blob: 0ef6822044d6e6b6330fbcc3844e40bfbec75781 [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
463 res = bytearray()
464 while limit < 0 or len(res) < limit:
465 b = self.read(nreadahead())
466 if not b:
467 break
468 res += b
469 if res.endswith(b"\n"):
470 break
471 return bytes(res)
472
473 def __iter__(self):
474 self._checkClosed()
475 return self
476
477 def __next__(self):
478 line = self.readline()
479 if not line:
480 raise StopIteration
481 return line
482
483 def readlines(self, hint=None):
484 """Return a list of lines from the stream.
485
486 hint can be specified to control the number of lines read: no more
487 lines will be read if the total size (in bytes/characters) of all
488 lines so far exceeds hint.
489 """
490 if hint is None or hint <= 0:
491 return list(self)
492 n = 0
493 lines = []
494 for line in self:
495 lines.append(line)
496 n += len(line)
497 if n >= hint:
498 break
499 return lines
500
501 def writelines(self, lines):
502 self._checkClosed()
503 for line in lines:
504 self.write(line)
505
506io.IOBase.register(IOBase)
507
508
509class RawIOBase(IOBase):
510
511 """Base class for raw binary I/O."""
512
513 # The read() method is implemented by calling readinto(); derived
514 # classes that want to support read() only need to implement
515 # readinto() as a primitive operation. In general, readinto() can be
516 # more efficient than read().
517
518 # (It would be tempting to also provide an implementation of
519 # readinto() in terms of read(), in case the latter is a more suitable
520 # primitive operation, but that would lead to nasty recursion in case
521 # a subclass doesn't implement either.)
522
523 def read(self, n: int = -1) -> bytes:
524 """Read and return up to n bytes.
525
526 Returns an empty bytes object on EOF, or None if the object is
527 set not to block and has no data to read.
528 """
529 if n is None:
530 n = -1
531 if n < 0:
532 return self.readall()
533 b = bytearray(n.__index__())
534 n = self.readinto(b)
535 del b[n:]
536 return bytes(b)
537
538 def readall(self):
539 """Read until EOF, using multiple read() call."""
540 res = bytearray()
541 while True:
542 data = self.read(DEFAULT_BUFFER_SIZE)
543 if not data:
544 break
545 res += data
546 return bytes(res)
547
548 def readinto(self, b: bytearray) -> int:
549 """Read up to len(b) bytes into b.
550
551 Returns number of bytes read (0 for EOF), or None if the object
552 is set not to block as has no data to read.
553 """
554 self._unsupported("readinto")
555
556 def write(self, b: bytes) -> int:
557 """Write the given buffer to the IO stream.
558
559 Returns the number of bytes written, which may be less than len(b).
560 """
561 self._unsupported("write")
562
563io.RawIOBase.register(RawIOBase)
564from _io import FileIO
565RawIOBase.register(FileIO)
566
567
568class BufferedIOBase(IOBase):
569
570 """Base class for buffered IO objects.
571
572 The main difference with RawIOBase is that the read() method
573 supports omitting the size argument, and does not have a default
574 implementation that defers to readinto().
575
576 In addition, read(), readinto() and write() may raise
577 BlockingIOError if the underlying raw stream is in non-blocking
578 mode and not ready; unlike their raw counterparts, they will never
579 return None.
580
581 A typical implementation should not inherit from a RawIOBase
582 implementation, but wrap one.
583 """
584
585 def read(self, n: int = None) -> bytes:
586 """Read and return up to n bytes.
587
588 If the argument is omitted, None, or negative, reads and
589 returns all data until EOF.
590
591 If the argument is positive, and the underlying raw stream is
592 not 'interactive', multiple raw reads may be issued to satisfy
593 the byte count (unless EOF is reached first). But for
594 interactive raw streams (XXX and for pipes?), at most one raw
595 read will be issued, and a short result does not imply that
596 EOF is imminent.
597
598 Returns an empty bytes array on EOF.
599
600 Raises BlockingIOError if the underlying raw stream has no
601 data at the moment.
602 """
603 self._unsupported("read")
604
605 def read1(self, n: int=None) -> bytes:
606 """Read up to n bytes with at most one read() system call."""
607 self._unsupported("read1")
608
609 def readinto(self, b: bytearray) -> int:
610 """Read up to len(b) bytes into b.
611
612 Like read(), this may issue multiple reads to the underlying raw
613 stream, unless the latter is 'interactive'.
614
615 Returns the number of bytes read (0 for EOF).
616
617 Raises BlockingIOError if the underlying raw stream has no
618 data at the moment.
619 """
620 # XXX This ought to work with anything that supports the buffer API
621 data = self.read(len(b))
622 n = len(data)
623 try:
624 b[:n] = data
625 except TypeError as err:
626 import array
627 if not isinstance(b, array.array):
628 raise err
629 b[:n] = array.array('b', data)
630 return n
631
632 def write(self, b: bytes) -> int:
633 """Write the given buffer to the IO stream.
634
635 Return the number of bytes written, which is never less than
636 len(b).
637
638 Raises BlockingIOError if the buffer is full and the
639 underlying raw stream cannot accept more data at the moment.
640 """
641 self._unsupported("write")
642
643io.BufferedIOBase.register(BufferedIOBase)
644
645
646class _BufferedIOMixin(BufferedIOBase):
647
648 """A mixin implementation of BufferedIOBase with an underlying raw stream.
649
650 This passes most requests on to the underlying raw stream. It
651 does *not* provide implementations of read(), readinto() or
652 write().
653 """
654
655 def __init__(self, raw):
656 self.raw = raw
657
658 ### Positioning ###
659
660 def seek(self, pos, whence=0):
661 new_position = self.raw.seek(pos, whence)
662 if new_position < 0:
663 raise IOError("seek() returned an invalid position")
664 return new_position
665
666 def tell(self):
667 pos = self.raw.tell()
668 if pos < 0:
669 raise IOError("tell() returned an invalid position")
670 return pos
671
672 def truncate(self, pos=None):
673 # Flush the stream. We're mixing buffered I/O with lower-level I/O,
674 # and a flush may be necessary to synch both views of the current
675 # file state.
676 self.flush()
677
678 if pos is None:
679 pos = self.tell()
680 # XXX: Should seek() be used, instead of passing the position
681 # XXX directly to truncate?
682 return self.raw.truncate(pos)
683
684 ### Flush and close ###
685
686 def flush(self):
687 self.raw.flush()
688
689 def close(self):
690 if not self.closed:
691 try:
692 self.flush()
693 except IOError:
694 pass # If flush() fails, just give up
695 self.raw.close()
696
697 ### Inquiries ###
698
699 def seekable(self):
700 return self.raw.seekable()
701
702 def readable(self):
703 return self.raw.readable()
704
705 def writable(self):
706 return self.raw.writable()
707
708 @property
709 def closed(self):
710 return self.raw.closed
711
712 @property
713 def name(self):
714 return self.raw.name
715
716 @property
717 def mode(self):
718 return self.raw.mode
719
720 ### Lower-level APIs ###
721
722 def fileno(self):
723 return self.raw.fileno()
724
725 def isatty(self):
726 return self.raw.isatty()
727
728
729class BytesIO(BufferedIOBase):
730
731 """Buffered I/O implementation using an in-memory bytes buffer."""
732
733 def __init__(self, initial_bytes=None):
734 buf = bytearray()
735 if initial_bytes is not None:
736 buf += initial_bytes
737 self._buffer = buf
738 self._pos = 0
739
740 def getvalue(self):
741 """Return the bytes value (contents) of the buffer
742 """
743 if self.closed:
744 raise ValueError("getvalue on closed file")
745 return bytes(self._buffer)
746
747 def read(self, n=None):
748 if self.closed:
749 raise ValueError("read from closed file")
750 if n is None:
751 n = -1
752 if n < 0:
753 n = len(self._buffer)
754 if len(self._buffer) <= self._pos:
755 return b""
756 newpos = min(len(self._buffer), self._pos + n)
757 b = self._buffer[self._pos : newpos]
758 self._pos = newpos
759 return bytes(b)
760
761 def read1(self, n):
762 """This is the same as read.
763 """
764 return self.read(n)
765
766 def write(self, b):
767 if self.closed:
768 raise ValueError("write to closed file")
769 if isinstance(b, str):
770 raise TypeError("can't write str to binary stream")
771 n = len(b)
772 if n == 0:
773 return 0
774 pos = self._pos
775 if pos > len(self._buffer):
776 # Inserts null bytes between the current end of the file
777 # and the new write position.
778 padding = b'\x00' * (pos - len(self._buffer))
779 self._buffer += padding
780 self._buffer[pos:pos + n] = b
781 self._pos += n
782 return n
783
784 def seek(self, pos, whence=0):
785 if self.closed:
786 raise ValueError("seek on closed file")
787 try:
788 pos = pos.__index__()
789 except AttributeError as err:
790 raise TypeError("an integer is required") from err
791 if whence == 0:
792 if pos < 0:
793 raise ValueError("negative seek position %r" % (pos,))
794 self._pos = pos
795 elif whence == 1:
796 self._pos = max(0, self._pos + pos)
797 elif whence == 2:
798 self._pos = max(0, len(self._buffer) + pos)
799 else:
800 raise ValueError("invalid whence value")
801 return self._pos
802
803 def tell(self):
804 if self.closed:
805 raise ValueError("tell on closed file")
806 return self._pos
807
808 def truncate(self, pos=None):
809 if self.closed:
810 raise ValueError("truncate on closed file")
811 if pos is None:
812 pos = self._pos
813 elif pos < 0:
814 raise ValueError("negative truncate position %r" % (pos,))
815 del self._buffer[pos:]
816 return self.seek(pos)
817
818 def readable(self):
819 return True
820
821 def writable(self):
822 return True
823
824 def seekable(self):
825 return True
826
827
828class BufferedReader(_BufferedIOMixin):
829
830 """BufferedReader(raw[, buffer_size])
831
832 A buffer for a readable, sequential BaseRawIO object.
833
834 The constructor creates a BufferedReader for the given readable raw
835 stream and buffer_size. If buffer_size is omitted, DEFAULT_BUFFER_SIZE
836 is used.
837 """
838
839 def __init__(self, raw, buffer_size=DEFAULT_BUFFER_SIZE):
840 """Create a new buffered reader using the given readable raw IO object.
841 """
842 raw._checkReadable()
843 _BufferedIOMixin.__init__(self, raw)
844 if buffer_size <= 0:
845 raise ValueError("invalid buffer size")
846 self.buffer_size = buffer_size
847 self._reset_read_buf()
848 self._read_lock = Lock()
849
850 def _reset_read_buf(self):
851 self._read_buf = b""
852 self._read_pos = 0
853
854 def read(self, n=None):
855 """Read n bytes.
856
857 Returns exactly n bytes of data unless the underlying raw IO
858 stream reaches EOF or if the call would block in non-blocking
859 mode. If n is negative, read until EOF or until read() would
860 block.
861 """
862 if n is not None and n < -1:
863 raise ValueError("invalid number of bytes to read")
864 with self._read_lock:
865 return self._read_unlocked(n)
866
867 def _read_unlocked(self, n=None):
868 nodata_val = b""
869 empty_values = (b"", None)
870 buf = self._read_buf
871 pos = self._read_pos
872
873 # Special case for when the number of bytes to read is unspecified.
874 if n is None or n == -1:
875 self._reset_read_buf()
876 chunks = [buf[pos:]] # Strip the consumed bytes.
877 current_size = 0
878 while True:
879 # Read until EOF or until read() would block.
880 chunk = self.raw.read()
881 if chunk in empty_values:
882 nodata_val = chunk
883 break
884 current_size += len(chunk)
885 chunks.append(chunk)
886 return b"".join(chunks) or nodata_val
887
888 # The number of bytes to read is specified, return at most n bytes.
889 avail = len(buf) - pos # Length of the available buffered data.
890 if n <= avail:
891 # Fast path: the data to read is fully buffered.
892 self._read_pos += n
893 return buf[pos:pos+n]
894 # Slow path: read from the stream until enough bytes are read,
895 # or until an EOF occurs or until read() would block.
896 chunks = [buf[pos:]]
897 wanted = max(self.buffer_size, n)
898 while avail < n:
899 chunk = self.raw.read(wanted)
900 if chunk in empty_values:
901 nodata_val = chunk
902 break
903 avail += len(chunk)
904 chunks.append(chunk)
905 # n is more then avail only when an EOF occurred or when
906 # read() would have blocked.
907 n = min(n, avail)
908 out = b"".join(chunks)
909 self._read_buf = out[n:] # Save the extra data in the buffer.
910 self._read_pos = 0
911 return out[:n] if out else nodata_val
912
913 def peek(self, n=0):
914 """Returns buffered bytes without advancing the position.
915
916 The argument indicates a desired minimal number of bytes; we
917 do at most one raw read to satisfy it. We never return more
918 than self.buffer_size.
919 """
920 with self._read_lock:
921 return self._peek_unlocked(n)
922
923 def _peek_unlocked(self, n=0):
924 want = min(n, self.buffer_size)
925 have = len(self._read_buf) - self._read_pos
926 if have < want or have <= 0:
927 to_read = self.buffer_size - have
928 current = self.raw.read(to_read)
929 if current:
930 self._read_buf = self._read_buf[self._read_pos:] + current
931 self._read_pos = 0
932 return self._read_buf[self._read_pos:]
933
934 def read1(self, n):
935 """Reads up to n bytes, with at most one read() system call."""
936 # Returns up to n bytes. If at least one byte is buffered, we
937 # only return buffered bytes. Otherwise, we do one raw read.
938 if n < 0:
939 raise ValueError("number of bytes to read must be positive")
940 if n == 0:
941 return b""
942 with self._read_lock:
943 self._peek_unlocked(1)
944 return self._read_unlocked(
945 min(n, len(self._read_buf) - self._read_pos))
946
947 def tell(self):
948 return _BufferedIOMixin.tell(self) - len(self._read_buf) + self._read_pos
949
950 def seek(self, pos, whence=0):
951 if not (0 <= whence <= 2):
952 raise ValueError("invalid whence value")
953 with self._read_lock:
954 if whence == 1:
955 pos -= len(self._read_buf) - self._read_pos
956 pos = _BufferedIOMixin.seek(self, pos, whence)
957 self._reset_read_buf()
958 return pos
959
960class BufferedWriter(_BufferedIOMixin):
961
962 """A buffer for a writeable sequential RawIO object.
963
964 The constructor creates a BufferedWriter for the given writeable raw
965 stream. If the buffer_size is not given, it defaults to
Benjamin Peterson59406a92009-03-26 17:10:29 +0000966 DEFAULT_BUFFER_SIZE.
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000967 """
968
Benjamin Peterson59406a92009-03-26 17:10:29 +0000969 _warning_stack_offset = 2
970
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000971 def __init__(self, raw,
972 buffer_size=DEFAULT_BUFFER_SIZE, max_buffer_size=None):
973 raw._checkWritable()
974 _BufferedIOMixin.__init__(self, raw)
975 if buffer_size <= 0:
976 raise ValueError("invalid buffer size")
Benjamin Peterson59406a92009-03-26 17:10:29 +0000977 if max_buffer_size is not None:
978 warnings.warn("max_buffer_size is deprecated", DeprecationWarning,
979 self._warning_stack_offset)
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000980 self.buffer_size = buffer_size
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000981 self._write_buf = bytearray()
982 self._write_lock = Lock()
983
984 def write(self, b):
985 if self.closed:
986 raise ValueError("write to closed file")
987 if isinstance(b, str):
988 raise TypeError("can't write str to binary stream")
989 with self._write_lock:
990 # XXX we can implement some more tricks to try and avoid
991 # partial writes
992 if len(self._write_buf) > self.buffer_size:
993 # We're full, so let's pre-flush the buffer
994 try:
995 self._flush_unlocked()
996 except BlockingIOError as e:
997 # We can't accept anything else.
998 # XXX Why not just let the exception pass through?
999 raise BlockingIOError(e.errno, e.strerror, 0)
1000 before = len(self._write_buf)
1001 self._write_buf.extend(b)
1002 written = len(self._write_buf) - before
1003 if len(self._write_buf) > self.buffer_size:
1004 try:
1005 self._flush_unlocked()
1006 except BlockingIOError as e:
Benjamin Peterson394ee002009-03-05 22:33:59 +00001007 if len(self._write_buf) > self.buffer_size:
1008 # We've hit the buffer_size. We have to accept a partial
1009 # write and cut back our buffer.
1010 overage = len(self._write_buf) - self.buffer_size
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001011 written -= overage
Benjamin Peterson394ee002009-03-05 22:33:59 +00001012 self._write_buf = self._write_buf[:self.buffer_size]
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001013 raise BlockingIOError(e.errno, e.strerror, written)
1014 return written
1015
1016 def truncate(self, pos=None):
1017 with self._write_lock:
1018 self._flush_unlocked()
1019 if pos is None:
1020 pos = self.raw.tell()
1021 return self.raw.truncate(pos)
1022
1023 def flush(self):
1024 with self._write_lock:
1025 self._flush_unlocked()
1026
1027 def _flush_unlocked(self):
1028 if self.closed:
1029 raise ValueError("flush of closed file")
1030 written = 0
1031 try:
1032 while self._write_buf:
1033 n = self.raw.write(self._write_buf)
1034 if n > len(self._write_buf) or n < 0:
1035 raise IOError("write() returned incorrect number of bytes")
1036 del self._write_buf[:n]
1037 written += n
1038 except BlockingIOError as e:
1039 n = e.characters_written
1040 del self._write_buf[:n]
1041 written += n
1042 raise BlockingIOError(e.errno, e.strerror, written)
1043
1044 def tell(self):
1045 return _BufferedIOMixin.tell(self) + len(self._write_buf)
1046
1047 def seek(self, pos, whence=0):
1048 if not (0 <= whence <= 2):
1049 raise ValueError("invalid whence")
1050 with self._write_lock:
1051 self._flush_unlocked()
1052 return _BufferedIOMixin.seek(self, pos, whence)
1053
1054
1055class BufferedRWPair(BufferedIOBase):
1056
1057 """A buffered reader and writer object together.
1058
1059 A buffered reader object and buffered writer object put together to
1060 form a sequential IO object that can read and write. This is typically
1061 used with a socket or two-way pipe.
1062
1063 reader and writer are RawIOBase objects that are readable and
1064 writeable respectively. If the buffer_size is omitted it defaults to
Benjamin Peterson59406a92009-03-26 17:10:29 +00001065 DEFAULT_BUFFER_SIZE.
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001066 """
1067
1068 # XXX The usefulness of this (compared to having two separate IO
1069 # objects) is questionable.
1070
1071 def __init__(self, reader, writer,
1072 buffer_size=DEFAULT_BUFFER_SIZE, max_buffer_size=None):
1073 """Constructor.
1074
1075 The arguments are two RawIO instances.
1076 """
Benjamin Peterson59406a92009-03-26 17:10:29 +00001077 if max_buffer_size is not None:
1078 warnings.warn("max_buffer_size is deprecated", DeprecationWarning, 2)
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001079 reader._checkReadable()
1080 writer._checkWritable()
1081 self.reader = BufferedReader(reader, buffer_size)
Benjamin Peterson59406a92009-03-26 17:10:29 +00001082 self.writer = BufferedWriter(writer, buffer_size)
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001083
1084 def read(self, n=None):
1085 if n is None:
1086 n = -1
1087 return self.reader.read(n)
1088
1089 def readinto(self, b):
1090 return self.reader.readinto(b)
1091
1092 def write(self, b):
1093 return self.writer.write(b)
1094
1095 def peek(self, n=0):
1096 return self.reader.peek(n)
1097
1098 def read1(self, n):
1099 return self.reader.read1(n)
1100
1101 def readable(self):
1102 return self.reader.readable()
1103
1104 def writable(self):
1105 return self.writer.writable()
1106
1107 def flush(self):
1108 return self.writer.flush()
1109
1110 def close(self):
1111 self.writer.close()
1112 self.reader.close()
1113
1114 def isatty(self):
1115 return self.reader.isatty() or self.writer.isatty()
1116
1117 @property
1118 def closed(self):
1119 return self.writer.closed
1120
1121
1122class BufferedRandom(BufferedWriter, BufferedReader):
1123
1124 """A buffered interface to random access streams.
1125
1126 The constructor creates a reader and writer for a seekable stream,
1127 raw, given in the first argument. If the buffer_size is omitted it
Benjamin Peterson59406a92009-03-26 17:10:29 +00001128 defaults to DEFAULT_BUFFER_SIZE.
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001129 """
1130
Benjamin Peterson59406a92009-03-26 17:10:29 +00001131 _warning_stack_offset = 3
1132
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001133 def __init__(self, raw,
1134 buffer_size=DEFAULT_BUFFER_SIZE, max_buffer_size=None):
1135 raw._checkSeekable()
1136 BufferedReader.__init__(self, raw, buffer_size)
1137 BufferedWriter.__init__(self, raw, buffer_size, max_buffer_size)
1138
1139 def seek(self, pos, whence=0):
1140 if not (0 <= whence <= 2):
1141 raise ValueError("invalid whence")
1142 self.flush()
1143 if self._read_buf:
1144 # Undo read ahead.
1145 with self._read_lock:
1146 self.raw.seek(self._read_pos - len(self._read_buf), 1)
1147 # First do the raw seek, then empty the read buffer, so that
1148 # if the raw seek fails, we don't lose buffered data forever.
1149 pos = self.raw.seek(pos, whence)
1150 with self._read_lock:
1151 self._reset_read_buf()
1152 if pos < 0:
1153 raise IOError("seek() returned invalid position")
1154 return pos
1155
1156 def tell(self):
1157 if self._write_buf:
1158 return BufferedWriter.tell(self)
1159 else:
1160 return BufferedReader.tell(self)
1161
1162 def truncate(self, pos=None):
1163 if pos is None:
1164 pos = self.tell()
1165 # Use seek to flush the read buffer.
1166 self.seek(pos)
1167 return BufferedWriter.truncate(self)
1168
1169 def read(self, n=None):
1170 if n is None:
1171 n = -1
1172 self.flush()
1173 return BufferedReader.read(self, n)
1174
1175 def readinto(self, b):
1176 self.flush()
1177 return BufferedReader.readinto(self, b)
1178
1179 def peek(self, n=0):
1180 self.flush()
1181 return BufferedReader.peek(self, n)
1182
1183 def read1(self, n):
1184 self.flush()
1185 return BufferedReader.read1(self, n)
1186
1187 def write(self, b):
1188 if self._read_buf:
1189 # Undo readahead
1190 with self._read_lock:
1191 self.raw.seek(self._read_pos - len(self._read_buf), 1)
1192 self._reset_read_buf()
1193 return BufferedWriter.write(self, b)
1194
1195
1196class TextIOBase(IOBase):
1197
1198 """Base class for text I/O.
1199
1200 This class provides a character and line based interface to stream
1201 I/O. There is no readinto method because Python's character strings
1202 are immutable. There is no public constructor.
1203 """
1204
1205 def read(self, n: int = -1) -> str:
1206 """Read at most n characters from stream.
1207
1208 Read from underlying buffer until we have n characters or we hit EOF.
1209 If n is negative or omitted, read until EOF.
1210 """
1211 self._unsupported("read")
1212
1213 def write(self, s: str) -> int:
1214 """Write string s to stream."""
1215 self._unsupported("write")
1216
1217 def truncate(self, pos: int = None) -> int:
1218 """Truncate size to pos."""
1219 self._unsupported("truncate")
1220
1221 def readline(self) -> str:
1222 """Read until newline or EOF.
1223
1224 Returns an empty string if EOF is hit immediately.
1225 """
1226 self._unsupported("readline")
1227
1228 @property
1229 def encoding(self):
1230 """Subclasses should override."""
1231 return None
1232
1233 @property
1234 def newlines(self):
1235 """Line endings translated so far.
1236
1237 Only line endings translated during reading are considered.
1238
1239 Subclasses should override.
1240 """
1241 return None
1242
1243io.TextIOBase.register(TextIOBase)
1244
1245
1246class IncrementalNewlineDecoder(codecs.IncrementalDecoder):
1247 r"""Codec used when reading a file in universal newlines mode. It wraps
1248 another incremental decoder, translating \r\n and \r into \n. It also
1249 records the types of newlines encountered. When used with
1250 translate=False, it ensures that the newline sequence is returned in
1251 one piece.
1252 """
1253 def __init__(self, decoder, translate, errors='strict'):
1254 codecs.IncrementalDecoder.__init__(self, errors=errors)
1255 self.translate = translate
1256 self.decoder = decoder
1257 self.seennl = 0
1258 self.pendingcr = False
1259
1260 def decode(self, input, final=False):
1261 # decode input (with the eventual \r from a previous pass)
1262 if self.decoder is None:
1263 output = input
1264 else:
1265 output = self.decoder.decode(input, final=final)
1266 if self.pendingcr and (output or final):
1267 output = "\r" + output
1268 self.pendingcr = False
1269
1270 # retain last \r even when not translating data:
1271 # then readline() is sure to get \r\n in one pass
1272 if output.endswith("\r") and not final:
1273 output = output[:-1]
1274 self.pendingcr = True
1275
1276 # Record which newlines are read
1277 crlf = output.count('\r\n')
1278 cr = output.count('\r') - crlf
1279 lf = output.count('\n') - crlf
1280 self.seennl |= (lf and self._LF) | (cr and self._CR) \
1281 | (crlf and self._CRLF)
1282
1283 if self.translate:
1284 if crlf:
1285 output = output.replace("\r\n", "\n")
1286 if cr:
1287 output = output.replace("\r", "\n")
1288
1289 return output
1290
1291 def getstate(self):
1292 if self.decoder is None:
1293 buf = b""
1294 flag = 0
1295 else:
1296 buf, flag = self.decoder.getstate()
1297 flag <<= 1
1298 if self.pendingcr:
1299 flag |= 1
1300 return buf, flag
1301
1302 def setstate(self, state):
1303 buf, flag = state
1304 self.pendingcr = bool(flag & 1)
1305 if self.decoder is not None:
1306 self.decoder.setstate((buf, flag >> 1))
1307
1308 def reset(self):
1309 self.seennl = 0
1310 self.pendingcr = False
1311 if self.decoder is not None:
1312 self.decoder.reset()
1313
1314 _LF = 1
1315 _CR = 2
1316 _CRLF = 4
1317
1318 @property
1319 def newlines(self):
1320 return (None,
1321 "\n",
1322 "\r",
1323 ("\r", "\n"),
1324 "\r\n",
1325 ("\n", "\r\n"),
1326 ("\r", "\r\n"),
1327 ("\r", "\n", "\r\n")
1328 )[self.seennl]
1329
1330
1331class TextIOWrapper(TextIOBase):
1332
1333 r"""Character and line based layer over a BufferedIOBase object, buffer.
1334
1335 encoding gives the name of the encoding that the stream will be
1336 decoded or encoded with. It defaults to locale.getpreferredencoding.
1337
1338 errors determines the strictness of encoding and decoding (see the
1339 codecs.register) and defaults to "strict".
1340
1341 newline can be None, '', '\n', '\r', or '\r\n'. It controls the
1342 handling of line endings. If it is None, universal newlines is
1343 enabled. With this enabled, on input, the lines endings '\n', '\r',
1344 or '\r\n' are translated to '\n' before being returned to the
1345 caller. Conversely, on output, '\n' is translated to the system
1346 default line seperator, os.linesep. If newline is any other of its
1347 legal values, that newline becomes the newline when the file is read
1348 and it is returned untranslated. On output, '\n' is converted to the
1349 newline.
1350
1351 If line_buffering is True, a call to flush is implied when a call to
1352 write contains a newline character.
1353 """
1354
1355 _CHUNK_SIZE = 2048
1356
1357 def __init__(self, buffer, encoding=None, errors=None, newline=None,
1358 line_buffering=False):
1359 if newline is not None and not isinstance(newline, str):
1360 raise TypeError("illegal newline type: %r" % (type(newline),))
1361 if newline not in (None, "", "\n", "\r", "\r\n"):
1362 raise ValueError("illegal newline value: %r" % (newline,))
1363 if encoding is None:
1364 try:
1365 encoding = os.device_encoding(buffer.fileno())
1366 except (AttributeError, UnsupportedOperation):
1367 pass
1368 if encoding is None:
1369 try:
1370 import locale
1371 except ImportError:
1372 # Importing locale may fail if Python is being built
1373 encoding = "ascii"
1374 else:
1375 encoding = locale.getpreferredencoding()
1376
1377 if not isinstance(encoding, str):
1378 raise ValueError("invalid encoding: %r" % encoding)
1379
1380 if errors is None:
1381 errors = "strict"
1382 else:
1383 if not isinstance(errors, str):
1384 raise ValueError("invalid errors: %r" % errors)
1385
1386 self.buffer = buffer
1387 self._line_buffering = line_buffering
1388 self._encoding = encoding
1389 self._errors = errors
1390 self._readuniversal = not newline
1391 self._readtranslate = newline is None
1392 self._readnl = newline
1393 self._writetranslate = newline != ''
1394 self._writenl = newline or os.linesep
1395 self._encoder = None
1396 self._decoder = None
1397 self._decoded_chars = '' # buffer for text returned from decoder
1398 self._decoded_chars_used = 0 # offset into _decoded_chars for read()
1399 self._snapshot = None # info for reconstructing decoder state
1400 self._seekable = self._telling = self.buffer.seekable()
1401
1402 # self._snapshot is either None, or a tuple (dec_flags, next_input)
1403 # where dec_flags is the second (integer) item of the decoder state
1404 # and next_input is the chunk of input bytes that comes next after the
1405 # snapshot point. We use this to reconstruct decoder states in tell().
1406
1407 # Naming convention:
1408 # - "bytes_..." for integer variables that count input bytes
1409 # - "chars_..." for integer variables that count decoded characters
1410
Benjamin Petersonc4c0eae2009-03-09 00:07:03 +00001411 def __repr__(self):
1412 return "<TextIOWrapper encoding={0}>".format(self.encoding)
1413
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001414 @property
1415 def encoding(self):
1416 return self._encoding
1417
1418 @property
1419 def errors(self):
1420 return self._errors
1421
1422 @property
1423 def line_buffering(self):
1424 return self._line_buffering
1425
1426 def seekable(self):
1427 return self._seekable
1428
1429 def readable(self):
1430 return self.buffer.readable()
1431
1432 def writable(self):
1433 return self.buffer.writable()
1434
1435 def flush(self):
1436 self.buffer.flush()
1437 self._telling = self._seekable
1438
1439 def close(self):
1440 try:
1441 self.flush()
Benjamin Peterson54f963e2009-04-14 22:02:08 +00001442 except IOError:
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001443 pass # If flush() fails, just give up
1444 self.buffer.close()
1445
1446 @property
1447 def closed(self):
1448 return self.buffer.closed
1449
1450 @property
1451 def name(self):
1452 return self.buffer.name
1453
1454 def fileno(self):
1455 return self.buffer.fileno()
1456
1457 def isatty(self):
1458 return self.buffer.isatty()
1459
1460 def write(self, s: str):
1461 if self.closed:
1462 raise ValueError("write to closed file")
1463 if not isinstance(s, str):
1464 raise TypeError("can't write %s to text stream" %
1465 s.__class__.__name__)
1466 length = len(s)
1467 haslf = (self._writetranslate or self._line_buffering) and "\n" in s
1468 if haslf and self._writetranslate and self._writenl != "\n":
1469 s = s.replace("\n", self._writenl)
1470 encoder = self._encoder or self._get_encoder()
1471 # XXX What if we were just reading?
1472 b = encoder.encode(s)
1473 self.buffer.write(b)
1474 if self._line_buffering and (haslf or "\r" in s):
1475 self.flush()
1476 self._snapshot = None
1477 if self._decoder:
1478 self._decoder.reset()
1479 return length
1480
1481 def _get_encoder(self):
1482 make_encoder = codecs.getincrementalencoder(self._encoding)
1483 self._encoder = make_encoder(self._errors)
1484 return self._encoder
1485
1486 def _get_decoder(self):
1487 make_decoder = codecs.getincrementaldecoder(self._encoding)
1488 decoder = make_decoder(self._errors)
1489 if self._readuniversal:
1490 decoder = IncrementalNewlineDecoder(decoder, self._readtranslate)
1491 self._decoder = decoder
1492 return decoder
1493
1494 # The following three methods implement an ADT for _decoded_chars.
1495 # Text returned from the decoder is buffered here until the client
1496 # requests it by calling our read() or readline() method.
1497 def _set_decoded_chars(self, chars):
1498 """Set the _decoded_chars buffer."""
1499 self._decoded_chars = chars
1500 self._decoded_chars_used = 0
1501
1502 def _get_decoded_chars(self, n=None):
1503 """Advance into the _decoded_chars buffer."""
1504 offset = self._decoded_chars_used
1505 if n is None:
1506 chars = self._decoded_chars[offset:]
1507 else:
1508 chars = self._decoded_chars[offset:offset + n]
1509 self._decoded_chars_used += len(chars)
1510 return chars
1511
1512 def _rewind_decoded_chars(self, n):
1513 """Rewind the _decoded_chars buffer."""
1514 if self._decoded_chars_used < n:
1515 raise AssertionError("rewind decoded_chars out of bounds")
1516 self._decoded_chars_used -= n
1517
1518 def _read_chunk(self):
1519 """
1520 Read and decode the next chunk of data from the BufferedReader.
1521 """
1522
1523 # The return value is True unless EOF was reached. The decoded
1524 # string is placed in self._decoded_chars (replacing its previous
1525 # value). The entire input chunk is sent to the decoder, though
1526 # some of it may remain buffered in the decoder, yet to be
1527 # converted.
1528
1529 if self._decoder is None:
1530 raise ValueError("no decoder")
1531
1532 if self._telling:
1533 # To prepare for tell(), we need to snapshot a point in the
1534 # file where the decoder's input buffer is empty.
1535
1536 dec_buffer, dec_flags = self._decoder.getstate()
1537 # Given this, we know there was a valid snapshot point
1538 # len(dec_buffer) bytes ago with decoder state (b'', dec_flags).
1539
1540 # Read a chunk, decode it, and put the result in self._decoded_chars.
1541 input_chunk = self.buffer.read1(self._CHUNK_SIZE)
1542 eof = not input_chunk
1543 self._set_decoded_chars(self._decoder.decode(input_chunk, eof))
1544
1545 if self._telling:
1546 # At the snapshot point, len(dec_buffer) bytes before the read,
1547 # the next input to be decoded is dec_buffer + input_chunk.
1548 self._snapshot = (dec_flags, dec_buffer + input_chunk)
1549
1550 return not eof
1551
1552 def _pack_cookie(self, position, dec_flags=0,
1553 bytes_to_feed=0, need_eof=0, chars_to_skip=0):
1554 # The meaning of a tell() cookie is: seek to position, set the
1555 # decoder flags to dec_flags, read bytes_to_feed bytes, feed them
1556 # into the decoder with need_eof as the EOF flag, then skip
1557 # chars_to_skip characters of the decoded result. For most simple
1558 # decoders, tell() will often just give a byte offset in the file.
1559 return (position | (dec_flags<<64) | (bytes_to_feed<<128) |
1560 (chars_to_skip<<192) | bool(need_eof)<<256)
1561
1562 def _unpack_cookie(self, bigint):
1563 rest, position = divmod(bigint, 1<<64)
1564 rest, dec_flags = divmod(rest, 1<<64)
1565 rest, bytes_to_feed = divmod(rest, 1<<64)
1566 need_eof, chars_to_skip = divmod(rest, 1<<64)
1567 return position, dec_flags, bytes_to_feed, need_eof, chars_to_skip
1568
1569 def tell(self):
1570 if not self._seekable:
1571 raise IOError("underlying stream is not seekable")
1572 if not self._telling:
1573 raise IOError("telling position disabled by next() call")
1574 self.flush()
1575 position = self.buffer.tell()
1576 decoder = self._decoder
1577 if decoder is None or self._snapshot is None:
1578 if self._decoded_chars:
1579 # This should never happen.
1580 raise AssertionError("pending decoded text")
1581 return position
1582
1583 # Skip backward to the snapshot point (see _read_chunk).
1584 dec_flags, next_input = self._snapshot
1585 position -= len(next_input)
1586
1587 # How many decoded characters have been used up since the snapshot?
1588 chars_to_skip = self._decoded_chars_used
1589 if chars_to_skip == 0:
1590 # We haven't moved from the snapshot point.
1591 return self._pack_cookie(position, dec_flags)
1592
1593 # Starting from the snapshot position, we will walk the decoder
1594 # forward until it gives us enough decoded characters.
1595 saved_state = decoder.getstate()
1596 try:
1597 # Note our initial start point.
1598 decoder.setstate((b'', dec_flags))
1599 start_pos = position
1600 start_flags, bytes_fed, chars_decoded = dec_flags, 0, 0
1601 need_eof = 0
1602
1603 # Feed the decoder one byte at a time. As we go, note the
1604 # nearest "safe start point" before the current location
1605 # (a point where the decoder has nothing buffered, so seek()
1606 # can safely start from there and advance to this location).
1607 next_byte = bytearray(1)
1608 for next_byte[0] in next_input:
1609 bytes_fed += 1
1610 chars_decoded += len(decoder.decode(next_byte))
1611 dec_buffer, dec_flags = decoder.getstate()
1612 if not dec_buffer and chars_decoded <= chars_to_skip:
1613 # Decoder buffer is empty, so this is a safe start point.
1614 start_pos += bytes_fed
1615 chars_to_skip -= chars_decoded
1616 start_flags, bytes_fed, chars_decoded = dec_flags, 0, 0
1617 if chars_decoded >= chars_to_skip:
1618 break
1619 else:
1620 # We didn't get enough decoded data; signal EOF to get more.
1621 chars_decoded += len(decoder.decode(b'', final=True))
1622 need_eof = 1
1623 if chars_decoded < chars_to_skip:
1624 raise IOError("can't reconstruct logical file position")
1625
1626 # The returned cookie corresponds to the last safe start point.
1627 return self._pack_cookie(
1628 start_pos, start_flags, bytes_fed, need_eof, chars_to_skip)
1629 finally:
1630 decoder.setstate(saved_state)
1631
1632 def truncate(self, pos=None):
1633 self.flush()
1634 if pos is None:
1635 pos = self.tell()
1636 self.seek(pos)
1637 return self.buffer.truncate()
1638
1639 def seek(self, cookie, whence=0):
1640 if self.closed:
1641 raise ValueError("tell on closed file")
1642 if not self._seekable:
1643 raise IOError("underlying stream is not seekable")
1644 if whence == 1: # seek relative to current position
1645 if cookie != 0:
1646 raise IOError("can't do nonzero cur-relative seeks")
1647 # Seeking to the current position should attempt to
1648 # sync the underlying buffer with the current position.
1649 whence = 0
1650 cookie = self.tell()
1651 if whence == 2: # seek relative to end of file
1652 if cookie != 0:
1653 raise IOError("can't do nonzero end-relative seeks")
1654 self.flush()
1655 position = self.buffer.seek(0, 2)
1656 self._set_decoded_chars('')
1657 self._snapshot = None
1658 if self._decoder:
1659 self._decoder.reset()
1660 return position
1661 if whence != 0:
1662 raise ValueError("invalid whence (%r, should be 0, 1 or 2)" %
1663 (whence,))
1664 if cookie < 0:
1665 raise ValueError("negative seek position %r" % (cookie,))
1666 self.flush()
1667
1668 # The strategy of seek() is to go back to the safe start point
1669 # and replay the effect of read(chars_to_skip) from there.
1670 start_pos, dec_flags, bytes_to_feed, need_eof, chars_to_skip = \
1671 self._unpack_cookie(cookie)
1672
1673 # Seek back to the safe start point.
1674 self.buffer.seek(start_pos)
1675 self._set_decoded_chars('')
1676 self._snapshot = None
1677
1678 # Restore the decoder to its state from the safe start point.
Benjamin Peterson9363a652009-03-05 00:42:09 +00001679 if cookie == 0 and self._decoder:
1680 self._decoder.reset()
1681 elif self._decoder or dec_flags or chars_to_skip:
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001682 self._decoder = self._decoder or self._get_decoder()
1683 self._decoder.setstate((b'', dec_flags))
1684 self._snapshot = (dec_flags, b'')
1685
1686 if chars_to_skip:
1687 # Just like _read_chunk, feed the decoder and save a snapshot.
1688 input_chunk = self.buffer.read(bytes_to_feed)
1689 self._set_decoded_chars(
1690 self._decoder.decode(input_chunk, need_eof))
1691 self._snapshot = (dec_flags, input_chunk)
1692
1693 # Skip chars_to_skip of the decoded characters.
1694 if len(self._decoded_chars) < chars_to_skip:
1695 raise IOError("can't restore logical file position")
1696 self._decoded_chars_used = chars_to_skip
1697
1698 return cookie
1699
1700 def read(self, n=None):
Benjamin Petersona1b49012009-03-31 23:11:32 +00001701 self._checkReadable()
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001702 if n is None:
1703 n = -1
1704 decoder = self._decoder or self._get_decoder()
1705 if n < 0:
1706 # Read everything.
1707 result = (self._get_decoded_chars() +
1708 decoder.decode(self.buffer.read(), final=True))
1709 self._set_decoded_chars('')
1710 self._snapshot = None
1711 return result
1712 else:
1713 # Keep reading chunks until we have n characters to return.
1714 eof = False
1715 result = self._get_decoded_chars(n)
1716 while len(result) < n and not eof:
1717 eof = not self._read_chunk()
1718 result += self._get_decoded_chars(n - len(result))
1719 return result
1720
1721 def __next__(self):
1722 self._telling = False
1723 line = self.readline()
1724 if not line:
1725 self._snapshot = None
1726 self._telling = self._seekable
1727 raise StopIteration
1728 return line
1729
1730 def readline(self, limit=None):
1731 if self.closed:
1732 raise ValueError("read from closed file")
1733 if limit is None:
1734 limit = -1
1735
1736 # Grab all the decoded text (we will rewind any extra bits later).
1737 line = self._get_decoded_chars()
1738
1739 start = 0
1740 # Make the decoder if it doesn't already exist.
1741 if not self._decoder:
1742 self._get_decoder()
1743
1744 pos = endpos = None
1745 while True:
1746 if self._readtranslate:
1747 # Newlines are already translated, only search for \n
1748 pos = line.find('\n', start)
1749 if pos >= 0:
1750 endpos = pos + 1
1751 break
1752 else:
1753 start = len(line)
1754
1755 elif self._readuniversal:
1756 # Universal newline search. Find any of \r, \r\n, \n
1757 # The decoder ensures that \r\n are not split in two pieces
1758
1759 # In C we'd look for these in parallel of course.
1760 nlpos = line.find("\n", start)
1761 crpos = line.find("\r", start)
1762 if crpos == -1:
1763 if nlpos == -1:
1764 # Nothing found
1765 start = len(line)
1766 else:
1767 # Found \n
1768 endpos = nlpos + 1
1769 break
1770 elif nlpos == -1:
1771 # Found lone \r
1772 endpos = crpos + 1
1773 break
1774 elif nlpos < crpos:
1775 # Found \n
1776 endpos = nlpos + 1
1777 break
1778 elif nlpos == crpos + 1:
1779 # Found \r\n
1780 endpos = crpos + 2
1781 break
1782 else:
1783 # Found \r
1784 endpos = crpos + 1
1785 break
1786 else:
1787 # non-universal
1788 pos = line.find(self._readnl)
1789 if pos >= 0:
1790 endpos = pos + len(self._readnl)
1791 break
1792
1793 if limit >= 0 and len(line) >= limit:
1794 endpos = limit # reached length limit
1795 break
1796
1797 # No line ending seen yet - get more data'
1798 while self._read_chunk():
1799 if self._decoded_chars:
1800 break
1801 if self._decoded_chars:
1802 line += self._get_decoded_chars()
1803 else:
1804 # end of file
1805 self._set_decoded_chars('')
1806 self._snapshot = None
1807 return line
1808
1809 if limit >= 0 and endpos > limit:
1810 endpos = limit # don't exceed limit
1811
1812 # Rewind _decoded_chars to just after the line ending we found.
1813 self._rewind_decoded_chars(len(line) - endpos)
1814 return line[:endpos]
1815
1816 @property
1817 def newlines(self):
1818 return self._decoder.newlines if self._decoder else None
1819
1820
1821class StringIO(TextIOWrapper):
1822 """Text I/O implementation using an in-memory buffer.
1823
1824 The initial_value argument sets the value of object. The newline
1825 argument is like the one of TextIOWrapper's constructor.
1826 """
1827
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001828 def __init__(self, initial_value="", newline="\n"):
1829 super(StringIO, self).__init__(BytesIO(),
1830 encoding="utf-8",
1831 errors="strict",
1832 newline=newline)
Antoine Pitrou11446482009-04-04 14:09:30 +00001833 # Issue #5645: make universal newlines semantics the same as in the
1834 # C version, even under Windows.
1835 if newline is None:
1836 self._writetranslate = False
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001837 if initial_value:
1838 if not isinstance(initial_value, str):
1839 initial_value = str(initial_value)
1840 self.write(initial_value)
1841 self.seek(0)
1842
1843 def getvalue(self):
1844 self.flush()
1845 return self.buffer.getvalue().decode(self._encoding, self._errors)
Benjamin Peterson9fd459a2009-03-09 00:09:44 +00001846
1847 def __repr__(self):
1848 # TextIOWrapper tells the encoding in its repr. In StringIO,
1849 # that's a implementation detail.
1850 return object.__repr__(self)
Benjamin Petersonb487e632009-03-21 03:08:31 +00001851
1852 @property
1853 def encoding(self):
1854 return None