blob: c9c42977007f56406c346ed0bbb5329c3a469406 [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
16from io import __all__
Benjamin Peterson8d5fd4e2009-04-02 01:03:26 +000017from io import SEEK_SET, SEEK_CUR, SEEK_END
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +000018
19# open() uses st_blksize whenever we can
20DEFAULT_BUFFER_SIZE = 8 * 1024 # bytes
21
22# NOTE: Base classes defined here are registered with the "official" ABCs
23# defined in io.py. We don't use real inheritance though, because we don't
24# want to inherit the C implementations.
25
26
27class BlockingIOError(IOError):
28
29 """Exception raised when I/O would block on a non-blocking I/O stream."""
30
31 def __init__(self, errno, strerror, characters_written=0):
32 super().__init__(errno, strerror)
33 if not isinstance(characters_written, int):
34 raise TypeError("characters_written must be a integer")
35 self.characters_written = characters_written
36
37
Benjamin Peterson95e392c2010-04-27 21:07:21 +000038def open(file: (str, bytes), mode: str = "r", buffering: int = -1,
Benjamin Peterson9990e8c2009-04-18 14:47:50 +000039 encoding: str = None, errors: str = None,
40 newline: str = None, closefd: bool = True) -> "IOBase":
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +000041
42 r"""Open file and return a stream. Raise IOError upon failure.
43
44 file is either a text or byte string giving the name (and the path
45 if the file isn't in the current working directory) of the file to
46 be opened or an integer file descriptor of the file to be
47 wrapped. (If a file descriptor is given, it is closed when the
48 returned I/O object is closed, unless closefd is set to False.)
49
50 mode is an optional string that specifies the mode in which the file
51 is opened. It defaults to 'r' which means open for reading in text
52 mode. Other common values are 'w' for writing (truncating the file if
53 it already exists), and 'a' for appending (which on some Unix systems,
54 means that all writes append to the end of the file regardless of the
55 current seek position). In text mode, if encoding is not specified the
56 encoding used is platform dependent. (For reading and writing raw
57 bytes use binary mode and leave encoding unspecified.) The available
58 modes are:
59
60 ========= ===============================================================
61 Character Meaning
62 --------- ---------------------------------------------------------------
63 'r' open for reading (default)
64 'w' open for writing, truncating the file first
65 'a' open for writing, appending to the end of the file if it exists
66 'b' binary mode
67 't' text mode (default)
68 '+' open a disk file for updating (reading and writing)
69 'U' universal newline mode (for backwards compatibility; unneeded
70 for new code)
71 ========= ===============================================================
72
73 The default mode is 'rt' (open for reading text). For binary random
74 access, the mode 'w+b' opens and truncates the file to 0 bytes, while
75 'r+b' opens the file without truncation.
76
77 Python distinguishes between files opened in binary and text modes,
78 even when the underlying operating system doesn't. Files opened in
79 binary mode (appending 'b' to the mode argument) return contents as
80 bytes objects without any decoding. In text mode (the default, or when
81 't' is appended to the mode argument), the contents of the file are
82 returned as strings, the bytes having been first decoded using a
83 platform-dependent encoding or using the specified encoding if given.
84
Antoine Pitroud5587bc2009-12-19 21:08:31 +000085 buffering is an optional integer used to set the buffering policy.
86 Pass 0 to switch buffering off (only allowed in binary mode), 1 to select
87 line buffering (only usable in text mode), and an integer > 1 to indicate
88 the size of a fixed-size chunk buffer. When no buffering argument is
89 given, the default buffering policy works as follows:
90
91 * Binary files are buffered in fixed-size chunks; the size of the buffer
92 is chosen using a heuristic trying to determine the underlying device's
93 "block size" and falling back on `io.DEFAULT_BUFFER_SIZE`.
94 On many systems, the buffer will typically be 4096 or 8192 bytes long.
95
96 * "Interactive" text files (files for which isatty() returns True)
97 use line buffering. Other text files use the policy described above
98 for binary files.
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +000099
100 encoding is the name of the encoding used to decode or encode the
101 file. This should only be used in text mode. The default encoding is
102 platform dependent, but any encoding supported by Python can be
103 passed. See the codecs module for the list of supported encodings.
104
105 errors is an optional string that specifies how encoding errors are to
106 be handled---this argument should not be used in binary mode. Pass
107 'strict' to raise a ValueError exception if there is an encoding error
108 (the default of None has the same effect), or pass 'ignore' to ignore
109 errors. (Note that ignoring encoding errors can lead to data loss.)
110 See the documentation for codecs.register for a list of the permitted
111 encoding error strings.
112
113 newline controls how universal newlines works (it only applies to text
114 mode). It can be None, '', '\n', '\r', and '\r\n'. It works as
115 follows:
116
117 * On input, if newline is None, universal newlines mode is
118 enabled. Lines in the input can end in '\n', '\r', or '\r\n', and
119 these are translated into '\n' before being returned to the
120 caller. If it is '', universal newline mode is enabled, but line
121 endings are returned to the caller untranslated. If it has any of
122 the other legal values, input lines are only terminated by the given
123 string, and the line ending is returned to the caller untranslated.
124
125 * On output, if newline is None, any '\n' characters written are
126 translated to the system default line separator, os.linesep. If
127 newline is '', no translation takes place. If newline is any of the
128 other legal values, any '\n' characters written are translated to
129 the given string.
130
131 If closefd is False, the underlying file descriptor will be kept open
132 when the file is closed. This does not work when a file name is given
133 and must be True in that case.
134
135 open() returns a file object whose type depends on the mode, and
136 through which the standard file operations such as reading and writing
137 are performed. When open() is used to open a file in a text mode ('w',
138 'r', 'wt', 'rt', etc.), it returns a TextIOWrapper. When used to open
139 a file in a binary mode, the returned class varies: in read binary
140 mode, it returns a BufferedReader; in write binary and append binary
141 modes, it returns a BufferedWriter, and in read/write mode, it returns
142 a BufferedRandom.
143
144 It is also possible to use a string or bytearray as a file for both
145 reading and writing. For strings StringIO can be used like a file
146 opened in a text mode, and for bytes a BytesIO can be used like a file
147 opened in a binary mode.
148 """
149 if not isinstance(file, (str, bytes, int)):
150 raise TypeError("invalid file: %r" % file)
151 if not isinstance(mode, str):
152 raise TypeError("invalid mode: %r" % mode)
Benjamin Peterson95e392c2010-04-27 21:07:21 +0000153 if not isinstance(buffering, int):
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000154 raise TypeError("invalid buffering: %r" % buffering)
155 if encoding is not None and not isinstance(encoding, str):
156 raise TypeError("invalid encoding: %r" % encoding)
157 if errors is not None and not isinstance(errors, str):
158 raise TypeError("invalid errors: %r" % errors)
159 modes = set(mode)
160 if modes - set("arwb+tU") or len(mode) > len(modes):
161 raise ValueError("invalid mode: %r" % mode)
162 reading = "r" in modes
163 writing = "w" in modes
164 appending = "a" in modes
165 updating = "+" in modes
166 text = "t" in modes
167 binary = "b" in modes
168 if "U" in modes:
169 if writing or appending:
170 raise ValueError("can't use U and writing mode at once")
171 reading = True
172 if text and binary:
173 raise ValueError("can't have text and binary mode at once")
174 if reading + writing + appending > 1:
175 raise ValueError("can't have read/write/append mode at once")
176 if not (reading or writing or appending):
177 raise ValueError("must have exactly one of read/write/append mode")
178 if binary and encoding is not None:
179 raise ValueError("binary mode doesn't take an encoding argument")
180 if binary and errors is not None:
181 raise ValueError("binary mode doesn't take an errors argument")
182 if binary and newline is not None:
183 raise ValueError("binary mode doesn't take a newline argument")
184 raw = FileIO(file,
185 (reading and "r" or "") +
186 (writing and "w" or "") +
187 (appending and "a" or "") +
188 (updating and "+" or ""),
189 closefd)
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000190 line_buffering = False
191 if buffering == 1 or buffering < 0 and raw.isatty():
192 buffering = -1
193 line_buffering = True
194 if buffering < 0:
195 buffering = DEFAULT_BUFFER_SIZE
196 try:
197 bs = os.fstat(raw.fileno()).st_blksize
198 except (os.error, AttributeError):
199 pass
200 else:
201 if bs > 1:
202 buffering = bs
203 if buffering < 0:
204 raise ValueError("invalid buffering size")
205 if buffering == 0:
206 if binary:
207 return raw
208 raise ValueError("can't have unbuffered text I/O")
209 if updating:
210 buffer = BufferedRandom(raw, buffering)
211 elif writing or appending:
212 buffer = BufferedWriter(raw, buffering)
213 elif reading:
214 buffer = BufferedReader(raw, buffering)
215 else:
216 raise ValueError("unknown mode: %r" % mode)
217 if binary:
218 return buffer
219 text = TextIOWrapper(buffer, encoding, errors, newline, line_buffering)
220 text.mode = mode
221 return text
222
223
224class DocDescriptor:
225 """Helper for builtins.open.__doc__
226 """
227 def __get__(self, obj, typ):
228 return (
229 "open(file, mode='r', buffering=None, encoding=None, "
230 "errors=None, newline=None, closefd=True)\n\n" +
231 open.__doc__)
232
233class OpenWrapper:
234 """Wrapper for builtins.open
235
236 Trick so that open won't become a bound method when stored
237 as a class variable (as dbm.dumb does).
238
239 See initstdio() in Python/pythonrun.c.
240 """
241 __doc__ = DocDescriptor()
242
243 def __new__(cls, *args, **kwargs):
244 return open(*args, **kwargs)
245
246
247class UnsupportedOperation(ValueError, IOError):
248 pass
249
250
251class IOBase(metaclass=abc.ABCMeta):
252
253 """The abstract base class for all I/O classes, acting on streams of
254 bytes. There is no public constructor.
255
256 This class provides dummy implementations for many methods that
257 derived classes can override selectively; the default implementations
258 represent a file that cannot be read, written or seeked.
259
260 Even though IOBase does not declare read, readinto, or write because
261 their signatures will vary, implementations and clients should
262 consider those methods part of the interface. Also, implementations
263 may raise a IOError when operations they do not support are called.
264
265 The basic type used for binary data read from or written to a file is
266 bytes. bytearrays are accepted too, and in some cases (such as
267 readinto) needed. Text I/O classes work with str data.
268
269 Note that calling any method (even inquiries) on a closed stream is
270 undefined. Implementations may raise IOError in this case.
271
272 IOBase (and its subclasses) support the iterator protocol, meaning
273 that an IOBase object can be iterated over yielding the lines in a
274 stream.
275
276 IOBase also supports the :keyword:`with` statement. In this example,
277 fp is closed after the suite of the with statement is complete:
278
279 with open('spam.txt', 'r') as fp:
280 fp.write('Spam and eggs!')
281 """
282
283 ### Internal ###
284
285 def _unsupported(self, name: str) -> IOError:
286 """Internal: raise an exception for unsupported operations."""
287 raise UnsupportedOperation("%s.%s() not supported" %
288 (self.__class__.__name__, name))
289
290 ### Positioning ###
291
292 def seek(self, pos: int, whence: int = 0) -> int:
293 """Change stream position.
294
295 Change the stream position to byte offset offset. offset is
296 interpreted relative to the position indicated by whence. Values
297 for whence are:
298
299 * 0 -- start of stream (the default); offset should be zero or positive
300 * 1 -- current stream position; offset may be negative
301 * 2 -- end of stream; offset is usually negative
302
303 Return the new absolute position.
304 """
305 self._unsupported("seek")
306
307 def tell(self) -> int:
308 """Return current stream position."""
309 return self.seek(0, 1)
310
311 def truncate(self, pos: int = None) -> int:
312 """Truncate file to size bytes.
313
314 Size defaults to the current IO position as reported by tell(). Return
315 the new size.
316 """
317 self._unsupported("truncate")
318
319 ### Flush and close ###
320
321 def flush(self) -> None:
322 """Flush write buffers, if applicable.
323
324 This is not implemented for read-only and non-blocking streams.
325 """
326 # XXX Should this return the number of bytes written???
327
328 __closed = False
329
330 def close(self) -> None:
331 """Flush and close the IO object.
332
333 This method has no effect if the file is already closed.
334 """
335 if not self.__closed:
336 try:
337 self.flush()
338 except IOError:
339 pass # If flush() fails, just give up
340 self.__closed = True
341
342 def __del__(self) -> None:
343 """Destructor. Calls close()."""
344 # The try/except block is in case this is called at program
345 # exit time, when it's possible that globals have already been
346 # deleted, and then the close() call might fail. Since
347 # there's nothing we can do about such failures and they annoy
348 # the end users, we suppress the traceback.
349 try:
350 self.close()
351 except:
352 pass
353
354 ### Inquiries ###
355
356 def seekable(self) -> bool:
357 """Return whether object supports random access.
358
359 If False, seek(), tell() and truncate() will raise IOError.
360 This method may need to do a test seek().
361 """
362 return False
363
364 def _checkSeekable(self, msg=None):
365 """Internal: raise an IOError if file is not seekable
366 """
367 if not self.seekable():
368 raise IOError("File or stream is not seekable."
369 if msg is None else msg)
370
371
372 def readable(self) -> bool:
373 """Return whether object was opened for reading.
374
375 If False, read() will raise IOError.
376 """
377 return False
378
379 def _checkReadable(self, msg=None):
380 """Internal: raise an IOError if file is not readable
381 """
382 if not self.readable():
383 raise IOError("File or stream is not readable."
384 if msg is None else msg)
385
386 def writable(self) -> bool:
387 """Return whether object was opened for writing.
388
389 If False, write() and truncate() will raise IOError.
390 """
391 return False
392
393 def _checkWritable(self, msg=None):
394 """Internal: raise an IOError if file is not writable
395 """
396 if not self.writable():
397 raise IOError("File or stream is not writable."
398 if msg is None else msg)
399
400 @property
401 def closed(self):
402 """closed: bool. True iff the file has been closed.
403
404 For backwards compatibility, this is a property, not a predicate.
405 """
406 return self.__closed
407
408 def _checkClosed(self, msg=None):
409 """Internal: raise an ValueError if file is closed
410 """
411 if self.closed:
412 raise ValueError("I/O operation on closed file."
413 if msg is None else msg)
414
415 ### Context manager ###
416
417 def __enter__(self) -> "IOBase": # That's a forward reference
418 """Context management protocol. Returns self."""
419 self._checkClosed()
420 return self
421
422 def __exit__(self, *args) -> None:
423 """Context management protocol. Calls close()"""
424 self.close()
425
426 ### Lower-level APIs ###
427
428 # XXX Should these be present even if unimplemented?
429
430 def fileno(self) -> int:
431 """Returns underlying file descriptor if one exists.
432
433 An IOError is raised if the IO object does not use a file descriptor.
434 """
435 self._unsupported("fileno")
436
437 def isatty(self) -> bool:
438 """Return whether this is an 'interactive' stream.
439
440 Return False if it can't be determined.
441 """
442 self._checkClosed()
443 return False
444
445 ### Readline[s] and writelines ###
446
447 def readline(self, limit: int = -1) -> bytes:
448 r"""Read and return a line from the stream.
449
450 If limit is specified, at most limit bytes will be read.
451
452 The line terminator is always b'\n' for binary files; for text
453 files, the newlines argument to open can be used to select the line
454 terminator(s) recognized.
455 """
456 # For backwards compatibility, a (slowish) readline().
457 if hasattr(self, "peek"):
458 def nreadahead():
459 readahead = self.peek(1)
460 if not readahead:
461 return 1
462 n = (readahead.find(b"\n") + 1) or len(readahead)
463 if limit >= 0:
464 n = min(n, limit)
465 return n
466 else:
467 def nreadahead():
468 return 1
469 if limit is None:
470 limit = -1
Benjamin Petersonb01138a2009-04-24 22:59:52 +0000471 elif not isinstance(limit, int):
472 raise TypeError("limit must be an integer")
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000473 res = bytearray()
474 while limit < 0 or len(res) < limit:
475 b = self.read(nreadahead())
476 if not b:
477 break
478 res += b
479 if res.endswith(b"\n"):
480 break
481 return bytes(res)
482
483 def __iter__(self):
484 self._checkClosed()
485 return self
486
487 def __next__(self):
488 line = self.readline()
489 if not line:
490 raise StopIteration
491 return line
492
493 def readlines(self, hint=None):
494 """Return a list of lines from the stream.
495
496 hint can be specified to control the number of lines read: no more
497 lines will be read if the total size (in bytes/characters) of all
498 lines so far exceeds hint.
499 """
500 if hint is None or hint <= 0:
501 return list(self)
502 n = 0
503 lines = []
504 for line in self:
505 lines.append(line)
506 n += len(line)
507 if n >= hint:
508 break
509 return lines
510
511 def writelines(self, lines):
512 self._checkClosed()
513 for line in lines:
514 self.write(line)
515
516io.IOBase.register(IOBase)
517
518
519class RawIOBase(IOBase):
520
521 """Base class for raw binary I/O."""
522
523 # The read() method is implemented by calling readinto(); derived
524 # classes that want to support read() only need to implement
525 # readinto() as a primitive operation. In general, readinto() can be
526 # more efficient than read().
527
528 # (It would be tempting to also provide an implementation of
529 # readinto() in terms of read(), in case the latter is a more suitable
530 # primitive operation, but that would lead to nasty recursion in case
531 # a subclass doesn't implement either.)
532
533 def read(self, n: int = -1) -> bytes:
534 """Read and return up to n bytes.
535
536 Returns an empty bytes object on EOF, or None if the object is
537 set not to block and has no data to read.
538 """
539 if n is None:
540 n = -1
541 if n < 0:
542 return self.readall()
543 b = bytearray(n.__index__())
544 n = self.readinto(b)
545 del b[n:]
546 return bytes(b)
547
548 def readall(self):
549 """Read until EOF, using multiple read() call."""
550 res = bytearray()
551 while True:
552 data = self.read(DEFAULT_BUFFER_SIZE)
553 if not data:
554 break
555 res += data
556 return bytes(res)
557
558 def readinto(self, b: bytearray) -> int:
559 """Read up to len(b) bytes into b.
560
561 Returns number of bytes read (0 for EOF), or None if the object
562 is set not to block as has no data to read.
563 """
564 self._unsupported("readinto")
565
566 def write(self, b: bytes) -> int:
567 """Write the given buffer to the IO stream.
568
569 Returns the number of bytes written, which may be less than len(b).
570 """
571 self._unsupported("write")
572
573io.RawIOBase.register(RawIOBase)
574from _io import FileIO
575RawIOBase.register(FileIO)
576
577
578class BufferedIOBase(IOBase):
579
580 """Base class for buffered IO objects.
581
582 The main difference with RawIOBase is that the read() method
583 supports omitting the size argument, and does not have a default
584 implementation that defers to readinto().
585
586 In addition, read(), readinto() and write() may raise
587 BlockingIOError if the underlying raw stream is in non-blocking
588 mode and not ready; unlike their raw counterparts, they will never
589 return None.
590
591 A typical implementation should not inherit from a RawIOBase
592 implementation, but wrap one.
593 """
594
595 def read(self, n: int = None) -> bytes:
596 """Read and return up to n bytes.
597
598 If the argument is omitted, None, or negative, reads and
599 returns all data until EOF.
600
601 If the argument is positive, and the underlying raw stream is
602 not 'interactive', multiple raw reads may be issued to satisfy
603 the byte count (unless EOF is reached first). But for
604 interactive raw streams (XXX and for pipes?), at most one raw
605 read will be issued, and a short result does not imply that
606 EOF is imminent.
607
608 Returns an empty bytes array on EOF.
609
610 Raises BlockingIOError if the underlying raw stream has no
611 data at the moment.
612 """
613 self._unsupported("read")
614
615 def read1(self, n: int=None) -> bytes:
616 """Read up to n bytes with at most one read() system call."""
617 self._unsupported("read1")
618
619 def readinto(self, b: bytearray) -> int:
620 """Read up to len(b) bytes into b.
621
622 Like read(), this may issue multiple reads to the underlying raw
623 stream, unless the latter is 'interactive'.
624
625 Returns the number of bytes read (0 for EOF).
626
627 Raises BlockingIOError if the underlying raw stream has no
628 data at the moment.
629 """
630 # XXX This ought to work with anything that supports the buffer API
631 data = self.read(len(b))
632 n = len(data)
633 try:
634 b[:n] = data
635 except TypeError as err:
636 import array
637 if not isinstance(b, array.array):
638 raise err
639 b[:n] = array.array('b', data)
640 return n
641
642 def write(self, b: bytes) -> int:
643 """Write the given buffer to the IO stream.
644
645 Return the number of bytes written, which is never less than
646 len(b).
647
648 Raises BlockingIOError if the buffer is full and the
649 underlying raw stream cannot accept more data at the moment.
650 """
651 self._unsupported("write")
652
Benjamin Petersond2e0c792009-05-01 20:40:59 +0000653 def detach(self) -> None:
654 """
655 Separate the underlying raw stream from the buffer and return it.
656
657 After the raw stream has been detached, the buffer is in an unusable
658 state.
659 """
660 self._unsupported("detach")
661
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000662io.BufferedIOBase.register(BufferedIOBase)
663
664
665class _BufferedIOMixin(BufferedIOBase):
666
667 """A mixin implementation of BufferedIOBase with an underlying raw stream.
668
669 This passes most requests on to the underlying raw stream. It
670 does *not* provide implementations of read(), readinto() or
671 write().
672 """
673
674 def __init__(self, raw):
675 self.raw = raw
676
677 ### Positioning ###
678
679 def seek(self, pos, whence=0):
680 new_position = self.raw.seek(pos, whence)
681 if new_position < 0:
682 raise IOError("seek() returned an invalid position")
683 return new_position
684
685 def tell(self):
686 pos = self.raw.tell()
687 if pos < 0:
688 raise IOError("tell() returned an invalid position")
689 return pos
690
691 def truncate(self, pos=None):
692 # Flush the stream. We're mixing buffered I/O with lower-level I/O,
693 # and a flush may be necessary to synch both views of the current
694 # file state.
695 self.flush()
696
697 if pos is None:
698 pos = self.tell()
699 # XXX: Should seek() be used, instead of passing the position
700 # XXX directly to truncate?
701 return self.raw.truncate(pos)
702
703 ### Flush and close ###
704
705 def flush(self):
706 self.raw.flush()
707
708 def close(self):
Benjamin Petersond2e0c792009-05-01 20:40:59 +0000709 if not self.closed and self.raw is not None:
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000710 try:
711 self.flush()
712 except IOError:
713 pass # If flush() fails, just give up
714 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):
Benjamin Petersond2e0c792009-05-01 20:40:59 +00001525 if self.buffer is not None:
1526 try:
1527 self.flush()
1528 except IOError:
1529 pass # If flush() fails, just give up
1530 self.buffer.close()
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001531
1532 @property
1533 def closed(self):
1534 return self.buffer.closed
1535
1536 @property
1537 def name(self):
1538 return self.buffer.name
1539
1540 def fileno(self):
1541 return self.buffer.fileno()
1542
1543 def isatty(self):
1544 return self.buffer.isatty()
1545
1546 def write(self, s: str):
1547 if self.closed:
1548 raise ValueError("write to closed file")
1549 if not isinstance(s, str):
1550 raise TypeError("can't write %s to text stream" %
1551 s.__class__.__name__)
1552 length = len(s)
1553 haslf = (self._writetranslate or self._line_buffering) and "\n" in s
1554 if haslf and self._writetranslate and self._writenl != "\n":
1555 s = s.replace("\n", self._writenl)
1556 encoder = self._encoder or self._get_encoder()
1557 # XXX What if we were just reading?
1558 b = encoder.encode(s)
1559 self.buffer.write(b)
1560 if self._line_buffering and (haslf or "\r" in s):
1561 self.flush()
1562 self._snapshot = None
1563 if self._decoder:
1564 self._decoder.reset()
1565 return length
1566
1567 def _get_encoder(self):
1568 make_encoder = codecs.getincrementalencoder(self._encoding)
1569 self._encoder = make_encoder(self._errors)
1570 return self._encoder
1571
1572 def _get_decoder(self):
1573 make_decoder = codecs.getincrementaldecoder(self._encoding)
1574 decoder = make_decoder(self._errors)
1575 if self._readuniversal:
1576 decoder = IncrementalNewlineDecoder(decoder, self._readtranslate)
1577 self._decoder = decoder
1578 return decoder
1579
1580 # The following three methods implement an ADT for _decoded_chars.
1581 # Text returned from the decoder is buffered here until the client
1582 # requests it by calling our read() or readline() method.
1583 def _set_decoded_chars(self, chars):
1584 """Set the _decoded_chars buffer."""
1585 self._decoded_chars = chars
1586 self._decoded_chars_used = 0
1587
1588 def _get_decoded_chars(self, n=None):
1589 """Advance into the _decoded_chars buffer."""
1590 offset = self._decoded_chars_used
1591 if n is None:
1592 chars = self._decoded_chars[offset:]
1593 else:
1594 chars = self._decoded_chars[offset:offset + n]
1595 self._decoded_chars_used += len(chars)
1596 return chars
1597
1598 def _rewind_decoded_chars(self, n):
1599 """Rewind the _decoded_chars buffer."""
1600 if self._decoded_chars_used < n:
1601 raise AssertionError("rewind decoded_chars out of bounds")
1602 self._decoded_chars_used -= n
1603
1604 def _read_chunk(self):
1605 """
1606 Read and decode the next chunk of data from the BufferedReader.
1607 """
1608
1609 # The return value is True unless EOF was reached. The decoded
1610 # string is placed in self._decoded_chars (replacing its previous
1611 # value). The entire input chunk is sent to the decoder, though
1612 # some of it may remain buffered in the decoder, yet to be
1613 # converted.
1614
1615 if self._decoder is None:
1616 raise ValueError("no decoder")
1617
1618 if self._telling:
1619 # To prepare for tell(), we need to snapshot a point in the
1620 # file where the decoder's input buffer is empty.
1621
1622 dec_buffer, dec_flags = self._decoder.getstate()
1623 # Given this, we know there was a valid snapshot point
1624 # len(dec_buffer) bytes ago with decoder state (b'', dec_flags).
1625
1626 # Read a chunk, decode it, and put the result in self._decoded_chars.
1627 input_chunk = self.buffer.read1(self._CHUNK_SIZE)
1628 eof = not input_chunk
1629 self._set_decoded_chars(self._decoder.decode(input_chunk, eof))
1630
1631 if self._telling:
1632 # At the snapshot point, len(dec_buffer) bytes before the read,
1633 # the next input to be decoded is dec_buffer + input_chunk.
1634 self._snapshot = (dec_flags, dec_buffer + input_chunk)
1635
1636 return not eof
1637
1638 def _pack_cookie(self, position, dec_flags=0,
1639 bytes_to_feed=0, need_eof=0, chars_to_skip=0):
1640 # The meaning of a tell() cookie is: seek to position, set the
1641 # decoder flags to dec_flags, read bytes_to_feed bytes, feed them
1642 # into the decoder with need_eof as the EOF flag, then skip
1643 # chars_to_skip characters of the decoded result. For most simple
1644 # decoders, tell() will often just give a byte offset in the file.
1645 return (position | (dec_flags<<64) | (bytes_to_feed<<128) |
1646 (chars_to_skip<<192) | bool(need_eof)<<256)
1647
1648 def _unpack_cookie(self, bigint):
1649 rest, position = divmod(bigint, 1<<64)
1650 rest, dec_flags = divmod(rest, 1<<64)
1651 rest, bytes_to_feed = divmod(rest, 1<<64)
1652 need_eof, chars_to_skip = divmod(rest, 1<<64)
1653 return position, dec_flags, bytes_to_feed, need_eof, chars_to_skip
1654
1655 def tell(self):
1656 if not self._seekable:
1657 raise IOError("underlying stream is not seekable")
1658 if not self._telling:
1659 raise IOError("telling position disabled by next() call")
1660 self.flush()
1661 position = self.buffer.tell()
1662 decoder = self._decoder
1663 if decoder is None or self._snapshot is None:
1664 if self._decoded_chars:
1665 # This should never happen.
1666 raise AssertionError("pending decoded text")
1667 return position
1668
1669 # Skip backward to the snapshot point (see _read_chunk).
1670 dec_flags, next_input = self._snapshot
1671 position -= len(next_input)
1672
1673 # How many decoded characters have been used up since the snapshot?
1674 chars_to_skip = self._decoded_chars_used
1675 if chars_to_skip == 0:
1676 # We haven't moved from the snapshot point.
1677 return self._pack_cookie(position, dec_flags)
1678
1679 # Starting from the snapshot position, we will walk the decoder
1680 # forward until it gives us enough decoded characters.
1681 saved_state = decoder.getstate()
1682 try:
1683 # Note our initial start point.
1684 decoder.setstate((b'', dec_flags))
1685 start_pos = position
1686 start_flags, bytes_fed, chars_decoded = dec_flags, 0, 0
1687 need_eof = 0
1688
1689 # Feed the decoder one byte at a time. As we go, note the
1690 # nearest "safe start point" before the current location
1691 # (a point where the decoder has nothing buffered, so seek()
1692 # can safely start from there and advance to this location).
1693 next_byte = bytearray(1)
1694 for next_byte[0] in next_input:
1695 bytes_fed += 1
1696 chars_decoded += len(decoder.decode(next_byte))
1697 dec_buffer, dec_flags = decoder.getstate()
1698 if not dec_buffer and chars_decoded <= chars_to_skip:
1699 # Decoder buffer is empty, so this is a safe start point.
1700 start_pos += bytes_fed
1701 chars_to_skip -= chars_decoded
1702 start_flags, bytes_fed, chars_decoded = dec_flags, 0, 0
1703 if chars_decoded >= chars_to_skip:
1704 break
1705 else:
1706 # We didn't get enough decoded data; signal EOF to get more.
1707 chars_decoded += len(decoder.decode(b'', final=True))
1708 need_eof = 1
1709 if chars_decoded < chars_to_skip:
1710 raise IOError("can't reconstruct logical file position")
1711
1712 # The returned cookie corresponds to the last safe start point.
1713 return self._pack_cookie(
1714 start_pos, start_flags, bytes_fed, need_eof, chars_to_skip)
1715 finally:
1716 decoder.setstate(saved_state)
1717
1718 def truncate(self, pos=None):
1719 self.flush()
1720 if pos is None:
1721 pos = self.tell()
Antoine Pitrou905a2ff2010-01-31 22:47:27 +00001722 return self.buffer.truncate(pos)
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001723
Benjamin Petersond2e0c792009-05-01 20:40:59 +00001724 def detach(self):
1725 if self.buffer is None:
1726 raise ValueError("buffer is already detached")
1727 self.flush()
1728 buffer = self.buffer
1729 self.buffer = None
1730 return buffer
1731
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001732 def seek(self, cookie, whence=0):
1733 if self.closed:
1734 raise ValueError("tell on closed file")
1735 if not self._seekable:
1736 raise IOError("underlying stream is not seekable")
1737 if whence == 1: # seek relative to current position
1738 if cookie != 0:
1739 raise IOError("can't do nonzero cur-relative seeks")
1740 # Seeking to the current position should attempt to
1741 # sync the underlying buffer with the current position.
1742 whence = 0
1743 cookie = self.tell()
1744 if whence == 2: # seek relative to end of file
1745 if cookie != 0:
1746 raise IOError("can't do nonzero end-relative seeks")
1747 self.flush()
1748 position = self.buffer.seek(0, 2)
1749 self._set_decoded_chars('')
1750 self._snapshot = None
1751 if self._decoder:
1752 self._decoder.reset()
1753 return position
1754 if whence != 0:
1755 raise ValueError("invalid whence (%r, should be 0, 1 or 2)" %
1756 (whence,))
1757 if cookie < 0:
1758 raise ValueError("negative seek position %r" % (cookie,))
1759 self.flush()
1760
1761 # The strategy of seek() is to go back to the safe start point
1762 # and replay the effect of read(chars_to_skip) from there.
1763 start_pos, dec_flags, bytes_to_feed, need_eof, chars_to_skip = \
1764 self._unpack_cookie(cookie)
1765
1766 # Seek back to the safe start point.
1767 self.buffer.seek(start_pos)
1768 self._set_decoded_chars('')
1769 self._snapshot = None
1770
1771 # Restore the decoder to its state from the safe start point.
Benjamin Peterson9363a652009-03-05 00:42:09 +00001772 if cookie == 0 and self._decoder:
1773 self._decoder.reset()
1774 elif self._decoder or dec_flags or chars_to_skip:
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001775 self._decoder = self._decoder or self._get_decoder()
1776 self._decoder.setstate((b'', dec_flags))
1777 self._snapshot = (dec_flags, b'')
1778
1779 if chars_to_skip:
1780 # Just like _read_chunk, feed the decoder and save a snapshot.
1781 input_chunk = self.buffer.read(bytes_to_feed)
1782 self._set_decoded_chars(
1783 self._decoder.decode(input_chunk, need_eof))
1784 self._snapshot = (dec_flags, input_chunk)
1785
1786 # Skip chars_to_skip of the decoded characters.
1787 if len(self._decoded_chars) < chars_to_skip:
1788 raise IOError("can't restore logical file position")
1789 self._decoded_chars_used = chars_to_skip
1790
Antoine Pitroue4501852009-05-14 18:55:55 +00001791 # Finally, reset the encoder (merely useful for proper BOM handling)
1792 try:
1793 encoder = self._encoder or self._get_encoder()
1794 except LookupError:
1795 # Sometimes the encoder doesn't exist
1796 pass
1797 else:
1798 if cookie != 0:
1799 encoder.setstate(0)
1800 else:
1801 encoder.reset()
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001802 return cookie
1803
1804 def read(self, n=None):
Benjamin Petersona1b49012009-03-31 23:11:32 +00001805 self._checkReadable()
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001806 if n is None:
1807 n = -1
1808 decoder = self._decoder or self._get_decoder()
Florent Xiclunab14930c2010-03-13 15:26:44 +00001809 try:
1810 n.__index__
1811 except AttributeError as err:
1812 raise TypeError("an integer is required") from err
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001813 if n < 0:
1814 # Read everything.
1815 result = (self._get_decoded_chars() +
1816 decoder.decode(self.buffer.read(), final=True))
1817 self._set_decoded_chars('')
1818 self._snapshot = None
1819 return result
1820 else:
1821 # Keep reading chunks until we have n characters to return.
1822 eof = False
1823 result = self._get_decoded_chars(n)
1824 while len(result) < n and not eof:
1825 eof = not self._read_chunk()
1826 result += self._get_decoded_chars(n - len(result))
1827 return result
1828
1829 def __next__(self):
1830 self._telling = False
1831 line = self.readline()
1832 if not line:
1833 self._snapshot = None
1834 self._telling = self._seekable
1835 raise StopIteration
1836 return line
1837
1838 def readline(self, limit=None):
1839 if self.closed:
1840 raise ValueError("read from closed file")
1841 if limit is None:
1842 limit = -1
Benjamin Petersonb01138a2009-04-24 22:59:52 +00001843 elif not isinstance(limit, int):
1844 raise TypeError("limit must be an integer")
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001845
1846 # Grab all the decoded text (we will rewind any extra bits later).
1847 line = self._get_decoded_chars()
1848
1849 start = 0
1850 # Make the decoder if it doesn't already exist.
1851 if not self._decoder:
1852 self._get_decoder()
1853
1854 pos = endpos = None
1855 while True:
1856 if self._readtranslate:
1857 # Newlines are already translated, only search for \n
1858 pos = line.find('\n', start)
1859 if pos >= 0:
1860 endpos = pos + 1
1861 break
1862 else:
1863 start = len(line)
1864
1865 elif self._readuniversal:
1866 # Universal newline search. Find any of \r, \r\n, \n
1867 # The decoder ensures that \r\n are not split in two pieces
1868
1869 # In C we'd look for these in parallel of course.
1870 nlpos = line.find("\n", start)
1871 crpos = line.find("\r", start)
1872 if crpos == -1:
1873 if nlpos == -1:
1874 # Nothing found
1875 start = len(line)
1876 else:
1877 # Found \n
1878 endpos = nlpos + 1
1879 break
1880 elif nlpos == -1:
1881 # Found lone \r
1882 endpos = crpos + 1
1883 break
1884 elif nlpos < crpos:
1885 # Found \n
1886 endpos = nlpos + 1
1887 break
1888 elif nlpos == crpos + 1:
1889 # Found \r\n
1890 endpos = crpos + 2
1891 break
1892 else:
1893 # Found \r
1894 endpos = crpos + 1
1895 break
1896 else:
1897 # non-universal
1898 pos = line.find(self._readnl)
1899 if pos >= 0:
1900 endpos = pos + len(self._readnl)
1901 break
1902
1903 if limit >= 0 and len(line) >= limit:
1904 endpos = limit # reached length limit
1905 break
1906
1907 # No line ending seen yet - get more data'
1908 while self._read_chunk():
1909 if self._decoded_chars:
1910 break
1911 if self._decoded_chars:
1912 line += self._get_decoded_chars()
1913 else:
1914 # end of file
1915 self._set_decoded_chars('')
1916 self._snapshot = None
1917 return line
1918
1919 if limit >= 0 and endpos > limit:
1920 endpos = limit # don't exceed limit
1921
1922 # Rewind _decoded_chars to just after the line ending we found.
1923 self._rewind_decoded_chars(len(line) - endpos)
1924 return line[:endpos]
1925
1926 @property
1927 def newlines(self):
1928 return self._decoder.newlines if self._decoder else None
1929
1930
1931class StringIO(TextIOWrapper):
1932 """Text I/O implementation using an in-memory buffer.
1933
1934 The initial_value argument sets the value of object. The newline
1935 argument is like the one of TextIOWrapper's constructor.
1936 """
1937
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001938 def __init__(self, initial_value="", newline="\n"):
1939 super(StringIO, self).__init__(BytesIO(),
1940 encoding="utf-8",
1941 errors="strict",
1942 newline=newline)
Antoine Pitrou11446482009-04-04 14:09:30 +00001943 # Issue #5645: make universal newlines semantics the same as in the
1944 # C version, even under Windows.
1945 if newline is None:
1946 self._writetranslate = False
Alexandre Vassalottid2bb18b2009-07-22 03:07:33 +00001947 if initial_value is not None:
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001948 if not isinstance(initial_value, str):
Alexandre Vassalottid2bb18b2009-07-22 03:07:33 +00001949 raise TypeError("initial_value must be str or None, not {0}"
1950 .format(type(initial_value).__name__))
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001951 initial_value = str(initial_value)
1952 self.write(initial_value)
1953 self.seek(0)
1954
1955 def getvalue(self):
1956 self.flush()
1957 return self.buffer.getvalue().decode(self._encoding, self._errors)
Benjamin Peterson9fd459a2009-03-09 00:09:44 +00001958
1959 def __repr__(self):
1960 # TextIOWrapper tells the encoding in its repr. In StringIO,
1961 # that's a implementation detail.
1962 return object.__repr__(self)
Benjamin Petersonb487e632009-03-21 03:08:31 +00001963
1964 @property
Benjamin Peterson0926ad12009-06-06 18:02:12 +00001965 def errors(self):
1966 return None
1967
1968 @property
Benjamin Petersonb487e632009-03-21 03:08:31 +00001969 def encoding(self):
1970 return None
Benjamin Petersond2e0c792009-05-01 20:40:59 +00001971
1972 def detach(self):
1973 # This doesn't make sense on StringIO.
1974 self._unsupported("detach")