188 | else {
|
---|
1189 | /* Must be from us: fgets didn't fill the
|
---|
1190 | * buffer and didn't find a newline, so it
|
---|
1191 | * must be the last and newline-free line of
|
---|
1192 | * the file.
|
---|
1193 | */
|
---|
1194 | assert(p > pvfree && *(p-1) == '\0');
|
---|
1195 | --p; /* don't include \0 from fgets */
|
---|
1196 | }
|
---|
1197 | v = PyString_FromStringAndSize(buf, p - buf);
|
---|
1198 | return v;
|
---|
1199 | }
|
---|
1200 | /* yuck: fgets overwrote all the newlines, i.e. the entire
|
---|
1201 | * buffer. So this line isn't over yet, or maybe it is but
|
---|
1202 | * we're exactly at EOF. If we haven't already, try using the
|
---|
1203 | * rest of the stack buffer.
|
---|
1204 | */
|
---|
1205 | assert(*(pvend-1) == '\0');
|
---|
1206 | if (pvfree == buf) {
|
---|
1207 | pvfree = pvend - 1; /* overwrite trailing null */
|
---|
1208 | total_v_size = MAXBUFSIZE;
|
---|
1209 | }
|
---|
1210 | else
|
---|
1211 | break;
|
---|
1212 | }
|
---|
1213 |
|
---|
1214 | /* The stack buffer isn't big enough; malloc a string object and read
|
---|
1215 | * into its buffer.
|
---|
1216 | */
|
---|
1217 | total_v_size = MAXBUFSIZE << 1;
|
---|
1218 | v = PyString_FromStringAndSize((char*)NULL, (int)total_v_size);
|
---|
1219 | if (v == NULL)
|
---|
1220 | return v;
|
---|
1221 | /* copy over everything except the last null byte */
|
---|
1222 | memcpy(BUF(v), buf, MAXBUFSIZE-1);
|
---|
1223 | pvfree = BUF(v) + MAXBUFSIZE - 1;
|
---|
1224 |
|
---|
1225 | /* Keep reading stuff into v; if it ever ends successfully, break
|
---|
1226 | * after setting p one beyond the end of the line. The code here is
|
---|
1227 | * very much like the code above, except reads into v's buffer; see
|
---|
1228 | * the code above for detailed comments about the logic.
|
---|
1229 | */
|
---|
1230 | for (;;) {
|
---|
1231 | FILE_BEGIN_ALLOW_THREADS(f)
|
---|
1232 | pvend = BUF(v) + total_v_size;
|
---|
1233 | nfree = pvend - pvfree;
|
---|
1234 | memset(pvfree, '\n', nfree);
|
---|
1235 | assert(nfree < INT_MAX);
|
---|
1236 | p = fgets(pvfree, (int)nfree, fp);
|
---|
1237 | FILE_END_ALLOW_THREADS(f)
|
---|
1238 |
|
---|
1239 | if (p == NULL) {
|
---|
1240 | clearerr(fp);
|
---|
1241 | if (PyErr_CheckSignals()) {
|
---|
1242 | Py_DECREF(v);
|
---|
1243 | return NULL;
|
---|
1244 | }
|
---|
1245 | p = pvfree;
|
---|
1246 | break;
|
---|
1247 | }
|
---|
1248 | p = memchr(pvfree, '\n', nfree);
|
---|
1249 | if (p != NULL) {
|
---|
1250 | if (p+1 < pvend && *(p+1) == '\0') {
|
---|
1251 | /* \n came from fgets */
|
---|
1252 | ++p;
|
---|
1253 | break;
|
---|
1254 | }
|
---|
1255 | /* \n came from us; last line of file, no newline */
|
---|
1256 | assert(p > pvfree && *(p-1) == '\0');
|
---|
1257 | --p;
|
---|
1258 | break;
|
---|
1259 | }
|
---|
1260 | /* expand buffer and try again */
|
---|
1261 | assert(*(pvend-1) == '\0');
|
---|
1262 | increment = total_v_size >> 2; /* mild exponential growth */
|
---|
1263 | prev_v_size = total_v_size;
|
---|
1264 | total_v_size += increment;
|
---|
1265 | /* check for overflow */
|
---|
1266 | if (total_v_size <= prev_v_size ||
|
---|
1267 | total_v_size > PY_SSIZE_T_MAX) {
|
---|
1268 | PyErr_SetString(PyExc_OverflowError,
|
---|
1269 | "line is longer than a Python string can hold");
|
---|
1270 | Py_DECREF(v);
|
---|
1271 | return NULL;
|
---|
1272 | }
|
---|
1273 | if (_PyString_Resize(&v, (int)total_v_size) < 0)
|
---|
1274 | return NULL;
|
---|
1275 | /* overwrite the trailing null byte */
|
---|
1276 | pvfree = BUF(v) + (prev_v_size - 1);
|
---|
1277 | }
|
---|
1278 | if (BUF(v) + total_v_size != p)
|
---|
1279 | _PyString_Resize(&v, p - BUF(v));
|
---|
1280 | return v;
|
---|
1281 | #undef INITBUFSIZE
|
---|
1282 | #undef MAXBUFSIZE
|
---|
1283 | }
|
---|
1284 | #endif /* ifdef USE_FGETS_IN_GETLINE */
|
---|
1285 |
|
---|
1286 | /* Internal routine to get a line.
|
---|
1287 | Size argument interpretation:
|
---|
1288 | > 0: max length;
|
---|
1289 | <= 0: read arbitrary line
|
---|
1290 | */
|
---|
1291 |
|
---|
1292 | static PyObject *
|
---|
1293 | get_line(PyFileObject *f, int n)
|
---|
1294 | {
|
---|
1295 | FILE *fp = f->f_fp;
|
---|
1296 | int c;
|
---|
1297 | char *buf, *end;
|
---|
1298 | size_t total_v_size; /* total # of slots in buffer */
|
---|
1299 | size_t used_v_size; /* # used slots in buffer */
|
---|
1300 | size_t increment; /* amount to increment the buffer */
|
---|
1301 | PyObject *v;
|
---|
1302 | int newlinetypes = f->f_newlinetypes;
|
---|
1303 | int skipnextlf = f->f_skipnextlf;
|
---|
1304 | int univ_newline = f->f_univ_newline;
|
---|
1305 |
|
---|
1306 | #if defined(USE_FGETS_IN_GETLINE)
|
---|
1307 | if (n <= 0 && !univ_newline )
|
---|
1308 | return getline_via_fgets(f, fp);
|
---|
1309 | #endif
|
---|
1310 | total_v_size = n > 0 ? n : 100;
|
---|
1311 | v = PyString_FromStringAndSize((char *)NULL, total_v_size);
|
---|
1312 | if (v == NULL)
|
---|
1313 | return NULL;
|
---|
1314 | buf = BUF(v);
|
---|
1315 | end = buf + total_v_size;
|
---|
1316 |
|
---|
1317 | for (;;) {
|
---|
1318 | FILE_BEGIN_ALLOW_THREADS(f)
|
---|
1319 | FLOCKFILE(fp);
|
---|
1320 | if (univ_newline) {
|
---|
1321 | c = 'x'; /* Shut up gcc warning */
|
---|
1322 | while ( buf != end && (c = GETC(fp)) != EOF ) {
|
---|
1323 | if (skipnextlf ) {
|
---|
1324 | skipnextlf = 0;
|
---|
1325 | if (c == '\n') {
|
---|
1326 | /* Seeing a \n here with
|
---|
1327 | * skipnextlf true means we
|
---|
1328 | * saw a \r before.
|
---|
1329 | */
|
---|
1330 | newlinetypes |= NEWLINE_CRLF;
|
---|
1331 | c = GETC(fp);
|
---|
1332 | if (c == EOF) break;
|
---|
1333 | } else {
|
---|
1334 | newlinetypes |= NEWLINE_CR;
|
---|
1335 | }
|
---|
1336 | }
|
---|
1337 | if (c == '\r') {
|
---|
1338 | skipnextlf = 1;
|
---|
1339 | c = '\n';
|
---|
1340 | } else if ( c == '\n')
|
---|
1341 | newlinetypes |= NEWLINE_LF;
|
---|
1342 | *buf++ = c;
|
---|
1343 | if (c == '\n') break;
|
---|
1344 | }
|
---|
1345 | if ( c == EOF && skipnextlf )
|
---|
1346 | newlinetypes |= NEWLINE_CR;
|
---|
1347 | } else /* If not universal newlines use the normal loop */
|
---|
1348 | while ((c = GETC(fp)) != EOF &&
|
---|
1349 | (*buf++ = c) != '\n' &&
|
---|
1350 | buf != end)
|
---|
1351 | ;
|
---|
1352 | FUNLOCKFILE(fp);
|
---|
1353 | FILE_END_ALLOW_THREADS(f)
|
---|
1354 | f->f_newlinetypes = newlinetypes;
|
---|
1355 | f->f_skipnextlf = skipnextlf;
|
---|
1356 | if (c == '\n')
|
---|
1357 | break;
|
---|
1358 | if (c == EOF) {
|
---|
1359 | if (ferror(fp)) {
|
---|
1360 | PyErr_SetFromErrno(PyExc_IOError);
|
---|
1361 | clearerr(fp);
|
---|
1362 | Py_DECREF(v);
|
---|
1363 | return NULL;
|
---|
1364 | }
|
---|
1365 | clearerr(fp);
|
---|
1366 | if (PyErr_CheckSignals()) {
|
---|
1367 | Py_DECREF(v);
|
---|
1368 | return NULL;
|
---|
1369 | }
|
---|
1370 | break;
|
---|
1371 | }
|
---|
1372 | /* Must be because buf == end */
|
---|
1373 | if (n > 0)
|
---|
1374 | break;
|
---|
1375 | used_v_size = total_v_size;
|
---|
1376 | increment = total_v_size >> 2; /* mild exponential growth */
|
---|
1377 | total_v_size += increment;
|
---|
1378 | if (total_v_size > PY_SSIZE_T_MAX) {
|
---|
1379 | PyErr_SetString(PyExc_OverflowError,
|
---|
1380 | "line is longer than a Python string can hold");
|
---|
1381 | Py_DECREF(v);
|
---|
1382 | return NULL;
|
---|
1383 | }
|
---|
1384 | if (_PyString_Resize(&v, total_v_size) < 0)
|
---|
1385 | return NULL;
|
---|
1386 | buf = BUF(v) + used_v_size;
|
---|
1387 | end = BUF(v) + total_v_size;
|
---|
1388 | }
|
---|
1389 |
|
---|
1390 | used_v_size = buf - BUF(v);
|
---|
1391 | if (used_v_size != total_v_size)
|
---|
1392 | _PyString_Resize(&v, used_v_size);
|
---|
1393 | return v;
|
---|
1394 | }
|
---|
1395 |
|
---|
1396 | /* External C interface */
|
---|
1397 |
|
---|
1398 | PyObject *
|
---|
1399 | PyFile_GetLine(PyObject *f, int n)
|
---|
1400 | {
|
---|
1401 | PyObject *result;
|
---|
1402 |
|
---|
1403 | if (f == NULL) {
|
---|
1404 | PyErr_BadInternalCall();
|
---|
1405 | return NULL;
|
---|
1406 | }
|
---|
1407 |
|
---|
1408 | if (PyFile_Check(f)) {
|
---|
1409 | PyFileObject *fo = (PyFileObject *)f;
|
---|
1410 | if (fo->f_fp == NULL)
|
---|
1411 | return err_closed();
|
---|
1412 | if (!fo->readable)
|
---|
1413 | return err_mode("reading");
|
---|
1414 | /* refuse to mix with f.next() */
|
---|
1415 | if (fo->f_buf != NULL &&
|
---|
1416 | (fo->f_bufend - fo->f_bufptr) > 0 &&
|
---|
1417 | fo->f_buf[0] != '\0')
|
---|
1418 | return err_iterbuffered();
|
---|
1419 | result = get_line(fo, n);
|
---|
1420 | }
|
---|
1421 | else {
|
---|
1422 | PyObject *reader;
|
---|
1423 | PyObject *args;
|
---|
1424 |
|
---|
1425 | reader = PyObject_GetAttrString(f, "readline");
|
---|
1426 | if (reader == NULL)
|
---|
1427 | return NULL;
|
---|
1428 | if (n <= 0)
|
---|
1429 | args = PyTuple_New(0);
|
---|
1430 | else
|
---|
1431 | args = Py_BuildValue("(i)", n);
|
---|
1432 | if (args == NULL) {
|
---|
1433 | Py_DECREF(reader);
|
---|
1434 | return NULL;
|
---|
1435 | }
|
---|
1436 | result = PyEval_CallObject(reader, args);
|
---|
1437 | Py_DECREF(reader);
|
---|
1438 | Py_DECREF(args);
|
---|
1439 | if (result != NULL && !PyString_Check(result) &&
|
---|
1440 | !PyUnicode_Check(result)) {
|
---|
1441 | Py_DECREF(result);
|
---|
1442 | result = NULL;
|
---|
1443 | PyErr_SetString(PyExc_TypeError,
|
---|
1444 | "object.readline() returned non-string");
|
---|
1445 | }
|
---|
1446 | }
|
---|
1447 |
|
---|
1448 | if (n < 0 && result != NULL && PyString_Check(result)) {
|
---|
1449 | char *s = PyString_AS_STRING(result);
|
---|
1450 | Py_ssize_t len = PyString_GET_SIZE(result);
|
---|
1451 | if (len == 0) {
|
---|
1452 | Py_DECREF(result);
|
---|
1453 | result = NULL;
|
---|
1454 | PyErr_SetString(PyExc_EOFError,
|
---|
1455 | "EOF when reading a line");
|
---|
1456 | }
|
---|
1457 | else if (s[len-1] == '\n') {
|
---|
1458 | if (result->ob_refcnt == 1)
|
---|
1459 | _PyString_Resize(&result, len-1);
|
---|
1460 | else {
|
---|
1461 | PyObject *v;
|
---|
1462 | v = PyString_FromStringAndSize(s, len-1);
|
---|
1463 | Py_DECREF(result);
|
---|
1464 | result = v;
|
---|
1465 | }
|
---|
1466 | }
|
---|
1467 | }
|
---|
1468 | #ifdef Py_USING_UNICODE
|
---|
1469 | if (n < 0 && result != NULL && PyUnicode_Check(result)) {
|
---|
1470 | Py_UNICODE *s = PyUnicode_AS_UNICODE(result);
|
---|
1471 | Py_ssize_t len = PyUnicode_GET_SIZE(result);
|
---|
1472 | if (len == 0) {
|
---|
1473 | Py_DECREF(result);
|
---|
1474 | result = NULL;
|
---|
1475 | PyErr_SetString(PyExc_EOFError,
|
---|
1476 | "EOF when reading a line");
|
---|
1477 | }
|
---|
1478 | else if (s[len-1] == '\n') {
|
---|
1479 | if (result->ob_refcnt == 1)
|
---|
1480 | PyUnicode_Resize(&result, len-1);
|
---|
1481 | else {
|
---|
1482 | PyObject *v;
|
---|
1483 | v = PyUnicode_FromUnicode(s, len-1);
|
---|
1484 | Py_DECREF(result);
|
---|
1485 | result = v;
|
---|
1486 | }
|
---|
1487 | }
|
---|
1488 | }
|
---|
1489 | #endif
|
---|
1490 | return result;
|
---|
1491 | }
|
---|
1492 |
|
---|
1493 | /* Python method */
|
---|
1494 |
|
---|
1495 | static PyObject *
|
---|
1496 | file_readline(PyFileObject *f, PyObject *args)
|
---|
1497 | {
|
---|
1498 | int n = -1;
|
---|
1499 |
|
---|
1500 | if (f->f_fp == NULL)
|
---|
1501 | return err_closed();
|
---|
1502 | if (!f->readable)
|
---|
1503 | return err_mode("reading");
|
---|
1504 | /* refuse to mix with f.next() */
|
---|
1505 | if (f->f_buf != NULL &&
|
---|
1506 | (f->f_bufend - f->f_bufptr) > 0 &&
|
---|
1507 | f->f_buf[0] != '\0')
|
---|
1508 | return err_iterbuffered();
|
---|
1509 | if (!PyArg_ParseTuple(args, "|i:readline", &n))
|
---|
1510 | return NULL;
|
---|
1511 | if (n == 0)
|
---|
1512 | return PyString_FromString("");
|
---|
1513 | if (n < 0)
|
---|
1514 | n = 0;
|
---|
1515 | return get_line(f, n);
|
---|
1516 | }
|
---|
1517 |
|
---|
1518 | static PyObject *
|
---|
1519 | file_readlines(PyFileObject *f, PyObject *args)
|
---|
1520 | {
|
---|
1521 | long sizehint = 0;
|
---|
1522 | PyObject *list = NULL;
|
---|
1523 | PyObject *line;
|
---|
1524 | char small_buffer[SMALLCHUNK];
|
---|
1525 | char *buffer = small_buffer;
|
---|
1526 | size_t buffersize = SMALLCHUNK;
|
---|
1527 | PyObject *big_buffer = NULL;
|
---|
1528 | size_t nfilled = 0;
|
---|
1529 | size_t nread;
|
---|
1530 | size_t totalread = 0;
|
---|
1531 | char *p, *q, *end;
|
---|
1532 | int err;
|
---|
1533 | int shortread = 0;
|
---|
1534 |
|
---|
1535 | if (f->f_fp == NULL)
|
---|
1536 | return err_closed();
|
---|
1537 | if (!f->readable)
|
---|
1538 | return err_mode("reading");
|
---|
1539 | /* refuse to mix with f.next() */
|
---|
1540 | if (f->f_buf != NULL &&
|
---|
1541 | (f->f_bufend - f->f_bufptr) > 0 &&
|
---|
1542 | f->f_buf[0] != '\0')
|
---|
1543 | return err_iterbuffered();
|
---|
1544 | if (!PyArg_ParseTuple(args, "|l:readlines", &sizehint))
|
---|
1545 | return NULL;
|
---|
1546 | if ((list = PyList_New(0)) == NULL)
|
---|
1547 | return NULL;
|
---|
1548 | for (;;) {
|
---|
1549 | if (shortread)
|
---|
1550 | nread = 0;
|
---|
1551 | else {
|
---|
1552 | FILE_BEGIN_ALLOW_THREADS(f)
|
---|
1553 | errno = 0;
|
---|
1554 | nread = Py_UniversalNewlineFread(buffer+nfilled,
|
---|
1555 | buffersize-nfilled, f->f_fp, (PyObject *)f);
|
---|
1556 | FILE_END_ALLOW_THREADS(f)
|
---|
1557 | shortread = (nread < buffersize-nfilled);
|
---|
1558 | }
|
---|
1559 | if (nread == 0) {
|
---|
1560 | sizehint = 0;
|
---|
1561 | if (!ferror(f->f_fp))
|
---|
1562 | break;
|
---|
1563 | PyErr_SetFromErrno(PyExc_IOError);
|
---|
1564 | clearerr(f->f_fp);
|
---|
1565 | goto error;
|
---|
1566 | }
|
---|
1567 | totalread += nread;
|
---|
1568 | p = (char *)memchr(buffer+nfilled, '\n', nread);
|
---|
1569 | if (p == NULL) {
|
---|
1570 | /* Need a larger buffer to fit this line */
|
---|
1571 | nfilled += nread;
|
---|
1572 | buffersize *= 2;
|
---|
1573 | if (buffersize > PY_SSIZE_T_MAX) {
|
---|
1574 | PyErr_SetString(PyExc_OverflowError,
|
---|
1575 | "line is longer than a Python string can hold");
|
---|
1576 | goto error;
|
---|
1577 | }
|
---|
1578 | if (big_buffer == NULL) {
|
---|
1579 | /* Create the big buffer */
|
---|
1580 | big_buffer = PyString_FromStringAndSize(
|
---|
1581 | NULL, buffersize);
|
---|
1582 | if (big_buffer == NULL)
|
---|
1583 | goto error;
|
---|
1584 | buffer = PyString_AS_STRING(big_buffer);
|
---|
1585 | memcpy(buffer, small_buffer, nfilled);
|
---|
1586 | }
|
---|
1587 | else {
|
---|
1588 | /* Grow the big buffer */
|
---|
1589 | if ( _PyString_Resize(&big_buffer, buffersize) < 0 )
|
---|
1590 | goto error;
|
---|
1591 | buffer = PyString_AS_STRING(big_buffer);
|
---|
1592 | }
|
---|
1593 | continue;
|
---|
1594 | }
|
---|
1595 | end = buffer+nfilled+nread;
|
---|
1596 | q = buffer;
|
---|
1597 | do {
|
---|
1598 | /* Process complete lines */
|
---|
1599 | p++;
|
---|
1600 | line = PyString_FromStringAndSize(q, p-q);
|
---|
1601 | if (line == NULL)
|
---|
1602 | goto error;
|
---|
1603 | err = PyList_Append(list, line);
|
---|
1604 | Py_DECREF(line);
|
---|
1605 | if (err != 0)
|
---|
1606 | goto error;
|
---|
1607 | q = p;
|
---|
1608 | p = (char *)memchr(q, '\n', end-q);
|
---|
1609 | } while (p != NULL);
|
---|
1610 | /* Move the remaining incomplete line to the start */
|
---|
1611 | nfilled = end-q;
|
---|
1612 | memmove(buffer, q, nfilled);
|
---|
1613 | if (sizehint > 0)
|
---|
1614 | if (totalread >= (size_t)sizehint)
|
---|
1615 | break;
|
---|
1616 | }
|
---|
1617 | if (nfilled != 0) {
|
---|
1618 | /* Partial last line */
|
---|
1619 | line = PyString_FromStringAndSize(buffer, nfilled);
|
---|
1620 | if (line == NULL)
|
---|
1621 | goto error;
|
---|
1622 | if (sizehint > 0) {
|
---|
1623 | /* Need to complete the last line */
|
---|
1624 | PyObject *rest = get_line(f, 0);
|
---|
1625 | if (rest == NULL) {
|
---|
1626 | Py_DECREF(line);
|
---|
1627 | goto error;
|
---|
1628 | }
|
---|
1629 | PyString_Concat(&line, rest);
|
---|
1630 | Py_DECREF(rest);
|
---|
1631 | if (line == NULL)
|
---|
1632 | goto error;
|
---|
1633 | }
|
---|
1634 | err = PyList_Append(list, line);
|
---|
1635 | Py_DECREF(line);
|
---|
1636 | if (err != 0)
|
---|
1637 | goto error;
|
---|
1638 | }
|
---|
1639 |
|
---|
1640 | cleanup:
|
---|
1641 | Py_XDECREF(big_buffer);
|
---|
1642 | return list;
|
---|
1643 |
|
---|
1644 | error:
|
---|
1645 | Py_CLEAR(list);
|
---|
1646 | goto cleanup;
|
---|
1647 | }
|
---|
1648 |
|
---|
1649 | static PyObject *
|
---|
1650 | file_write(PyFileObject *f, PyObject *args)
|
---|
1651 | {
|
---|
1652 | Py_buffer pbuf;
|
---|
1653 | char *s;
|
---|
1654 | Py_ssize_t n, n2;
|
---|
1655 | if (f->f_fp == NULL)
|
---|
1656 | return err_closed();
|
---|
1657 | if (!f->writable)
|
---|
1658 | return err_mode("writing");
|
---|
1659 | if (f->f_binary) {
|
---|
1660 | if (!PyArg_ParseTuple(args, "s*", &pbuf))
|
---|
1661 | return NULL;
|
---|
1662 | s = pbuf.buf;
|
---|
1663 | n = pbuf.len;
|
---|
1664 | } else
|
---|
1665 | if (!PyArg_ParseTuple(args, "t#", &s, &n))
|
---|
1666 | return NULL;
|
---|
1667 | f->f_softspace = 0;
|
---|
1668 | FILE_BEGIN_ALLOW_THREADS(f)
|
---|
1669 | errno = 0;
|
---|
1670 | n2 = fwrite(s, 1, n, f->f_fp);
|
---|
1671 | FILE_END_ALLOW_THREADS(f)
|
---|
1672 | if (f->f_binary)
|
---|
1673 | PyBuffer_Release(&pbuf);
|
---|
1674 | if (n2 != n) {
|
---|
1675 | PyErr_SetFromErrno(PyExc_IOError);
|
---|
1676 | clearerr(f->f_fp);
|
---|
1677 | return NULL;
|
---|
1678 | }
|
---|
1679 | Py_INCREF(Py_None);
|
---|
1680 | return Py_None;
|
---|
1681 | }
|
---|
1682 |
|
---|
1683 | static PyObject *
|
---|
1684 | file_writelines(PyFileObject *f, PyObject *seq)
|
---|
1685 | {
|
---|
1686 | #define CHUNKSIZE 1000
|
---|
1687 | PyObject *list, *line;
|
---|
1688 | PyObject *it; /* iter(seq) */
|
---|
1689 | PyObject *result;
|
---|
1690 | int index, islist;
|
---|
1691 | Py_ssize_t i, j, nwritten, len;
|
---|
1692 |
|
---|
1693 | assert(seq != NULL);
|
---|
1694 | if (f->f_fp == NULL)
|
---|
1695 | return err_closed();
|
---|
1696 | if (!f->writable)
|
---|
1697 | return err_mode("writing");
|
---|
1698 |
|
---|
1699 | result = NULL;
|
---|
1700 | list = NULL;
|
---|
1701 | islist = PyList_Check(seq);
|
---|
1702 | if (islist)
|
---|
1703 | it = NULL;
|
---|
1704 | else {
|
---|
1705 | it = PyObject_GetIter(seq);
|
---|
1706 | if (it == NULL) {
|
---|
1707 | PyErr_SetString(PyExc_TypeError,
|
---|
1708 | "writelines() requires an iterable argument");
|
---|
1709 | return NULL;
|
---|
1710 | }
|
---|
1711 | /* From here on, fail by going to error, to reclaim "it". */
|
---|
1712 | list = PyList_New(CHUNKSIZE);
|
---|
1713 | if (list == NULL)
|
---|
1714 | goto error;
|
---|
1715 | }
|
---|
1716 |
|
---|
1717 | /* Strategy: slurp CHUNKSIZE lines into a private list,
|
---|
1718 | checking that they are all strings, then write that list
|
---|
1719 | without holding the interpreter lock, then come back for more. */
|
---|
1720 | for (index = 0; ; index += CHUNKSIZE) {
|
---|
1721 | if (islist) {
|
---|
1722 | Py_XDECREF(list);
|
---|
1723 | list = PyList_GetSlice(seq, index, index+CHUNKSIZE);
|
---|
1724 | if (list == NULL)
|
---|
1725 | goto error;
|
---|
1726 | j = PyList_GET_SIZE(list);
|
---|
1727 | }
|
---|
1728 | else {
|
---|
1729 | for (j = 0; j < CHUNKSIZE; j++) {
|
---|
1730 | line = PyIter_Next(it);
|
---|
1731 | if (line == NULL) {
|
---|
1732 | if (PyErr_Occurred())
|
---|
1733 | goto error;
|
---|
1734 | break;
|
---|
1735 | }
|
---|
1736 | PyList_SetItem(list, j, line);
|
---|
1737 | }
|
---|
1738 | }
|
---|
1739 | if (j == 0)
|
---|
1740 | break;
|
---|
1741 |
|
---|
1742 | /* Check that all entries are indeed strings. If not,
|
---|
1743 | apply the same rules as for file.write() and
|
---|
1744 | convert the results to strings. This is slow, but
|
---|
1745 | seems to be the only way since all conversion APIs
|
---|
1746 | could potentially execute Python code. */
|
---|
1747 | for (i = 0; i < j; i++) {
|
---|
1748 | PyObject *v = PyList_GET_ITEM(list, i);
|
---|
1749 | if (!PyString_Check(v)) {
|
---|
1750 | const char *buffer;
|
---|
1751 | if (((f->f_binary &&
|
---|
1752 | PyObject_AsReadBuffer(v,
|
---|
1753 | (const void**)&buffer,
|
---|
1754 | &len)) ||
|
---|
1755 | PyObject_AsCharBuffer(v,
|
---|
1756 | &buffer,
|
---|
1757 | &len))) {
|
---|
1758 | PyErr_SetString(PyExc_TypeError,
|
---|
1759 | "writelines() argument must be a sequence of strings");
|
---|
1760 | goto error;
|
---|
1761 | }
|
---|
1762 | line = PyString_FromStringAndSize(buffer,
|
---|
1763 | len);
|
---|
1764 | if (line == NULL)
|
---|
1765 | goto error;
|
---|
1766 | Py_DECREF(v);
|
---|
1767 | PyList_SET_ITEM(list, i, line);
|
---|
1768 | }
|
---|
1769 | }
|
---|
1770 |
|
---|
1771 | /* Since we are releasing the global lock, the
|
---|
1772 | following code may *not* execute Python code. */
|
---|
1773 | f->f_softspace = 0;
|
---|
1774 | FILE_BEGIN_ALLOW_THREADS(f)
|
---|
1775 | errno = 0;
|
---|
1776 | for (i = 0; i < j; i++) {
|
---|
1777 | line = PyList_GET_ITEM(list, i);
|
---|
1778 | len = PyString_GET_SIZE(line);
|
---|
1779 | nwritten = fwrite(PyString_AS_STRING(line),
|
---|
1780 | 1, len, f->f_fp);
|
---|
1781 | if (nwritten != len) {
|
---|
1782 | FILE_ABORT_ALLOW_THREADS(f)
|
---|
1783 | PyErr_SetFromErrno(PyExc_IOError);
|
---|
1784 | clearerr(f->f_fp);
|
---|
1785 | goto error;
|
---|
1786 | }
|
---|
1787 | }
|
---|
1788 | FILE_END_ALLOW_THREADS(f)
|
---|
1789 |
|
---|
1790 | if (j < CHUNKSIZE)
|
---|
1791 | break;
|
---|
1792 | }
|
---|
1793 |
|
---|
1794 | Py_INCREF(Py_None);
|
---|
1795 | result = Py_None;
|
---|
1796 | error:
|
---|
1797 | Py_XDECREF(list);
|
---|
1798 | Py_XDECREF(it);
|
---|
1799 | return result;
|
---|
1800 | #undef CHUNKSIZE
|
---|
1801 | }
|
---|
1802 |
|
---|
1803 | static PyObject *
|
---|
1804 | file_self(PyFileObject *f)
|
---|
1805 | {
|
---|
1806 | if (f->f_fp == NULL)
|
---|
1807 | return err_closed();
|
---|
1808 | Py_INCREF(f);
|
---|
1809 | return (PyObject *)f;
|
---|
1810 | }
|
---|
1811 |
|
---|
1812 | static PyObject *
|
---|
1813 | file_xreadlines(PyFileObject *f)
|
---|
1814 | {
|
---|
1815 | if (PyErr_WarnPy3k("f.xreadlines() not supported in 3.x, "
|
---|
1816 | "try 'for line in f' instead", 1) < 0)
|
---|
1817 | return NULL;
|
---|
1818 | return file_self(f);
|
---|
1819 | }
|
---|
1820 |
|
---|
1821 | static PyObject *
|
---|
1822 | file_exit(PyObject *f, PyObject *args)
|
---|
1823 | {
|
---|
1824 | PyObject *ret = PyObject_CallMethod(f, "close", NULL);
|
---|
1825 | if (!ret)
|
---|
1826 | /* If error occurred, pass through */
|
---|
1827 | return NULL;
|
---|
1828 | Py_DECREF(ret);
|
---|
1829 | /* We cannot return the result of close since a true
|
---|
1830 | * value will be interpreted as "yes, swallow the
|
---|
1831 | * exception if one was raised inside the with block". */
|
---|
1832 | Py_RETURN_NONE;
|
---|
1833 | }
|
---|
1834 |
|
---|
1835 | PyDoc_STRVAR(readline_doc,
|
---|
1836 | "readline([size]) -> next line from the file, as a string.\n"
|
---|
1837 | "\n"
|
---|
1838 | "Retain newline. A non-negative size argument limits the maximum\n"
|
---|
1839 | "number of bytes to return (an incomplete line may be returned then).\n"
|
---|
1840 | "Return an empty string at EOF.");
|
---|
1841 |
|
---|
1842 | PyDoc_STRVAR(read_doc,
|
---|
1843 | "read([size]) -> read at most size bytes, returned as a string.\n"
|
---|
1844 | "\n"
|
---|
1845 | "If the size argument is negative or omitted, read until EOF is reached.\n"
|
---|
1846 | "Notice that when in non-blocking mode, less data than what was requested\n"
|
---|
1847 | "may be returned, even if no size parameter was given.");
|
---|
1848 |
|
---|
1849 | PyDoc_STRVAR(write_doc,
|
---|
1850 | "write(str) -> None. Write string str to file.\n"
|
---|
1851 | "\n"
|
---|
1852 | "Note that due to buffering, flush() or close() may be needed before\n"
|
---|
1853 | "the file on disk reflects the data written.");
|
---|
1854 |
|
---|
1855 | PyDoc_STRVAR(fileno_doc,
|
---|
1856 | "fileno() -> integer \"file descriptor\".\n"
|
---|
1857 | "\n"
|
---|
1858 | "This is needed for lower-level file interfaces, such os.read().");
|
---|
1859 |
|
---|
1860 | PyDoc_STRVAR(seek_doc,
|
---|
1861 | "seek(offset[, whence]) -> None. Move to new file position.\n"
|
---|
1862 | "\n"
|
---|
1863 | "Argument offset is a byte count. Optional argument whence defaults to\n"
|
---|
1864 | "0 (offset from start of file, offset should be >= 0); other values are 1\n"
|
---|
1865 | "(move relative to current position, positive or negative), and 2 (move\n"
|
---|
1866 | "relative to end of file, usually negative, although many platforms allow\n"
|
---|
1867 | "seeking beyond the end of a file). If the file is opened in text mode,\n"
|
---|
1868 | "only offsets returned by tell() are legal. Use of other offsets causes\n"
|
---|
1869 | "undefined behavior."
|
---|
1870 | "\n"
|
---|
1871 | "Note that not all file objects are seekable.");
|
---|
1872 |
|
---|
1873 | #ifdef HAVE_FTRUNCATE
|
---|
1874 | PyDoc_STRVAR(truncate_doc,
|
---|
1875 | "truncate([size]) -> None. Truncate the file to at most size bytes.\n"
|
---|
1876 | "\n"
|
---|
1877 | "Size defaults to the current file position, as returned by tell().");
|
---|
1878 | #endif
|
---|
1879 |
|
---|
1880 | PyDoc_STRVAR(tell_doc,
|
---|
1881 | "tell() -> current file position, an integer (may be a long integer).");
|
---|
1882 |
|
---|
1883 | PyDoc_STRVAR(readinto_doc,
|
---|
1884 | "readinto() -> Undocumented. Don't use this; it may go away.");
|
---|
1885 |
|
---|
1886 | PyDoc_STRVAR(readlines_doc,
|
---|
1887 | "readlines([size]) -> list of strings, each a line from the file.\n"
|
---|
1888 | "\n"
|
---|
1889 | "Call readline() repeatedly and return a list of the lines so read.\n"
|
---|
1890 | "The optional size argument, if given, is an approximate bound on the\n"
|
---|
1891 | "total number of bytes in the lines returned.");
|
---|
1892 |
|
---|
1893 | PyDoc_STRVAR(xreadlines_doc,
|
---|
1894 | "xreadlines() -> returns self.\n"
|
---|
1895 | "\n"
|
---|
1896 | "For backward compatibility. File objects now include the performance\n"
|
---|
1897 | "optimizations previously implemented in the xreadlines module.");
|
---|
1898 |
|
---|
1899 | PyDoc_STRVAR(writelines_doc,
|
---|
1900 | "writelines(sequence_of_strings) -> None. Write the strings to the file.\n"
|
---|
1901 | "\n"
|
---|
1902 | "Note that newlines are not added. The sequence can be any iterable object\n"
|
---|
1903 | "producing strings. This is equivalent to calling write() for each string.");
|
---|
1904 |
|
---|
1905 | PyDoc_STRVAR(flush_doc,
|
---|
1906 | "flush() -> None. Flush the internal I/O buffer.");
|
---|
1907 |
|
---|
1908 | PyDoc_STRVAR(close_doc,
|
---|
1909 | "close() -> None or (perhaps) an integer. Close the file.\n"
|
---|
1910 | "\n"
|
---|
1911 | "Sets data attribute .closed to True. A closed file cannot be used for\n"
|
---|
1912 | "further I/O operations. close() may be called more than once without\n"
|
---|
1913 | "error. Some kinds of file objects (for example, opened by popen())\n"
|
---|
1914 | "may return an exit status upon closing.");
|
---|
1915 |
|
---|
1916 | PyDoc_STRVAR(isatty_doc,
|
---|
1917 | "isatty() -> true or false. True if the file is connected to a tty device.");
|
---|
1918 |
|
---|
1919 | PyDoc_STRVAR(enter_doc,
|
---|
1920 | "__enter__() -> self.");
|
---|
1921 |
|
---|
1922 | PyDoc_STRVAR(exit_doc,
|
---|
1923 | "__exit__(*excinfo) -> None. Closes the file.");
|
---|
1924 |
|
---|
1925 | static PyMethodDef file_methods[] = {
|
---|
1926 | {"readline", (PyCFunction)file_readline, METH_VARARGS, readline_doc},
|
---|
1927 | {"read", (PyCFunction)file_read, METH_VARARGS, read_doc},
|
---|
1928 | {"write", (PyCFunction)file_write, METH_VARARGS, write_doc},
|
---|
1929 | {"fileno", (PyCFunction)file_fileno, METH_NOARGS, fileno_doc},
|
---|
1930 | {"seek", (PyCFunction)file_seek, METH_VARARGS, seek_doc},
|
---|
1931 | #ifdef HAVE_FTRUNCATE
|
---|
1932 | {"truncate", (PyCFunction)file_truncate, METH_VARARGS, truncate_doc},
|
---|
1933 | #endif
|
---|
1934 | {"tell", (PyCFunction)file_tell, METH_NOARGS, tell_doc},
|
---|
1935 | {"readinto", (PyCFunction)file_readinto, METH_VARARGS, readinto_doc},
|
---|
1936 | {"readlines", (PyCFunction)file_readlines, METH_VARARGS, readlines_doc},
|
---|
1937 | {"xreadlines",(PyCFunction)file_xreadlines, METH_NOARGS, xreadlines_doc},
|
---|
1938 | {"writelines",(PyCFunction)file_writelines, METH_O, writelines_doc},
|
---|
1939 | {"flush", (PyCFunction)file_flush, METH_NOARGS, flush_doc},
|
---|
1940 | {"close", (PyCFunction)file_close, METH_NOARGS, close_doc},
|
---|
1941 | {"isatty", (PyCFunction)file_isatty, METH_NOARGS, isatty_doc},
|
---|
1942 | {"__enter__", (PyCFunction)file_self, METH_NOARGS, enter_doc},
|
---|
1943 | {"__exit__", (PyCFunction)file_exit, METH_VARARGS, exit_doc},
|
---|
1944 | {NULL, NULL} /* sentinel */
|
---|
1945 | };
|
---|
1946 |
|
---|
1947 | #define OFF(x) offsetof(PyFileObject, x)
|
---|
1948 |
|
---|
1949 | static PyMemberDef file_memberlist[] = {
|
---|
1950 | {"mode", T_OBJECT, OFF(f_mode), RO,
|
---|
1951 | "file mode ('r', 'U', 'w', 'a', possibly with 'b' or '+' added)"},
|
---|
1952 | {"name", T_OBJECT, OFF(f_name), RO,
|
---|
1953 | "file name"},
|
---|
1954 | {"encoding", T_OBJECT, OFF(f_encoding), RO,
|
---|
1955 | "file encoding"},
|
---|
1956 | {"errors", T_OBJECT, OFF(f_errors), RO,
|
---|
1957 | "Unicode error handler"},
|
---|
1958 | /* getattr(f, "closed") is implemented without this table */
|
---|
1959 | {NULL} /* Sentinel */
|
---|
1960 | };
|
---|
1961 |
|
---|
1962 | static PyObject *
|
---|
1963 | get_closed(PyFileObject *f, void *closure)
|
---|
1964 | {
|
---|
1965 | return PyBool_FromLong((long)(f->f_fp == 0));
|
---|
1966 | }
|
---|
1967 | static PyObject *
|
---|
1968 | get_newlines(PyFileObject *f, void *closure)
|
---|
1969 | {
|
---|
1970 | switch (f->f_newlinetypes) {
|
---|
1971 | case NEWLINE_UNKNOWN:
|
---|
1972 | Py_INCREF(Py_None);
|
---|
1973 | return Py_None;
|
---|
1974 | case NEWLINE_CR:
|
---|
1975 | return PyString_FromString("\r");
|
---|
1976 | case NEWLINE_LF:
|
---|
1977 | return PyString_FromString("\n");
|
---|
1978 | case NEWLINE_CR|NEWLINE_LF:
|
---|
1979 | return Py_BuildValue("(ss)", "\r", "\n");
|
---|
1980 | case NEWLINE_CRLF:
|
---|
1981 | return PyString_FromString("\r\n");
|
---|
1982 | case NEWLINE_CR|NEWLINE_CRLF:
|
---|
1983 | return Py_BuildValue("(ss)", "\r", "\r\n");
|
---|
1984 | case NEWLINE_LF|NEWLINE_CRLF:
|
---|
1985 | return Py_BuildValue("(ss)", "\n", "\r\n");
|
---|
1986 | case NEWLINE_CR|NEWLINE_LF|NEWLINE_CRLF:
|
---|
1987 | return Py_BuildValue("(sss)", "\r", "\n", "\r\n");
|
---|
1988 | default:
|
---|
1989 | PyErr_Format(PyExc_SystemError,
|
---|
1990 | "Unknown newlines value 0x%x\n",
|
---|
1991 | f->f_newlinetypes);
|
---|
1992 | return NULL;
|
---|
1993 | }
|
---|
1994 | }
|
---|
1995 |
|
---|
1996 | static PyObject *
|
---|
1997 | get_softspace(PyFileObject *f, void *closure)
|
---|
1998 | {
|
---|
1999 | if (PyErr_WarnPy3k("file.softspace not supported in 3.x", 1) < 0)
|
---|
2000 | return NULL;
|
---|
2001 | return PyInt_FromLong(f->f_softspace);
|
---|
2002 | }
|
---|
2003 |
|
---|
2004 | static int
|
---|
2005 | set_softspace(PyFileObject *f, PyObject *value)
|
---|
2006 | {
|
---|
2007 | int new;
|
---|
2008 | if (PyErr_WarnPy3k("file.softspace not supported in 3.x", 1) < 0)
|
---|
2009 | return -1;
|
---|
2010 |
|
---|
2011 | if (value == NULL) {
|
---|
2012 | PyErr_SetString(PyExc_TypeError,
|
---|
2013 | "can't delete softspace attribute");
|
---|
2014 | return -1;
|
---|
2015 | }
|
---|
2016 |
|
---|
2017 | new = PyInt_AsLong(value);
|
---|
2018 | if (new == -1 && PyErr_Occurred())
|
---|
2019 | return -1;
|
---|
2020 | f->f_softspace = new;
|
---|
2021 | return 0;
|
---|
2022 | }
|
---|
2023 |
|
---|
2024 | static PyGetSetDef file_getsetlist[] = {
|
---|
2025 | {"closed", (getter)get_closed, NULL, "True if the file is closed"},
|
---|
2026 | {"newlines", (getter)get_newlines, NULL,
|
---|
2027 | "end-of-line convention used in this file"},
|
---|
2028 | {"softspace", (getter)get_softspace, (setter)set_softspace,
|
---|
2029 | "flag indicating that a space needs to be printed; used by print"},
|
---|
2030 | {0},
|
---|
2031 | };
|
---|
2032 |
|
---|
2033 | static void
|
---|
2034 | drop_readahead(PyFileObject *f)
|
---|
2035 | {
|
---|
2036 | if (f->f_buf != NULL) {
|
---|
2037 | PyMem_Free(f->f_buf);
|
---|
2038 | f->f_buf = NULL;
|
---|
2039 | }
|
---|
2040 | }
|
---|
2041 |
|
---|
2042 | /* Make sure that file has a readahead buffer with at least one byte
|
---|
2043 | (unless at EOF) and no more than bufsize. Returns negative value on
|
---|
2044 | error, will set MemoryError if bufsize bytes cannot be allocated. */
|
---|
2045 | static int
|
---|
2046 | readahead(PyFileObject *f, int bufsize)
|
---|
2047 | {
|
---|
2048 | Py_ssize_t chunksize;
|
---|
2049 |
|
---|
2050 | if (f->f_buf != NULL) {
|
---|
2051 | if( (f->f_bufend - f->f_bufptr) >= 1)
|
---|
2052 | return 0;
|
---|
2053 | else
|
---|
2054 | drop_readahead(f);
|
---|
2055 | }
|
---|
2056 | if ((f->f_buf = (char *)PyMem_Malloc(bufsize)) == NULL) {
|
---|
2057 | PyErr_NoMemory();
|
---|
2058 | return -1;
|
---|
2059 | }
|
---|
2060 | FILE_BEGIN_ALLOW_THREADS(f)
|
---|
2061 | errno = 0;
|
---|
2062 | chunksize = Py_UniversalNewlineFread(
|
---|
2063 | f->f_buf, bufsize, f->f_fp, (PyObject *)f);
|
---|
2064 | FILE_END_ALLOW_THREADS(f)
|
---|
2065 | if (chunksize == 0) {
|
---|
2066 | if (ferror(f->f_fp)) {
|
---|
2067 | PyErr_SetFromErrno(PyExc_IOError);
|
---|
2068 | clearerr(f->f_fp);
|
---|
2069 | drop_readahead(f);
|
---|
2070 | return -1;
|
---|
2071 | }
|
---|
2072 | }
|
---|
2073 | f->f_bufptr = f->f_buf;
|
---|
2074 | f->f_bufend = f->f_buf + chunksize;
|
---|
2075 | return 0;
|
---|
2076 | }
|
---|
2077 |
|
---|
2078 | /* Used by file_iternext. The returned string will start with 'skip'
|
---|
2079 | uninitialized bytes followed by the remainder of the line. Don't be
|
---|
2080 | horrified by the recursive call: maximum recursion depth is limited by
|
---|
2081 | logarithmic buffer growth to about 50 even when reading a 1gb line. */
|
---|
2082 |
|
---|
2083 | static PyStringObject *
|
---|
2084 | readahead_get_line_skip(PyFileObject *f, int skip, int bufsize)
|
---|
2085 | {
|
---|
2086 | PyStringObject* s;
|
---|
2087 | char *bufptr;
|
---|
2088 | char *buf;
|
---|
2089 | Py_ssize_t len;
|
---|
2090 |
|
---|
2091 | if (f->f_buf == NULL)
|
---|
2092 | if (readahead(f, bufsize) < 0)
|
---|
2093 | return NULL;
|
---|
2094 |
|
---|
2095 | len = f->f_bufend - f->f_bufptr;
|
---|
2096 | if (len == 0)
|
---|
2097 | return (PyStringObject *)
|
---|
2098 | PyString_FromStringAndSize(NULL, skip);
|
---|
2099 | bufptr = (char *)memchr(f->f_bufptr, '\n', len);
|
---|
2100 | if (bufptr != NULL) {
|
---|
2101 | bufptr++; /* Count the '\n' */
|
---|
2102 | len = bufptr - f->f_bufptr;
|
---|
2103 | s = (PyStringObject *)
|
---|
2104 | PyString_FromStringAndSize(NULL, skip+len);
|
---|
2105 | if (s == NULL)
|
---|
2106 | return NULL;
|
---|
2107 | memcpy(PyString_AS_STRING(s)+skip, f->f_bufptr, len);
|
---|
2108 | f->f_bufptr = bufptr;
|
---|
2109 | if (bufptr == f->f_bufend)
|
---|
2110 | drop_readahead(f);
|
---|
2111 | } else {
|
---|
2112 | bufptr = f->f_bufptr;
|
---|
2113 | buf = f->f_buf;
|
---|
2114 | f->f_buf = NULL; /* Force new readahead buffer */
|
---|
2115 | assert(skip+len < INT_MAX);
|
---|
2116 | s = readahead_get_line_skip(
|
---|
2117 | f, (int)(skip+len), bufsize + (bufsize>>2) );
|
---|
2118 | if (s == NULL) {
|
---|
2119 | PyMem_Free(buf);
|
---|
2120 | return NULL;
|
---|
2121 | }
|
---|
2122 | memcpy(PyString_AS_STRING(s)+skip, bufptr, len);
|
---|
2123 | PyMem_Free(buf);
|
---|
2124 | }
|
---|
2125 | return s;
|
---|
2126 | }
|
---|
2127 |
|
---|
2128 | /* A larger buffer size may actually decrease performance. */
|
---|
2129 | #define READAHEAD_BUFSIZE 8192
|
---|
2130 |
|
---|
2131 | static PyObject *
|
---|
2132 | file_iternext(PyFileObject *f)
|
---|
2133 | {
|
---|
2134 | PyStringObject* l;
|
---|
2135 |
|
---|
2136 | if (f->f_fp == NULL)
|
---|
2137 | return err_closed();
|
---|
2138 | if (!f->readable)
|
---|
2139 | return err_mode("reading");
|
---|
2140 |
|
---|
2141 | l = readahead_get_line_skip(f, 0, READAHEAD_BUFSIZE);
|
---|
2142 | if (l == NULL || PyString_GET_SIZE(l) == 0) {
|
---|
2143 | Py_XDECREF(l);
|
---|
2144 | return NULL;
|
---|
2145 | }
|
---|
2146 | return (PyObject *)l;
|
---|
2147 | }
|
---|
2148 |
|
---|
2149 |
|
---|
2150 | static PyObject *
|
---|
2151 | file_new(PyTypeObject *type, PyObject *args, PyObject *kwds)
|
---|
2152 | {
|
---|
2153 | PyObject *self;
|
---|
2154 | static PyObject *not_yet_string;
|
---|
2155 |
|
---|
2156 | assert(type != NULL && type->tp_alloc != NULL);
|
---|
2157 |
|
---|
2158 | if (not_yet_string == NULL) {
|
---|
2159 | not_yet_string = PyString_InternFromString("<uninitialized file>");
|
---|
2160 | if (not_yet_string == NULL)
|
---|
2161 | return NULL;
|
---|
2162 | }
|
---|
2163 |
|
---|
2164 | self = type->tp_alloc(type, 0);
|
---|
2165 | if (self != NULL) {
|
---|
2166 | /* Always fill in the name and mode, so that nobody else
|
---|
2167 | needs to special-case NULLs there. */
|
---|
2168 | Py_INCREF(not_yet_string);
|
---|
2169 | ((PyFileObject *)self)->f_name = not_yet_string;
|
---|
2170 | Py_INCREF(not_yet_string);
|
---|
2171 | ((PyFileObject *)self)->f_mode = not_yet_string;
|
---|
2172 | Py_INCREF(Py_None);
|
---|
2173 | ((PyFileObject *)self)->f_encoding = Py_None;
|
---|
2174 | Py_INCREF(Py_None);
|
---|
2175 | ((PyFileObject *)self)->f_errors = Py_None;
|
---|
2176 | ((PyFileObject *)self)->weakreflist = NULL;
|
---|
2177 | ((PyFileObject *)self)->unlocked_count = 0;
|
---|
2178 | }
|
---|
2179 | return self;
|
---|
2180 | }
|
---|
2181 |
|
---|
2182 | static int
|
---|
2183 | file_init(PyObject *self, PyObject *args, PyObject *kwds)
|
---|
2184 | {
|
---|
2185 | PyFileObject *foself = (PyFileObject *)self;
|
---|
2186 | int ret = 0;
|
---|
2187 | static char *kwlist[] = {"name", "mode", "buffering", 0};
|
---|
2188 | char *name = NULL;
|
---|
2189 | char *mode = "r";
|
---|
2190 | int bufsize = -1;
|
---|
2191 | int wideargument = 0;
|
---|
2192 |
|
---|
2193 | assert(PyFile_Check(self));
|
---|
2194 | if (foself->f_fp != NULL) {
|
---|
2195 | /* Have to close the existing file first. */
|
---|
2196 | PyObject *closeresult = file_close(foself);
|
---|
2197 | if (closeresult == NULL)
|
---|
2198 | return -1;
|
---|
2199 | Py_DECREF(closeresult);
|
---|
2200 | }
|
---|
2201 |
|
---|
2202 | #ifdef Py_WIN_WIDE_FILENAMES
|
---|
2203 | if (GetVersion() < 0x80000000) { /* On NT, so wide API available */
|
---|
2204 | PyObject *po;
|
---|
2205 | if (PyArg_ParseTupleAndKeywords(args, kwds, "U|si:file",
|
---|
2206 | kwlist, &po, &mode, &bufsize)) {
|
---|
2207 | wideargument = 1;
|
---|
2208 | if (fill_file_fields(foself, NULL, po, mode,
|
---|
2209 | fclose) == NULL)
|
---|
2210 | goto Error;
|
---|
2211 | } else {
|
---|
2212 | /* Drop the argument parsing error as narrow
|
---|
2213 | strings are also valid. */
|
---|
2214 | PyErr_Clear();
|
---|
2215 | }
|
---|
2216 | }
|
---|
2217 | #endif
|
---|
2218 |
|
---|
2219 | if (!wideargument) {
|
---|
2220 | PyObject *o_name;
|
---|
2221 |
|
---|
2222 | if (!PyArg_ParseTupleAndKeywords(args, kwds, "et|si:file", kwlist,
|
---|
2223 | Py_FileSystemDefaultEncoding,
|
---|
2224 | &name,
|
---|
2225 | &mode, &bufsize))
|
---|
2226 | return -1;
|
---|
2227 |
|
---|
2228 | /* We parse again to get the name as a PyObject */
|
---|
2229 | if (!PyArg_ParseTupleAndKeywords(args, kwds, "O|si:file",
|
---|
2230 | kwlist, &o_name, &mode,
|
---|
2231 | &bufsize))
|
---|
2232 | goto Error;
|
---|
2233 |
|
---|
2234 | if (fill_file_fields(foself, NULL, o_name, mode,
|
---|
2235 | fclose) == NULL)
|
---|
2236 | goto Error;
|
---|
2237 | }
|
---|
2238 | if (open_the_file(foself, name, mode) == NULL)
|
---|
2239 | goto Error;
|
---|
2240 | foself->f_setbuf = NULL;
|
---|
2241 | PyFile_SetBufSize(self, bufsize);
|
---|
2242 | goto Done;
|
---|
2243 |
|
---|
2244 | Error:
|
---|
2245 | ret = -1;
|
---|
2246 | /* fall through */
|
---|
2247 | Done:
|
---|
2248 | PyMem_Free(name); /* free the encoded string */
|
---|
2249 | return ret;
|
---|
2250 | }
|
---|
2251 |
|
---|
2252 | PyDoc_VAR(file_doc) =
|
---|
2253 | PyDoc_STR(
|
---|
2254 | "file(name[, mode[, buffering]]) -> file object\n"
|
---|
2255 | "\n"
|
---|
2256 | "Open a file. The mode can be 'r', 'w' or 'a' for reading (default),\n"
|
---|
2257 | "writing or appending. The file will be created if it doesn't exist\n"
|
---|
2258 | "when opened for writing or appending; it will be truncated when\n"
|
---|
2259 | "opened for writing. Add a 'b' to the mode for binary files.\n"
|
---|
2260 | "Add a '+' to the mode to allow simultaneous reading and writing.\n"
|
---|
2261 | "If the buffering argument is given, 0 means unbuffered, 1 means line\n"
|
---|
2262 | "buffered, and larger numbers specify the buffer size. The preferred way\n"
|
---|
2263 | "to open a file is with the builtin open() function.\n"
|
---|
2264 | )
|
---|
2265 | PyDoc_STR(
|
---|
2266 | "Add a 'U' to mode to open the file for input with universal newline\n"
|
---|
2267 | "support. Any line ending in the input file will be seen as a '\\n'\n"
|
---|
2268 | "in Python. Also, a file so opened gains the attribute 'newlines';\n"
|
---|
2269 | "the value for this attribute is one of None (no newline read yet),\n"
|
---|
2270 | "'\\r', '\\n', '\\r\\n' or a tuple containing all the newline types seen.\n"
|
---|
2271 | "\n"
|
---|
2272 | "'U' cannot be combined with 'w' or '+' mode.\n"
|
---|
2273 | );
|
---|
2274 |
|
---|
2275 | PyTypeObject PyFile_Type = {
|
---|
2276 | PyVarObject_HEAD_INIT(&PyType_Type, 0)
|
---|
2277 | "file",
|
---|
2278 | sizeof(PyFileObject),
|
---|
2279 | 0,
|
---|
2280 | (destructor)file_dealloc, /* tp_dealloc */
|
---|
2281 | 0, /* tp_print */
|
---|
2282 | 0, /* tp_getattr */
|
---|
2283 | 0, /* tp_setattr */
|
---|
2284 | 0, /* tp_compare */
|
---|
2285 | (reprfunc)file_repr, /* tp_repr */
|
---|
2286 | 0, /* tp_as_number */
|
---|
2287 | 0, /* tp_as_sequence */
|
---|
2288 | 0, /* tp_as_mapping */
|
---|
2289 | 0, /* tp_hash */
|
---|
2290 | 0, /* tp_call */
|
---|
2291 | 0, /* tp_str */
|
---|
2292 | PyObject_GenericGetAttr, /* tp_getattro */
|
---|
2293 | /* softspace is writable: we must supply tp_setattro */
|
---|
2294 | PyObject_GenericSetAttr, /* tp_setattro */
|
---|
2295 | 0, /* tp_as_buffer */
|
---|
2296 | Py_TPFLAGS_DEFAULT | Py_TPFLAGS_BASETYPE | Py_TPFLAGS_HAVE_WEAKREFS, /* tp_flags */
|
---|
2297 | file_doc, /* tp_doc */
|
---|
2298 | 0, /* tp_traverse */
|
---|
2299 | 0, /* tp_clear */
|
---|
2300 | 0, /* tp_richcompare */
|
---|
2301 | offsetof(PyFileObject, weakreflist), /* tp_weaklistoffset */
|
---|
2302 | (getiterfunc)file_self, /* tp_iter */
|
---|
2303 | (iternextfunc)file_iternext, /* tp_iternext */
|
---|
2304 | file_methods, /* tp_methods */
|
---|
2305 | file_memberlist, /* tp_members */
|
---|
2306 | file_getsetlist, /* tp_getset */
|
---|
2307 | 0, /* tp_base */
|
---|
2308 | 0, /* tp_dict */
|
---|
2309 | 0, /* tp_descr_get */
|
---|
2310 | 0, /* tp_descr_set */
|
---|
2311 | 0, /* tp_dictoffset */
|
---|
2312 | file_init, /* tp_init */
|
---|
2313 | PyType_GenericAlloc, /* tp_alloc */
|
---|
2314 | file_new, /* tp_new */
|
---|
2315 | PyObject_Del, /* tp_free */
|
---|
2316 | };
|
---|
2317 |
|
---|
2318 | /* Interface for the 'soft space' between print items. */
|
---|
2319 |
|
---|
2320 | int
|
---|
2321 | PyFile_SoftSpace(PyObject *f, int newflag)
|
---|
2322 | {
|
---|
2323 | long oldflag = 0;
|
---|
2324 | if (f == NULL) {
|
---|
2325 | /* Do nothing */
|
---|
2326 | }
|
---|
2327 | else if (PyFile_Check(f)) {
|
---|
2328 | oldflag = ((PyFileObject *)f)->f_softspace;
|
---|
2329 | ((PyFileObject *)f)->f_softspace = newflag;
|
---|
2330 | }
|
---|
2331 | else {
|
---|
2332 | PyObject *v;
|
---|
2333 | v = PyObject_GetAttrString(f, "softspace");
|
---|
2334 | if (v == NULL)
|
---|
2335 | PyErr_Clear();
|
---|
2336 | else {
|
---|
2337 | if (PyInt_Check(v))
|
---|
2338 | oldflag = PyInt_AsLong(v);
|
---|
2339 | assert(oldflag < INT_MAX);
|
---|
2340 | Py_DECREF(v);
|
---|
2341 | }
|
---|
2342 | v = PyInt_FromLong((long)newflag);
|
---|
2343 | if (v == NULL)
|
---|
2344 | PyErr_Clear();
|
---|
2345 | else {
|
---|
2346 | if (PyObject_SetAttrString(f, "softspace", v) != 0)
|
---|
2347 | PyErr_Clear();
|
---|
2348 | Py_DECREF(v);
|
---|
2349 | }
|
---|
2350 | }
|
---|
2351 | return (int)oldflag;
|
---|
2352 | }
|
---|
2353 |
|
---|
2354 | /* Interfaces to write objects/strings to file-like objects */
|
---|
2355 |
|
---|
2356 | int
|
---|
2357 | PyFile_WriteObject(PyObject *v, PyObject *f, int flags)
|
---|
2358 | {
|
---|
2359 | PyObject *writer, *value, *args, *result;
|
---|
2360 | if (f == NULL) {
|
---|
2361 | PyErr_SetString(PyExc_TypeError, "writeobject with NULL file");
|
---|
2362 | return -1;
|
---|
2363 | }
|
---|
2364 | else if (PyFile_Check(f)) {
|
---|
2365 | PyFileObject *fobj = (PyFileObject *) f;
|
---|
2366 | #ifdef Py_USING_UNICODE
|
---|
2367 | PyObject *enc = fobj->f_encoding;
|
---|
2368 | int result;
|
---|
2369 | #endif
|
---|
2370 | if (fobj->f_fp == NULL) {
|
---|
2371 | err_closed();
|
---|
2372 | return -1;
|
---|
2373 | }
|
---|
2374 | #ifdef Py_USING_UNICODE
|
---|
2375 | if ((flags & Py_PRINT_RAW) &&
|
---|
2376 | PyUnicode_Check(v) && enc != Py_None) {
|
---|
2377 | char *cenc = PyString_AS_STRING(enc);
|
---|
2378 | char *errors = fobj->f_errors == Py_None ?
|
---|
2379 | "strict" : PyString_AS_STRING(fobj->f_errors);
|
---|
2380 | value = PyUnicode_AsEncodedString(v, cenc, errors);
|
---|
2381 | if (value == NULL)
|
---|
2382 | return -1;
|
---|
2383 | } else {
|
---|
2384 | value = v;
|
---|
2385 | Py_INCREF(value);
|
---|
2386 | }
|
---|
2387 | result = file_PyObject_Print(value, fobj, flags);
|
---|
2388 | Py_DECREF(value);
|
---|
2389 | return result;
|
---|
2390 | #else
|
---|
2391 | return file_PyObject_Print(v, fobj, flags);
|
---|
2392 | #endif
|
---|
2393 | }
|
---|
2394 | writer = PyObject_GetAttrString(f, "write");
|
---|
2395 | if (writer == NULL)
|
---|
2396 | return -1;
|
---|
2397 | if (flags & Py_PRINT_RAW) {
|
---|
2398 | if (PyUnicode_Check(v)) {
|
---|
2399 | value = v;
|
---|
2400 | Py_INCREF(value);
|
---|
2401 | } else
|
---|
2402 | value = PyObject_Str(v);
|
---|
2403 | }
|
---|
2404 | else
|
---|
2405 | value = PyObject_Repr(v);
|
---|
2406 | if (value == NULL) {
|
---|
2407 | Py_DECREF(writer);
|
---|
2408 | return -1;
|
---|
2409 | }
|
---|
2410 | args = PyTuple_Pack(1, value);
|
---|
2411 | if (args == NULL) {
|
---|
2412 | Py_DECREF(value);
|
---|
2413 | Py_DECREF(writer);
|
---|
2414 | return -1;
|
---|
2415 | }
|
---|
2416 | result = PyEval_CallObject(writer, args);
|
---|
2417 | Py_DECREF(args);
|
---|
2418 | Py_DECREF(value);
|
---|
2419 | Py_DECREF(writer);
|
---|
2420 | if (result == NULL)
|
---|
2421 | return -1;
|
---|
2422 | Py_DECREF(result);
|
---|
2423 | return 0;
|
---|
2424 | }
|
---|
2425 |
|
---|
2426 | int
|
---|
2427 | PyFile_WriteString(const char *s, PyObject *f)
|
---|
2428 | {
|
---|
2429 |
|
---|
2430 | if (f == NULL) {
|
---|
2431 | /* Should be caused by a pre-existing error */
|
---|
2432 | if (!PyErr_Occurred())
|
---|
2433 | PyErr_SetString(PyExc_SystemError,
|
---|
2434 | "null file for PyFile_WriteString");
|
---|
2435 | return -1;
|
---|
2436 | }
|
---|
2437 | else if (PyFile_Check(f)) {
|
---|
2438 | PyFileObject *fobj = (PyFileObject *) f;
|
---|
2439 | FILE *fp = PyFile_AsFile(f);
|
---|
2440 | if (fp == NULL) {
|
---|
2441 | err_closed();
|
---|
2442 | return -1;
|
---|
2443 | }
|
---|
2444 | FILE_BEGIN_ALLOW_THREADS(fobj)
|
---|
2445 | fputs(s, fp);
|
---|
2446 | FILE_END_ALLOW_THREADS(fobj)
|
---|
2447 | return 0;
|
---|
2448 | }
|
---|
2449 | else if (!PyErr_Occurred()) {
|
---|
2450 | PyObject *v = PyString_FromString(s);
|
---|
2451 | int err;
|
---|
2452 | if (v == NULL)
|
---|
2453 | return -1;
|
---|
2454 | err = PyFile_WriteObject(v, f, Py_PRINT_RAW);
|
---|
2455 | Py_DECREF(v);
|
---|
2456 | return err;
|
---|
2457 | }
|
---|
2458 | else
|
---|
2459 | return -1;
|
---|
2460 | }
|
---|
2461 |
|
---|
2462 | /* Try to get a file-descriptor from a Python object. If the object
|
---|
2463 | is an integer or long integer, its value is returned. If not, the
|
---|
2464 | object's fileno() method is called if it exists; the method must return
|
---|
2465 | an integer or long integer, which is returned as the file descriptor value.
|
---|
2466 | -1 is returned on failure.
|
---|
2467 | */
|
---|
2468 |
|
---|
2469 | int PyObject_AsFileDescriptor(PyObject *o)
|
---|
2470 | {
|
---|
2471 | int fd;
|
---|
2472 | PyObject *meth;
|
---|
2473 |
|
---|
2474 | if (PyInt_Check(o)) {
|
---|
2475 | fd = PyInt_AsLong(o);
|
---|
2476 | }
|
---|
2477 | else if (PyLong_Check(o)) {
|
---|
2478 | fd = PyLong_AsLong(o);
|
---|
2479 | }
|
---|
2480 | else if ((meth = PyObject_GetAttrString(o, "fileno")) != NULL)
|
---|
2481 | {
|
---|
2482 | PyObject *fno = PyEval_CallObject(meth, NULL);
|
---|
2483 | Py_DECREF(meth);
|
---|
2484 | if (fno == NULL)
|
---|
2485 | return -1;
|
---|
2486 |
|
---|
2487 | if (PyInt_Check(fno)) {
|
---|
2488 | fd = PyInt_AsLong(fno);
|
---|
2489 | Py_DECREF(fno);
|
---|
2490 | }
|
---|
2491 | else if (PyLong_Check(fno)) {
|
---|
2492 | fd = PyLong_AsLong(fno);
|
---|
2493 | Py_DECREF(fno);
|
---|
2494 | }
|
---|
2495 | else {
|
---|
2496 | PyErr_SetString(PyExc_TypeError,
|
---|
2497 | "fileno() returned a non-integer");
|
---|
2498 | Py_DECREF(fno);
|
---|
2499 | return -1;
|
---|
2500 | }
|
---|
2501 | }
|
---|
2502 | else {
|
---|
2503 | PyErr_SetString(PyExc_TypeError,
|
---|
2504 | "argument must be an int, or have a fileno() method.");
|
---|
2505 | return -1;
|
---|
2506 | }
|
---|
2507 |
|
---|
2508 | if (fd < 0) {
|
---|
2509 | PyErr_Format(PyExc_ValueError,
|
---|
2510 | "file descriptor cannot be a negative integer (%i)",
|
---|
2511 | fd);
|
---|
2512 | return -1;
|
---|
2513 | }
|
---|
2514 | return fd;
|
---|
2515 | }
|
---|
2516 |
|
---|
2517 | /* From here on we need access to the real fgets and fread */
|
---|
2518 | #undef fgets
|
---|
2519 | #undef fread
|
---|
2520 |
|
---|
2521 | /*
|
---|
2522 | ** Py_UniversalNewlineFgets is an fgets variation that understands
|
---|
2523 | ** all of \r, \n and \r\n conventions.
|
---|
2524 | ** The stream should be opened in binary mode.
|
---|
2525 | ** If fobj is NULL the routine always does newline conversion, and
|
---|
2526 | ** it may peek one char ahead to gobble the second char in \r\n.
|
---|
2527 | ** If fobj is non-NULL it must be a PyFileObject. In this case there
|
---|
2528 | ** is no readahead but in stead a flag is used to skip a following
|
---|
2529 | ** \n on the next read. Also, if the file is open in binary mode
|
---|
2530 | ** the whole conversion is skipped. Finally, the routine keeps track of
|
---|
2531 | ** the different types of newlines seen.
|
---|
2532 | ** Note that we need no error handling: fgets() treats error and eof
|
---|
2533 | ** identically.
|
---|
2534 | */
|
---|
2535 | char *
|
---|
2536 | Py_UniversalNewlineFgets(char *buf, int n, FILE *stream, PyObject *fobj)
|
---|
2537 | {
|
---|
2538 | char *p = buf;
|
---|
2539 | int c;
|
---|
2540 | int newlinetypes = 0;
|
---|
2541 | int skipnextlf = 0;
|
---|
2542 | int univ_newline = 1;
|
---|
2543 |
|
---|
2544 | if (fobj) {
|
---|
2545 | if (!PyFile_Check(fobj)) {
|
---|
2546 | errno = ENXIO; /* What can you do... */
|
---|
2547 | return NULL;
|
---|
2548 | }
|
---|
2549 | univ_newline = ((PyFileObject *)fobj)->f_univ_newline;
|
---|
2550 | if ( !univ_newline )
|
---|
2551 | return fgets(buf, n, stream);
|
---|
2552 | newlinetypes = ((PyFileObject *)fobj)->f_newlinetypes;
|
---|
2553 | skipnextlf = ((PyFileObject *)fobj)->f_skipnextlf;
|
---|
2554 | }
|
---|
2555 | FLOCKFILE(stream);
|
---|
2556 | c = 'x'; /* Shut up gcc warning */
|
---|
2557 | while (--n > 0 && (c = GETC(stream)) != EOF ) {
|
---|
2558 | if (skipnextlf ) {
|
---|
2559 | skipnextlf = 0;
|
---|
2560 | if (c == '\n') {
|
---|
2561 | /* Seeing a \n here with skipnextlf true
|
---|
2562 | ** means we saw a \r before.
|
---|
2563 | */
|
---|
2564 | newlinetypes |= NEWLINE_CRLF;
|
---|
2565 | c = GETC(stream);
|
---|
2566 | if (c == EOF) break;
|
---|
2567 | } else {
|
---|
2568 | /*
|
---|
2569 | ** Note that c == EOF also brings us here,
|
---|
2570 | ** so we're okay if the last char in the file
|
---|
2571 | ** is a CR.
|
---|
2572 | */
|
---|
2573 | newlinetypes |= NEWLINE_CR;
|
---|
2574 | }
|
---|
2575 | }
|
---|
2576 | if (c == '\r') {
|
---|
2577 | /* A \r is translated into a \n, and we skip
|
---|
2578 | ** an adjacent \n, if any. We don't set the
|
---|
2579 | ** newlinetypes flag until we've seen the next char.
|
---|
2580 | */
|
---|
2581 | skipnextlf = 1;
|
---|
2582 | c = '\n';
|
---|
2583 | } else if ( c == '\n') {
|
---|
2584 | newlinetypes |= NEWLINE_LF;
|
---|
2585 | }
|
---|
2586 | *p++ = c;
|
---|
2587 | if (c == '\n') break;
|
---|
2588 | }
|
---|
2589 | if ( c == EOF && skipnextlf )
|
---|
2590 | newlinetypes |= NEWLINE_CR;
|
---|
2591 | FUNLOCKFILE(stream);
|
---|
2592 | *p = '\0';
|
---|
2593 | if (fobj) {
|
---|
2594 | ((PyFileObject *)fobj)->f_newlinetypes = newlinetypes;
|
---|
2595 | ((PyFileObject *)fobj)->f_skipnextlf = skipnextlf;
|
---|
2596 | } else if ( skipnextlf ) {
|
---|
2597 | /* If we have no file object we cannot save the
|
---|
2598 | ** skipnextlf flag. We have to readahead, which
|
---|
2599 | ** will cause a pause if we're reading from an
|
---|
2600 | ** interactive stream, but that is very unlikely
|
---|
2601 | ** unless we're doing something silly like
|
---|
2602 | ** execfile("/dev/tty").
|
---|
2603 | */
|
---|
2604 | c = GETC(stream);
|
---|
2605 | if ( c != '\n' )
|
---|
2606 | ungetc(c, stream);
|
---|
2607 | }
|
---|
2608 | if (p == buf)
|
---|
2609 | return NULL;
|
---|
2610 | return buf;
|
---|
2611 | }
|
---|
2612 |
|
---|
2613 | /*
|
---|
2614 | ** Py_UniversalNewlineFread is an fread variation that understands
|
---|
2615 | ** all of \r, \n and \r\n conventions.
|
---|
2616 | ** The stream should be opened in binary mode.
|
---|
2617 | ** fobj must be a PyFileObject. In this case there
|
---|
2618 | ** is no readahead but in stead a flag is used to skip a following
|
---|
2619 | ** \n on the next read. Also, if the file is open in binary mode
|
---|
2620 | ** the whole conversion is skipped. Finally, the routine keeps track of
|
---|
2621 | ** the different types of newlines seen.
|
---|
2622 | */
|
---|
2623 | size_t
|
---|
2624 | Py_UniversalNewlineFread(char *buf, size_t n,
|
---|
2625 | FILE *stream, PyObject *fobj)
|
---|
2626 | {
|
---|
2627 | char *dst = buf;
|
---|
2628 | PyFileObject *f = (PyFileObject *)fobj;
|
---|
2629 | int newlinetypes, skipnextlf;
|
---|
2630 |
|
---|
2631 | assert(buf != NULL);
|
---|
2632 | assert(stream != NULL);
|
---|
2633 |
|
---|
2634 | if (!fobj || !PyFile_Check(fobj)) {
|
---|
2635 | errno = ENXIO; /* What can you do... */
|
---|
2636 | return 0;
|
---|
2637 | }
|
---|
2638 | if (!f->f_univ_newline)
|
---|
2639 | return fread(buf, 1, n, stream);
|
---|
2640 | newlinetypes = f->f_newlinetypes;
|
---|
2641 | skipnextlf = f->f_skipnextlf;
|
---|
2642 | /* Invariant: n is the number of bytes remaining to be filled
|
---|
2643 | * in the buffer.
|
---|
2644 | */
|
---|
2645 | while (n) {
|
---|
2646 | size_t nread;
|
---|
2647 | int shortread;
|
---|
2648 | char *src = dst;
|
---|
2649 |
|
---|
2650 | nread = fread(dst, 1, n, stream);
|
---|
2651 | assert(nread <= n);
|
---|
2652 | if (nread == 0)
|
---|
2653 | break;
|
---|
2654 |
|
---|
2655 | n -= nread; /* assuming 1 byte out for each in; will adjust */
|
---|
2656 | shortread = n != 0; /* true iff EOF or error */
|
---|
2657 | while (nread--) {
|
---|
2658 | char c = *src++;
|
---|
2659 | if (c == '\r') {
|
---|
2660 | /* Save as LF and set flag to skip next LF. */
|
---|
2661 | *dst++ = '\n';
|
---|
2662 | skipnextlf = 1;
|
---|
2663 | }
|
---|
2664 | else if (skipnextlf && c == '\n') {
|
---|
2665 | /* Skip LF, and remember we saw CR LF. */
|
---|
2666 | skipnextlf = 0;
|
---|
2667 | newlinetypes |= NEWLINE_CRLF;
|
---|
2668 | ++n;
|
---|
2669 | }
|
---|
2670 | else {
|
---|
2671 | /* Normal char to be stored in buffer. Also
|
---|
2672 | * update the newlinetypes flag if either this
|
---|
2673 | * is an LF or the previous char was a CR.
|
---|
2674 | */
|
---|
2675 | if (c == '\n')
|
---|
2676 | newlinetypes |= NEWLINE_LF;
|
---|
2677 | else if (skipnextlf)
|
---|
2678 | newlinetypes |= NEWLINE_CR;
|
---|
2679 | *dst++ = c;
|
---|
2680 | skipnextlf = 0;
|
---|
2681 | }
|
---|
2682 | }
|
---|
2683 | if (shortread) {
|
---|
2684 | /* If this is EOF, update type flags. */
|
---|
2685 | if (skipnextlf && feof(stream))
|
---|
2686 | newlinetypes |= NEWLINE_CR;
|
---|
2687 | break;
|
---|
2688 | }
|
---|
2689 | }
|
---|
2690 | f->f_newlinetypes = newlinetypes;
|
---|
2691 | f->f_skipnextlf = skipnextlf;
|
---|
2692 | return dst - buf;
|
---|
2693 | }
|
---|
2694 |
|
---|
2695 | #ifdef __cplusplus
|
---|
2696 | }
|
---|
2697 | #endif
|
---|