blob: 12ae4b6a2be5620ea1f929bb1dfe1f898d2d32ae [file] [log] [blame]
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001"""
2Python implementation of the io module.
3"""
4
5import os
6import abc
7import codecs
Benjamin Peterson59406a92009-03-26 17:10:29 +00008import warnings
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00009# Import _thread instead of threading to reduce startup cost
10try:
11 from _thread import allocate_lock as Lock
12except ImportError:
13 from _dummy_thread import allocate_lock as Lock
14
15import io
Benjamin Petersonc3be11a2010-04-27 21:24:03 +000016from io import (__all__, SEEK_SET, SEEK_CUR, SEEK_END)
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +000017
18# open() uses st_blksize whenever we can
19DEFAULT_BUFFER_SIZE = 8 * 1024 # bytes
20
21# NOTE: Base classes defined here are registered with the "official" ABCs
22# defined in io.py. We don't use real inheritance though, because we don't
23# want to inherit the C implementations.
24
25
26class BlockingIOError(IOError):
27
28 """Exception raised when I/O would block on a non-blocking I/O stream."""
29
30 def __init__(self, errno, strerror, characters_written=0):
31 super().__init__(errno, strerror)
32 if not isinstance(characters_written, int):
33 raise TypeError("characters_written must be a integer")
34 self.characters_written = characters_written
35
36
Benjamin Peterson95e392c2010-04-27 21:07:21 +000037def open(file: (str, bytes), mode: str = "r", buffering: int = -1,
Benjamin Peterson9990e8c2009-04-18 14:47:50 +000038 encoding: str = None, errors: str = None,
39 newline: str = None, closefd: bool = True) -> "IOBase":
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +000040
41 r"""Open file and return a stream. Raise IOError upon failure.
42
43 file is either a text or byte string giving the name (and the path
44 if the file isn't in the current working directory) of the file to
45 be opened or an integer file descriptor of the file to be
46 wrapped. (If a file descriptor is given, it is closed when the
47 returned I/O object is closed, unless closefd is set to False.)
48
49 mode is an optional string that specifies the mode in which the file
50 is opened. It defaults to 'r' which means open for reading in text
51 mode. Other common values are 'w' for writing (truncating the file if
52 it already exists), and 'a' for appending (which on some Unix systems,
53 means that all writes append to the end of the file regardless of the
54 current seek position). In text mode, if encoding is not specified the
55 encoding used is platform dependent. (For reading and writing raw
56 bytes use binary mode and leave encoding unspecified.) The available
57 modes are:
58
59 ========= ===============================================================
60 Character Meaning
61 --------- ---------------------------------------------------------------
62 'r' open for reading (default)
63 'w' open for writing, truncating the file first
64 'a' open for writing, appending to the end of the file if it exists
65 'b' binary mode
66 't' text mode (default)
67 '+' open a disk file for updating (reading and writing)
68 'U' universal newline mode (for backwards compatibility; unneeded
69 for new code)
70 ========= ===============================================================
71
72 The default mode is 'rt' (open for reading text). For binary random
73 access, the mode 'w+b' opens and truncates the file to 0 bytes, while
74 'r+b' opens the file without truncation.
75
76 Python distinguishes between files opened in binary and text modes,
77 even when the underlying operating system doesn't. Files opened in
78 binary mode (appending 'b' to the mode argument) return contents as
79 bytes objects without any decoding. In text mode (the default, or when
80 't' is appended to the mode argument), the contents of the file are
81 returned as strings, the bytes having been first decoded using a
82 platform-dependent encoding or using the specified encoding if given.
83
Antoine Pitroud5587bc2009-12-19 21:08:31 +000084 buffering is an optional integer used to set the buffering policy.
85 Pass 0 to switch buffering off (only allowed in binary mode), 1 to select
86 line buffering (only usable in text mode), and an integer > 1 to indicate
87 the size of a fixed-size chunk buffer. When no buffering argument is
88 given, the default buffering policy works as follows:
89
90 * Binary files are buffered in fixed-size chunks; the size of the buffer
91 is chosen using a heuristic trying to determine the underlying device's
92 "block size" and falling back on `io.DEFAULT_BUFFER_SIZE`.
93 On many systems, the buffer will typically be 4096 or 8192 bytes long.
94
95 * "Interactive" text files (files for which isatty() returns True)
96 use line buffering. Other text files use the policy described above
97 for binary files.
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +000098
99 encoding is the name of the encoding used to decode or encode the
100 file. This should only be used in text mode. The default encoding is
101 platform dependent, but any encoding supported by Python can be
102 passed. See the codecs module for the list of supported encodings.
103
104 errors is an optional string that specifies how encoding errors are to
105 be handled---this argument should not be used in binary mode. Pass
106 'strict' to raise a ValueError exception if there is an encoding error
107 (the default of None has the same effect), or pass 'ignore' to ignore
108 errors. (Note that ignoring encoding errors can lead to data loss.)
109 See the documentation for codecs.register for a list of the permitted
110 encoding error strings.
111
112 newline controls how universal newlines works (it only applies to text
113 mode). It can be None, '', '\n', '\r', and '\r\n'. It works as
114 follows:
115
116 * On input, if newline is None, universal newlines mode is
117 enabled. Lines in the input can end in '\n', '\r', or '\r\n', and
118 these are translated into '\n' before being returned to the
119 caller. If it is '', universal newline mode is enabled, but line
120 endings are returned to the caller untranslated. If it has any of
121 the other legal values, input lines are only terminated by the given
122 string, and the line ending is returned to the caller untranslated.
123
124 * On output, if newline is None, any '\n' characters written are
125 translated to the system default line separator, os.linesep. If
126 newline is '', no translation takes place. If newline is any of the
127 other legal values, any '\n' characters written are translated to
128 the given string.
129
130 If closefd is False, the underlying file descriptor will be kept open
131 when the file is closed. This does not work when a file name is given
132 and must be True in that case.
133
134 open() returns a file object whose type depends on the mode, and
135 through which the standard file operations such as reading and writing
136 are performed. When open() is used to open a file in a text mode ('w',
137 'r', 'wt', 'rt', etc.), it returns a TextIOWrapper. When used to open
138 a file in a binary mode, the returned class varies: in read binary
139 mode, it returns a BufferedReader; in write binary and append binary
140 modes, it returns a BufferedWriter, and in read/write mode, it returns
141 a BufferedRandom.
142
143 It is also possible to use a string or bytearray as a file for both
144 reading and writing. For strings StringIO can be used like a file
145 opened in a text mode, and for bytes a BytesIO can be used like a file
146 opened in a binary mode.
147 """
148 if not isinstance(file, (str, bytes, int)):
149 raise TypeError("invalid file: %r" % file)
150 if not isinstance(mode, str):
151 raise TypeError("invalid mode: %r" % mode)
Benjamin Peterson95e392c2010-04-27 21:07:21 +0000152 if not isinstance(buffering, int):
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000153 raise TypeError("invalid buffering: %r" % buffering)
154 if encoding is not None and not isinstance(encoding, str):
155 raise TypeError("invalid encoding: %r" % encoding)
156 if errors is not None and not isinstance(errors, str):
157 raise TypeError("invalid errors: %r" % errors)
158 modes = set(mode)
159 if modes - set("arwb+tU") or len(mode) > len(modes):
160 raise ValueError("invalid mode: %r" % mode)
161 reading = "r" in modes
162 writing = "w" in modes
163 appending = "a" in modes
164 updating = "+" in modes
165 text = "t" in modes
166 binary = "b" in modes
167 if "U" in modes:
168 if writing or appending:
169 raise ValueError("can't use U and writing mode at once")
170 reading = True
171 if text and binary:
172 raise ValueError("can't have text and binary mode at once")
173 if reading + writing + appending > 1:
174 raise ValueError("can't have read/write/append mode at once")
175 if not (reading or writing or appending):
176 raise ValueError("must have exactly one of read/write/append mode")
177 if binary and encoding is not None:
178 raise ValueError("binary mode doesn't take an encoding argument")
179 if binary and errors is not None:
180 raise ValueError("binary mode doesn't take an errors argument")
181 if binary and newline is not None:
182 raise ValueError("binary mode doesn't take a newline argument")
183 raw = FileIO(file,
184 (reading and "r" or "") +
185 (writing and "w" or "") +
186 (appending and "a" or "") +
187 (updating and "+" or ""),
188 closefd)
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000189 line_buffering = False
190 if buffering == 1 or buffering < 0 and raw.isatty():
191 buffering = -1
192 line_buffering = True
193 if buffering < 0:
194 buffering = DEFAULT_BUFFER_SIZE
195 try:
196 bs = os.fstat(raw.fileno()).st_blksize
197 except (os.error, AttributeError):
198 pass
199 else:
200 if bs > 1:
201 buffering = bs
202 if buffering < 0:
203 raise ValueError("invalid buffering size")
204 if buffering == 0:
205 if binary:
206 return raw
207 raise ValueError("can't have unbuffered text I/O")
208 if updating:
209 buffer = BufferedRandom(raw, buffering)
210 elif writing or appending:
211 buffer = BufferedWriter(raw, buffering)
212 elif reading:
213 buffer = BufferedReader(raw, buffering)
214 else:
215 raise ValueError("unknown mode: %r" % mode)
216 if binary:
217 return buffer
218 text = TextIOWrapper(buffer, encoding, errors, newline, line_buffering)
219 text.mode = mode
220 return text
221
222
223class DocDescriptor:
224 """Helper for builtins.open.__doc__
225 """
226 def __get__(self, obj, typ):
227 return (
Benjamin Petersonc3be11a2010-04-27 21:24:03 +0000228 "open(file, mode='r', buffering=-1, encoding=None, "
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000229 "errors=None, newline=None, closefd=True)\n\n" +
230 open.__doc__)
231
232class OpenWrapper:
233 """Wrapper for builtins.open
234
235 Trick so that open won't become a bound method when stored
236 as a class variable (as dbm.dumb does).
237
238 See initstdio() in Python/pythonrun.c.
239 """
240 __doc__ = DocDescriptor()
241
242 def __new__(cls, *args, **kwargs):
243 return open(*args, **kwargs)
244
245
Antoine Pitrou0d739d72010-09-05 23:01:12 +0000246# In normal operation, both `UnsupportedOperation`s should be bound to the
247# same object.
248try:
249 UnsupportedOperation = io.UnsupportedOperation
250except AttributeError:
251 class UnsupportedOperation(ValueError, IOError):
252 pass
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000253
254
255class IOBase(metaclass=abc.ABCMeta):
256
257 """The abstract base class for all I/O classes, acting on streams of
258 bytes. There is no public constructor.
259
260 This class provides dummy implementations for many methods that
261 derived classes can override selectively; the default implementations
262 represent a file that cannot be read, written or seeked.
263
264 Even though IOBase does not declare read, readinto, or write because
265 their signatures will vary, implementations and clients should
266 consider those methods part of the interface. Also, implementations
267 may raise a IOError when operations they do not support are called.
268
269 The basic type used for binary data read from or written to a file is
270 bytes. bytearrays are accepted too, and in some cases (such as
271 readinto) needed. Text I/O classes work with str data.
272
273 Note that calling any method (even inquiries) on a closed stream is
274 undefined. Implementations may raise IOError in this case.
275
276 IOBase (and its subclasses) support the iterator protocol, meaning
277 that an IOBase object can be iterated over yielding the lines in a
278 stream.
279
280 IOBase also supports the :keyword:`with` statement. In this example,
281 fp is closed after the suite of the with statement is complete:
282
283 with open('spam.txt', 'r') as fp:
284 fp.write('Spam and eggs!')
285 """
286
287 ### Internal ###
288
289 def _unsupported(self, name: str) -> IOError:
290 """Internal: raise an exception for unsupported operations."""
291 raise UnsupportedOperation("%s.%s() not supported" %
292 (self.__class__.__name__, name))
293
294 ### Positioning ###
295
296 def seek(self, pos: int, whence: int = 0) -> int:
297 """Change stream position.
298
299 Change the stream position to byte offset offset. offset is
300 interpreted relative to the position indicated by whence. Values
301 for whence are:
302
303 * 0 -- start of stream (the default); offset should be zero or positive
304 * 1 -- current stream position; offset may be negative
305 * 2 -- end of stream; offset is usually negative
306
307 Return the new absolute position.
308 """
309 self._unsupported("seek")
310
311 def tell(self) -> int:
312 """Return current stream position."""
313 return self.seek(0, 1)
314
315 def truncate(self, pos: int = None) -> int:
316 """Truncate file to size bytes.
317
318 Size defaults to the current IO position as reported by tell(). Return
319 the new size.
320 """
321 self._unsupported("truncate")
322
323 ### Flush and close ###
324
325 def flush(self) -> None:
326 """Flush write buffers, if applicable.
327
328 This is not implemented for read-only and non-blocking streams.
329 """
Antoine Pitrou6be88762010-05-03 16:48:20 +0000330 self._checkClosed()
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000331 # XXX Should this return the number of bytes written???
332
333 __closed = False
334
335 def close(self) -> None:
336 """Flush and close the IO object.
337
338 This method has no effect if the file is already closed.
339 """
340 if not self.__closed:
Antoine Pitrou6be88762010-05-03 16:48:20 +0000341 self.flush()
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000342 self.__closed = True
343
344 def __del__(self) -> None:
345 """Destructor. Calls close()."""
346 # The try/except block is in case this is called at program
347 # exit time, when it's possible that globals have already been
348 # deleted, and then the close() call might fail. Since
349 # there's nothing we can do about such failures and they annoy
350 # the end users, we suppress the traceback.
351 try:
352 self.close()
353 except:
354 pass
355
356 ### Inquiries ###
357
358 def seekable(self) -> bool:
359 """Return whether object supports random access.
360
361 If False, seek(), tell() and truncate() will raise IOError.
362 This method may need to do a test seek().
363 """
364 return False
365
366 def _checkSeekable(self, msg=None):
367 """Internal: raise an IOError if file is not seekable
368 """
369 if not self.seekable():
Antoine Pitrou0d739d72010-09-05 23:01:12 +0000370 raise UnsupportedOperation("File or stream is not seekable."
371 if msg is None else msg)
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000372
373 def readable(self) -> bool:
374 """Return whether object was opened for reading.
375
376 If False, read() will raise IOError.
377 """
378 return False
379
380 def _checkReadable(self, msg=None):
381 """Internal: raise an IOError if file is not readable
382 """
383 if not self.readable():
Antoine Pitrou0d739d72010-09-05 23:01:12 +0000384 raise UnsupportedOperation("File or stream is not readable."
385 if msg is None else msg)
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000386
387 def writable(self) -> bool:
388 """Return whether object was opened for writing.
389
390 If False, write() and truncate() will raise IOError.
391 """
392 return False
393
394 def _checkWritable(self, msg=None):
395 """Internal: raise an IOError if file is not writable
396 """
397 if not self.writable():
Antoine Pitrou0d739d72010-09-05 23:01:12 +0000398 raise UnsupportedOperation("File or stream is not writable."
399 if msg is None else msg)
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000400
401 @property
402 def closed(self):
403 """closed: bool. True iff the file has been closed.
404
405 For backwards compatibility, this is a property, not a predicate.
406 """
407 return self.__closed
408
409 def _checkClosed(self, msg=None):
410 """Internal: raise an ValueError if file is closed
411 """
412 if self.closed:
413 raise ValueError("I/O operation on closed file."
414 if msg is None else msg)
415
416 ### Context manager ###
417
418 def __enter__(self) -> "IOBase": # That's a forward reference
419 """Context management protocol. Returns self."""
420 self._checkClosed()
421 return self
422
423 def __exit__(self, *args) -> None:
424 """Context management protocol. Calls close()"""
425 self.close()
426
427 ### Lower-level APIs ###
428
429 # XXX Should these be present even if unimplemented?
430
431 def fileno(self) -> int:
432 """Returns underlying file descriptor if one exists.
433
434 An IOError is raised if the IO object does not use a file descriptor.
435 """
436 self._unsupported("fileno")
437
438 def isatty(self) -> bool:
439 """Return whether this is an 'interactive' stream.
440
441 Return False if it can't be determined.
442 """
443 self._checkClosed()
444 return False
445
446 ### Readline[s] and writelines ###
447
448 def readline(self, limit: int = -1) -> bytes:
449 r"""Read and return a line from the stream.
450
451 If limit is specified, at most limit bytes will be read.
452
453 The line terminator is always b'\n' for binary files; for text
454 files, the newlines argument to open can be used to select the line
455 terminator(s) recognized.
456 """
457 # For backwards compatibility, a (slowish) readline().
458 if hasattr(self, "peek"):
459 def nreadahead():
460 readahead = self.peek(1)
461 if not readahead:
462 return 1
463 n = (readahead.find(b"\n") + 1) or len(readahead)
464 if limit >= 0:
465 n = min(n, limit)
466 return n
467 else:
468 def nreadahead():
469 return 1
470 if limit is None:
471 limit = -1
Benjamin Petersonb01138a2009-04-24 22:59:52 +0000472 elif not isinstance(limit, int):
473 raise TypeError("limit must be an integer")
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000474 res = bytearray()
475 while limit < 0 or len(res) < limit:
476 b = self.read(nreadahead())
477 if not b:
478 break
479 res += b
480 if res.endswith(b"\n"):
481 break
482 return bytes(res)
483
484 def __iter__(self):
485 self._checkClosed()
486 return self
487
488 def __next__(self):
489 line = self.readline()
490 if not line:
491 raise StopIteration
492 return line
493
494 def readlines(self, hint=None):
495 """Return a list of lines from the stream.
496
497 hint can be specified to control the number of lines read: no more
498 lines will be read if the total size (in bytes/characters) of all
499 lines so far exceeds hint.
500 """
501 if hint is None or hint <= 0:
502 return list(self)
503 n = 0
504 lines = []
505 for line in self:
506 lines.append(line)
507 n += len(line)
508 if n >= hint:
509 break
510 return lines
511
512 def writelines(self, lines):
513 self._checkClosed()
514 for line in lines:
515 self.write(line)
516
517io.IOBase.register(IOBase)
518
519
520class RawIOBase(IOBase):
521
522 """Base class for raw binary I/O."""
523
524 # The read() method is implemented by calling readinto(); derived
525 # classes that want to support read() only need to implement
526 # readinto() as a primitive operation. In general, readinto() can be
527 # more efficient than read().
528
529 # (It would be tempting to also provide an implementation of
530 # readinto() in terms of read(), in case the latter is a more suitable
531 # primitive operation, but that would lead to nasty recursion in case
532 # a subclass doesn't implement either.)
533
534 def read(self, n: int = -1) -> bytes:
535 """Read and return up to n bytes.
536
537 Returns an empty bytes object on EOF, or None if the object is
538 set not to block and has no data to read.
539 """
540 if n is None:
541 n = -1
542 if n < 0:
543 return self.readall()
544 b = bytearray(n.__index__())
545 n = self.readinto(b)
546 del b[n:]
547 return bytes(b)
548
549 def readall(self):
550 """Read until EOF, using multiple read() call."""
551 res = bytearray()
552 while True:
553 data = self.read(DEFAULT_BUFFER_SIZE)
554 if not data:
555 break
556 res += data
557 return bytes(res)
558
559 def readinto(self, b: bytearray) -> int:
560 """Read up to len(b) bytes into b.
561
562 Returns number of bytes read (0 for EOF), or None if the object
563 is set not to block as has no data to read.
564 """
565 self._unsupported("readinto")
566
567 def write(self, b: bytes) -> int:
568 """Write the given buffer to the IO stream.
569
570 Returns the number of bytes written, which may be less than len(b).
571 """
572 self._unsupported("write")
573
574io.RawIOBase.register(RawIOBase)
575from _io import FileIO
576RawIOBase.register(FileIO)
577
578
579class BufferedIOBase(IOBase):
580
581 """Base class for buffered IO objects.
582
583 The main difference with RawIOBase is that the read() method
584 supports omitting the size argument, and does not have a default
585 implementation that defers to readinto().
586
587 In addition, read(), readinto() and write() may raise
588 BlockingIOError if the underlying raw stream is in non-blocking
589 mode and not ready; unlike their raw counterparts, they will never
590 return None.
591
592 A typical implementation should not inherit from a RawIOBase
593 implementation, but wrap one.
594 """
595
596 def read(self, n: int = None) -> bytes:
597 """Read and return up to n bytes.
598
599 If the argument is omitted, None, or negative, reads and
600 returns all data until EOF.
601
602 If the argument is positive, and the underlying raw stream is
603 not 'interactive', multiple raw reads may be issued to satisfy
604 the byte count (unless EOF is reached first). But for
605 interactive raw streams (XXX and for pipes?), at most one raw
606 read will be issued, and a short result does not imply that
607 EOF is imminent.
608
609 Returns an empty bytes array on EOF.
610
611 Raises BlockingIOError if the underlying raw stream has no
612 data at the moment.
613 """
614 self._unsupported("read")
615
616 def read1(self, n: int=None) -> bytes:
617 """Read up to n bytes with at most one read() system call."""
618 self._unsupported("read1")
619
620 def readinto(self, b: bytearray) -> int:
621 """Read up to len(b) bytes into b.
622
623 Like read(), this may issue multiple reads to the underlying raw
624 stream, unless the latter is 'interactive'.
625
626 Returns the number of bytes read (0 for EOF).
627
628 Raises BlockingIOError if the underlying raw stream has no
629 data at the moment.
630 """
631 # XXX This ought to work with anything that supports the buffer API
632 data = self.read(len(b))
633 n = len(data)
634 try:
635 b[:n] = data
636 except TypeError as err:
637 import array
638 if not isinstance(b, array.array):
639 raise err
640 b[:n] = array.array('b', data)
641 return n
642
643 def write(self, b: bytes) -> int:
644 """Write the given buffer to the IO stream.
645
646 Return the number of bytes written, which is never less than
647 len(b).
648
649 Raises BlockingIOError if the buffer is full and the
650 underlying raw stream cannot accept more data at the moment.
651 """
652 self._unsupported("write")
653
Benjamin Petersond2e0c792009-05-01 20:40:59 +0000654 def detach(self) -> None:
655 """
656 Separate the underlying raw stream from the buffer and return it.
657
658 After the raw stream has been detached, the buffer is in an unusable
659 state.
660 """
661 self._unsupported("detach")
662
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000663io.BufferedIOBase.register(BufferedIOBase)
664
665
666class _BufferedIOMixin(BufferedIOBase):
667
668 """A mixin implementation of BufferedIOBase with an underlying raw stream.
669
670 This passes most requests on to the underlying raw stream. It
671 does *not* provide implementations of read(), readinto() or
672 write().
673 """
674
675 def __init__(self, raw):
676 self.raw = raw
677
678 ### Positioning ###
679
680 def seek(self, pos, whence=0):
681 new_position = self.raw.seek(pos, whence)
682 if new_position < 0:
683 raise IOError("seek() returned an invalid position")
684 return new_position
685
686 def tell(self):
687 pos = self.raw.tell()
688 if pos < 0:
689 raise IOError("tell() returned an invalid position")
690 return pos
691
692 def truncate(self, pos=None):
693 # Flush the stream. We're mixing buffered I/O with lower-level I/O,
694 # and a flush may be necessary to synch both views of the current
695 # file state.
696 self.flush()
697
698 if pos is None:
699 pos = self.tell()
700 # XXX: Should seek() be used, instead of passing the position
701 # XXX directly to truncate?
702 return self.raw.truncate(pos)
703
704 ### Flush and close ###
705
706 def flush(self):
Antoine Pitrou6be88762010-05-03 16:48:20 +0000707 if self.closed:
708 raise ValueError("flush of closed file")
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000709 self.raw.flush()
710
711 def close(self):
Antoine Pitrou6be88762010-05-03 16:48:20 +0000712 if self.raw is not None and not self.closed:
713 self.flush()
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000714 self.raw.close()
715
Benjamin Petersond2e0c792009-05-01 20:40:59 +0000716 def detach(self):
717 if self.raw is None:
718 raise ValueError("raw stream already detached")
719 self.flush()
720 raw = self.raw
721 self.raw = None
722 return raw
723
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000724 ### Inquiries ###
725
726 def seekable(self):
727 return self.raw.seekable()
728
729 def readable(self):
730 return self.raw.readable()
731
732 def writable(self):
733 return self.raw.writable()
734
735 @property
736 def closed(self):
737 return self.raw.closed
738
739 @property
740 def name(self):
741 return self.raw.name
742
743 @property
744 def mode(self):
745 return self.raw.mode
746
Antoine Pitrou716c4442009-05-23 19:04:03 +0000747 def __repr__(self):
748 clsname = self.__class__.__name__
749 try:
750 name = self.name
751 except AttributeError:
752 return "<_pyio.{0}>".format(clsname)
753 else:
754 return "<_pyio.{0} name={1!r}>".format(clsname, name)
755
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000756 ### Lower-level APIs ###
757
758 def fileno(self):
759 return self.raw.fileno()
760
761 def isatty(self):
762 return self.raw.isatty()
763
764
765class BytesIO(BufferedIOBase):
766
767 """Buffered I/O implementation using an in-memory bytes buffer."""
768
769 def __init__(self, initial_bytes=None):
770 buf = bytearray()
771 if initial_bytes is not None:
772 buf += initial_bytes
773 self._buffer = buf
774 self._pos = 0
775
Alexandre Vassalotticf76e1a2009-07-22 03:24:36 +0000776 def __getstate__(self):
777 if self.closed:
778 raise ValueError("__getstate__ on closed file")
779 return self.__dict__.copy()
780
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000781 def getvalue(self):
782 """Return the bytes value (contents) of the buffer
783 """
784 if self.closed:
785 raise ValueError("getvalue on closed file")
786 return bytes(self._buffer)
787
788 def read(self, n=None):
789 if self.closed:
790 raise ValueError("read from closed file")
791 if n is None:
792 n = -1
793 if n < 0:
794 n = len(self._buffer)
795 if len(self._buffer) <= self._pos:
796 return b""
797 newpos = min(len(self._buffer), self._pos + n)
798 b = self._buffer[self._pos : newpos]
799 self._pos = newpos
800 return bytes(b)
801
802 def read1(self, n):
803 """This is the same as read.
804 """
805 return self.read(n)
806
807 def write(self, b):
808 if self.closed:
809 raise ValueError("write to closed file")
810 if isinstance(b, str):
811 raise TypeError("can't write str to binary stream")
812 n = len(b)
813 if n == 0:
814 return 0
815 pos = self._pos
816 if pos > len(self._buffer):
817 # Inserts null bytes between the current end of the file
818 # and the new write position.
819 padding = b'\x00' * (pos - len(self._buffer))
820 self._buffer += padding
821 self._buffer[pos:pos + n] = b
822 self._pos += n
823 return n
824
825 def seek(self, pos, whence=0):
826 if self.closed:
827 raise ValueError("seek on closed file")
828 try:
Florent Xiclunab14930c2010-03-13 15:26:44 +0000829 pos.__index__
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000830 except AttributeError as err:
831 raise TypeError("an integer is required") from err
832 if whence == 0:
833 if pos < 0:
834 raise ValueError("negative seek position %r" % (pos,))
835 self._pos = pos
836 elif whence == 1:
837 self._pos = max(0, self._pos + pos)
838 elif whence == 2:
839 self._pos = max(0, len(self._buffer) + pos)
840 else:
841 raise ValueError("invalid whence value")
842 return self._pos
843
844 def tell(self):
845 if self.closed:
846 raise ValueError("tell on closed file")
847 return self._pos
848
849 def truncate(self, pos=None):
850 if self.closed:
851 raise ValueError("truncate on closed file")
852 if pos is None:
853 pos = self._pos
Florent Xiclunab14930c2010-03-13 15:26:44 +0000854 else:
855 try:
856 pos.__index__
857 except AttributeError as err:
858 raise TypeError("an integer is required") from err
859 if pos < 0:
860 raise ValueError("negative truncate position %r" % (pos,))
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000861 del self._buffer[pos:]
Antoine Pitrou905a2ff2010-01-31 22:47:27 +0000862 return pos
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000863
864 def readable(self):
865 return True
866
867 def writable(self):
868 return True
869
870 def seekable(self):
871 return True
872
873
874class BufferedReader(_BufferedIOMixin):
875
876 """BufferedReader(raw[, buffer_size])
877
878 A buffer for a readable, sequential BaseRawIO object.
879
880 The constructor creates a BufferedReader for the given readable raw
881 stream and buffer_size. If buffer_size is omitted, DEFAULT_BUFFER_SIZE
882 is used.
883 """
884
885 def __init__(self, raw, buffer_size=DEFAULT_BUFFER_SIZE):
886 """Create a new buffered reader using the given readable raw IO object.
887 """
Antoine Pitroucf4c7492009-04-19 00:09:36 +0000888 if not raw.readable():
889 raise IOError('"raw" argument must be readable.')
890
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000891 _BufferedIOMixin.__init__(self, raw)
892 if buffer_size <= 0:
893 raise ValueError("invalid buffer size")
894 self.buffer_size = buffer_size
895 self._reset_read_buf()
896 self._read_lock = Lock()
897
898 def _reset_read_buf(self):
899 self._read_buf = b""
900 self._read_pos = 0
901
902 def read(self, n=None):
903 """Read n bytes.
904
905 Returns exactly n bytes of data unless the underlying raw IO
906 stream reaches EOF or if the call would block in non-blocking
907 mode. If n is negative, read until EOF or until read() would
908 block.
909 """
910 if n is not None and n < -1:
911 raise ValueError("invalid number of bytes to read")
912 with self._read_lock:
913 return self._read_unlocked(n)
914
915 def _read_unlocked(self, n=None):
916 nodata_val = b""
917 empty_values = (b"", None)
918 buf = self._read_buf
919 pos = self._read_pos
920
921 # Special case for when the number of bytes to read is unspecified.
922 if n is None or n == -1:
923 self._reset_read_buf()
924 chunks = [buf[pos:]] # Strip the consumed bytes.
925 current_size = 0
926 while True:
927 # Read until EOF or until read() would block.
928 chunk = self.raw.read()
929 if chunk in empty_values:
930 nodata_val = chunk
931 break
932 current_size += len(chunk)
933 chunks.append(chunk)
934 return b"".join(chunks) or nodata_val
935
936 # The number of bytes to read is specified, return at most n bytes.
937 avail = len(buf) - pos # Length of the available buffered data.
938 if n <= avail:
939 # Fast path: the data to read is fully buffered.
940 self._read_pos += n
941 return buf[pos:pos+n]
942 # Slow path: read from the stream until enough bytes are read,
943 # or until an EOF occurs or until read() would block.
944 chunks = [buf[pos:]]
945 wanted = max(self.buffer_size, n)
946 while avail < n:
947 chunk = self.raw.read(wanted)
948 if chunk in empty_values:
949 nodata_val = chunk
950 break
951 avail += len(chunk)
952 chunks.append(chunk)
953 # n is more then avail only when an EOF occurred or when
954 # read() would have blocked.
955 n = min(n, avail)
956 out = b"".join(chunks)
957 self._read_buf = out[n:] # Save the extra data in the buffer.
958 self._read_pos = 0
959 return out[:n] if out else nodata_val
960
961 def peek(self, n=0):
962 """Returns buffered bytes without advancing the position.
963
964 The argument indicates a desired minimal number of bytes; we
965 do at most one raw read to satisfy it. We never return more
966 than self.buffer_size.
967 """
968 with self._read_lock:
969 return self._peek_unlocked(n)
970
971 def _peek_unlocked(self, n=0):
972 want = min(n, self.buffer_size)
973 have = len(self._read_buf) - self._read_pos
974 if have < want or have <= 0:
975 to_read = self.buffer_size - have
976 current = self.raw.read(to_read)
977 if current:
978 self._read_buf = self._read_buf[self._read_pos:] + current
979 self._read_pos = 0
980 return self._read_buf[self._read_pos:]
981
982 def read1(self, n):
983 """Reads up to n bytes, with at most one read() system call."""
984 # Returns up to n bytes. If at least one byte is buffered, we
985 # only return buffered bytes. Otherwise, we do one raw read.
986 if n < 0:
987 raise ValueError("number of bytes to read must be positive")
988 if n == 0:
989 return b""
990 with self._read_lock:
991 self._peek_unlocked(1)
992 return self._read_unlocked(
993 min(n, len(self._read_buf) - self._read_pos))
994
995 def tell(self):
996 return _BufferedIOMixin.tell(self) - len(self._read_buf) + self._read_pos
997
998 def seek(self, pos, whence=0):
999 if not (0 <= whence <= 2):
1000 raise ValueError("invalid whence value")
1001 with self._read_lock:
1002 if whence == 1:
1003 pos -= len(self._read_buf) - self._read_pos
1004 pos = _BufferedIOMixin.seek(self, pos, whence)
1005 self._reset_read_buf()
1006 return pos
1007
1008class BufferedWriter(_BufferedIOMixin):
1009
1010 """A buffer for a writeable sequential RawIO object.
1011
1012 The constructor creates a BufferedWriter for the given writeable raw
1013 stream. If the buffer_size is not given, it defaults to
Benjamin Peterson59406a92009-03-26 17:10:29 +00001014 DEFAULT_BUFFER_SIZE.
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001015 """
1016
Benjamin Peterson59406a92009-03-26 17:10:29 +00001017 _warning_stack_offset = 2
1018
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001019 def __init__(self, raw,
1020 buffer_size=DEFAULT_BUFFER_SIZE, max_buffer_size=None):
Antoine Pitroucf4c7492009-04-19 00:09:36 +00001021 if not raw.writable():
1022 raise IOError('"raw" argument must be writable.')
1023
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001024 _BufferedIOMixin.__init__(self, raw)
1025 if buffer_size <= 0:
1026 raise ValueError("invalid buffer size")
Benjamin Peterson59406a92009-03-26 17:10:29 +00001027 if max_buffer_size is not None:
1028 warnings.warn("max_buffer_size is deprecated", DeprecationWarning,
1029 self._warning_stack_offset)
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001030 self.buffer_size = buffer_size
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001031 self._write_buf = bytearray()
1032 self._write_lock = Lock()
1033
1034 def write(self, b):
1035 if self.closed:
1036 raise ValueError("write to closed file")
1037 if isinstance(b, str):
1038 raise TypeError("can't write str to binary stream")
1039 with self._write_lock:
1040 # XXX we can implement some more tricks to try and avoid
1041 # partial writes
1042 if len(self._write_buf) > self.buffer_size:
1043 # We're full, so let's pre-flush the buffer
1044 try:
1045 self._flush_unlocked()
1046 except BlockingIOError as e:
1047 # We can't accept anything else.
1048 # XXX Why not just let the exception pass through?
1049 raise BlockingIOError(e.errno, e.strerror, 0)
1050 before = len(self._write_buf)
1051 self._write_buf.extend(b)
1052 written = len(self._write_buf) - before
1053 if len(self._write_buf) > self.buffer_size:
1054 try:
1055 self._flush_unlocked()
1056 except BlockingIOError as e:
Benjamin Peterson394ee002009-03-05 22:33:59 +00001057 if len(self._write_buf) > self.buffer_size:
1058 # We've hit the buffer_size. We have to accept a partial
1059 # write and cut back our buffer.
1060 overage = len(self._write_buf) - self.buffer_size
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001061 written -= overage
Benjamin Peterson394ee002009-03-05 22:33:59 +00001062 self._write_buf = self._write_buf[:self.buffer_size]
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001063 raise BlockingIOError(e.errno, e.strerror, written)
1064 return written
1065
1066 def truncate(self, pos=None):
1067 with self._write_lock:
1068 self._flush_unlocked()
1069 if pos is None:
1070 pos = self.raw.tell()
1071 return self.raw.truncate(pos)
1072
1073 def flush(self):
1074 with self._write_lock:
1075 self._flush_unlocked()
1076
1077 def _flush_unlocked(self):
1078 if self.closed:
1079 raise ValueError("flush of closed file")
1080 written = 0
1081 try:
1082 while self._write_buf:
1083 n = self.raw.write(self._write_buf)
1084 if n > len(self._write_buf) or n < 0:
1085 raise IOError("write() returned incorrect number of bytes")
1086 del self._write_buf[:n]
1087 written += n
1088 except BlockingIOError as e:
1089 n = e.characters_written
1090 del self._write_buf[:n]
1091 written += n
1092 raise BlockingIOError(e.errno, e.strerror, written)
1093
1094 def tell(self):
1095 return _BufferedIOMixin.tell(self) + len(self._write_buf)
1096
1097 def seek(self, pos, whence=0):
1098 if not (0 <= whence <= 2):
1099 raise ValueError("invalid whence")
1100 with self._write_lock:
1101 self._flush_unlocked()
1102 return _BufferedIOMixin.seek(self, pos, whence)
1103
1104
1105class BufferedRWPair(BufferedIOBase):
1106
1107 """A buffered reader and writer object together.
1108
1109 A buffered reader object and buffered writer object put together to
1110 form a sequential IO object that can read and write. This is typically
1111 used with a socket or two-way pipe.
1112
1113 reader and writer are RawIOBase objects that are readable and
1114 writeable respectively. If the buffer_size is omitted it defaults to
Benjamin Peterson59406a92009-03-26 17:10:29 +00001115 DEFAULT_BUFFER_SIZE.
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001116 """
1117
1118 # XXX The usefulness of this (compared to having two separate IO
1119 # objects) is questionable.
1120
1121 def __init__(self, reader, writer,
1122 buffer_size=DEFAULT_BUFFER_SIZE, max_buffer_size=None):
1123 """Constructor.
1124
1125 The arguments are two RawIO instances.
1126 """
Benjamin Peterson59406a92009-03-26 17:10:29 +00001127 if max_buffer_size is not None:
1128 warnings.warn("max_buffer_size is deprecated", DeprecationWarning, 2)
Antoine Pitroucf4c7492009-04-19 00:09:36 +00001129
1130 if not reader.readable():
1131 raise IOError('"reader" argument must be readable.')
1132
1133 if not writer.writable():
1134 raise IOError('"writer" argument must be writable.')
1135
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001136 self.reader = BufferedReader(reader, buffer_size)
Benjamin Peterson59406a92009-03-26 17:10:29 +00001137 self.writer = BufferedWriter(writer, buffer_size)
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001138
1139 def read(self, n=None):
1140 if n is None:
1141 n = -1
1142 return self.reader.read(n)
1143
1144 def readinto(self, b):
1145 return self.reader.readinto(b)
1146
1147 def write(self, b):
1148 return self.writer.write(b)
1149
1150 def peek(self, n=0):
1151 return self.reader.peek(n)
1152
1153 def read1(self, n):
1154 return self.reader.read1(n)
1155
1156 def readable(self):
1157 return self.reader.readable()
1158
1159 def writable(self):
1160 return self.writer.writable()
1161
1162 def flush(self):
1163 return self.writer.flush()
1164
1165 def close(self):
1166 self.writer.close()
1167 self.reader.close()
1168
1169 def isatty(self):
1170 return self.reader.isatty() or self.writer.isatty()
1171
1172 @property
1173 def closed(self):
1174 return self.writer.closed
1175
1176
1177class BufferedRandom(BufferedWriter, BufferedReader):
1178
1179 """A buffered interface to random access streams.
1180
1181 The constructor creates a reader and writer for a seekable stream,
1182 raw, given in the first argument. If the buffer_size is omitted it
Benjamin Peterson59406a92009-03-26 17:10:29 +00001183 defaults to DEFAULT_BUFFER_SIZE.
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001184 """
1185
Benjamin Peterson59406a92009-03-26 17:10:29 +00001186 _warning_stack_offset = 3
1187
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001188 def __init__(self, raw,
1189 buffer_size=DEFAULT_BUFFER_SIZE, max_buffer_size=None):
1190 raw._checkSeekable()
1191 BufferedReader.__init__(self, raw, buffer_size)
1192 BufferedWriter.__init__(self, raw, buffer_size, max_buffer_size)
1193
1194 def seek(self, pos, whence=0):
1195 if not (0 <= whence <= 2):
1196 raise ValueError("invalid whence")
1197 self.flush()
1198 if self._read_buf:
1199 # Undo read ahead.
1200 with self._read_lock:
1201 self.raw.seek(self._read_pos - len(self._read_buf), 1)
1202 # First do the raw seek, then empty the read buffer, so that
1203 # if the raw seek fails, we don't lose buffered data forever.
1204 pos = self.raw.seek(pos, whence)
1205 with self._read_lock:
1206 self._reset_read_buf()
1207 if pos < 0:
1208 raise IOError("seek() returned invalid position")
1209 return pos
1210
1211 def tell(self):
1212 if self._write_buf:
1213 return BufferedWriter.tell(self)
1214 else:
1215 return BufferedReader.tell(self)
1216
1217 def truncate(self, pos=None):
1218 if pos is None:
1219 pos = self.tell()
1220 # Use seek to flush the read buffer.
Antoine Pitrou905a2ff2010-01-31 22:47:27 +00001221 return BufferedWriter.truncate(self, pos)
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001222
1223 def read(self, n=None):
1224 if n is None:
1225 n = -1
1226 self.flush()
1227 return BufferedReader.read(self, n)
1228
1229 def readinto(self, b):
1230 self.flush()
1231 return BufferedReader.readinto(self, b)
1232
1233 def peek(self, n=0):
1234 self.flush()
1235 return BufferedReader.peek(self, n)
1236
1237 def read1(self, n):
1238 self.flush()
1239 return BufferedReader.read1(self, n)
1240
1241 def write(self, b):
1242 if self._read_buf:
1243 # Undo readahead
1244 with self._read_lock:
1245 self.raw.seek(self._read_pos - len(self._read_buf), 1)
1246 self._reset_read_buf()
1247 return BufferedWriter.write(self, b)
1248
1249
1250class TextIOBase(IOBase):
1251
1252 """Base class for text I/O.
1253
1254 This class provides a character and line based interface to stream
1255 I/O. There is no readinto method because Python's character strings
1256 are immutable. There is no public constructor.
1257 """
1258
1259 def read(self, n: int = -1) -> str:
1260 """Read at most n characters from stream.
1261
1262 Read from underlying buffer until we have n characters or we hit EOF.
1263 If n is negative or omitted, read until EOF.
1264 """
1265 self._unsupported("read")
1266
1267 def write(self, s: str) -> int:
1268 """Write string s to stream."""
1269 self._unsupported("write")
1270
1271 def truncate(self, pos: int = None) -> int:
1272 """Truncate size to pos."""
1273 self._unsupported("truncate")
1274
1275 def readline(self) -> str:
1276 """Read until newline or EOF.
1277
1278 Returns an empty string if EOF is hit immediately.
1279 """
1280 self._unsupported("readline")
1281
Benjamin Petersond2e0c792009-05-01 20:40:59 +00001282 def detach(self) -> None:
1283 """
1284 Separate the underlying buffer from the TextIOBase and return it.
1285
1286 After the underlying buffer has been detached, the TextIO is in an
1287 unusable state.
1288 """
1289 self._unsupported("detach")
1290
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001291 @property
1292 def encoding(self):
1293 """Subclasses should override."""
1294 return None
1295
1296 @property
1297 def newlines(self):
1298 """Line endings translated so far.
1299
1300 Only line endings translated during reading are considered.
1301
1302 Subclasses should override.
1303 """
1304 return None
1305
Benjamin Peterson0926ad12009-06-06 18:02:12 +00001306 @property
1307 def errors(self):
1308 """Error setting of the decoder or encoder.
1309
1310 Subclasses should override."""
1311 return None
1312
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001313io.TextIOBase.register(TextIOBase)
1314
1315
1316class IncrementalNewlineDecoder(codecs.IncrementalDecoder):
1317 r"""Codec used when reading a file in universal newlines mode. It wraps
1318 another incremental decoder, translating \r\n and \r into \n. It also
1319 records the types of newlines encountered. When used with
1320 translate=False, it ensures that the newline sequence is returned in
1321 one piece.
1322 """
1323 def __init__(self, decoder, translate, errors='strict'):
1324 codecs.IncrementalDecoder.__init__(self, errors=errors)
1325 self.translate = translate
1326 self.decoder = decoder
1327 self.seennl = 0
1328 self.pendingcr = False
1329
1330 def decode(self, input, final=False):
1331 # decode input (with the eventual \r from a previous pass)
1332 if self.decoder is None:
1333 output = input
1334 else:
1335 output = self.decoder.decode(input, final=final)
1336 if self.pendingcr and (output or final):
1337 output = "\r" + output
1338 self.pendingcr = False
1339
1340 # retain last \r even when not translating data:
1341 # then readline() is sure to get \r\n in one pass
1342 if output.endswith("\r") and not final:
1343 output = output[:-1]
1344 self.pendingcr = True
1345
1346 # Record which newlines are read
1347 crlf = output.count('\r\n')
1348 cr = output.count('\r') - crlf
1349 lf = output.count('\n') - crlf
1350 self.seennl |= (lf and self._LF) | (cr and self._CR) \
1351 | (crlf and self._CRLF)
1352
1353 if self.translate:
1354 if crlf:
1355 output = output.replace("\r\n", "\n")
1356 if cr:
1357 output = output.replace("\r", "\n")
1358
1359 return output
1360
1361 def getstate(self):
1362 if self.decoder is None:
1363 buf = b""
1364 flag = 0
1365 else:
1366 buf, flag = self.decoder.getstate()
1367 flag <<= 1
1368 if self.pendingcr:
1369 flag |= 1
1370 return buf, flag
1371
1372 def setstate(self, state):
1373 buf, flag = state
1374 self.pendingcr = bool(flag & 1)
1375 if self.decoder is not None:
1376 self.decoder.setstate((buf, flag >> 1))
1377
1378 def reset(self):
1379 self.seennl = 0
1380 self.pendingcr = False
1381 if self.decoder is not None:
1382 self.decoder.reset()
1383
1384 _LF = 1
1385 _CR = 2
1386 _CRLF = 4
1387
1388 @property
1389 def newlines(self):
1390 return (None,
1391 "\n",
1392 "\r",
1393 ("\r", "\n"),
1394 "\r\n",
1395 ("\n", "\r\n"),
1396 ("\r", "\r\n"),
1397 ("\r", "\n", "\r\n")
1398 )[self.seennl]
1399
1400
1401class TextIOWrapper(TextIOBase):
1402
1403 r"""Character and line based layer over a BufferedIOBase object, buffer.
1404
1405 encoding gives the name of the encoding that the stream will be
1406 decoded or encoded with. It defaults to locale.getpreferredencoding.
1407
1408 errors determines the strictness of encoding and decoding (see the
1409 codecs.register) and defaults to "strict".
1410
1411 newline can be None, '', '\n', '\r', or '\r\n'. It controls the
1412 handling of line endings. If it is None, universal newlines is
1413 enabled. With this enabled, on input, the lines endings '\n', '\r',
1414 or '\r\n' are translated to '\n' before being returned to the
1415 caller. Conversely, on output, '\n' is translated to the system
1416 default line seperator, os.linesep. If newline is any other of its
1417 legal values, that newline becomes the newline when the file is read
1418 and it is returned untranslated. On output, '\n' is converted to the
1419 newline.
1420
1421 If line_buffering is True, a call to flush is implied when a call to
1422 write contains a newline character.
1423 """
1424
1425 _CHUNK_SIZE = 2048
1426
1427 def __init__(self, buffer, encoding=None, errors=None, newline=None,
1428 line_buffering=False):
1429 if newline is not None and not isinstance(newline, str):
1430 raise TypeError("illegal newline type: %r" % (type(newline),))
1431 if newline not in (None, "", "\n", "\r", "\r\n"):
1432 raise ValueError("illegal newline value: %r" % (newline,))
1433 if encoding is None:
1434 try:
1435 encoding = os.device_encoding(buffer.fileno())
1436 except (AttributeError, UnsupportedOperation):
1437 pass
1438 if encoding is None:
1439 try:
1440 import locale
1441 except ImportError:
1442 # Importing locale may fail if Python is being built
1443 encoding = "ascii"
1444 else:
1445 encoding = locale.getpreferredencoding()
1446
1447 if not isinstance(encoding, str):
1448 raise ValueError("invalid encoding: %r" % encoding)
1449
1450 if errors is None:
1451 errors = "strict"
1452 else:
1453 if not isinstance(errors, str):
1454 raise ValueError("invalid errors: %r" % errors)
1455
1456 self.buffer = buffer
1457 self._line_buffering = line_buffering
1458 self._encoding = encoding
1459 self._errors = errors
1460 self._readuniversal = not newline
1461 self._readtranslate = newline is None
1462 self._readnl = newline
1463 self._writetranslate = newline != ''
1464 self._writenl = newline or os.linesep
1465 self._encoder = None
1466 self._decoder = None
1467 self._decoded_chars = '' # buffer for text returned from decoder
1468 self._decoded_chars_used = 0 # offset into _decoded_chars for read()
1469 self._snapshot = None # info for reconstructing decoder state
1470 self._seekable = self._telling = self.buffer.seekable()
1471
Antoine Pitroue4501852009-05-14 18:55:55 +00001472 if self._seekable and self.writable():
1473 position = self.buffer.tell()
1474 if position != 0:
1475 try:
1476 self._get_encoder().setstate(0)
1477 except LookupError:
1478 # Sometimes the encoder doesn't exist
1479 pass
1480
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001481 # self._snapshot is either None, or a tuple (dec_flags, next_input)
1482 # where dec_flags is the second (integer) item of the decoder state
1483 # and next_input is the chunk of input bytes that comes next after the
1484 # snapshot point. We use this to reconstruct decoder states in tell().
1485
1486 # Naming convention:
1487 # - "bytes_..." for integer variables that count input bytes
1488 # - "chars_..." for integer variables that count decoded characters
1489
Benjamin Petersonc4c0eae2009-03-09 00:07:03 +00001490 def __repr__(self):
Antoine Pitrou716c4442009-05-23 19:04:03 +00001491 try:
1492 name = self.name
1493 except AttributeError:
1494 return "<_pyio.TextIOWrapper encoding={0!r}>".format(self.encoding)
1495 else:
1496 return "<_pyio.TextIOWrapper name={0!r} encoding={1!r}>".format(
1497 name, self.encoding)
Benjamin Petersonc4c0eae2009-03-09 00:07:03 +00001498
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001499 @property
1500 def encoding(self):
1501 return self._encoding
1502
1503 @property
1504 def errors(self):
1505 return self._errors
1506
1507 @property
1508 def line_buffering(self):
1509 return self._line_buffering
1510
1511 def seekable(self):
1512 return self._seekable
1513
1514 def readable(self):
1515 return self.buffer.readable()
1516
1517 def writable(self):
1518 return self.buffer.writable()
1519
1520 def flush(self):
1521 self.buffer.flush()
1522 self._telling = self._seekable
1523
1524 def close(self):
Antoine Pitrou6be88762010-05-03 16:48:20 +00001525 if self.buffer is not None and not self.closed:
1526 self.flush()
Benjamin Petersond2e0c792009-05-01 20:40:59 +00001527 self.buffer.close()
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001528
1529 @property
1530 def closed(self):
1531 return self.buffer.closed
1532
1533 @property
1534 def name(self):
1535 return self.buffer.name
1536
1537 def fileno(self):
1538 return self.buffer.fileno()
1539
1540 def isatty(self):
1541 return self.buffer.isatty()
1542
1543 def write(self, s: str):
1544 if self.closed:
1545 raise ValueError("write to closed file")
1546 if not isinstance(s, str):
1547 raise TypeError("can't write %s to text stream" %
1548 s.__class__.__name__)
1549 length = len(s)
1550 haslf = (self._writetranslate or self._line_buffering) and "\n" in s
1551 if haslf and self._writetranslate and self._writenl != "\n":
1552 s = s.replace("\n", self._writenl)
1553 encoder = self._encoder or self._get_encoder()
1554 # XXX What if we were just reading?
1555 b = encoder.encode(s)
1556 self.buffer.write(b)
1557 if self._line_buffering and (haslf or "\r" in s):
1558 self.flush()
1559 self._snapshot = None
1560 if self._decoder:
1561 self._decoder.reset()
1562 return length
1563
1564 def _get_encoder(self):
1565 make_encoder = codecs.getincrementalencoder(self._encoding)
1566 self._encoder = make_encoder(self._errors)
1567 return self._encoder
1568
1569 def _get_decoder(self):
1570 make_decoder = codecs.getincrementaldecoder(self._encoding)
1571 decoder = make_decoder(self._errors)
1572 if self._readuniversal:
1573 decoder = IncrementalNewlineDecoder(decoder, self._readtranslate)
1574 self._decoder = decoder
1575 return decoder
1576
1577 # The following three methods implement an ADT for _decoded_chars.
1578 # Text returned from the decoder is buffered here until the client
1579 # requests it by calling our read() or readline() method.
1580 def _set_decoded_chars(self, chars):
1581 """Set the _decoded_chars buffer."""
1582 self._decoded_chars = chars
1583 self._decoded_chars_used = 0
1584
1585 def _get_decoded_chars(self, n=None):
1586 """Advance into the _decoded_chars buffer."""
1587 offset = self._decoded_chars_used
1588 if n is None:
1589 chars = self._decoded_chars[offset:]
1590 else:
1591 chars = self._decoded_chars[offset:offset + n]
1592 self._decoded_chars_used += len(chars)
1593 return chars
1594
1595 def _rewind_decoded_chars(self, n):
1596 """Rewind the _decoded_chars buffer."""
1597 if self._decoded_chars_used < n:
1598 raise AssertionError("rewind decoded_chars out of bounds")
1599 self._decoded_chars_used -= n
1600
1601 def _read_chunk(self):
1602 """
1603 Read and decode the next chunk of data from the BufferedReader.
1604 """
1605
1606 # The return value is True unless EOF was reached. The decoded
1607 # string is placed in self._decoded_chars (replacing its previous
1608 # value). The entire input chunk is sent to the decoder, though
1609 # some of it may remain buffered in the decoder, yet to be
1610 # converted.
1611
1612 if self._decoder is None:
1613 raise ValueError("no decoder")
1614
1615 if self._telling:
1616 # To prepare for tell(), we need to snapshot a point in the
1617 # file where the decoder's input buffer is empty.
1618
1619 dec_buffer, dec_flags = self._decoder.getstate()
1620 # Given this, we know there was a valid snapshot point
1621 # len(dec_buffer) bytes ago with decoder state (b'', dec_flags).
1622
1623 # Read a chunk, decode it, and put the result in self._decoded_chars.
1624 input_chunk = self.buffer.read1(self._CHUNK_SIZE)
1625 eof = not input_chunk
1626 self._set_decoded_chars(self._decoder.decode(input_chunk, eof))
1627
1628 if self._telling:
1629 # At the snapshot point, len(dec_buffer) bytes before the read,
1630 # the next input to be decoded is dec_buffer + input_chunk.
1631 self._snapshot = (dec_flags, dec_buffer + input_chunk)
1632
1633 return not eof
1634
1635 def _pack_cookie(self, position, dec_flags=0,
1636 bytes_to_feed=0, need_eof=0, chars_to_skip=0):
1637 # The meaning of a tell() cookie is: seek to position, set the
1638 # decoder flags to dec_flags, read bytes_to_feed bytes, feed them
1639 # into the decoder with need_eof as the EOF flag, then skip
1640 # chars_to_skip characters of the decoded result. For most simple
1641 # decoders, tell() will often just give a byte offset in the file.
1642 return (position | (dec_flags<<64) | (bytes_to_feed<<128) |
1643 (chars_to_skip<<192) | bool(need_eof)<<256)
1644
1645 def _unpack_cookie(self, bigint):
1646 rest, position = divmod(bigint, 1<<64)
1647 rest, dec_flags = divmod(rest, 1<<64)
1648 rest, bytes_to_feed = divmod(rest, 1<<64)
1649 need_eof, chars_to_skip = divmod(rest, 1<<64)
1650 return position, dec_flags, bytes_to_feed, need_eof, chars_to_skip
1651
1652 def tell(self):
1653 if not self._seekable:
Antoine Pitrou0d739d72010-09-05 23:01:12 +00001654 raise UnsupportedOperation("underlying stream is not seekable")
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001655 if not self._telling:
1656 raise IOError("telling position disabled by next() call")
1657 self.flush()
1658 position = self.buffer.tell()
1659 decoder = self._decoder
1660 if decoder is None or self._snapshot is None:
1661 if self._decoded_chars:
1662 # This should never happen.
1663 raise AssertionError("pending decoded text")
1664 return position
1665
1666 # Skip backward to the snapshot point (see _read_chunk).
1667 dec_flags, next_input = self._snapshot
1668 position -= len(next_input)
1669
1670 # How many decoded characters have been used up since the snapshot?
1671 chars_to_skip = self._decoded_chars_used
1672 if chars_to_skip == 0:
1673 # We haven't moved from the snapshot point.
1674 return self._pack_cookie(position, dec_flags)
1675
1676 # Starting from the snapshot position, we will walk the decoder
1677 # forward until it gives us enough decoded characters.
1678 saved_state = decoder.getstate()
1679 try:
1680 # Note our initial start point.
1681 decoder.setstate((b'', dec_flags))
1682 start_pos = position
1683 start_flags, bytes_fed, chars_decoded = dec_flags, 0, 0
1684 need_eof = 0
1685
1686 # Feed the decoder one byte at a time. As we go, note the
1687 # nearest "safe start point" before the current location
1688 # (a point where the decoder has nothing buffered, so seek()
1689 # can safely start from there and advance to this location).
1690 next_byte = bytearray(1)
1691 for next_byte[0] in next_input:
1692 bytes_fed += 1
1693 chars_decoded += len(decoder.decode(next_byte))
1694 dec_buffer, dec_flags = decoder.getstate()
1695 if not dec_buffer and chars_decoded <= chars_to_skip:
1696 # Decoder buffer is empty, so this is a safe start point.
1697 start_pos += bytes_fed
1698 chars_to_skip -= chars_decoded
1699 start_flags, bytes_fed, chars_decoded = dec_flags, 0, 0
1700 if chars_decoded >= chars_to_skip:
1701 break
1702 else:
1703 # We didn't get enough decoded data; signal EOF to get more.
1704 chars_decoded += len(decoder.decode(b'', final=True))
1705 need_eof = 1
1706 if chars_decoded < chars_to_skip:
1707 raise IOError("can't reconstruct logical file position")
1708
1709 # The returned cookie corresponds to the last safe start point.
1710 return self._pack_cookie(
1711 start_pos, start_flags, bytes_fed, need_eof, chars_to_skip)
1712 finally:
1713 decoder.setstate(saved_state)
1714
1715 def truncate(self, pos=None):
1716 self.flush()
1717 if pos is None:
1718 pos = self.tell()
Antoine Pitrou905a2ff2010-01-31 22:47:27 +00001719 return self.buffer.truncate(pos)
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001720
Benjamin Petersond2e0c792009-05-01 20:40:59 +00001721 def detach(self):
1722 if self.buffer is None:
1723 raise ValueError("buffer is already detached")
1724 self.flush()
1725 buffer = self.buffer
1726 self.buffer = None
1727 return buffer
1728
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001729 def seek(self, cookie, whence=0):
1730 if self.closed:
1731 raise ValueError("tell on closed file")
1732 if not self._seekable:
Antoine Pitrou0d739d72010-09-05 23:01:12 +00001733 raise UnsupportedOperation("underlying stream is not seekable")
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001734 if whence == 1: # seek relative to current position
1735 if cookie != 0:
Antoine Pitrou0d739d72010-09-05 23:01:12 +00001736 raise UnsupportedOperation("can't do nonzero cur-relative seeks")
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001737 # Seeking to the current position should attempt to
1738 # sync the underlying buffer with the current position.
1739 whence = 0
1740 cookie = self.tell()
1741 if whence == 2: # seek relative to end of file
1742 if cookie != 0:
Antoine Pitrou0d739d72010-09-05 23:01:12 +00001743 raise UnsupportedOperation("can't do nonzero end-relative seeks")
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001744 self.flush()
1745 position = self.buffer.seek(0, 2)
1746 self._set_decoded_chars('')
1747 self._snapshot = None
1748 if self._decoder:
1749 self._decoder.reset()
1750 return position
1751 if whence != 0:
1752 raise ValueError("invalid whence (%r, should be 0, 1 or 2)" %
1753 (whence,))
1754 if cookie < 0:
1755 raise ValueError("negative seek position %r" % (cookie,))
1756 self.flush()
1757
1758 # The strategy of seek() is to go back to the safe start point
1759 # and replay the effect of read(chars_to_skip) from there.
1760 start_pos, dec_flags, bytes_to_feed, need_eof, chars_to_skip = \
1761 self._unpack_cookie(cookie)
1762
1763 # Seek back to the safe start point.
1764 self.buffer.seek(start_pos)
1765 self._set_decoded_chars('')
1766 self._snapshot = None
1767
1768 # Restore the decoder to its state from the safe start point.
Benjamin Peterson9363a652009-03-05 00:42:09 +00001769 if cookie == 0 and self._decoder:
1770 self._decoder.reset()
1771 elif self._decoder or dec_flags or chars_to_skip:
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001772 self._decoder = self._decoder or self._get_decoder()
1773 self._decoder.setstate((b'', dec_flags))
1774 self._snapshot = (dec_flags, b'')
1775
1776 if chars_to_skip:
1777 # Just like _read_chunk, feed the decoder and save a snapshot.
1778 input_chunk = self.buffer.read(bytes_to_feed)
1779 self._set_decoded_chars(
1780 self._decoder.decode(input_chunk, need_eof))
1781 self._snapshot = (dec_flags, input_chunk)
1782
1783 # Skip chars_to_skip of the decoded characters.
1784 if len(self._decoded_chars) < chars_to_skip:
1785 raise IOError("can't restore logical file position")
1786 self._decoded_chars_used = chars_to_skip
1787
Antoine Pitroue4501852009-05-14 18:55:55 +00001788 # Finally, reset the encoder (merely useful for proper BOM handling)
1789 try:
1790 encoder = self._encoder or self._get_encoder()
1791 except LookupError:
1792 # Sometimes the encoder doesn't exist
1793 pass
1794 else:
1795 if cookie != 0:
1796 encoder.setstate(0)
1797 else:
1798 encoder.reset()
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001799 return cookie
1800
1801 def read(self, n=None):
Benjamin Petersona1b49012009-03-31 23:11:32 +00001802 self._checkReadable()
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001803 if n is None:
1804 n = -1
1805 decoder = self._decoder or self._get_decoder()
Florent Xiclunab14930c2010-03-13 15:26:44 +00001806 try:
1807 n.__index__
1808 except AttributeError as err:
1809 raise TypeError("an integer is required") from err
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001810 if n < 0:
1811 # Read everything.
1812 result = (self._get_decoded_chars() +
1813 decoder.decode(self.buffer.read(), final=True))
1814 self._set_decoded_chars('')
1815 self._snapshot = None
1816 return result
1817 else:
1818 # Keep reading chunks until we have n characters to return.
1819 eof = False
1820 result = self._get_decoded_chars(n)
1821 while len(result) < n and not eof:
1822 eof = not self._read_chunk()
1823 result += self._get_decoded_chars(n - len(result))
1824 return result
1825
1826 def __next__(self):
1827 self._telling = False
1828 line = self.readline()
1829 if not line:
1830 self._snapshot = None
1831 self._telling = self._seekable
1832 raise StopIteration
1833 return line
1834
1835 def readline(self, limit=None):
1836 if self.closed:
1837 raise ValueError("read from closed file")
1838 if limit is None:
1839 limit = -1
Benjamin Petersonb01138a2009-04-24 22:59:52 +00001840 elif not isinstance(limit, int):
1841 raise TypeError("limit must be an integer")
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001842
1843 # Grab all the decoded text (we will rewind any extra bits later).
1844 line = self._get_decoded_chars()
1845
1846 start = 0
1847 # Make the decoder if it doesn't already exist.
1848 if not self._decoder:
1849 self._get_decoder()
1850
1851 pos = endpos = None
1852 while True:
1853 if self._readtranslate:
1854 # Newlines are already translated, only search for \n
1855 pos = line.find('\n', start)
1856 if pos >= 0:
1857 endpos = pos + 1
1858 break
1859 else:
1860 start = len(line)
1861
1862 elif self._readuniversal:
1863 # Universal newline search. Find any of \r, \r\n, \n
1864 # The decoder ensures that \r\n are not split in two pieces
1865
1866 # In C we'd look for these in parallel of course.
1867 nlpos = line.find("\n", start)
1868 crpos = line.find("\r", start)
1869 if crpos == -1:
1870 if nlpos == -1:
1871 # Nothing found
1872 start = len(line)
1873 else:
1874 # Found \n
1875 endpos = nlpos + 1
1876 break
1877 elif nlpos == -1:
1878 # Found lone \r
1879 endpos = crpos + 1
1880 break
1881 elif nlpos < crpos:
1882 # Found \n
1883 endpos = nlpos + 1
1884 break
1885 elif nlpos == crpos + 1:
1886 # Found \r\n
1887 endpos = crpos + 2
1888 break
1889 else:
1890 # Found \r
1891 endpos = crpos + 1
1892 break
1893 else:
1894 # non-universal
1895 pos = line.find(self._readnl)
1896 if pos >= 0:
1897 endpos = pos + len(self._readnl)
1898 break
1899
1900 if limit >= 0 and len(line) >= limit:
1901 endpos = limit # reached length limit
1902 break
1903
1904 # No line ending seen yet - get more data'
1905 while self._read_chunk():
1906 if self._decoded_chars:
1907 break
1908 if self._decoded_chars:
1909 line += self._get_decoded_chars()
1910 else:
1911 # end of file
1912 self._set_decoded_chars('')
1913 self._snapshot = None
1914 return line
1915
1916 if limit >= 0 and endpos > limit:
1917 endpos = limit # don't exceed limit
1918
1919 # Rewind _decoded_chars to just after the line ending we found.
1920 self._rewind_decoded_chars(len(line) - endpos)
1921 return line[:endpos]
1922
1923 @property
1924 def newlines(self):
1925 return self._decoder.newlines if self._decoder else None
1926
1927
1928class StringIO(TextIOWrapper):
1929 """Text I/O implementation using an in-memory buffer.
1930
1931 The initial_value argument sets the value of object. The newline
1932 argument is like the one of TextIOWrapper's constructor.
1933 """
1934
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001935 def __init__(self, initial_value="", newline="\n"):
1936 super(StringIO, self).__init__(BytesIO(),
1937 encoding="utf-8",
1938 errors="strict",
1939 newline=newline)
Antoine Pitrou11446482009-04-04 14:09:30 +00001940 # Issue #5645: make universal newlines semantics the same as in the
1941 # C version, even under Windows.
1942 if newline is None:
1943 self._writetranslate = False
Alexandre Vassalottid2bb18b2009-07-22 03:07:33 +00001944 if initial_value is not None:
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001945 if not isinstance(initial_value, str):
Alexandre Vassalottid2bb18b2009-07-22 03:07:33 +00001946 raise TypeError("initial_value must be str or None, not {0}"
1947 .format(type(initial_value).__name__))
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001948 initial_value = str(initial_value)
1949 self.write(initial_value)
1950 self.seek(0)
1951
1952 def getvalue(self):
1953 self.flush()
1954 return self.buffer.getvalue().decode(self._encoding, self._errors)
Benjamin Peterson9fd459a2009-03-09 00:09:44 +00001955
1956 def __repr__(self):
1957 # TextIOWrapper tells the encoding in its repr. In StringIO,
1958 # that's a implementation detail.
1959 return object.__repr__(self)
Benjamin Petersonb487e632009-03-21 03:08:31 +00001960
1961 @property
Benjamin Peterson0926ad12009-06-06 18:02:12 +00001962 def errors(self):
1963 return None
1964
1965 @property
Benjamin Petersonb487e632009-03-21 03:08:31 +00001966 def encoding(self):
1967 return None
Benjamin Petersond2e0c792009-05-01 20:40:59 +00001968
1969 def detach(self):
1970 # This doesn't make sense on StringIO.
1971 self._unsupported("detach")