Antoine Pitrou | 1969059 | 2009-06-12 20:14:08 +0000 | [diff] [blame] | 1 | #define PY_SSIZE_T_CLEAN |
| 2 | #include "Python.h" |
| 3 | #include "structmember.h" |
| 4 | #include "_iomodule.h" |
| 5 | |
| 6 | /* Implementation note: the buffer is always at least one character longer |
| 7 | than the enclosed string, for proper functioning of _PyIO_find_line_ending. |
| 8 | */ |
| 9 | |
| 10 | typedef struct { |
| 11 | PyObject_HEAD |
| 12 | Py_UNICODE *buf; |
| 13 | Py_ssize_t pos; |
| 14 | Py_ssize_t string_size; |
| 15 | size_t buf_size; |
| 16 | |
| 17 | char ok; /* initialized? */ |
| 18 | char closed; |
| 19 | char readuniversal; |
| 20 | char readtranslate; |
| 21 | PyObject *decoder; |
| 22 | PyObject *readnl; |
| 23 | PyObject *writenl; |
| 24 | |
| 25 | PyObject *dict; |
| 26 | PyObject *weakreflist; |
| 27 | } stringio; |
| 28 | |
| 29 | #define CHECK_INITIALIZED(self) \ |
| 30 | if (self->ok <= 0) { \ |
| 31 | PyErr_SetString(PyExc_ValueError, \ |
| 32 | "I/O operation on uninitialized object"); \ |
| 33 | return NULL; \ |
| 34 | } |
| 35 | |
| 36 | #define CHECK_CLOSED(self) \ |
| 37 | if (self->closed) { \ |
| 38 | PyErr_SetString(PyExc_ValueError, \ |
| 39 | "I/O operation on closed file"); \ |
| 40 | return NULL; \ |
| 41 | } |
| 42 | |
| 43 | PyDoc_STRVAR(stringio_doc, |
| 44 | "Text I/O implementation using an in-memory buffer.\n" |
| 45 | "\n" |
| 46 | "The initial_value argument sets the value of object. The newline\n" |
| 47 | "argument is like the one of TextIOWrapper's constructor."); |
| 48 | |
| 49 | |
| 50 | /* Internal routine for changing the size, in terms of characters, of the |
| 51 | buffer of StringIO objects. The caller should ensure that the 'size' |
| 52 | argument is non-negative. Returns 0 on success, -1 otherwise. */ |
| 53 | static int |
| 54 | resize_buffer(stringio *self, size_t size) |
| 55 | { |
| 56 | /* Here, unsigned types are used to avoid dealing with signed integer |
| 57 | overflow, which is undefined in C. */ |
| 58 | size_t alloc = self->buf_size; |
| 59 | Py_UNICODE *new_buf = NULL; |
| 60 | |
| 61 | assert(self->buf != NULL); |
| 62 | |
| 63 | /* Reserve one more char for line ending detection. */ |
| 64 | size = size + 1; |
| 65 | /* For simplicity, stay in the range of the signed type. Anyway, Python |
| 66 | doesn't allow strings to be longer than this. */ |
| 67 | if (size > PY_SSIZE_T_MAX) |
| 68 | goto overflow; |
| 69 | |
| 70 | if (size < alloc / 2) { |
| 71 | /* Major downsize; resize down to exact size. */ |
| 72 | alloc = size + 1; |
| 73 | } |
| 74 | else if (size < alloc) { |
| 75 | /* Within allocated size; quick exit */ |
| 76 | return 0; |
| 77 | } |
| 78 | else if (size <= alloc * 1.125) { |
| 79 | /* Moderate upsize; overallocate similar to list_resize() */ |
| 80 | alloc = size + (size >> 3) + (size < 9 ? 3 : 6); |
| 81 | } |
| 82 | else { |
| 83 | /* Major upsize; resize up to exact size */ |
| 84 | alloc = size + 1; |
| 85 | } |
| 86 | |
| 87 | if (alloc > ((size_t)-1) / sizeof(Py_UNICODE)) |
| 88 | goto overflow; |
| 89 | new_buf = (Py_UNICODE *)PyMem_Realloc(self->buf, |
| 90 | alloc * sizeof(Py_UNICODE)); |
| 91 | if (new_buf == NULL) { |
| 92 | PyErr_NoMemory(); |
| 93 | return -1; |
| 94 | } |
| 95 | self->buf_size = alloc; |
| 96 | self->buf = new_buf; |
| 97 | |
| 98 | return 0; |
| 99 | |
| 100 | overflow: |
| 101 | PyErr_SetString(PyExc_OverflowError, |
| 102 | "new buffer size too large"); |
| 103 | return -1; |
| 104 | } |
| 105 | |
| 106 | /* Internal routine for writing a whole PyUnicode object to the buffer of a |
| 107 | StringIO object. Returns 0 on success, or -1 on error. */ |
| 108 | static Py_ssize_t |
| 109 | write_str(stringio *self, PyObject *obj) |
| 110 | { |
| 111 | Py_UNICODE *str; |
| 112 | Py_ssize_t len; |
| 113 | PyObject *decoded = NULL; |
| 114 | assert(self->buf != NULL); |
| 115 | assert(self->pos >= 0); |
| 116 | |
| 117 | if (self->decoder != NULL) { |
| 118 | decoded = _PyIncrementalNewlineDecoder_decode( |
| 119 | self->decoder, obj, 1 /* always final */); |
| 120 | } |
| 121 | else { |
| 122 | decoded = obj; |
| 123 | Py_INCREF(decoded); |
| 124 | } |
| 125 | if (self->writenl) { |
| 126 | PyObject *translated = PyUnicode_Replace( |
| 127 | decoded, _PyIO_str_nl, self->writenl, -1); |
| 128 | Py_DECREF(decoded); |
| 129 | decoded = translated; |
| 130 | } |
| 131 | if (decoded == NULL) |
| 132 | return -1; |
| 133 | |
| 134 | assert(PyUnicode_Check(decoded)); |
| 135 | str = PyUnicode_AS_UNICODE(decoded); |
| 136 | len = PyUnicode_GET_SIZE(decoded); |
| 137 | |
| 138 | assert(len >= 0); |
| 139 | |
| 140 | /* This overflow check is not strictly necessary. However, it avoids us to |
| 141 | deal with funky things like comparing an unsigned and a signed |
| 142 | integer. */ |
| 143 | if (self->pos > PY_SSIZE_T_MAX - len) { |
| 144 | PyErr_SetString(PyExc_OverflowError, |
| 145 | "new position too large"); |
| 146 | goto fail; |
| 147 | } |
| 148 | if (self->pos + len > self->string_size) { |
| 149 | if (resize_buffer(self, self->pos + len) < 0) |
| 150 | goto fail; |
| 151 | } |
| 152 | |
| 153 | if (self->pos > self->string_size) { |
| 154 | /* In case of overseek, pad with null bytes the buffer region between |
| 155 | the end of stream and the current position. |
| 156 | |
| 157 | 0 lo string_size hi |
| 158 | | |<---used--->|<----------available----------->| |
| 159 | | | <--to pad-->|<---to write---> | |
Ezio Melotti | c2077b0 | 2011-03-16 12:34:31 +0200 | [diff] [blame] | 160 | 0 buf position |
Antoine Pitrou | 1969059 | 2009-06-12 20:14:08 +0000 | [diff] [blame] | 161 | |
| 162 | */ |
| 163 | memset(self->buf + self->string_size, '\0', |
| 164 | (self->pos - self->string_size) * sizeof(Py_UNICODE)); |
| 165 | } |
| 166 | |
| 167 | /* Copy the data to the internal buffer, overwriting some of the |
| 168 | existing data if self->pos < self->string_size. */ |
| 169 | memcpy(self->buf + self->pos, str, len * sizeof(Py_UNICODE)); |
| 170 | self->pos += len; |
| 171 | |
| 172 | /* Set the new length of the internal string if it has changed. */ |
| 173 | if (self->string_size < self->pos) { |
| 174 | self->string_size = self->pos; |
| 175 | } |
| 176 | |
| 177 | Py_DECREF(decoded); |
| 178 | return 0; |
| 179 | |
| 180 | fail: |
| 181 | Py_XDECREF(decoded); |
| 182 | return -1; |
| 183 | } |
| 184 | |
| 185 | PyDoc_STRVAR(stringio_getvalue_doc, |
| 186 | "Retrieve the entire contents of the object."); |
| 187 | |
| 188 | static PyObject * |
| 189 | stringio_getvalue(stringio *self) |
| 190 | { |
| 191 | CHECK_INITIALIZED(self); |
| 192 | CHECK_CLOSED(self); |
| 193 | return PyUnicode_FromUnicode(self->buf, self->string_size); |
| 194 | } |
| 195 | |
| 196 | PyDoc_STRVAR(stringio_tell_doc, |
| 197 | "Tell the current file position."); |
| 198 | |
| 199 | static PyObject * |
| 200 | stringio_tell(stringio *self) |
| 201 | { |
| 202 | CHECK_INITIALIZED(self); |
| 203 | CHECK_CLOSED(self); |
| 204 | return PyLong_FromSsize_t(self->pos); |
| 205 | } |
| 206 | |
| 207 | PyDoc_STRVAR(stringio_read_doc, |
| 208 | "Read at most n characters, returned as a string.\n" |
| 209 | "\n" |
| 210 | "If the argument is negative or omitted, read until EOF\n" |
| 211 | "is reached. Return an empty string at EOF.\n"); |
| 212 | |
| 213 | static PyObject * |
| 214 | stringio_read(stringio *self, PyObject *args) |
| 215 | { |
| 216 | Py_ssize_t size, n; |
| 217 | Py_UNICODE *output; |
| 218 | PyObject *arg = Py_None; |
| 219 | |
| 220 | CHECK_INITIALIZED(self); |
| 221 | if (!PyArg_ParseTuple(args, "|O:read", &arg)) |
| 222 | return NULL; |
| 223 | CHECK_CLOSED(self); |
| 224 | |
| 225 | if (PyNumber_Check(arg)) { |
| 226 | size = PyNumber_AsSsize_t(arg, PyExc_OverflowError); |
| 227 | if (size == -1 && PyErr_Occurred()) |
| 228 | return NULL; |
| 229 | } |
| 230 | else if (arg == Py_None) { |
| 231 | /* Read until EOF is reached, by default. */ |
| 232 | size = -1; |
| 233 | } |
| 234 | else { |
| 235 | PyErr_Format(PyExc_TypeError, "integer argument expected, got '%s'", |
| 236 | Py_TYPE(arg)->tp_name); |
| 237 | return NULL; |
| 238 | } |
| 239 | |
| 240 | /* adjust invalid sizes */ |
| 241 | n = self->string_size - self->pos; |
| 242 | if (size < 0 || size > n) { |
| 243 | size = n; |
| 244 | if (size < 0) |
| 245 | size = 0; |
| 246 | } |
| 247 | |
| 248 | output = self->buf + self->pos; |
| 249 | self->pos += size; |
| 250 | return PyUnicode_FromUnicode(output, size); |
| 251 | } |
| 252 | |
| 253 | /* Internal helper, used by stringio_readline and stringio_iternext */ |
| 254 | static PyObject * |
| 255 | _stringio_readline(stringio *self, Py_ssize_t limit) |
| 256 | { |
| 257 | Py_UNICODE *start, *end, old_char; |
| 258 | Py_ssize_t len, consumed; |
| 259 | |
| 260 | /* In case of overseek, return the empty string */ |
| 261 | if (self->pos >= self->string_size) |
| 262 | return PyUnicode_FromString(""); |
| 263 | |
| 264 | start = self->buf + self->pos; |
| 265 | if (limit < 0 || limit > self->string_size - self->pos) |
| 266 | limit = self->string_size - self->pos; |
| 267 | |
| 268 | end = start + limit; |
| 269 | old_char = *end; |
| 270 | *end = '\0'; |
| 271 | len = _PyIO_find_line_ending( |
| 272 | self->readtranslate, self->readuniversal, self->readnl, |
| 273 | start, end, &consumed); |
| 274 | *end = old_char; |
| 275 | /* If we haven't found any line ending, we just return everything |
| 276 | (`consumed` is ignored). */ |
| 277 | if (len < 0) |
| 278 | len = limit; |
| 279 | self->pos += len; |
| 280 | return PyUnicode_FromUnicode(start, len); |
| 281 | } |
| 282 | |
| 283 | PyDoc_STRVAR(stringio_readline_doc, |
| 284 | "Read until newline or EOF.\n" |
| 285 | "\n" |
| 286 | "Returns an empty string if EOF is hit immediately.\n"); |
| 287 | |
| 288 | static PyObject * |
| 289 | stringio_readline(stringio *self, PyObject *args) |
| 290 | { |
| 291 | PyObject *arg = Py_None; |
| 292 | Py_ssize_t limit = -1; |
| 293 | |
| 294 | CHECK_INITIALIZED(self); |
| 295 | if (!PyArg_ParseTuple(args, "|O:readline", &arg)) |
| 296 | return NULL; |
| 297 | CHECK_CLOSED(self); |
| 298 | |
| 299 | if (PyNumber_Check(arg)) { |
| 300 | limit = PyNumber_AsSsize_t(arg, PyExc_OverflowError); |
| 301 | if (limit == -1 && PyErr_Occurred()) |
| 302 | return NULL; |
| 303 | } |
| 304 | else if (arg != Py_None) { |
| 305 | PyErr_Format(PyExc_TypeError, "integer argument expected, got '%s'", |
| 306 | Py_TYPE(arg)->tp_name); |
| 307 | return NULL; |
| 308 | } |
| 309 | return _stringio_readline(self, limit); |
| 310 | } |
| 311 | |
| 312 | static PyObject * |
| 313 | stringio_iternext(stringio *self) |
| 314 | { |
| 315 | PyObject *line; |
| 316 | |
| 317 | CHECK_INITIALIZED(self); |
| 318 | CHECK_CLOSED(self); |
| 319 | |
| 320 | if (Py_TYPE(self) == &PyStringIO_Type) { |
| 321 | /* Skip method call overhead for speed */ |
| 322 | line = _stringio_readline(self, -1); |
| 323 | } |
| 324 | else { |
| 325 | /* XXX is subclassing StringIO really supported? */ |
| 326 | line = PyObject_CallMethodObjArgs((PyObject *)self, |
| 327 | _PyIO_str_readline, NULL); |
| 328 | if (line && !PyUnicode_Check(line)) { |
| 329 | PyErr_Format(PyExc_IOError, |
| 330 | "readline() should have returned an str object, " |
| 331 | "not '%.200s'", Py_TYPE(line)->tp_name); |
| 332 | Py_DECREF(line); |
| 333 | return NULL; |
| 334 | } |
| 335 | } |
| 336 | |
| 337 | if (line == NULL) |
| 338 | return NULL; |
| 339 | |
| 340 | if (PyUnicode_GET_SIZE(line) == 0) { |
| 341 | /* Reached EOF */ |
| 342 | Py_DECREF(line); |
| 343 | return NULL; |
| 344 | } |
| 345 | |
| 346 | return line; |
| 347 | } |
| 348 | |
| 349 | PyDoc_STRVAR(stringio_truncate_doc, |
| 350 | "Truncate size to pos.\n" |
| 351 | "\n" |
| 352 | "The pos argument defaults to the current file position, as\n" |
Antoine Pitrou | f3fa074 | 2010-01-31 22:26:04 +0000 | [diff] [blame] | 353 | "returned by tell(). The current file position is unchanged.\n" |
Antoine Pitrou | 1969059 | 2009-06-12 20:14:08 +0000 | [diff] [blame] | 354 | "Returns the new absolute position.\n"); |
| 355 | |
| 356 | static PyObject * |
| 357 | stringio_truncate(stringio *self, PyObject *args) |
| 358 | { |
| 359 | Py_ssize_t size; |
| 360 | PyObject *arg = Py_None; |
| 361 | |
| 362 | CHECK_INITIALIZED(self); |
| 363 | if (!PyArg_ParseTuple(args, "|O:truncate", &arg)) |
| 364 | return NULL; |
| 365 | CHECK_CLOSED(self); |
| 366 | |
| 367 | if (PyNumber_Check(arg)) { |
| 368 | size = PyNumber_AsSsize_t(arg, PyExc_OverflowError); |
| 369 | if (size == -1 && PyErr_Occurred()) |
| 370 | return NULL; |
| 371 | } |
| 372 | else if (arg == Py_None) { |
| 373 | /* Truncate to current position if no argument is passed. */ |
| 374 | size = self->pos; |
| 375 | } |
| 376 | else { |
| 377 | PyErr_Format(PyExc_TypeError, "integer argument expected, got '%s'", |
| 378 | Py_TYPE(arg)->tp_name); |
| 379 | return NULL; |
| 380 | } |
| 381 | |
| 382 | if (size < 0) { |
| 383 | PyErr_Format(PyExc_ValueError, |
| 384 | "Negative size value %zd", size); |
| 385 | return NULL; |
| 386 | } |
| 387 | |
| 388 | if (size < self->string_size) { |
| 389 | if (resize_buffer(self, size) < 0) |
| 390 | return NULL; |
| 391 | self->string_size = size; |
| 392 | } |
Antoine Pitrou | 1969059 | 2009-06-12 20:14:08 +0000 | [diff] [blame] | 393 | |
| 394 | return PyLong_FromSsize_t(size); |
| 395 | } |
| 396 | |
| 397 | PyDoc_STRVAR(stringio_seek_doc, |
| 398 | "Change stream position.\n" |
| 399 | "\n" |
| 400 | "Seek to character offset pos relative to position indicated by whence:\n" |
| 401 | " 0 Start of stream (the default). pos should be >= 0;\n" |
| 402 | " 1 Current position - pos must be 0;\n" |
| 403 | " 2 End of stream - pos must be 0.\n" |
| 404 | "Returns the new absolute position.\n"); |
| 405 | |
| 406 | static PyObject * |
| 407 | stringio_seek(stringio *self, PyObject *args) |
| 408 | { |
| 409 | PyObject *posobj; |
| 410 | Py_ssize_t pos; |
| 411 | int mode = 0; |
| 412 | |
| 413 | CHECK_INITIALIZED(self); |
| 414 | if (!PyArg_ParseTuple(args, "O|i:seek", &posobj, &mode)) |
| 415 | return NULL; |
| 416 | |
| 417 | pos = PyNumber_AsSsize_t(posobj, PyExc_OverflowError); |
| 418 | if (pos == -1 && PyErr_Occurred()) |
| 419 | return NULL; |
| 420 | |
| 421 | CHECK_CLOSED(self); |
| 422 | |
| 423 | if (mode != 0 && mode != 1 && mode != 2) { |
| 424 | PyErr_Format(PyExc_ValueError, |
| 425 | "Invalid whence (%i, should be 0, 1 or 2)", mode); |
| 426 | return NULL; |
| 427 | } |
| 428 | else if (pos < 0 && mode == 0) { |
| 429 | PyErr_Format(PyExc_ValueError, |
| 430 | "Negative seek position %zd", pos); |
| 431 | return NULL; |
| 432 | } |
| 433 | else if (mode != 0 && pos != 0) { |
| 434 | PyErr_SetString(PyExc_IOError, |
| 435 | "Can't do nonzero cur-relative seeks"); |
| 436 | return NULL; |
| 437 | } |
| 438 | |
| 439 | /* mode 0: offset relative to beginning of the string. |
| 440 | mode 1: no change to current position. |
| 441 | mode 2: change position to end of file. */ |
| 442 | if (mode == 1) { |
| 443 | pos = self->pos; |
| 444 | } |
| 445 | else if (mode == 2) { |
| 446 | pos = self->string_size; |
| 447 | } |
| 448 | |
| 449 | self->pos = pos; |
| 450 | |
| 451 | return PyLong_FromSsize_t(self->pos); |
| 452 | } |
| 453 | |
| 454 | PyDoc_STRVAR(stringio_write_doc, |
| 455 | "Write string to file.\n" |
| 456 | "\n" |
| 457 | "Returns the number of characters written, which is always equal to\n" |
| 458 | "the length of the string.\n"); |
| 459 | |
| 460 | static PyObject * |
| 461 | stringio_write(stringio *self, PyObject *obj) |
| 462 | { |
| 463 | Py_ssize_t size; |
| 464 | |
| 465 | CHECK_INITIALIZED(self); |
| 466 | if (!PyUnicode_Check(obj)) { |
Eli Bendersky | eaf312a | 2011-07-22 14:39:55 +0300 | [diff] [blame] | 467 | PyErr_Format(PyExc_TypeError, "unicode argument expected, got '%s'", |
Antoine Pitrou | 1969059 | 2009-06-12 20:14:08 +0000 | [diff] [blame] | 468 | Py_TYPE(obj)->tp_name); |
| 469 | return NULL; |
| 470 | } |
| 471 | CHECK_CLOSED(self); |
| 472 | size = PyUnicode_GET_SIZE(obj); |
| 473 | |
| 474 | if (size > 0 && write_str(self, obj) < 0) |
| 475 | return NULL; |
| 476 | |
| 477 | return PyLong_FromSsize_t(size); |
| 478 | } |
| 479 | |
| 480 | PyDoc_STRVAR(stringio_close_doc, |
| 481 | "Close the IO object. Attempting any further operation after the\n" |
| 482 | "object is closed will raise a ValueError.\n" |
| 483 | "\n" |
| 484 | "This method has no effect if the file is already closed.\n"); |
| 485 | |
| 486 | static PyObject * |
| 487 | stringio_close(stringio *self) |
| 488 | { |
| 489 | self->closed = 1; |
| 490 | /* Free up some memory */ |
| 491 | if (resize_buffer(self, 0) < 0) |
| 492 | return NULL; |
| 493 | Py_CLEAR(self->readnl); |
| 494 | Py_CLEAR(self->writenl); |
| 495 | Py_CLEAR(self->decoder); |
| 496 | Py_RETURN_NONE; |
| 497 | } |
| 498 | |
| 499 | static int |
| 500 | stringio_traverse(stringio *self, visitproc visit, void *arg) |
| 501 | { |
| 502 | Py_VISIT(self->dict); |
| 503 | return 0; |
| 504 | } |
| 505 | |
| 506 | static int |
| 507 | stringio_clear(stringio *self) |
| 508 | { |
| 509 | Py_CLEAR(self->dict); |
| 510 | return 0; |
| 511 | } |
| 512 | |
| 513 | static void |
| 514 | stringio_dealloc(stringio *self) |
| 515 | { |
| 516 | _PyObject_GC_UNTRACK(self); |
Antoine Pitrou | f98a267 | 2009-10-24 11:59:41 +0000 | [diff] [blame] | 517 | self->ok = 0; |
| 518 | if (self->buf) { |
| 519 | PyMem_Free(self->buf); |
| 520 | self->buf = NULL; |
| 521 | } |
Antoine Pitrou | 1969059 | 2009-06-12 20:14:08 +0000 | [diff] [blame] | 522 | Py_CLEAR(self->readnl); |
| 523 | Py_CLEAR(self->writenl); |
| 524 | Py_CLEAR(self->decoder); |
Antoine Pitrou | f98a267 | 2009-10-24 11:59:41 +0000 | [diff] [blame] | 525 | Py_CLEAR(self->dict); |
Antoine Pitrou | 1969059 | 2009-06-12 20:14:08 +0000 | [diff] [blame] | 526 | if (self->weakreflist != NULL) |
| 527 | PyObject_ClearWeakRefs((PyObject *) self); |
| 528 | Py_TYPE(self)->tp_free(self); |
| 529 | } |
| 530 | |
| 531 | static PyObject * |
| 532 | stringio_new(PyTypeObject *type, PyObject *args, PyObject *kwds) |
| 533 | { |
| 534 | stringio *self; |
| 535 | |
| 536 | assert(type != NULL && type->tp_alloc != NULL); |
| 537 | self = (stringio *)type->tp_alloc(type, 0); |
| 538 | if (self == NULL) |
| 539 | return NULL; |
| 540 | |
Antoine Pitrou | fa94e80 | 2009-10-24 12:23:18 +0000 | [diff] [blame] | 541 | /* tp_alloc initializes all the fields to zero. So we don't have to |
| 542 | initialize them here. */ |
| 543 | |
Antoine Pitrou | 1969059 | 2009-06-12 20:14:08 +0000 | [diff] [blame] | 544 | self->buf = (Py_UNICODE *)PyMem_Malloc(0); |
| 545 | if (self->buf == NULL) { |
| 546 | Py_DECREF(self); |
| 547 | return PyErr_NoMemory(); |
| 548 | } |
| 549 | |
| 550 | return (PyObject *)self; |
| 551 | } |
| 552 | |
| 553 | static int |
| 554 | stringio_init(stringio *self, PyObject *args, PyObject *kwds) |
| 555 | { |
| 556 | char *kwlist[] = {"initial_value", "newline", NULL}; |
| 557 | PyObject *value = NULL; |
| 558 | char *newline = "\n"; |
| 559 | |
| 560 | if (!PyArg_ParseTupleAndKeywords(args, kwds, "|Oz:__init__", kwlist, |
| 561 | &value, &newline)) |
| 562 | return -1; |
| 563 | |
| 564 | if (newline && newline[0] != '\0' |
| 565 | && !(newline[0] == '\n' && newline[1] == '\0') |
| 566 | && !(newline[0] == '\r' && newline[1] == '\0') |
| 567 | && !(newline[0] == '\r' && newline[1] == '\n' && newline[2] == '\0')) { |
| 568 | PyErr_Format(PyExc_ValueError, |
| 569 | "illegal newline value: %s", newline); |
| 570 | return -1; |
| 571 | } |
| 572 | if (value && value != Py_None && !PyUnicode_Check(value)) { |
Antoine Pitrou | fa94e80 | 2009-10-24 12:23:18 +0000 | [diff] [blame] | 573 | PyErr_Format(PyExc_TypeError, |
| 574 | "initial_value must be unicode or None, not %.200s", |
Antoine Pitrou | 1969059 | 2009-06-12 20:14:08 +0000 | [diff] [blame] | 575 | Py_TYPE(value)->tp_name); |
| 576 | return -1; |
| 577 | } |
| 578 | |
| 579 | self->ok = 0; |
| 580 | |
| 581 | Py_CLEAR(self->readnl); |
| 582 | Py_CLEAR(self->writenl); |
| 583 | Py_CLEAR(self->decoder); |
| 584 | |
| 585 | if (newline) { |
| 586 | self->readnl = PyString_FromString(newline); |
| 587 | if (self->readnl == NULL) |
| 588 | return -1; |
| 589 | } |
| 590 | self->readuniversal = (newline == NULL || newline[0] == '\0'); |
| 591 | self->readtranslate = (newline == NULL); |
| 592 | /* If newline == "", we don't translate anything. |
| 593 | If newline == "\n" or newline == None, we translate to "\n", which is |
| 594 | a no-op. |
| 595 | (for newline == None, TextIOWrapper translates to os.sepline, but it |
| 596 | is pointless for StringIO) |
| 597 | */ |
| 598 | if (newline != NULL && newline[0] == '\r') { |
| 599 | self->writenl = PyUnicode_FromString(newline); |
| 600 | } |
| 601 | |
| 602 | if (self->readuniversal) { |
| 603 | self->decoder = PyObject_CallFunction( |
| 604 | (PyObject *)&PyIncrementalNewlineDecoder_Type, |
| 605 | "Oi", Py_None, (int) self->readtranslate); |
| 606 | if (self->decoder == NULL) |
| 607 | return -1; |
| 608 | } |
| 609 | |
| 610 | /* Now everything is set up, resize buffer to size of initial value, |
| 611 | and copy it */ |
| 612 | self->string_size = 0; |
| 613 | if (value && value != Py_None) { |
| 614 | Py_ssize_t len = PyUnicode_GetSize(value); |
| 615 | /* This is a heuristic, for newline translation might change |
| 616 | the string length. */ |
| 617 | if (resize_buffer(self, len) < 0) |
| 618 | return -1; |
| 619 | self->pos = 0; |
| 620 | if (write_str(self, value) < 0) |
| 621 | return -1; |
| 622 | } |
| 623 | else { |
| 624 | if (resize_buffer(self, 0) < 0) |
| 625 | return -1; |
| 626 | } |
| 627 | self->pos = 0; |
| 628 | |
| 629 | self->closed = 0; |
| 630 | self->ok = 1; |
| 631 | return 0; |
| 632 | } |
| 633 | |
| 634 | /* Properties and pseudo-properties */ |
Antoine Pitrou | c5eec0e | 2012-09-05 20:11:49 +0200 | [diff] [blame] | 635 | |
| 636 | PyDoc_STRVAR(stringio_readable_doc, |
| 637 | "readable() -> bool. Returns True if the IO object can be read."); |
| 638 | |
| 639 | PyDoc_STRVAR(stringio_writable_doc, |
| 640 | "writable() -> bool. Returns True if the IO object can be written."); |
| 641 | |
| 642 | PyDoc_STRVAR(stringio_seekable_doc, |
| 643 | "seekable() -> bool. Returns True if the IO object can be seeked."); |
| 644 | |
Antoine Pitrou | 1969059 | 2009-06-12 20:14:08 +0000 | [diff] [blame] | 645 | static PyObject * |
| 646 | stringio_seekable(stringio *self, PyObject *args) |
| 647 | { |
| 648 | CHECK_INITIALIZED(self); |
Antoine Pitrou | c5eec0e | 2012-09-05 20:11:49 +0200 | [diff] [blame] | 649 | CHECK_CLOSED(self); |
Antoine Pitrou | 1969059 | 2009-06-12 20:14:08 +0000 | [diff] [blame] | 650 | Py_RETURN_TRUE; |
| 651 | } |
| 652 | |
| 653 | static PyObject * |
| 654 | stringio_readable(stringio *self, PyObject *args) |
| 655 | { |
| 656 | CHECK_INITIALIZED(self); |
Antoine Pitrou | c5eec0e | 2012-09-05 20:11:49 +0200 | [diff] [blame] | 657 | CHECK_CLOSED(self); |
Antoine Pitrou | 1969059 | 2009-06-12 20:14:08 +0000 | [diff] [blame] | 658 | Py_RETURN_TRUE; |
| 659 | } |
| 660 | |
| 661 | static PyObject * |
| 662 | stringio_writable(stringio *self, PyObject *args) |
| 663 | { |
| 664 | CHECK_INITIALIZED(self); |
Antoine Pitrou | c5eec0e | 2012-09-05 20:11:49 +0200 | [diff] [blame] | 665 | CHECK_CLOSED(self); |
Antoine Pitrou | 1969059 | 2009-06-12 20:14:08 +0000 | [diff] [blame] | 666 | Py_RETURN_TRUE; |
| 667 | } |
| 668 | |
Antoine Pitrou | fa94e80 | 2009-10-24 12:23:18 +0000 | [diff] [blame] | 669 | /* Pickling support. |
| 670 | |
| 671 | The implementation of __getstate__ is similar to the one for BytesIO, |
| 672 | except that we also save the newline parameter. For __setstate__ and unlike |
| 673 | BytesIO, we call __init__ to restore the object's state. Doing so allows us |
| 674 | to avoid decoding the complex newline state while keeping the object |
| 675 | representation compact. |
| 676 | |
| 677 | See comment in bytesio.c regarding why only pickle protocols and onward are |
| 678 | supported. |
| 679 | */ |
| 680 | |
| 681 | static PyObject * |
| 682 | stringio_getstate(stringio *self) |
| 683 | { |
| 684 | PyObject *initvalue = stringio_getvalue(self); |
| 685 | PyObject *dict; |
| 686 | PyObject *state; |
| 687 | |
| 688 | if (initvalue == NULL) |
| 689 | return NULL; |
| 690 | if (self->dict == NULL) { |
| 691 | Py_INCREF(Py_None); |
| 692 | dict = Py_None; |
| 693 | } |
| 694 | else { |
| 695 | dict = PyDict_Copy(self->dict); |
| 696 | if (dict == NULL) |
| 697 | return NULL; |
| 698 | } |
| 699 | |
| 700 | state = Py_BuildValue("(OOnN)", initvalue, |
| 701 | self->readnl ? self->readnl : Py_None, |
| 702 | self->pos, dict); |
| 703 | Py_DECREF(initvalue); |
| 704 | return state; |
| 705 | } |
| 706 | |
| 707 | static PyObject * |
| 708 | stringio_setstate(stringio *self, PyObject *state) |
| 709 | { |
| 710 | PyObject *initarg; |
| 711 | PyObject *position_obj; |
| 712 | PyObject *dict; |
| 713 | Py_ssize_t pos; |
| 714 | |
| 715 | assert(state != NULL); |
| 716 | CHECK_CLOSED(self); |
| 717 | |
| 718 | /* We allow the state tuple to be longer than 4, because we may need |
| 719 | someday to extend the object's state without breaking |
| 720 | backward-compatibility. */ |
| 721 | if (!PyTuple_Check(state) || Py_SIZE(state) < 4) { |
| 722 | PyErr_Format(PyExc_TypeError, |
| 723 | "%.200s.__setstate__ argument should be 4-tuple, got %.200s", |
| 724 | Py_TYPE(self)->tp_name, Py_TYPE(state)->tp_name); |
| 725 | return NULL; |
| 726 | } |
| 727 | |
| 728 | /* Initialize the object's state. */ |
| 729 | initarg = PyTuple_GetSlice(state, 0, 2); |
| 730 | if (initarg == NULL) |
| 731 | return NULL; |
| 732 | if (stringio_init(self, initarg, NULL) < 0) { |
| 733 | Py_DECREF(initarg); |
| 734 | return NULL; |
| 735 | } |
| 736 | Py_DECREF(initarg); |
| 737 | |
| 738 | /* Restore the buffer state. Even if __init__ did initialize the buffer, |
| 739 | we have to initialize it again since __init__ may translates the |
| 740 | newlines in the inital_value string. We clearly do not want that |
| 741 | because the string value in the state tuple has already been translated |
| 742 | once by __init__. So we do not take any chance and replace object's |
| 743 | buffer completely. */ |
| 744 | { |
| 745 | Py_UNICODE *buf = PyUnicode_AS_UNICODE(PyTuple_GET_ITEM(state, 0)); |
| 746 | Py_ssize_t bufsize = PyUnicode_GET_SIZE(PyTuple_GET_ITEM(state, 0)); |
| 747 | if (resize_buffer(self, bufsize) < 0) |
| 748 | return NULL; |
| 749 | memcpy(self->buf, buf, bufsize * sizeof(Py_UNICODE)); |
| 750 | self->string_size = bufsize; |
| 751 | } |
| 752 | |
| 753 | /* Set carefully the position value. Alternatively, we could use the seek |
| 754 | method instead of modifying self->pos directly to better protect the |
| 755 | object internal state against errneous (or malicious) inputs. */ |
| 756 | position_obj = PyTuple_GET_ITEM(state, 2); |
| 757 | if (!PyIndex_Check(position_obj)) { |
| 758 | PyErr_Format(PyExc_TypeError, |
| 759 | "third item of state must be an integer, got %.200s", |
| 760 | Py_TYPE(position_obj)->tp_name); |
| 761 | return NULL; |
| 762 | } |
| 763 | pos = PyNumber_AsSsize_t(position_obj, PyExc_OverflowError); |
| 764 | if (pos == -1 && PyErr_Occurred()) |
| 765 | return NULL; |
| 766 | if (pos < 0) { |
| 767 | PyErr_SetString(PyExc_ValueError, |
| 768 | "position value cannot be negative"); |
| 769 | return NULL; |
| 770 | } |
| 771 | self->pos = pos; |
| 772 | |
| 773 | /* Set the dictionary of the instance variables. */ |
| 774 | dict = PyTuple_GET_ITEM(state, 3); |
| 775 | if (dict != Py_None) { |
| 776 | if (!PyDict_Check(dict)) { |
| 777 | PyErr_Format(PyExc_TypeError, |
| 778 | "fourth item of state should be a dict, got a %.200s", |
| 779 | Py_TYPE(dict)->tp_name); |
| 780 | return NULL; |
| 781 | } |
| 782 | if (self->dict) { |
| 783 | /* Alternatively, we could replace the internal dictionary |
| 784 | completely. However, it seems more practical to just update it. */ |
| 785 | if (PyDict_Update(self->dict, dict) < 0) |
| 786 | return NULL; |
| 787 | } |
| 788 | else { |
| 789 | Py_INCREF(dict); |
| 790 | self->dict = dict; |
| 791 | } |
| 792 | } |
| 793 | |
| 794 | Py_RETURN_NONE; |
| 795 | } |
| 796 | |
| 797 | |
Antoine Pitrou | 1969059 | 2009-06-12 20:14:08 +0000 | [diff] [blame] | 798 | static PyObject * |
Antoine Pitrou | 1969059 | 2009-06-12 20:14:08 +0000 | [diff] [blame] | 799 | stringio_closed(stringio *self, void *context) |
| 800 | { |
| 801 | CHECK_INITIALIZED(self); |
| 802 | return PyBool_FromLong(self->closed); |
| 803 | } |
| 804 | |
| 805 | static PyObject * |
| 806 | stringio_line_buffering(stringio *self, void *context) |
| 807 | { |
| 808 | CHECK_INITIALIZED(self); |
| 809 | CHECK_CLOSED(self); |
| 810 | Py_RETURN_FALSE; |
| 811 | } |
| 812 | |
| 813 | static PyObject * |
| 814 | stringio_newlines(stringio *self, void *context) |
| 815 | { |
| 816 | CHECK_INITIALIZED(self); |
| 817 | CHECK_CLOSED(self); |
| 818 | if (self->decoder == NULL) |
| 819 | Py_RETURN_NONE; |
| 820 | return PyObject_GetAttr(self->decoder, _PyIO_str_newlines); |
| 821 | } |
| 822 | |
| 823 | static struct PyMethodDef stringio_methods[] = { |
| 824 | {"close", (PyCFunction)stringio_close, METH_NOARGS, stringio_close_doc}, |
Antoine Pitrou | 5cd2d8c | 2010-09-02 22:23:19 +0000 | [diff] [blame] | 825 | {"getvalue", (PyCFunction)stringio_getvalue, METH_NOARGS, stringio_getvalue_doc}, |
Antoine Pitrou | 1969059 | 2009-06-12 20:14:08 +0000 | [diff] [blame] | 826 | {"read", (PyCFunction)stringio_read, METH_VARARGS, stringio_read_doc}, |
| 827 | {"readline", (PyCFunction)stringio_readline, METH_VARARGS, stringio_readline_doc}, |
| 828 | {"tell", (PyCFunction)stringio_tell, METH_NOARGS, stringio_tell_doc}, |
| 829 | {"truncate", (PyCFunction)stringio_truncate, METH_VARARGS, stringio_truncate_doc}, |
| 830 | {"seek", (PyCFunction)stringio_seek, METH_VARARGS, stringio_seek_doc}, |
| 831 | {"write", (PyCFunction)stringio_write, METH_O, stringio_write_doc}, |
Antoine Pitrou | fa94e80 | 2009-10-24 12:23:18 +0000 | [diff] [blame] | 832 | |
Antoine Pitrou | c5eec0e | 2012-09-05 20:11:49 +0200 | [diff] [blame] | 833 | {"seekable", (PyCFunction)stringio_seekable, METH_NOARGS, stringio_seekable_doc}, |
| 834 | {"readable", (PyCFunction)stringio_readable, METH_NOARGS, stringio_readable_doc}, |
| 835 | {"writable", (PyCFunction)stringio_writable, METH_NOARGS, stringio_writable_doc}, |
Antoine Pitrou | fa94e80 | 2009-10-24 12:23:18 +0000 | [diff] [blame] | 836 | |
| 837 | {"__getstate__", (PyCFunction)stringio_getstate, METH_NOARGS}, |
| 838 | {"__setstate__", (PyCFunction)stringio_setstate, METH_O}, |
Antoine Pitrou | 1969059 | 2009-06-12 20:14:08 +0000 | [diff] [blame] | 839 | {NULL, NULL} /* sentinel */ |
| 840 | }; |
| 841 | |
| 842 | static PyGetSetDef stringio_getset[] = { |
| 843 | {"closed", (getter)stringio_closed, NULL, NULL}, |
| 844 | {"newlines", (getter)stringio_newlines, NULL, NULL}, |
| 845 | /* (following comments straight off of the original Python wrapper:) |
| 846 | XXX Cruft to support the TextIOWrapper API. This would only |
| 847 | be meaningful if StringIO supported the buffer attribute. |
| 848 | Hopefully, a better solution, than adding these pseudo-attributes, |
| 849 | will be found. |
| 850 | */ |
Antoine Pitrou | 1969059 | 2009-06-12 20:14:08 +0000 | [diff] [blame] | 851 | {"line_buffering", (getter)stringio_line_buffering, NULL, NULL}, |
| 852 | {NULL} |
| 853 | }; |
| 854 | |
| 855 | PyTypeObject PyStringIO_Type = { |
| 856 | PyVarObject_HEAD_INIT(NULL, 0) |
| 857 | "_io.StringIO", /*tp_name*/ |
| 858 | sizeof(stringio), /*tp_basicsize*/ |
| 859 | 0, /*tp_itemsize*/ |
| 860 | (destructor)stringio_dealloc, /*tp_dealloc*/ |
| 861 | 0, /*tp_print*/ |
| 862 | 0, /*tp_getattr*/ |
| 863 | 0, /*tp_setattr*/ |
| 864 | 0, /*tp_reserved*/ |
| 865 | 0, /*tp_repr*/ |
| 866 | 0, /*tp_as_number*/ |
| 867 | 0, /*tp_as_sequence*/ |
| 868 | 0, /*tp_as_mapping*/ |
| 869 | 0, /*tp_hash*/ |
| 870 | 0, /*tp_call*/ |
| 871 | 0, /*tp_str*/ |
| 872 | 0, /*tp_getattro*/ |
| 873 | 0, /*tp_setattro*/ |
| 874 | 0, /*tp_as_buffer*/ |
| 875 | Py_TPFLAGS_DEFAULT | Py_TPFLAGS_BASETYPE |
| 876 | | Py_TPFLAGS_HAVE_GC, /*tp_flags*/ |
| 877 | stringio_doc, /*tp_doc*/ |
| 878 | (traverseproc)stringio_traverse, /*tp_traverse*/ |
| 879 | (inquiry)stringio_clear, /*tp_clear*/ |
| 880 | 0, /*tp_richcompare*/ |
| 881 | offsetof(stringio, weakreflist), /*tp_weaklistoffset*/ |
| 882 | 0, /*tp_iter*/ |
| 883 | (iternextfunc)stringio_iternext, /*tp_iternext*/ |
| 884 | stringio_methods, /*tp_methods*/ |
| 885 | 0, /*tp_members*/ |
| 886 | stringio_getset, /*tp_getset*/ |
| 887 | 0, /*tp_base*/ |
| 888 | 0, /*tp_dict*/ |
| 889 | 0, /*tp_descr_get*/ |
| 890 | 0, /*tp_descr_set*/ |
| 891 | offsetof(stringio, dict), /*tp_dictoffset*/ |
| 892 | (initproc)stringio_init, /*tp_init*/ |
| 893 | 0, /*tp_alloc*/ |
| 894 | stringio_new, /*tp_new*/ |
| 895 | }; |