blob: c1cdf439e1282a552a9ed2c9c478ecebd0c3d70a [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 Peterson9990e8c2009-04-18 14:47:50 +000038def open(file: (str, bytes), mode: str = "r", buffering: int = None,
39 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
85 buffering is an optional integer used to set the buffering policy. By
86 default full buffering is on. Pass 0 to switch buffering off (only
87 allowed in binary mode), 1 to set line buffering, and an integer > 1
88 for full buffering.
89
90 encoding is the name of the encoding used to decode or encode the
91 file. This should only be used in text mode. The default encoding is
92 platform dependent, but any encoding supported by Python can be
93 passed. See the codecs module for the list of supported encodings.
94
95 errors is an optional string that specifies how encoding errors are to
96 be handled---this argument should not be used in binary mode. Pass
97 'strict' to raise a ValueError exception if there is an encoding error
98 (the default of None has the same effect), or pass 'ignore' to ignore
99 errors. (Note that ignoring encoding errors can lead to data loss.)
100 See the documentation for codecs.register for a list of the permitted
101 encoding error strings.
102
103 newline controls how universal newlines works (it only applies to text
104 mode). It can be None, '', '\n', '\r', and '\r\n'. It works as
105 follows:
106
107 * On input, if newline is None, universal newlines mode is
108 enabled. Lines in the input can end in '\n', '\r', or '\r\n', and
109 these are translated into '\n' before being returned to the
110 caller. If it is '', universal newline mode is enabled, but line
111 endings are returned to the caller untranslated. If it has any of
112 the other legal values, input lines are only terminated by the given
113 string, and the line ending is returned to the caller untranslated.
114
115 * On output, if newline is None, any '\n' characters written are
116 translated to the system default line separator, os.linesep. If
117 newline is '', no translation takes place. If newline is any of the
118 other legal values, any '\n' characters written are translated to
119 the given string.
120
121 If closefd is False, the underlying file descriptor will be kept open
122 when the file is closed. This does not work when a file name is given
123 and must be True in that case.
124
125 open() returns a file object whose type depends on the mode, and
126 through which the standard file operations such as reading and writing
127 are performed. When open() is used to open a file in a text mode ('w',
128 'r', 'wt', 'rt', etc.), it returns a TextIOWrapper. When used to open
129 a file in a binary mode, the returned class varies: in read binary
130 mode, it returns a BufferedReader; in write binary and append binary
131 modes, it returns a BufferedWriter, and in read/write mode, it returns
132 a BufferedRandom.
133
134 It is also possible to use a string or bytearray as a file for both
135 reading and writing. For strings StringIO can be used like a file
136 opened in a text mode, and for bytes a BytesIO can be used like a file
137 opened in a binary mode.
138 """
139 if not isinstance(file, (str, bytes, int)):
140 raise TypeError("invalid file: %r" % file)
141 if not isinstance(mode, str):
142 raise TypeError("invalid mode: %r" % mode)
143 if buffering is not None and not isinstance(buffering, int):
144 raise TypeError("invalid buffering: %r" % buffering)
145 if encoding is not None and not isinstance(encoding, str):
146 raise TypeError("invalid encoding: %r" % encoding)
147 if errors is not None and not isinstance(errors, str):
148 raise TypeError("invalid errors: %r" % errors)
149 modes = set(mode)
150 if modes - set("arwb+tU") or len(mode) > len(modes):
151 raise ValueError("invalid mode: %r" % mode)
152 reading = "r" in modes
153 writing = "w" in modes
154 appending = "a" in modes
155 updating = "+" in modes
156 text = "t" in modes
157 binary = "b" in modes
158 if "U" in modes:
159 if writing or appending:
160 raise ValueError("can't use U and writing mode at once")
161 reading = True
162 if text and binary:
163 raise ValueError("can't have text and binary mode at once")
164 if reading + writing + appending > 1:
165 raise ValueError("can't have read/write/append mode at once")
166 if not (reading or writing or appending):
167 raise ValueError("must have exactly one of read/write/append mode")
168 if binary and encoding is not None:
169 raise ValueError("binary mode doesn't take an encoding argument")
170 if binary and errors is not None:
171 raise ValueError("binary mode doesn't take an errors argument")
172 if binary and newline is not None:
173 raise ValueError("binary mode doesn't take a newline argument")
174 raw = FileIO(file,
175 (reading and "r" or "") +
176 (writing and "w" or "") +
177 (appending and "a" or "") +
178 (updating and "+" or ""),
179 closefd)
180 if buffering is None:
181 buffering = -1
182 line_buffering = False
183 if buffering == 1 or buffering < 0 and raw.isatty():
184 buffering = -1
185 line_buffering = True
186 if buffering < 0:
187 buffering = DEFAULT_BUFFER_SIZE
188 try:
189 bs = os.fstat(raw.fileno()).st_blksize
190 except (os.error, AttributeError):
191 pass
192 else:
193 if bs > 1:
194 buffering = bs
195 if buffering < 0:
196 raise ValueError("invalid buffering size")
197 if buffering == 0:
198 if binary:
199 return raw
200 raise ValueError("can't have unbuffered text I/O")
201 if updating:
202 buffer = BufferedRandom(raw, buffering)
203 elif writing or appending:
204 buffer = BufferedWriter(raw, buffering)
205 elif reading:
206 buffer = BufferedReader(raw, buffering)
207 else:
208 raise ValueError("unknown mode: %r" % mode)
209 if binary:
210 return buffer
211 text = TextIOWrapper(buffer, encoding, errors, newline, line_buffering)
212 text.mode = mode
213 return text
214
215
216class DocDescriptor:
217 """Helper for builtins.open.__doc__
218 """
219 def __get__(self, obj, typ):
220 return (
221 "open(file, mode='r', buffering=None, encoding=None, "
222 "errors=None, newline=None, closefd=True)\n\n" +
223 open.__doc__)
224
225class OpenWrapper:
226 """Wrapper for builtins.open
227
228 Trick so that open won't become a bound method when stored
229 as a class variable (as dbm.dumb does).
230
231 See initstdio() in Python/pythonrun.c.
232 """
233 __doc__ = DocDescriptor()
234
235 def __new__(cls, *args, **kwargs):
236 return open(*args, **kwargs)
237
238
239class UnsupportedOperation(ValueError, IOError):
240 pass
241
242
243class IOBase(metaclass=abc.ABCMeta):
244
245 """The abstract base class for all I/O classes, acting on streams of
246 bytes. There is no public constructor.
247
248 This class provides dummy implementations for many methods that
249 derived classes can override selectively; the default implementations
250 represent a file that cannot be read, written or seeked.
251
252 Even though IOBase does not declare read, readinto, or write because
253 their signatures will vary, implementations and clients should
254 consider those methods part of the interface. Also, implementations
255 may raise a IOError when operations they do not support are called.
256
257 The basic type used for binary data read from or written to a file is
258 bytes. bytearrays are accepted too, and in some cases (such as
259 readinto) needed. Text I/O classes work with str data.
260
261 Note that calling any method (even inquiries) on a closed stream is
262 undefined. Implementations may raise IOError in this case.
263
264 IOBase (and its subclasses) support the iterator protocol, meaning
265 that an IOBase object can be iterated over yielding the lines in a
266 stream.
267
268 IOBase also supports the :keyword:`with` statement. In this example,
269 fp is closed after the suite of the with statement is complete:
270
271 with open('spam.txt', 'r') as fp:
272 fp.write('Spam and eggs!')
273 """
274
275 ### Internal ###
276
277 def _unsupported(self, name: str) -> IOError:
278 """Internal: raise an exception for unsupported operations."""
279 raise UnsupportedOperation("%s.%s() not supported" %
280 (self.__class__.__name__, name))
281
282 ### Positioning ###
283
284 def seek(self, pos: int, whence: int = 0) -> int:
285 """Change stream position.
286
287 Change the stream position to byte offset offset. offset is
288 interpreted relative to the position indicated by whence. Values
289 for whence are:
290
291 * 0 -- start of stream (the default); offset should be zero or positive
292 * 1 -- current stream position; offset may be negative
293 * 2 -- end of stream; offset is usually negative
294
295 Return the new absolute position.
296 """
297 self._unsupported("seek")
298
299 def tell(self) -> int:
300 """Return current stream position."""
301 return self.seek(0, 1)
302
303 def truncate(self, pos: int = None) -> int:
304 """Truncate file to size bytes.
305
306 Size defaults to the current IO position as reported by tell(). Return
307 the new size.
308 """
309 self._unsupported("truncate")
310
311 ### Flush and close ###
312
313 def flush(self) -> None:
314 """Flush write buffers, if applicable.
315
316 This is not implemented for read-only and non-blocking streams.
317 """
318 # XXX Should this return the number of bytes written???
319
320 __closed = False
321
322 def close(self) -> None:
323 """Flush and close the IO object.
324
325 This method has no effect if the file is already closed.
326 """
327 if not self.__closed:
328 try:
329 self.flush()
330 except IOError:
331 pass # If flush() fails, just give up
332 self.__closed = True
333
334 def __del__(self) -> None:
335 """Destructor. Calls close()."""
336 # The try/except block is in case this is called at program
337 # exit time, when it's possible that globals have already been
338 # deleted, and then the close() call might fail. Since
339 # there's nothing we can do about such failures and they annoy
340 # the end users, we suppress the traceback.
341 try:
342 self.close()
343 except:
344 pass
345
346 ### Inquiries ###
347
348 def seekable(self) -> bool:
349 """Return whether object supports random access.
350
351 If False, seek(), tell() and truncate() will raise IOError.
352 This method may need to do a test seek().
353 """
354 return False
355
356 def _checkSeekable(self, msg=None):
357 """Internal: raise an IOError if file is not seekable
358 """
359 if not self.seekable():
360 raise IOError("File or stream is not seekable."
361 if msg is None else msg)
362
363
364 def readable(self) -> bool:
365 """Return whether object was opened for reading.
366
367 If False, read() will raise IOError.
368 """
369 return False
370
371 def _checkReadable(self, msg=None):
372 """Internal: raise an IOError if file is not readable
373 """
374 if not self.readable():
375 raise IOError("File or stream is not readable."
376 if msg is None else msg)
377
378 def writable(self) -> bool:
379 """Return whether object was opened for writing.
380
381 If False, write() and truncate() will raise IOError.
382 """
383 return False
384
385 def _checkWritable(self, msg=None):
386 """Internal: raise an IOError if file is not writable
387 """
388 if not self.writable():
389 raise IOError("File or stream is not writable."
390 if msg is None else msg)
391
392 @property
393 def closed(self):
394 """closed: bool. True iff the file has been closed.
395
396 For backwards compatibility, this is a property, not a predicate.
397 """
398 return self.__closed
399
400 def _checkClosed(self, msg=None):
401 """Internal: raise an ValueError if file is closed
402 """
403 if self.closed:
404 raise ValueError("I/O operation on closed file."
405 if msg is None else msg)
406
407 ### Context manager ###
408
409 def __enter__(self) -> "IOBase": # That's a forward reference
410 """Context management protocol. Returns self."""
411 self._checkClosed()
412 return self
413
414 def __exit__(self, *args) -> None:
415 """Context management protocol. Calls close()"""
416 self.close()
417
418 ### Lower-level APIs ###
419
420 # XXX Should these be present even if unimplemented?
421
422 def fileno(self) -> int:
423 """Returns underlying file descriptor if one exists.
424
425 An IOError is raised if the IO object does not use a file descriptor.
426 """
427 self._unsupported("fileno")
428
429 def isatty(self) -> bool:
430 """Return whether this is an 'interactive' stream.
431
432 Return False if it can't be determined.
433 """
434 self._checkClosed()
435 return False
436
437 ### Readline[s] and writelines ###
438
439 def readline(self, limit: int = -1) -> bytes:
440 r"""Read and return a line from the stream.
441
442 If limit is specified, at most limit bytes will be read.
443
444 The line terminator is always b'\n' for binary files; for text
445 files, the newlines argument to open can be used to select the line
446 terminator(s) recognized.
447 """
448 # For backwards compatibility, a (slowish) readline().
449 if hasattr(self, "peek"):
450 def nreadahead():
451 readahead = self.peek(1)
452 if not readahead:
453 return 1
454 n = (readahead.find(b"\n") + 1) or len(readahead)
455 if limit >= 0:
456 n = min(n, limit)
457 return n
458 else:
459 def nreadahead():
460 return 1
461 if limit is None:
462 limit = -1
Benjamin Petersonb01138a2009-04-24 22:59:52 +0000463 elif not isinstance(limit, int):
464 raise TypeError("limit must be an integer")
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000465 res = bytearray()
466 while limit < 0 or len(res) < limit:
467 b = self.read(nreadahead())
468 if not b:
469 break
470 res += b
471 if res.endswith(b"\n"):
472 break
473 return bytes(res)
474
475 def __iter__(self):
476 self._checkClosed()
477 return self
478
479 def __next__(self):
480 line = self.readline()
481 if not line:
482 raise StopIteration
483 return line
484
485 def readlines(self, hint=None):
486 """Return a list of lines from the stream.
487
488 hint can be specified to control the number of lines read: no more
489 lines will be read if the total size (in bytes/characters) of all
490 lines so far exceeds hint.
491 """
492 if hint is None or hint <= 0:
493 return list(self)
494 n = 0
495 lines = []
496 for line in self:
497 lines.append(line)
498 n += len(line)
499 if n >= hint:
500 break
501 return lines
502
503 def writelines(self, lines):
504 self._checkClosed()
505 for line in lines:
506 self.write(line)
507
508io.IOBase.register(IOBase)
509
510
511class RawIOBase(IOBase):
512
513 """Base class for raw binary I/O."""
514
515 # The read() method is implemented by calling readinto(); derived
516 # classes that want to support read() only need to implement
517 # readinto() as a primitive operation. In general, readinto() can be
518 # more efficient than read().
519
520 # (It would be tempting to also provide an implementation of
521 # readinto() in terms of read(), in case the latter is a more suitable
522 # primitive operation, but that would lead to nasty recursion in case
523 # a subclass doesn't implement either.)
524
525 def read(self, n: int = -1) -> bytes:
526 """Read and return up to n bytes.
527
528 Returns an empty bytes object on EOF, or None if the object is
529 set not to block and has no data to read.
530 """
531 if n is None:
532 n = -1
533 if n < 0:
534 return self.readall()
535 b = bytearray(n.__index__())
536 n = self.readinto(b)
537 del b[n:]
538 return bytes(b)
539
540 def readall(self):
541 """Read until EOF, using multiple read() call."""
542 res = bytearray()
543 while True:
544 data = self.read(DEFAULT_BUFFER_SIZE)
545 if not data:
546 break
547 res += data
548 return bytes(res)
549
550 def readinto(self, b: bytearray) -> int:
551 """Read up to len(b) bytes into b.
552
553 Returns number of bytes read (0 for EOF), or None if the object
554 is set not to block as has no data to read.
555 """
556 self._unsupported("readinto")
557
558 def write(self, b: bytes) -> int:
559 """Write the given buffer to the IO stream.
560
561 Returns the number of bytes written, which may be less than len(b).
562 """
563 self._unsupported("write")
564
565io.RawIOBase.register(RawIOBase)
566from _io import FileIO
567RawIOBase.register(FileIO)
568
569
570class BufferedIOBase(IOBase):
571
572 """Base class for buffered IO objects.
573
574 The main difference with RawIOBase is that the read() method
575 supports omitting the size argument, and does not have a default
576 implementation that defers to readinto().
577
578 In addition, read(), readinto() and write() may raise
579 BlockingIOError if the underlying raw stream is in non-blocking
580 mode and not ready; unlike their raw counterparts, they will never
581 return None.
582
583 A typical implementation should not inherit from a RawIOBase
584 implementation, but wrap one.
585 """
586
587 def read(self, n: int = None) -> bytes:
588 """Read and return up to n bytes.
589
590 If the argument is omitted, None, or negative, reads and
591 returns all data until EOF.
592
593 If the argument is positive, and the underlying raw stream is
594 not 'interactive', multiple raw reads may be issued to satisfy
595 the byte count (unless EOF is reached first). But for
596 interactive raw streams (XXX and for pipes?), at most one raw
597 read will be issued, and a short result does not imply that
598 EOF is imminent.
599
600 Returns an empty bytes array on EOF.
601
602 Raises BlockingIOError if the underlying raw stream has no
603 data at the moment.
604 """
605 self._unsupported("read")
606
607 def read1(self, n: int=None) -> bytes:
608 """Read up to n bytes with at most one read() system call."""
609 self._unsupported("read1")
610
611 def readinto(self, b: bytearray) -> int:
612 """Read up to len(b) bytes into b.
613
614 Like read(), this may issue multiple reads to the underlying raw
615 stream, unless the latter is 'interactive'.
616
617 Returns the number of bytes read (0 for EOF).
618
619 Raises BlockingIOError if the underlying raw stream has no
620 data at the moment.
621 """
622 # XXX This ought to work with anything that supports the buffer API
623 data = self.read(len(b))
624 n = len(data)
625 try:
626 b[:n] = data
627 except TypeError as err:
628 import array
629 if not isinstance(b, array.array):
630 raise err
631 b[:n] = array.array('b', data)
632 return n
633
634 def write(self, b: bytes) -> int:
635 """Write the given buffer to the IO stream.
636
637 Return the number of bytes written, which is never less than
638 len(b).
639
640 Raises BlockingIOError if the buffer is full and the
641 underlying raw stream cannot accept more data at the moment.
642 """
643 self._unsupported("write")
644
Benjamin Petersond2e0c792009-05-01 20:40:59 +0000645 def detach(self) -> None:
646 """
647 Separate the underlying raw stream from the buffer and return it.
648
649 After the raw stream has been detached, the buffer is in an unusable
650 state.
651 """
652 self._unsupported("detach")
653
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000654io.BufferedIOBase.register(BufferedIOBase)
655
656
657class _BufferedIOMixin(BufferedIOBase):
658
659 """A mixin implementation of BufferedIOBase with an underlying raw stream.
660
661 This passes most requests on to the underlying raw stream. It
662 does *not* provide implementations of read(), readinto() or
663 write().
664 """
665
666 def __init__(self, raw):
667 self.raw = raw
668
669 ### Positioning ###
670
671 def seek(self, pos, whence=0):
672 new_position = self.raw.seek(pos, whence)
673 if new_position < 0:
674 raise IOError("seek() returned an invalid position")
675 return new_position
676
677 def tell(self):
678 pos = self.raw.tell()
679 if pos < 0:
680 raise IOError("tell() returned an invalid position")
681 return pos
682
683 def truncate(self, pos=None):
684 # Flush the stream. We're mixing buffered I/O with lower-level I/O,
685 # and a flush may be necessary to synch both views of the current
686 # file state.
687 self.flush()
688
689 if pos is None:
690 pos = self.tell()
691 # XXX: Should seek() be used, instead of passing the position
692 # XXX directly to truncate?
693 return self.raw.truncate(pos)
694
695 ### Flush and close ###
696
697 def flush(self):
698 self.raw.flush()
699
700 def close(self):
Benjamin Petersond2e0c792009-05-01 20:40:59 +0000701 if not self.closed and self.raw is not None:
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000702 try:
703 self.flush()
704 except IOError:
705 pass # If flush() fails, just give up
706 self.raw.close()
707
Benjamin Petersond2e0c792009-05-01 20:40:59 +0000708 def detach(self):
709 if self.raw is None:
710 raise ValueError("raw stream already detached")
711 self.flush()
712 raw = self.raw
713 self.raw = None
714 return raw
715
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000716 ### Inquiries ###
717
718 def seekable(self):
719 return self.raw.seekable()
720
721 def readable(self):
722 return self.raw.readable()
723
724 def writable(self):
725 return self.raw.writable()
726
727 @property
728 def closed(self):
729 return self.raw.closed
730
731 @property
732 def name(self):
733 return self.raw.name
734
735 @property
736 def mode(self):
737 return self.raw.mode
738
Antoine Pitrou716c4442009-05-23 19:04:03 +0000739 def __repr__(self):
740 clsname = self.__class__.__name__
741 try:
742 name = self.name
743 except AttributeError:
744 return "<_pyio.{0}>".format(clsname)
745 else:
746 return "<_pyio.{0} name={1!r}>".format(clsname, name)
747
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000748 ### Lower-level APIs ###
749
750 def fileno(self):
751 return self.raw.fileno()
752
753 def isatty(self):
754 return self.raw.isatty()
755
756
757class BytesIO(BufferedIOBase):
758
759 """Buffered I/O implementation using an in-memory bytes buffer."""
760
761 def __init__(self, initial_bytes=None):
762 buf = bytearray()
763 if initial_bytes is not None:
764 buf += initial_bytes
765 self._buffer = buf
766 self._pos = 0
767
768 def getvalue(self):
769 """Return the bytes value (contents) of the buffer
770 """
771 if self.closed:
772 raise ValueError("getvalue on closed file")
773 return bytes(self._buffer)
774
775 def read(self, n=None):
776 if self.closed:
777 raise ValueError("read from closed file")
778 if n is None:
779 n = -1
780 if n < 0:
781 n = len(self._buffer)
782 if len(self._buffer) <= self._pos:
783 return b""
784 newpos = min(len(self._buffer), self._pos + n)
785 b = self._buffer[self._pos : newpos]
786 self._pos = newpos
787 return bytes(b)
788
789 def read1(self, n):
790 """This is the same as read.
791 """
792 return self.read(n)
793
794 def write(self, b):
795 if self.closed:
796 raise ValueError("write to closed file")
797 if isinstance(b, str):
798 raise TypeError("can't write str to binary stream")
799 n = len(b)
800 if n == 0:
801 return 0
802 pos = self._pos
803 if pos > len(self._buffer):
804 # Inserts null bytes between the current end of the file
805 # and the new write position.
806 padding = b'\x00' * (pos - len(self._buffer))
807 self._buffer += padding
808 self._buffer[pos:pos + n] = b
809 self._pos += n
810 return n
811
812 def seek(self, pos, whence=0):
813 if self.closed:
814 raise ValueError("seek on closed file")
815 try:
816 pos = pos.__index__()
817 except AttributeError as err:
818 raise TypeError("an integer is required") from err
819 if whence == 0:
820 if pos < 0:
821 raise ValueError("negative seek position %r" % (pos,))
822 self._pos = pos
823 elif whence == 1:
824 self._pos = max(0, self._pos + pos)
825 elif whence == 2:
826 self._pos = max(0, len(self._buffer) + pos)
827 else:
828 raise ValueError("invalid whence value")
829 return self._pos
830
831 def tell(self):
832 if self.closed:
833 raise ValueError("tell on closed file")
834 return self._pos
835
836 def truncate(self, pos=None):
837 if self.closed:
838 raise ValueError("truncate on closed file")
839 if pos is None:
840 pos = self._pos
841 elif pos < 0:
842 raise ValueError("negative truncate position %r" % (pos,))
843 del self._buffer[pos:]
844 return self.seek(pos)
845
846 def readable(self):
847 return True
848
849 def writable(self):
850 return True
851
852 def seekable(self):
853 return True
854
855
856class BufferedReader(_BufferedIOMixin):
857
858 """BufferedReader(raw[, buffer_size])
859
860 A buffer for a readable, sequential BaseRawIO object.
861
862 The constructor creates a BufferedReader for the given readable raw
863 stream and buffer_size. If buffer_size is omitted, DEFAULT_BUFFER_SIZE
864 is used.
865 """
866
867 def __init__(self, raw, buffer_size=DEFAULT_BUFFER_SIZE):
868 """Create a new buffered reader using the given readable raw IO object.
869 """
Antoine Pitroucf4c7492009-04-19 00:09:36 +0000870 if not raw.readable():
871 raise IOError('"raw" argument must be readable.')
872
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000873 _BufferedIOMixin.__init__(self, raw)
874 if buffer_size <= 0:
875 raise ValueError("invalid buffer size")
876 self.buffer_size = buffer_size
877 self._reset_read_buf()
878 self._read_lock = Lock()
879
880 def _reset_read_buf(self):
881 self._read_buf = b""
882 self._read_pos = 0
883
884 def read(self, n=None):
885 """Read n bytes.
886
887 Returns exactly n bytes of data unless the underlying raw IO
888 stream reaches EOF or if the call would block in non-blocking
889 mode. If n is negative, read until EOF or until read() would
890 block.
891 """
892 if n is not None and n < -1:
893 raise ValueError("invalid number of bytes to read")
894 with self._read_lock:
895 return self._read_unlocked(n)
896
897 def _read_unlocked(self, n=None):
898 nodata_val = b""
899 empty_values = (b"", None)
900 buf = self._read_buf
901 pos = self._read_pos
902
903 # Special case for when the number of bytes to read is unspecified.
904 if n is None or n == -1:
905 self._reset_read_buf()
906 chunks = [buf[pos:]] # Strip the consumed bytes.
907 current_size = 0
908 while True:
909 # Read until EOF or until read() would block.
910 chunk = self.raw.read()
911 if chunk in empty_values:
912 nodata_val = chunk
913 break
914 current_size += len(chunk)
915 chunks.append(chunk)
916 return b"".join(chunks) or nodata_val
917
918 # The number of bytes to read is specified, return at most n bytes.
919 avail = len(buf) - pos # Length of the available buffered data.
920 if n <= avail:
921 # Fast path: the data to read is fully buffered.
922 self._read_pos += n
923 return buf[pos:pos+n]
924 # Slow path: read from the stream until enough bytes are read,
925 # or until an EOF occurs or until read() would block.
926 chunks = [buf[pos:]]
927 wanted = max(self.buffer_size, n)
928 while avail < n:
929 chunk = self.raw.read(wanted)
930 if chunk in empty_values:
931 nodata_val = chunk
932 break
933 avail += len(chunk)
934 chunks.append(chunk)
935 # n is more then avail only when an EOF occurred or when
936 # read() would have blocked.
937 n = min(n, avail)
938 out = b"".join(chunks)
939 self._read_buf = out[n:] # Save the extra data in the buffer.
940 self._read_pos = 0
941 return out[:n] if out else nodata_val
942
943 def peek(self, n=0):
944 """Returns buffered bytes without advancing the position.
945
946 The argument indicates a desired minimal number of bytes; we
947 do at most one raw read to satisfy it. We never return more
948 than self.buffer_size.
949 """
950 with self._read_lock:
951 return self._peek_unlocked(n)
952
953 def _peek_unlocked(self, n=0):
954 want = min(n, self.buffer_size)
955 have = len(self._read_buf) - self._read_pos
956 if have < want or have <= 0:
957 to_read = self.buffer_size - have
958 current = self.raw.read(to_read)
959 if current:
960 self._read_buf = self._read_buf[self._read_pos:] + current
961 self._read_pos = 0
962 return self._read_buf[self._read_pos:]
963
964 def read1(self, n):
965 """Reads up to n bytes, with at most one read() system call."""
966 # Returns up to n bytes. If at least one byte is buffered, we
967 # only return buffered bytes. Otherwise, we do one raw read.
968 if n < 0:
969 raise ValueError("number of bytes to read must be positive")
970 if n == 0:
971 return b""
972 with self._read_lock:
973 self._peek_unlocked(1)
974 return self._read_unlocked(
975 min(n, len(self._read_buf) - self._read_pos))
976
977 def tell(self):
978 return _BufferedIOMixin.tell(self) - len(self._read_buf) + self._read_pos
979
980 def seek(self, pos, whence=0):
981 if not (0 <= whence <= 2):
982 raise ValueError("invalid whence value")
983 with self._read_lock:
984 if whence == 1:
985 pos -= len(self._read_buf) - self._read_pos
986 pos = _BufferedIOMixin.seek(self, pos, whence)
987 self._reset_read_buf()
988 return pos
989
990class BufferedWriter(_BufferedIOMixin):
991
992 """A buffer for a writeable sequential RawIO object.
993
994 The constructor creates a BufferedWriter for the given writeable raw
995 stream. If the buffer_size is not given, it defaults to
Benjamin Peterson59406a92009-03-26 17:10:29 +0000996 DEFAULT_BUFFER_SIZE.
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000997 """
998
Benjamin Peterson59406a92009-03-26 17:10:29 +0000999 _warning_stack_offset = 2
1000
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001001 def __init__(self, raw,
1002 buffer_size=DEFAULT_BUFFER_SIZE, max_buffer_size=None):
Antoine Pitroucf4c7492009-04-19 00:09:36 +00001003 if not raw.writable():
1004 raise IOError('"raw" argument must be writable.')
1005
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001006 _BufferedIOMixin.__init__(self, raw)
1007 if buffer_size <= 0:
1008 raise ValueError("invalid buffer size")
Benjamin Peterson59406a92009-03-26 17:10:29 +00001009 if max_buffer_size is not None:
1010 warnings.warn("max_buffer_size is deprecated", DeprecationWarning,
1011 self._warning_stack_offset)
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001012 self.buffer_size = buffer_size
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001013 self._write_buf = bytearray()
1014 self._write_lock = Lock()
1015
1016 def write(self, b):
1017 if self.closed:
1018 raise ValueError("write to closed file")
1019 if isinstance(b, str):
1020 raise TypeError("can't write str to binary stream")
1021 with self._write_lock:
1022 # XXX we can implement some more tricks to try and avoid
1023 # partial writes
1024 if len(self._write_buf) > self.buffer_size:
1025 # We're full, so let's pre-flush the buffer
1026 try:
1027 self._flush_unlocked()
1028 except BlockingIOError as e:
1029 # We can't accept anything else.
1030 # XXX Why not just let the exception pass through?
1031 raise BlockingIOError(e.errno, e.strerror, 0)
1032 before = len(self._write_buf)
1033 self._write_buf.extend(b)
1034 written = len(self._write_buf) - before
1035 if len(self._write_buf) > self.buffer_size:
1036 try:
1037 self._flush_unlocked()
1038 except BlockingIOError as e:
Benjamin Peterson394ee002009-03-05 22:33:59 +00001039 if len(self._write_buf) > self.buffer_size:
1040 # We've hit the buffer_size. We have to accept a partial
1041 # write and cut back our buffer.
1042 overage = len(self._write_buf) - self.buffer_size
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001043 written -= overage
Benjamin Peterson394ee002009-03-05 22:33:59 +00001044 self._write_buf = self._write_buf[:self.buffer_size]
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001045 raise BlockingIOError(e.errno, e.strerror, written)
1046 return written
1047
1048 def truncate(self, pos=None):
1049 with self._write_lock:
1050 self._flush_unlocked()
1051 if pos is None:
1052 pos = self.raw.tell()
1053 return self.raw.truncate(pos)
1054
1055 def flush(self):
1056 with self._write_lock:
1057 self._flush_unlocked()
1058
1059 def _flush_unlocked(self):
1060 if self.closed:
1061 raise ValueError("flush of closed file")
1062 written = 0
1063 try:
1064 while self._write_buf:
1065 n = self.raw.write(self._write_buf)
1066 if n > len(self._write_buf) or n < 0:
1067 raise IOError("write() returned incorrect number of bytes")
1068 del self._write_buf[:n]
1069 written += n
1070 except BlockingIOError as e:
1071 n = e.characters_written
1072 del self._write_buf[:n]
1073 written += n
1074 raise BlockingIOError(e.errno, e.strerror, written)
1075
1076 def tell(self):
1077 return _BufferedIOMixin.tell(self) + len(self._write_buf)
1078
1079 def seek(self, pos, whence=0):
1080 if not (0 <= whence <= 2):
1081 raise ValueError("invalid whence")
1082 with self._write_lock:
1083 self._flush_unlocked()
1084 return _BufferedIOMixin.seek(self, pos, whence)
1085
1086
1087class BufferedRWPair(BufferedIOBase):
1088
1089 """A buffered reader and writer object together.
1090
1091 A buffered reader object and buffered writer object put together to
1092 form a sequential IO object that can read and write. This is typically
1093 used with a socket or two-way pipe.
1094
1095 reader and writer are RawIOBase objects that are readable and
1096 writeable respectively. If the buffer_size is omitted it defaults to
Benjamin Peterson59406a92009-03-26 17:10:29 +00001097 DEFAULT_BUFFER_SIZE.
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001098 """
1099
1100 # XXX The usefulness of this (compared to having two separate IO
1101 # objects) is questionable.
1102
1103 def __init__(self, reader, writer,
1104 buffer_size=DEFAULT_BUFFER_SIZE, max_buffer_size=None):
1105 """Constructor.
1106
1107 The arguments are two RawIO instances.
1108 """
Benjamin Peterson59406a92009-03-26 17:10:29 +00001109 if max_buffer_size is not None:
1110 warnings.warn("max_buffer_size is deprecated", DeprecationWarning, 2)
Antoine Pitroucf4c7492009-04-19 00:09:36 +00001111
1112 if not reader.readable():
1113 raise IOError('"reader" argument must be readable.')
1114
1115 if not writer.writable():
1116 raise IOError('"writer" argument must be writable.')
1117
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001118 self.reader = BufferedReader(reader, buffer_size)
Benjamin Peterson59406a92009-03-26 17:10:29 +00001119 self.writer = BufferedWriter(writer, buffer_size)
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001120
1121 def read(self, n=None):
1122 if n is None:
1123 n = -1
1124 return self.reader.read(n)
1125
1126 def readinto(self, b):
1127 return self.reader.readinto(b)
1128
1129 def write(self, b):
1130 return self.writer.write(b)
1131
1132 def peek(self, n=0):
1133 return self.reader.peek(n)
1134
1135 def read1(self, n):
1136 return self.reader.read1(n)
1137
1138 def readable(self):
1139 return self.reader.readable()
1140
1141 def writable(self):
1142 return self.writer.writable()
1143
1144 def flush(self):
1145 return self.writer.flush()
1146
1147 def close(self):
1148 self.writer.close()
1149 self.reader.close()
1150
1151 def isatty(self):
1152 return self.reader.isatty() or self.writer.isatty()
1153
1154 @property
1155 def closed(self):
1156 return self.writer.closed
1157
1158
1159class BufferedRandom(BufferedWriter, BufferedReader):
1160
1161 """A buffered interface to random access streams.
1162
1163 The constructor creates a reader and writer for a seekable stream,
1164 raw, given in the first argument. If the buffer_size is omitted it
Benjamin Peterson59406a92009-03-26 17:10:29 +00001165 defaults to DEFAULT_BUFFER_SIZE.
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001166 """
1167
Benjamin Peterson59406a92009-03-26 17:10:29 +00001168 _warning_stack_offset = 3
1169
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001170 def __init__(self, raw,
1171 buffer_size=DEFAULT_BUFFER_SIZE, max_buffer_size=None):
1172 raw._checkSeekable()
1173 BufferedReader.__init__(self, raw, buffer_size)
1174 BufferedWriter.__init__(self, raw, buffer_size, max_buffer_size)
1175
1176 def seek(self, pos, whence=0):
1177 if not (0 <= whence <= 2):
1178 raise ValueError("invalid whence")
1179 self.flush()
1180 if self._read_buf:
1181 # Undo read ahead.
1182 with self._read_lock:
1183 self.raw.seek(self._read_pos - len(self._read_buf), 1)
1184 # First do the raw seek, then empty the read buffer, so that
1185 # if the raw seek fails, we don't lose buffered data forever.
1186 pos = self.raw.seek(pos, whence)
1187 with self._read_lock:
1188 self._reset_read_buf()
1189 if pos < 0:
1190 raise IOError("seek() returned invalid position")
1191 return pos
1192
1193 def tell(self):
1194 if self._write_buf:
1195 return BufferedWriter.tell(self)
1196 else:
1197 return BufferedReader.tell(self)
1198
1199 def truncate(self, pos=None):
1200 if pos is None:
1201 pos = self.tell()
1202 # Use seek to flush the read buffer.
1203 self.seek(pos)
1204 return BufferedWriter.truncate(self)
1205
1206 def read(self, n=None):
1207 if n is None:
1208 n = -1
1209 self.flush()
1210 return BufferedReader.read(self, n)
1211
1212 def readinto(self, b):
1213 self.flush()
1214 return BufferedReader.readinto(self, b)
1215
1216 def peek(self, n=0):
1217 self.flush()
1218 return BufferedReader.peek(self, n)
1219
1220 def read1(self, n):
1221 self.flush()
1222 return BufferedReader.read1(self, n)
1223
1224 def write(self, b):
1225 if self._read_buf:
1226 # Undo readahead
1227 with self._read_lock:
1228 self.raw.seek(self._read_pos - len(self._read_buf), 1)
1229 self._reset_read_buf()
1230 return BufferedWriter.write(self, b)
1231
1232
1233class TextIOBase(IOBase):
1234
1235 """Base class for text I/O.
1236
1237 This class provides a character and line based interface to stream
1238 I/O. There is no readinto method because Python's character strings
1239 are immutable. There is no public constructor.
1240 """
1241
1242 def read(self, n: int = -1) -> str:
1243 """Read at most n characters from stream.
1244
1245 Read from underlying buffer until we have n characters or we hit EOF.
1246 If n is negative or omitted, read until EOF.
1247 """
1248 self._unsupported("read")
1249
1250 def write(self, s: str) -> int:
1251 """Write string s to stream."""
1252 self._unsupported("write")
1253
1254 def truncate(self, pos: int = None) -> int:
1255 """Truncate size to pos."""
1256 self._unsupported("truncate")
1257
1258 def readline(self) -> str:
1259 """Read until newline or EOF.
1260
1261 Returns an empty string if EOF is hit immediately.
1262 """
1263 self._unsupported("readline")
1264
Benjamin Petersond2e0c792009-05-01 20:40:59 +00001265 def detach(self) -> None:
1266 """
1267 Separate the underlying buffer from the TextIOBase and return it.
1268
1269 After the underlying buffer has been detached, the TextIO is in an
1270 unusable state.
1271 """
1272 self._unsupported("detach")
1273
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001274 @property
1275 def encoding(self):
1276 """Subclasses should override."""
1277 return None
1278
1279 @property
1280 def newlines(self):
1281 """Line endings translated so far.
1282
1283 Only line endings translated during reading are considered.
1284
1285 Subclasses should override.
1286 """
1287 return None
1288
1289io.TextIOBase.register(TextIOBase)
1290
1291
1292class IncrementalNewlineDecoder(codecs.IncrementalDecoder):
1293 r"""Codec used when reading a file in universal newlines mode. It wraps
1294 another incremental decoder, translating \r\n and \r into \n. It also
1295 records the types of newlines encountered. When used with
1296 translate=False, it ensures that the newline sequence is returned in
1297 one piece.
1298 """
1299 def __init__(self, decoder, translate, errors='strict'):
1300 codecs.IncrementalDecoder.__init__(self, errors=errors)
1301 self.translate = translate
1302 self.decoder = decoder
1303 self.seennl = 0
1304 self.pendingcr = False
1305
1306 def decode(self, input, final=False):
1307 # decode input (with the eventual \r from a previous pass)
1308 if self.decoder is None:
1309 output = input
1310 else:
1311 output = self.decoder.decode(input, final=final)
1312 if self.pendingcr and (output or final):
1313 output = "\r" + output
1314 self.pendingcr = False
1315
1316 # retain last \r even when not translating data:
1317 # then readline() is sure to get \r\n in one pass
1318 if output.endswith("\r") and not final:
1319 output = output[:-1]
1320 self.pendingcr = True
1321
1322 # Record which newlines are read
1323 crlf = output.count('\r\n')
1324 cr = output.count('\r') - crlf
1325 lf = output.count('\n') - crlf
1326 self.seennl |= (lf and self._LF) | (cr and self._CR) \
1327 | (crlf and self._CRLF)
1328
1329 if self.translate:
1330 if crlf:
1331 output = output.replace("\r\n", "\n")
1332 if cr:
1333 output = output.replace("\r", "\n")
1334
1335 return output
1336
1337 def getstate(self):
1338 if self.decoder is None:
1339 buf = b""
1340 flag = 0
1341 else:
1342 buf, flag = self.decoder.getstate()
1343 flag <<= 1
1344 if self.pendingcr:
1345 flag |= 1
1346 return buf, flag
1347
1348 def setstate(self, state):
1349 buf, flag = state
1350 self.pendingcr = bool(flag & 1)
1351 if self.decoder is not None:
1352 self.decoder.setstate((buf, flag >> 1))
1353
1354 def reset(self):
1355 self.seennl = 0
1356 self.pendingcr = False
1357 if self.decoder is not None:
1358 self.decoder.reset()
1359
1360 _LF = 1
1361 _CR = 2
1362 _CRLF = 4
1363
1364 @property
1365 def newlines(self):
1366 return (None,
1367 "\n",
1368 "\r",
1369 ("\r", "\n"),
1370 "\r\n",
1371 ("\n", "\r\n"),
1372 ("\r", "\r\n"),
1373 ("\r", "\n", "\r\n")
1374 )[self.seennl]
1375
1376
1377class TextIOWrapper(TextIOBase):
1378
1379 r"""Character and line based layer over a BufferedIOBase object, buffer.
1380
1381 encoding gives the name of the encoding that the stream will be
1382 decoded or encoded with. It defaults to locale.getpreferredencoding.
1383
1384 errors determines the strictness of encoding and decoding (see the
1385 codecs.register) and defaults to "strict".
1386
1387 newline can be None, '', '\n', '\r', or '\r\n'. It controls the
1388 handling of line endings. If it is None, universal newlines is
1389 enabled. With this enabled, on input, the lines endings '\n', '\r',
1390 or '\r\n' are translated to '\n' before being returned to the
1391 caller. Conversely, on output, '\n' is translated to the system
1392 default line seperator, os.linesep. If newline is any other of its
1393 legal values, that newline becomes the newline when the file is read
1394 and it is returned untranslated. On output, '\n' is converted to the
1395 newline.
1396
1397 If line_buffering is True, a call to flush is implied when a call to
1398 write contains a newline character.
1399 """
1400
1401 _CHUNK_SIZE = 2048
1402
1403 def __init__(self, buffer, encoding=None, errors=None, newline=None,
1404 line_buffering=False):
1405 if newline is not None and not isinstance(newline, str):
1406 raise TypeError("illegal newline type: %r" % (type(newline),))
1407 if newline not in (None, "", "\n", "\r", "\r\n"):
1408 raise ValueError("illegal newline value: %r" % (newline,))
1409 if encoding is None:
1410 try:
1411 encoding = os.device_encoding(buffer.fileno())
1412 except (AttributeError, UnsupportedOperation):
1413 pass
1414 if encoding is None:
1415 try:
1416 import locale
1417 except ImportError:
1418 # Importing locale may fail if Python is being built
1419 encoding = "ascii"
1420 else:
1421 encoding = locale.getpreferredencoding()
1422
1423 if not isinstance(encoding, str):
1424 raise ValueError("invalid encoding: %r" % encoding)
1425
1426 if errors is None:
1427 errors = "strict"
1428 else:
1429 if not isinstance(errors, str):
1430 raise ValueError("invalid errors: %r" % errors)
1431
1432 self.buffer = buffer
1433 self._line_buffering = line_buffering
1434 self._encoding = encoding
1435 self._errors = errors
1436 self._readuniversal = not newline
1437 self._readtranslate = newline is None
1438 self._readnl = newline
1439 self._writetranslate = newline != ''
1440 self._writenl = newline or os.linesep
1441 self._encoder = None
1442 self._decoder = None
1443 self._decoded_chars = '' # buffer for text returned from decoder
1444 self._decoded_chars_used = 0 # offset into _decoded_chars for read()
1445 self._snapshot = None # info for reconstructing decoder state
1446 self._seekable = self._telling = self.buffer.seekable()
1447
Antoine Pitroue4501852009-05-14 18:55:55 +00001448 if self._seekable and self.writable():
1449 position = self.buffer.tell()
1450 if position != 0:
1451 try:
1452 self._get_encoder().setstate(0)
1453 except LookupError:
1454 # Sometimes the encoder doesn't exist
1455 pass
1456
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001457 # self._snapshot is either None, or a tuple (dec_flags, next_input)
1458 # where dec_flags is the second (integer) item of the decoder state
1459 # and next_input is the chunk of input bytes that comes next after the
1460 # snapshot point. We use this to reconstruct decoder states in tell().
1461
1462 # Naming convention:
1463 # - "bytes_..." for integer variables that count input bytes
1464 # - "chars_..." for integer variables that count decoded characters
1465
Benjamin Petersonc4c0eae2009-03-09 00:07:03 +00001466 def __repr__(self):
Antoine Pitrou716c4442009-05-23 19:04:03 +00001467 try:
1468 name = self.name
1469 except AttributeError:
1470 return "<_pyio.TextIOWrapper encoding={0!r}>".format(self.encoding)
1471 else:
1472 return "<_pyio.TextIOWrapper name={0!r} encoding={1!r}>".format(
1473 name, self.encoding)
Benjamin Petersonc4c0eae2009-03-09 00:07:03 +00001474
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001475 @property
1476 def encoding(self):
1477 return self._encoding
1478
1479 @property
1480 def errors(self):
1481 return self._errors
1482
1483 @property
1484 def line_buffering(self):
1485 return self._line_buffering
1486
1487 def seekable(self):
1488 return self._seekable
1489
1490 def readable(self):
1491 return self.buffer.readable()
1492
1493 def writable(self):
1494 return self.buffer.writable()
1495
1496 def flush(self):
1497 self.buffer.flush()
1498 self._telling = self._seekable
1499
1500 def close(self):
Benjamin Petersond2e0c792009-05-01 20:40:59 +00001501 if self.buffer is not None:
1502 try:
1503 self.flush()
1504 except IOError:
1505 pass # If flush() fails, just give up
1506 self.buffer.close()
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001507
1508 @property
1509 def closed(self):
1510 return self.buffer.closed
1511
1512 @property
1513 def name(self):
1514 return self.buffer.name
1515
1516 def fileno(self):
1517 return self.buffer.fileno()
1518
1519 def isatty(self):
1520 return self.buffer.isatty()
1521
1522 def write(self, s: str):
1523 if self.closed:
1524 raise ValueError("write to closed file")
1525 if not isinstance(s, str):
1526 raise TypeError("can't write %s to text stream" %
1527 s.__class__.__name__)
1528 length = len(s)
1529 haslf = (self._writetranslate or self._line_buffering) and "\n" in s
1530 if haslf and self._writetranslate and self._writenl != "\n":
1531 s = s.replace("\n", self._writenl)
1532 encoder = self._encoder or self._get_encoder()
1533 # XXX What if we were just reading?
1534 b = encoder.encode(s)
1535 self.buffer.write(b)
1536 if self._line_buffering and (haslf or "\r" in s):
1537 self.flush()
1538 self._snapshot = None
1539 if self._decoder:
1540 self._decoder.reset()
1541 return length
1542
1543 def _get_encoder(self):
1544 make_encoder = codecs.getincrementalencoder(self._encoding)
1545 self._encoder = make_encoder(self._errors)
1546 return self._encoder
1547
1548 def _get_decoder(self):
1549 make_decoder = codecs.getincrementaldecoder(self._encoding)
1550 decoder = make_decoder(self._errors)
1551 if self._readuniversal:
1552 decoder = IncrementalNewlineDecoder(decoder, self._readtranslate)
1553 self._decoder = decoder
1554 return decoder
1555
1556 # The following three methods implement an ADT for _decoded_chars.
1557 # Text returned from the decoder is buffered here until the client
1558 # requests it by calling our read() or readline() method.
1559 def _set_decoded_chars(self, chars):
1560 """Set the _decoded_chars buffer."""
1561 self._decoded_chars = chars
1562 self._decoded_chars_used = 0
1563
1564 def _get_decoded_chars(self, n=None):
1565 """Advance into the _decoded_chars buffer."""
1566 offset = self._decoded_chars_used
1567 if n is None:
1568 chars = self._decoded_chars[offset:]
1569 else:
1570 chars = self._decoded_chars[offset:offset + n]
1571 self._decoded_chars_used += len(chars)
1572 return chars
1573
1574 def _rewind_decoded_chars(self, n):
1575 """Rewind the _decoded_chars buffer."""
1576 if self._decoded_chars_used < n:
1577 raise AssertionError("rewind decoded_chars out of bounds")
1578 self._decoded_chars_used -= n
1579
1580 def _read_chunk(self):
1581 """
1582 Read and decode the next chunk of data from the BufferedReader.
1583 """
1584
1585 # The return value is True unless EOF was reached. The decoded
1586 # string is placed in self._decoded_chars (replacing its previous
1587 # value). The entire input chunk is sent to the decoder, though
1588 # some of it may remain buffered in the decoder, yet to be
1589 # converted.
1590
1591 if self._decoder is None:
1592 raise ValueError("no decoder")
1593
1594 if self._telling:
1595 # To prepare for tell(), we need to snapshot a point in the
1596 # file where the decoder's input buffer is empty.
1597
1598 dec_buffer, dec_flags = self._decoder.getstate()
1599 # Given this, we know there was a valid snapshot point
1600 # len(dec_buffer) bytes ago with decoder state (b'', dec_flags).
1601
1602 # Read a chunk, decode it, and put the result in self._decoded_chars.
1603 input_chunk = self.buffer.read1(self._CHUNK_SIZE)
1604 eof = not input_chunk
1605 self._set_decoded_chars(self._decoder.decode(input_chunk, eof))
1606
1607 if self._telling:
1608 # At the snapshot point, len(dec_buffer) bytes before the read,
1609 # the next input to be decoded is dec_buffer + input_chunk.
1610 self._snapshot = (dec_flags, dec_buffer + input_chunk)
1611
1612 return not eof
1613
1614 def _pack_cookie(self, position, dec_flags=0,
1615 bytes_to_feed=0, need_eof=0, chars_to_skip=0):
1616 # The meaning of a tell() cookie is: seek to position, set the
1617 # decoder flags to dec_flags, read bytes_to_feed bytes, feed them
1618 # into the decoder with need_eof as the EOF flag, then skip
1619 # chars_to_skip characters of the decoded result. For most simple
1620 # decoders, tell() will often just give a byte offset in the file.
1621 return (position | (dec_flags<<64) | (bytes_to_feed<<128) |
1622 (chars_to_skip<<192) | bool(need_eof)<<256)
1623
1624 def _unpack_cookie(self, bigint):
1625 rest, position = divmod(bigint, 1<<64)
1626 rest, dec_flags = divmod(rest, 1<<64)
1627 rest, bytes_to_feed = divmod(rest, 1<<64)
1628 need_eof, chars_to_skip = divmod(rest, 1<<64)
1629 return position, dec_flags, bytes_to_feed, need_eof, chars_to_skip
1630
1631 def tell(self):
1632 if not self._seekable:
1633 raise IOError("underlying stream is not seekable")
1634 if not self._telling:
1635 raise IOError("telling position disabled by next() call")
1636 self.flush()
1637 position = self.buffer.tell()
1638 decoder = self._decoder
1639 if decoder is None or self._snapshot is None:
1640 if self._decoded_chars:
1641 # This should never happen.
1642 raise AssertionError("pending decoded text")
1643 return position
1644
1645 # Skip backward to the snapshot point (see _read_chunk).
1646 dec_flags, next_input = self._snapshot
1647 position -= len(next_input)
1648
1649 # How many decoded characters have been used up since the snapshot?
1650 chars_to_skip = self._decoded_chars_used
1651 if chars_to_skip == 0:
1652 # We haven't moved from the snapshot point.
1653 return self._pack_cookie(position, dec_flags)
1654
1655 # Starting from the snapshot position, we will walk the decoder
1656 # forward until it gives us enough decoded characters.
1657 saved_state = decoder.getstate()
1658 try:
1659 # Note our initial start point.
1660 decoder.setstate((b'', dec_flags))
1661 start_pos = position
1662 start_flags, bytes_fed, chars_decoded = dec_flags, 0, 0
1663 need_eof = 0
1664
1665 # Feed the decoder one byte at a time. As we go, note the
1666 # nearest "safe start point" before the current location
1667 # (a point where the decoder has nothing buffered, so seek()
1668 # can safely start from there and advance to this location).
1669 next_byte = bytearray(1)
1670 for next_byte[0] in next_input:
1671 bytes_fed += 1
1672 chars_decoded += len(decoder.decode(next_byte))
1673 dec_buffer, dec_flags = decoder.getstate()
1674 if not dec_buffer and chars_decoded <= chars_to_skip:
1675 # Decoder buffer is empty, so this is a safe start point.
1676 start_pos += bytes_fed
1677 chars_to_skip -= chars_decoded
1678 start_flags, bytes_fed, chars_decoded = dec_flags, 0, 0
1679 if chars_decoded >= chars_to_skip:
1680 break
1681 else:
1682 # We didn't get enough decoded data; signal EOF to get more.
1683 chars_decoded += len(decoder.decode(b'', final=True))
1684 need_eof = 1
1685 if chars_decoded < chars_to_skip:
1686 raise IOError("can't reconstruct logical file position")
1687
1688 # The returned cookie corresponds to the last safe start point.
1689 return self._pack_cookie(
1690 start_pos, start_flags, bytes_fed, need_eof, chars_to_skip)
1691 finally:
1692 decoder.setstate(saved_state)
1693
1694 def truncate(self, pos=None):
1695 self.flush()
1696 if pos is None:
1697 pos = self.tell()
1698 self.seek(pos)
1699 return self.buffer.truncate()
1700
Benjamin Petersond2e0c792009-05-01 20:40:59 +00001701 def detach(self):
1702 if self.buffer is None:
1703 raise ValueError("buffer is already detached")
1704 self.flush()
1705 buffer = self.buffer
1706 self.buffer = None
1707 return buffer
1708
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001709 def seek(self, cookie, whence=0):
1710 if self.closed:
1711 raise ValueError("tell on closed file")
1712 if not self._seekable:
1713 raise IOError("underlying stream is not seekable")
1714 if whence == 1: # seek relative to current position
1715 if cookie != 0:
1716 raise IOError("can't do nonzero cur-relative seeks")
1717 # Seeking to the current position should attempt to
1718 # sync the underlying buffer with the current position.
1719 whence = 0
1720 cookie = self.tell()
1721 if whence == 2: # seek relative to end of file
1722 if cookie != 0:
1723 raise IOError("can't do nonzero end-relative seeks")
1724 self.flush()
1725 position = self.buffer.seek(0, 2)
1726 self._set_decoded_chars('')
1727 self._snapshot = None
1728 if self._decoder:
1729 self._decoder.reset()
1730 return position
1731 if whence != 0:
1732 raise ValueError("invalid whence (%r, should be 0, 1 or 2)" %
1733 (whence,))
1734 if cookie < 0:
1735 raise ValueError("negative seek position %r" % (cookie,))
1736 self.flush()
1737
1738 # The strategy of seek() is to go back to the safe start point
1739 # and replay the effect of read(chars_to_skip) from there.
1740 start_pos, dec_flags, bytes_to_feed, need_eof, chars_to_skip = \
1741 self._unpack_cookie(cookie)
1742
1743 # Seek back to the safe start point.
1744 self.buffer.seek(start_pos)
1745 self._set_decoded_chars('')
1746 self._snapshot = None
1747
1748 # Restore the decoder to its state from the safe start point.
Benjamin Peterson9363a652009-03-05 00:42:09 +00001749 if cookie == 0 and self._decoder:
1750 self._decoder.reset()
1751 elif self._decoder or dec_flags or chars_to_skip:
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001752 self._decoder = self._decoder or self._get_decoder()
1753 self._decoder.setstate((b'', dec_flags))
1754 self._snapshot = (dec_flags, b'')
1755
1756 if chars_to_skip:
1757 # Just like _read_chunk, feed the decoder and save a snapshot.
1758 input_chunk = self.buffer.read(bytes_to_feed)
1759 self._set_decoded_chars(
1760 self._decoder.decode(input_chunk, need_eof))
1761 self._snapshot = (dec_flags, input_chunk)
1762
1763 # Skip chars_to_skip of the decoded characters.
1764 if len(self._decoded_chars) < chars_to_skip:
1765 raise IOError("can't restore logical file position")
1766 self._decoded_chars_used = chars_to_skip
1767
Antoine Pitroue4501852009-05-14 18:55:55 +00001768 # Finally, reset the encoder (merely useful for proper BOM handling)
1769 try:
1770 encoder = self._encoder or self._get_encoder()
1771 except LookupError:
1772 # Sometimes the encoder doesn't exist
1773 pass
1774 else:
1775 if cookie != 0:
1776 encoder.setstate(0)
1777 else:
1778 encoder.reset()
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001779 return cookie
1780
1781 def read(self, n=None):
Benjamin Petersona1b49012009-03-31 23:11:32 +00001782 self._checkReadable()
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001783 if n is None:
1784 n = -1
1785 decoder = self._decoder or self._get_decoder()
1786 if n < 0:
1787 # Read everything.
1788 result = (self._get_decoded_chars() +
1789 decoder.decode(self.buffer.read(), final=True))
1790 self._set_decoded_chars('')
1791 self._snapshot = None
1792 return result
1793 else:
1794 # Keep reading chunks until we have n characters to return.
1795 eof = False
1796 result = self._get_decoded_chars(n)
1797 while len(result) < n and not eof:
1798 eof = not self._read_chunk()
1799 result += self._get_decoded_chars(n - len(result))
1800 return result
1801
1802 def __next__(self):
1803 self._telling = False
1804 line = self.readline()
1805 if not line:
1806 self._snapshot = None
1807 self._telling = self._seekable
1808 raise StopIteration
1809 return line
1810
1811 def readline(self, limit=None):
1812 if self.closed:
1813 raise ValueError("read from closed file")
1814 if limit is None:
1815 limit = -1
Benjamin Petersonb01138a2009-04-24 22:59:52 +00001816 elif not isinstance(limit, int):
1817 raise TypeError("limit must be an integer")
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001818
1819 # Grab all the decoded text (we will rewind any extra bits later).
1820 line = self._get_decoded_chars()
1821
1822 start = 0
1823 # Make the decoder if it doesn't already exist.
1824 if not self._decoder:
1825 self._get_decoder()
1826
1827 pos = endpos = None
1828 while True:
1829 if self._readtranslate:
1830 # Newlines are already translated, only search for \n
1831 pos = line.find('\n', start)
1832 if pos >= 0:
1833 endpos = pos + 1
1834 break
1835 else:
1836 start = len(line)
1837
1838 elif self._readuniversal:
1839 # Universal newline search. Find any of \r, \r\n, \n
1840 # The decoder ensures that \r\n are not split in two pieces
1841
1842 # In C we'd look for these in parallel of course.
1843 nlpos = line.find("\n", start)
1844 crpos = line.find("\r", start)
1845 if crpos == -1:
1846 if nlpos == -1:
1847 # Nothing found
1848 start = len(line)
1849 else:
1850 # Found \n
1851 endpos = nlpos + 1
1852 break
1853 elif nlpos == -1:
1854 # Found lone \r
1855 endpos = crpos + 1
1856 break
1857 elif nlpos < crpos:
1858 # Found \n
1859 endpos = nlpos + 1
1860 break
1861 elif nlpos == crpos + 1:
1862 # Found \r\n
1863 endpos = crpos + 2
1864 break
1865 else:
1866 # Found \r
1867 endpos = crpos + 1
1868 break
1869 else:
1870 # non-universal
1871 pos = line.find(self._readnl)
1872 if pos >= 0:
1873 endpos = pos + len(self._readnl)
1874 break
1875
1876 if limit >= 0 and len(line) >= limit:
1877 endpos = limit # reached length limit
1878 break
1879
1880 # No line ending seen yet - get more data'
1881 while self._read_chunk():
1882 if self._decoded_chars:
1883 break
1884 if self._decoded_chars:
1885 line += self._get_decoded_chars()
1886 else:
1887 # end of file
1888 self._set_decoded_chars('')
1889 self._snapshot = None
1890 return line
1891
1892 if limit >= 0 and endpos > limit:
1893 endpos = limit # don't exceed limit
1894
1895 # Rewind _decoded_chars to just after the line ending we found.
1896 self._rewind_decoded_chars(len(line) - endpos)
1897 return line[:endpos]
1898
1899 @property
1900 def newlines(self):
1901 return self._decoder.newlines if self._decoder else None
1902
1903
1904class StringIO(TextIOWrapper):
1905 """Text I/O implementation using an in-memory buffer.
1906
1907 The initial_value argument sets the value of object. The newline
1908 argument is like the one of TextIOWrapper's constructor.
1909 """
1910
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001911 def __init__(self, initial_value="", newline="\n"):
1912 super(StringIO, self).__init__(BytesIO(),
1913 encoding="utf-8",
1914 errors="strict",
1915 newline=newline)
Antoine Pitrou11446482009-04-04 14:09:30 +00001916 # Issue #5645: make universal newlines semantics the same as in the
1917 # C version, even under Windows.
1918 if newline is None:
1919 self._writetranslate = False
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001920 if initial_value:
1921 if not isinstance(initial_value, str):
1922 initial_value = str(initial_value)
1923 self.write(initial_value)
1924 self.seek(0)
1925
1926 def getvalue(self):
1927 self.flush()
1928 return self.buffer.getvalue().decode(self._encoding, self._errors)
Benjamin Peterson9fd459a2009-03-09 00:09:44 +00001929
1930 def __repr__(self):
1931 # TextIOWrapper tells the encoding in its repr. In StringIO,
1932 # that's a implementation detail.
1933 return object.__repr__(self)
Benjamin Petersonb487e632009-03-21 03:08:31 +00001934
1935 @property
1936 def encoding(self):
1937 return None
Benjamin Petersond2e0c792009-05-01 20:40:59 +00001938
1939 def detach(self):
1940 # This doesn't make sense on StringIO.
1941 self._unsupported("detach")