blob: 8960ce92557b93afa731906194f7f4c9c57b8396 [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
Antoine Pitroud5587bc2009-12-19 21:08:31 +000085 buffering is an optional integer used to set the buffering policy.
86 Pass 0 to switch buffering off (only allowed in binary mode), 1 to select
87 line buffering (only usable in text mode), and an integer > 1 to indicate
88 the size of a fixed-size chunk buffer. When no buffering argument is
89 given, the default buffering policy works as follows:
90
91 * Binary files are buffered in fixed-size chunks; the size of the buffer
92 is chosen using a heuristic trying to determine the underlying device's
93 "block size" and falling back on `io.DEFAULT_BUFFER_SIZE`.
94 On many systems, the buffer will typically be 4096 or 8192 bytes long.
95
96 * "Interactive" text files (files for which isatty() returns True)
97 use line buffering. Other text files use the policy described above
98 for binary files.
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +000099
100 encoding is the name of the encoding used to decode or encode the
101 file. This should only be used in text mode. The default encoding is
102 platform dependent, but any encoding supported by Python can be
103 passed. See the codecs module for the list of supported encodings.
104
105 errors is an optional string that specifies how encoding errors are to
106 be handled---this argument should not be used in binary mode. Pass
107 'strict' to raise a ValueError exception if there is an encoding error
108 (the default of None has the same effect), or pass 'ignore' to ignore
109 errors. (Note that ignoring encoding errors can lead to data loss.)
110 See the documentation for codecs.register for a list of the permitted
111 encoding error strings.
112
113 newline controls how universal newlines works (it only applies to text
114 mode). It can be None, '', '\n', '\r', and '\r\n'. It works as
115 follows:
116
117 * On input, if newline is None, universal newlines mode is
118 enabled. Lines in the input can end in '\n', '\r', or '\r\n', and
119 these are translated into '\n' before being returned to the
120 caller. If it is '', universal newline mode is enabled, but line
121 endings are returned to the caller untranslated. If it has any of
122 the other legal values, input lines are only terminated by the given
123 string, and the line ending is returned to the caller untranslated.
124
125 * On output, if newline is None, any '\n' characters written are
126 translated to the system default line separator, os.linesep. If
127 newline is '', no translation takes place. If newline is any of the
128 other legal values, any '\n' characters written are translated to
129 the given string.
130
131 If closefd is False, the underlying file descriptor will be kept open
132 when the file is closed. This does not work when a file name is given
133 and must be True in that case.
134
135 open() returns a file object whose type depends on the mode, and
136 through which the standard file operations such as reading and writing
137 are performed. When open() is used to open a file in a text mode ('w',
138 'r', 'wt', 'rt', etc.), it returns a TextIOWrapper. When used to open
139 a file in a binary mode, the returned class varies: in read binary
140 mode, it returns a BufferedReader; in write binary and append binary
141 modes, it returns a BufferedWriter, and in read/write mode, it returns
142 a BufferedRandom.
143
144 It is also possible to use a string or bytearray as a file for both
145 reading and writing. For strings StringIO can be used like a file
146 opened in a text mode, and for bytes a BytesIO can be used like a file
147 opened in a binary mode.
148 """
149 if not isinstance(file, (str, bytes, int)):
150 raise TypeError("invalid file: %r" % file)
151 if not isinstance(mode, str):
152 raise TypeError("invalid mode: %r" % mode)
153 if buffering is not None and not isinstance(buffering, int):
154 raise TypeError("invalid buffering: %r" % buffering)
155 if encoding is not None and not isinstance(encoding, str):
156 raise TypeError("invalid encoding: %r" % encoding)
157 if errors is not None and not isinstance(errors, str):
158 raise TypeError("invalid errors: %r" % errors)
159 modes = set(mode)
160 if modes - set("arwb+tU") or len(mode) > len(modes):
161 raise ValueError("invalid mode: %r" % mode)
162 reading = "r" in modes
163 writing = "w" in modes
164 appending = "a" in modes
165 updating = "+" in modes
166 text = "t" in modes
167 binary = "b" in modes
168 if "U" in modes:
169 if writing or appending:
170 raise ValueError("can't use U and writing mode at once")
171 reading = True
172 if text and binary:
173 raise ValueError("can't have text and binary mode at once")
174 if reading + writing + appending > 1:
175 raise ValueError("can't have read/write/append mode at once")
176 if not (reading or writing or appending):
177 raise ValueError("must have exactly one of read/write/append mode")
178 if binary and encoding is not None:
179 raise ValueError("binary mode doesn't take an encoding argument")
180 if binary and errors is not None:
181 raise ValueError("binary mode doesn't take an errors argument")
182 if binary and newline is not None:
183 raise ValueError("binary mode doesn't take a newline argument")
184 raw = FileIO(file,
185 (reading and "r" or "") +
186 (writing and "w" or "") +
187 (appending and "a" or "") +
188 (updating and "+" or ""),
189 closefd)
190 if buffering is None:
191 buffering = -1
192 line_buffering = False
193 if buffering == 1 or buffering < 0 and raw.isatty():
194 buffering = -1
195 line_buffering = True
196 if buffering < 0:
197 buffering = DEFAULT_BUFFER_SIZE
198 try:
199 bs = os.fstat(raw.fileno()).st_blksize
200 except (os.error, AttributeError):
201 pass
202 else:
203 if bs > 1:
204 buffering = bs
205 if buffering < 0:
206 raise ValueError("invalid buffering size")
207 if buffering == 0:
208 if binary:
209 return raw
210 raise ValueError("can't have unbuffered text I/O")
211 if updating:
212 buffer = BufferedRandom(raw, buffering)
213 elif writing or appending:
214 buffer = BufferedWriter(raw, buffering)
215 elif reading:
216 buffer = BufferedReader(raw, buffering)
217 else:
218 raise ValueError("unknown mode: %r" % mode)
219 if binary:
220 return buffer
221 text = TextIOWrapper(buffer, encoding, errors, newline, line_buffering)
222 text.mode = mode
223 return text
224
225
226class DocDescriptor:
227 """Helper for builtins.open.__doc__
228 """
229 def __get__(self, obj, typ):
230 return (
231 "open(file, mode='r', buffering=None, encoding=None, "
232 "errors=None, newline=None, closefd=True)\n\n" +
233 open.__doc__)
234
235class OpenWrapper:
236 """Wrapper for builtins.open
237
238 Trick so that open won't become a bound method when stored
239 as a class variable (as dbm.dumb does).
240
241 See initstdio() in Python/pythonrun.c.
242 """
243 __doc__ = DocDescriptor()
244
245 def __new__(cls, *args, **kwargs):
246 return open(*args, **kwargs)
247
248
249class UnsupportedOperation(ValueError, IOError):
250 pass
251
252
253class IOBase(metaclass=abc.ABCMeta):
254
255 """The abstract base class for all I/O classes, acting on streams of
256 bytes. There is no public constructor.
257
258 This class provides dummy implementations for many methods that
259 derived classes can override selectively; the default implementations
260 represent a file that cannot be read, written or seeked.
261
262 Even though IOBase does not declare read, readinto, or write because
263 their signatures will vary, implementations and clients should
264 consider those methods part of the interface. Also, implementations
265 may raise a IOError when operations they do not support are called.
266
267 The basic type used for binary data read from or written to a file is
268 bytes. bytearrays are accepted too, and in some cases (such as
269 readinto) needed. Text I/O classes work with str data.
270
271 Note that calling any method (even inquiries) on a closed stream is
272 undefined. Implementations may raise IOError in this case.
273
274 IOBase (and its subclasses) support the iterator protocol, meaning
275 that an IOBase object can be iterated over yielding the lines in a
276 stream.
277
278 IOBase also supports the :keyword:`with` statement. In this example,
279 fp is closed after the suite of the with statement is complete:
280
281 with open('spam.txt', 'r') as fp:
282 fp.write('Spam and eggs!')
283 """
284
285 ### Internal ###
286
287 def _unsupported(self, name: str) -> IOError:
288 """Internal: raise an exception for unsupported operations."""
289 raise UnsupportedOperation("%s.%s() not supported" %
290 (self.__class__.__name__, name))
291
292 ### Positioning ###
293
294 def seek(self, pos: int, whence: int = 0) -> int:
295 """Change stream position.
296
297 Change the stream position to byte offset offset. offset is
298 interpreted relative to the position indicated by whence. Values
299 for whence are:
300
301 * 0 -- start of stream (the default); offset should be zero or positive
302 * 1 -- current stream position; offset may be negative
303 * 2 -- end of stream; offset is usually negative
304
305 Return the new absolute position.
306 """
307 self._unsupported("seek")
308
309 def tell(self) -> int:
310 """Return current stream position."""
311 return self.seek(0, 1)
312
313 def truncate(self, pos: int = None) -> int:
314 """Truncate file to size bytes.
315
316 Size defaults to the current IO position as reported by tell(). Return
317 the new size.
318 """
319 self._unsupported("truncate")
320
321 ### Flush and close ###
322
323 def flush(self) -> None:
324 """Flush write buffers, if applicable.
325
326 This is not implemented for read-only and non-blocking streams.
327 """
328 # XXX Should this return the number of bytes written???
329
330 __closed = False
331
332 def close(self) -> None:
333 """Flush and close the IO object.
334
335 This method has no effect if the file is already closed.
336 """
337 if not self.__closed:
338 try:
339 self.flush()
340 except IOError:
341 pass # If flush() fails, just give up
342 self.__closed = True
343
344 def __del__(self) -> None:
345 """Destructor. Calls close()."""
346 # The try/except block is in case this is called at program
347 # exit time, when it's possible that globals have already been
348 # deleted, and then the close() call might fail. Since
349 # there's nothing we can do about such failures and they annoy
350 # the end users, we suppress the traceback.
351 try:
352 self.close()
353 except:
354 pass
355
356 ### Inquiries ###
357
358 def seekable(self) -> bool:
359 """Return whether object supports random access.
360
361 If False, seek(), tell() and truncate() will raise IOError.
362 This method may need to do a test seek().
363 """
364 return False
365
366 def _checkSeekable(self, msg=None):
367 """Internal: raise an IOError if file is not seekable
368 """
369 if not self.seekable():
370 raise IOError("File or stream is not seekable."
371 if msg is None else msg)
372
373
374 def readable(self) -> bool:
375 """Return whether object was opened for reading.
376
377 If False, read() will raise IOError.
378 """
379 return False
380
381 def _checkReadable(self, msg=None):
382 """Internal: raise an IOError if file is not readable
383 """
384 if not self.readable():
385 raise IOError("File or stream is not readable."
386 if msg is None else msg)
387
388 def writable(self) -> bool:
389 """Return whether object was opened for writing.
390
391 If False, write() and truncate() will raise IOError.
392 """
393 return False
394
395 def _checkWritable(self, msg=None):
396 """Internal: raise an IOError if file is not writable
397 """
398 if not self.writable():
399 raise IOError("File or stream is not writable."
400 if msg is None else msg)
401
402 @property
403 def closed(self):
404 """closed: bool. True iff the file has been closed.
405
406 For backwards compatibility, this is a property, not a predicate.
407 """
408 return self.__closed
409
410 def _checkClosed(self, msg=None):
411 """Internal: raise an ValueError if file is closed
412 """
413 if self.closed:
414 raise ValueError("I/O operation on closed file."
415 if msg is None else msg)
416
417 ### Context manager ###
418
419 def __enter__(self) -> "IOBase": # That's a forward reference
420 """Context management protocol. Returns self."""
421 self._checkClosed()
422 return self
423
424 def __exit__(self, *args) -> None:
425 """Context management protocol. Calls close()"""
426 self.close()
427
428 ### Lower-level APIs ###
429
430 # XXX Should these be present even if unimplemented?
431
432 def fileno(self) -> int:
433 """Returns underlying file descriptor if one exists.
434
435 An IOError is raised if the IO object does not use a file descriptor.
436 """
437 self._unsupported("fileno")
438
439 def isatty(self) -> bool:
440 """Return whether this is an 'interactive' stream.
441
442 Return False if it can't be determined.
443 """
444 self._checkClosed()
445 return False
446
447 ### Readline[s] and writelines ###
448
449 def readline(self, limit: int = -1) -> bytes:
450 r"""Read and return a line from the stream.
451
452 If limit is specified, at most limit bytes will be read.
453
454 The line terminator is always b'\n' for binary files; for text
455 files, the newlines argument to open can be used to select the line
456 terminator(s) recognized.
457 """
458 # For backwards compatibility, a (slowish) readline().
459 if hasattr(self, "peek"):
460 def nreadahead():
461 readahead = self.peek(1)
462 if not readahead:
463 return 1
464 n = (readahead.find(b"\n") + 1) or len(readahead)
465 if limit >= 0:
466 n = min(n, limit)
467 return n
468 else:
469 def nreadahead():
470 return 1
471 if limit is None:
472 limit = -1
Benjamin Petersonb01138a2009-04-24 22:59:52 +0000473 elif not isinstance(limit, int):
474 raise TypeError("limit must be an integer")
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000475 res = bytearray()
476 while limit < 0 or len(res) < limit:
477 b = self.read(nreadahead())
478 if not b:
479 break
480 res += b
481 if res.endswith(b"\n"):
482 break
483 return bytes(res)
484
485 def __iter__(self):
486 self._checkClosed()
487 return self
488
489 def __next__(self):
490 line = self.readline()
491 if not line:
492 raise StopIteration
493 return line
494
495 def readlines(self, hint=None):
496 """Return a list of lines from the stream.
497
498 hint can be specified to control the number of lines read: no more
499 lines will be read if the total size (in bytes/characters) of all
500 lines so far exceeds hint.
501 """
502 if hint is None or hint <= 0:
503 return list(self)
504 n = 0
505 lines = []
506 for line in self:
507 lines.append(line)
508 n += len(line)
509 if n >= hint:
510 break
511 return lines
512
513 def writelines(self, lines):
514 self._checkClosed()
515 for line in lines:
516 self.write(line)
517
518io.IOBase.register(IOBase)
519
520
521class RawIOBase(IOBase):
522
523 """Base class for raw binary I/O."""
524
525 # The read() method is implemented by calling readinto(); derived
526 # classes that want to support read() only need to implement
527 # readinto() as a primitive operation. In general, readinto() can be
528 # more efficient than read().
529
530 # (It would be tempting to also provide an implementation of
531 # readinto() in terms of read(), in case the latter is a more suitable
532 # primitive operation, but that would lead to nasty recursion in case
533 # a subclass doesn't implement either.)
534
535 def read(self, n: int = -1) -> bytes:
536 """Read and return up to n bytes.
537
538 Returns an empty bytes object on EOF, or None if the object is
539 set not to block and has no data to read.
540 """
541 if n is None:
542 n = -1
543 if n < 0:
544 return self.readall()
545 b = bytearray(n.__index__())
546 n = self.readinto(b)
547 del b[n:]
548 return bytes(b)
549
550 def readall(self):
551 """Read until EOF, using multiple read() call."""
552 res = bytearray()
553 while True:
554 data = self.read(DEFAULT_BUFFER_SIZE)
555 if not data:
556 break
557 res += data
558 return bytes(res)
559
560 def readinto(self, b: bytearray) -> int:
561 """Read up to len(b) bytes into b.
562
563 Returns number of bytes read (0 for EOF), or None if the object
564 is set not to block as has no data to read.
565 """
566 self._unsupported("readinto")
567
568 def write(self, b: bytes) -> int:
569 """Write the given buffer to the IO stream.
570
571 Returns the number of bytes written, which may be less than len(b).
572 """
573 self._unsupported("write")
574
575io.RawIOBase.register(RawIOBase)
576from _io import FileIO
577RawIOBase.register(FileIO)
578
579
580class BufferedIOBase(IOBase):
581
582 """Base class for buffered IO objects.
583
584 The main difference with RawIOBase is that the read() method
585 supports omitting the size argument, and does not have a default
586 implementation that defers to readinto().
587
588 In addition, read(), readinto() and write() may raise
589 BlockingIOError if the underlying raw stream is in non-blocking
590 mode and not ready; unlike their raw counterparts, they will never
591 return None.
592
593 A typical implementation should not inherit from a RawIOBase
594 implementation, but wrap one.
595 """
596
597 def read(self, n: int = None) -> bytes:
598 """Read and return up to n bytes.
599
600 If the argument is omitted, None, or negative, reads and
601 returns all data until EOF.
602
603 If the argument is positive, and the underlying raw stream is
604 not 'interactive', multiple raw reads may be issued to satisfy
605 the byte count (unless EOF is reached first). But for
606 interactive raw streams (XXX and for pipes?), at most one raw
607 read will be issued, and a short result does not imply that
608 EOF is imminent.
609
610 Returns an empty bytes array on EOF.
611
612 Raises BlockingIOError if the underlying raw stream has no
613 data at the moment.
614 """
615 self._unsupported("read")
616
617 def read1(self, n: int=None) -> bytes:
618 """Read up to n bytes with at most one read() system call."""
619 self._unsupported("read1")
620
621 def readinto(self, b: bytearray) -> int:
622 """Read up to len(b) bytes into b.
623
624 Like read(), this may issue multiple reads to the underlying raw
625 stream, unless the latter is 'interactive'.
626
627 Returns the number of bytes read (0 for EOF).
628
629 Raises BlockingIOError if the underlying raw stream has no
630 data at the moment.
631 """
632 # XXX This ought to work with anything that supports the buffer API
633 data = self.read(len(b))
634 n = len(data)
635 try:
636 b[:n] = data
637 except TypeError as err:
638 import array
639 if not isinstance(b, array.array):
640 raise err
641 b[:n] = array.array('b', data)
642 return n
643
644 def write(self, b: bytes) -> int:
645 """Write the given buffer to the IO stream.
646
647 Return the number of bytes written, which is never less than
648 len(b).
649
650 Raises BlockingIOError if the buffer is full and the
651 underlying raw stream cannot accept more data at the moment.
652 """
653 self._unsupported("write")
654
Benjamin Petersond2e0c792009-05-01 20:40:59 +0000655 def detach(self) -> None:
656 """
657 Separate the underlying raw stream from the buffer and return it.
658
659 After the raw stream has been detached, the buffer is in an unusable
660 state.
661 """
662 self._unsupported("detach")
663
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000664io.BufferedIOBase.register(BufferedIOBase)
665
666
667class _BufferedIOMixin(BufferedIOBase):
668
669 """A mixin implementation of BufferedIOBase with an underlying raw stream.
670
671 This passes most requests on to the underlying raw stream. It
672 does *not* provide implementations of read(), readinto() or
673 write().
674 """
675
676 def __init__(self, raw):
677 self.raw = raw
678
679 ### Positioning ###
680
681 def seek(self, pos, whence=0):
682 new_position = self.raw.seek(pos, whence)
683 if new_position < 0:
684 raise IOError("seek() returned an invalid position")
685 return new_position
686
687 def tell(self):
688 pos = self.raw.tell()
689 if pos < 0:
690 raise IOError("tell() returned an invalid position")
691 return pos
692
693 def truncate(self, pos=None):
694 # Flush the stream. We're mixing buffered I/O with lower-level I/O,
695 # and a flush may be necessary to synch both views of the current
696 # file state.
697 self.flush()
698
699 if pos is None:
700 pos = self.tell()
701 # XXX: Should seek() be used, instead of passing the position
702 # XXX directly to truncate?
703 return self.raw.truncate(pos)
704
705 ### Flush and close ###
706
707 def flush(self):
708 self.raw.flush()
709
710 def close(self):
Benjamin Petersond2e0c792009-05-01 20:40:59 +0000711 if not self.closed and self.raw is not None:
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000712 try:
713 self.flush()
714 except IOError:
715 pass # If flush() fails, just give up
716 self.raw.close()
717
Benjamin Petersond2e0c792009-05-01 20:40:59 +0000718 def detach(self):
719 if self.raw is None:
720 raise ValueError("raw stream already detached")
721 self.flush()
722 raw = self.raw
723 self.raw = None
724 return raw
725
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000726 ### Inquiries ###
727
728 def seekable(self):
729 return self.raw.seekable()
730
731 def readable(self):
732 return self.raw.readable()
733
734 def writable(self):
735 return self.raw.writable()
736
737 @property
738 def closed(self):
739 return self.raw.closed
740
741 @property
742 def name(self):
743 return self.raw.name
744
745 @property
746 def mode(self):
747 return self.raw.mode
748
Antoine Pitrou716c4442009-05-23 19:04:03 +0000749 def __repr__(self):
750 clsname = self.__class__.__name__
751 try:
752 name = self.name
753 except AttributeError:
754 return "<_pyio.{0}>".format(clsname)
755 else:
756 return "<_pyio.{0} name={1!r}>".format(clsname, name)
757
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000758 ### Lower-level APIs ###
759
760 def fileno(self):
761 return self.raw.fileno()
762
763 def isatty(self):
764 return self.raw.isatty()
765
766
767class BytesIO(BufferedIOBase):
768
769 """Buffered I/O implementation using an in-memory bytes buffer."""
770
771 def __init__(self, initial_bytes=None):
772 buf = bytearray()
773 if initial_bytes is not None:
774 buf += initial_bytes
775 self._buffer = buf
776 self._pos = 0
777
Alexandre Vassalotticf76e1a2009-07-22 03:24:36 +0000778 def __getstate__(self):
779 if self.closed:
780 raise ValueError("__getstate__ on closed file")
781 return self.__dict__.copy()
782
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000783 def getvalue(self):
784 """Return the bytes value (contents) of the buffer
785 """
786 if self.closed:
787 raise ValueError("getvalue on closed file")
788 return bytes(self._buffer)
789
790 def read(self, n=None):
791 if self.closed:
792 raise ValueError("read from closed file")
793 if n is None:
794 n = -1
795 if n < 0:
796 n = len(self._buffer)
797 if len(self._buffer) <= self._pos:
798 return b""
799 newpos = min(len(self._buffer), self._pos + n)
800 b = self._buffer[self._pos : newpos]
801 self._pos = newpos
802 return bytes(b)
803
804 def read1(self, n):
805 """This is the same as read.
806 """
807 return self.read(n)
808
809 def write(self, b):
810 if self.closed:
811 raise ValueError("write to closed file")
812 if isinstance(b, str):
813 raise TypeError("can't write str to binary stream")
814 n = len(b)
815 if n == 0:
816 return 0
817 pos = self._pos
818 if pos > len(self._buffer):
819 # Inserts null bytes between the current end of the file
820 # and the new write position.
821 padding = b'\x00' * (pos - len(self._buffer))
822 self._buffer += padding
823 self._buffer[pos:pos + n] = b
824 self._pos += n
825 return n
826
827 def seek(self, pos, whence=0):
828 if self.closed:
829 raise ValueError("seek on closed file")
830 try:
831 pos = pos.__index__()
832 except AttributeError as err:
833 raise TypeError("an integer is required") from err
834 if whence == 0:
835 if pos < 0:
836 raise ValueError("negative seek position %r" % (pos,))
837 self._pos = pos
838 elif whence == 1:
839 self._pos = max(0, self._pos + pos)
840 elif whence == 2:
841 self._pos = max(0, len(self._buffer) + pos)
842 else:
843 raise ValueError("invalid whence value")
844 return self._pos
845
846 def tell(self):
847 if self.closed:
848 raise ValueError("tell on closed file")
849 return self._pos
850
851 def truncate(self, pos=None):
852 if self.closed:
853 raise ValueError("truncate on closed file")
854 if pos is None:
855 pos = self._pos
856 elif pos < 0:
857 raise ValueError("negative truncate position %r" % (pos,))
858 del self._buffer[pos:]
Antoine Pitrou905a2ff2010-01-31 22:47:27 +0000859 return pos
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000860
861 def readable(self):
862 return True
863
864 def writable(self):
865 return True
866
867 def seekable(self):
868 return True
869
870
871class BufferedReader(_BufferedIOMixin):
872
873 """BufferedReader(raw[, buffer_size])
874
875 A buffer for a readable, sequential BaseRawIO object.
876
877 The constructor creates a BufferedReader for the given readable raw
878 stream and buffer_size. If buffer_size is omitted, DEFAULT_BUFFER_SIZE
879 is used.
880 """
881
882 def __init__(self, raw, buffer_size=DEFAULT_BUFFER_SIZE):
883 """Create a new buffered reader using the given readable raw IO object.
884 """
Antoine Pitroucf4c7492009-04-19 00:09:36 +0000885 if not raw.readable():
886 raise IOError('"raw" argument must be readable.')
887
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000888 _BufferedIOMixin.__init__(self, raw)
889 if buffer_size <= 0:
890 raise ValueError("invalid buffer size")
891 self.buffer_size = buffer_size
892 self._reset_read_buf()
893 self._read_lock = Lock()
894
895 def _reset_read_buf(self):
896 self._read_buf = b""
897 self._read_pos = 0
898
899 def read(self, n=None):
900 """Read n bytes.
901
902 Returns exactly n bytes of data unless the underlying raw IO
903 stream reaches EOF or if the call would block in non-blocking
904 mode. If n is negative, read until EOF or until read() would
905 block.
906 """
907 if n is not None and n < -1:
908 raise ValueError("invalid number of bytes to read")
909 with self._read_lock:
910 return self._read_unlocked(n)
911
912 def _read_unlocked(self, n=None):
913 nodata_val = b""
914 empty_values = (b"", None)
915 buf = self._read_buf
916 pos = self._read_pos
917
918 # Special case for when the number of bytes to read is unspecified.
919 if n is None or n == -1:
920 self._reset_read_buf()
921 chunks = [buf[pos:]] # Strip the consumed bytes.
922 current_size = 0
923 while True:
924 # Read until EOF or until read() would block.
925 chunk = self.raw.read()
926 if chunk in empty_values:
927 nodata_val = chunk
928 break
929 current_size += len(chunk)
930 chunks.append(chunk)
931 return b"".join(chunks) or nodata_val
932
933 # The number of bytes to read is specified, return at most n bytes.
934 avail = len(buf) - pos # Length of the available buffered data.
935 if n <= avail:
936 # Fast path: the data to read is fully buffered.
937 self._read_pos += n
938 return buf[pos:pos+n]
939 # Slow path: read from the stream until enough bytes are read,
940 # or until an EOF occurs or until read() would block.
941 chunks = [buf[pos:]]
942 wanted = max(self.buffer_size, n)
943 while avail < n:
944 chunk = self.raw.read(wanted)
945 if chunk in empty_values:
946 nodata_val = chunk
947 break
948 avail += len(chunk)
949 chunks.append(chunk)
950 # n is more then avail only when an EOF occurred or when
951 # read() would have blocked.
952 n = min(n, avail)
953 out = b"".join(chunks)
954 self._read_buf = out[n:] # Save the extra data in the buffer.
955 self._read_pos = 0
956 return out[:n] if out else nodata_val
957
958 def peek(self, n=0):
959 """Returns buffered bytes without advancing the position.
960
961 The argument indicates a desired minimal number of bytes; we
962 do at most one raw read to satisfy it. We never return more
963 than self.buffer_size.
964 """
965 with self._read_lock:
966 return self._peek_unlocked(n)
967
968 def _peek_unlocked(self, n=0):
969 want = min(n, self.buffer_size)
970 have = len(self._read_buf) - self._read_pos
971 if have < want or have <= 0:
972 to_read = self.buffer_size - have
973 current = self.raw.read(to_read)
974 if current:
975 self._read_buf = self._read_buf[self._read_pos:] + current
976 self._read_pos = 0
977 return self._read_buf[self._read_pos:]
978
979 def read1(self, n):
980 """Reads up to n bytes, with at most one read() system call."""
981 # Returns up to n bytes. If at least one byte is buffered, we
982 # only return buffered bytes. Otherwise, we do one raw read.
983 if n < 0:
984 raise ValueError("number of bytes to read must be positive")
985 if n == 0:
986 return b""
987 with self._read_lock:
988 self._peek_unlocked(1)
989 return self._read_unlocked(
990 min(n, len(self._read_buf) - self._read_pos))
991
992 def tell(self):
993 return _BufferedIOMixin.tell(self) - len(self._read_buf) + self._read_pos
994
995 def seek(self, pos, whence=0):
996 if not (0 <= whence <= 2):
997 raise ValueError("invalid whence value")
998 with self._read_lock:
999 if whence == 1:
1000 pos -= len(self._read_buf) - self._read_pos
1001 pos = _BufferedIOMixin.seek(self, pos, whence)
1002 self._reset_read_buf()
1003 return pos
1004
1005class BufferedWriter(_BufferedIOMixin):
1006
1007 """A buffer for a writeable sequential RawIO object.
1008
1009 The constructor creates a BufferedWriter for the given writeable raw
1010 stream. If the buffer_size is not given, it defaults to
Benjamin Peterson59406a92009-03-26 17:10:29 +00001011 DEFAULT_BUFFER_SIZE.
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001012 """
1013
Benjamin Peterson59406a92009-03-26 17:10:29 +00001014 _warning_stack_offset = 2
1015
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001016 def __init__(self, raw,
1017 buffer_size=DEFAULT_BUFFER_SIZE, max_buffer_size=None):
Antoine Pitroucf4c7492009-04-19 00:09:36 +00001018 if not raw.writable():
1019 raise IOError('"raw" argument must be writable.')
1020
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001021 _BufferedIOMixin.__init__(self, raw)
1022 if buffer_size <= 0:
1023 raise ValueError("invalid buffer size")
Benjamin Peterson59406a92009-03-26 17:10:29 +00001024 if max_buffer_size is not None:
1025 warnings.warn("max_buffer_size is deprecated", DeprecationWarning,
1026 self._warning_stack_offset)
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001027 self.buffer_size = buffer_size
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001028 self._write_buf = bytearray()
1029 self._write_lock = Lock()
1030
1031 def write(self, b):
1032 if self.closed:
1033 raise ValueError("write to closed file")
1034 if isinstance(b, str):
1035 raise TypeError("can't write str to binary stream")
1036 with self._write_lock:
1037 # XXX we can implement some more tricks to try and avoid
1038 # partial writes
1039 if len(self._write_buf) > self.buffer_size:
1040 # We're full, so let's pre-flush the buffer
1041 try:
1042 self._flush_unlocked()
1043 except BlockingIOError as e:
1044 # We can't accept anything else.
1045 # XXX Why not just let the exception pass through?
1046 raise BlockingIOError(e.errno, e.strerror, 0)
1047 before = len(self._write_buf)
1048 self._write_buf.extend(b)
1049 written = len(self._write_buf) - before
1050 if len(self._write_buf) > self.buffer_size:
1051 try:
1052 self._flush_unlocked()
1053 except BlockingIOError as e:
Benjamin Peterson394ee002009-03-05 22:33:59 +00001054 if len(self._write_buf) > self.buffer_size:
1055 # We've hit the buffer_size. We have to accept a partial
1056 # write and cut back our buffer.
1057 overage = len(self._write_buf) - self.buffer_size
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001058 written -= overage
Benjamin Peterson394ee002009-03-05 22:33:59 +00001059 self._write_buf = self._write_buf[:self.buffer_size]
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001060 raise BlockingIOError(e.errno, e.strerror, written)
1061 return written
1062
1063 def truncate(self, pos=None):
1064 with self._write_lock:
1065 self._flush_unlocked()
1066 if pos is None:
1067 pos = self.raw.tell()
1068 return self.raw.truncate(pos)
1069
1070 def flush(self):
1071 with self._write_lock:
1072 self._flush_unlocked()
1073
1074 def _flush_unlocked(self):
1075 if self.closed:
1076 raise ValueError("flush of closed file")
1077 written = 0
1078 try:
1079 while self._write_buf:
1080 n = self.raw.write(self._write_buf)
1081 if n > len(self._write_buf) or n < 0:
1082 raise IOError("write() returned incorrect number of bytes")
1083 del self._write_buf[:n]
1084 written += n
1085 except BlockingIOError as e:
1086 n = e.characters_written
1087 del self._write_buf[:n]
1088 written += n
1089 raise BlockingIOError(e.errno, e.strerror, written)
1090
1091 def tell(self):
1092 return _BufferedIOMixin.tell(self) + len(self._write_buf)
1093
1094 def seek(self, pos, whence=0):
1095 if not (0 <= whence <= 2):
1096 raise ValueError("invalid whence")
1097 with self._write_lock:
1098 self._flush_unlocked()
1099 return _BufferedIOMixin.seek(self, pos, whence)
1100
1101
1102class BufferedRWPair(BufferedIOBase):
1103
1104 """A buffered reader and writer object together.
1105
1106 A buffered reader object and buffered writer object put together to
1107 form a sequential IO object that can read and write. This is typically
1108 used with a socket or two-way pipe.
1109
1110 reader and writer are RawIOBase objects that are readable and
1111 writeable respectively. If the buffer_size is omitted it defaults to
Benjamin Peterson59406a92009-03-26 17:10:29 +00001112 DEFAULT_BUFFER_SIZE.
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001113 """
1114
1115 # XXX The usefulness of this (compared to having two separate IO
1116 # objects) is questionable.
1117
1118 def __init__(self, reader, writer,
1119 buffer_size=DEFAULT_BUFFER_SIZE, max_buffer_size=None):
1120 """Constructor.
1121
1122 The arguments are two RawIO instances.
1123 """
Benjamin Peterson59406a92009-03-26 17:10:29 +00001124 if max_buffer_size is not None:
1125 warnings.warn("max_buffer_size is deprecated", DeprecationWarning, 2)
Antoine Pitroucf4c7492009-04-19 00:09:36 +00001126
1127 if not reader.readable():
1128 raise IOError('"reader" argument must be readable.')
1129
1130 if not writer.writable():
1131 raise IOError('"writer" argument must be writable.')
1132
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001133 self.reader = BufferedReader(reader, buffer_size)
Benjamin Peterson59406a92009-03-26 17:10:29 +00001134 self.writer = BufferedWriter(writer, buffer_size)
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001135
1136 def read(self, n=None):
1137 if n is None:
1138 n = -1
1139 return self.reader.read(n)
1140
1141 def readinto(self, b):
1142 return self.reader.readinto(b)
1143
1144 def write(self, b):
1145 return self.writer.write(b)
1146
1147 def peek(self, n=0):
1148 return self.reader.peek(n)
1149
1150 def read1(self, n):
1151 return self.reader.read1(n)
1152
1153 def readable(self):
1154 return self.reader.readable()
1155
1156 def writable(self):
1157 return self.writer.writable()
1158
1159 def flush(self):
1160 return self.writer.flush()
1161
1162 def close(self):
1163 self.writer.close()
1164 self.reader.close()
1165
1166 def isatty(self):
1167 return self.reader.isatty() or self.writer.isatty()
1168
1169 @property
1170 def closed(self):
1171 return self.writer.closed
1172
1173
1174class BufferedRandom(BufferedWriter, BufferedReader):
1175
1176 """A buffered interface to random access streams.
1177
1178 The constructor creates a reader and writer for a seekable stream,
1179 raw, given in the first argument. If the buffer_size is omitted it
Benjamin Peterson59406a92009-03-26 17:10:29 +00001180 defaults to DEFAULT_BUFFER_SIZE.
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001181 """
1182
Benjamin Peterson59406a92009-03-26 17:10:29 +00001183 _warning_stack_offset = 3
1184
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001185 def __init__(self, raw,
1186 buffer_size=DEFAULT_BUFFER_SIZE, max_buffer_size=None):
1187 raw._checkSeekable()
1188 BufferedReader.__init__(self, raw, buffer_size)
1189 BufferedWriter.__init__(self, raw, buffer_size, max_buffer_size)
1190
1191 def seek(self, pos, whence=0):
1192 if not (0 <= whence <= 2):
1193 raise ValueError("invalid whence")
1194 self.flush()
1195 if self._read_buf:
1196 # Undo read ahead.
1197 with self._read_lock:
1198 self.raw.seek(self._read_pos - len(self._read_buf), 1)
1199 # First do the raw seek, then empty the read buffer, so that
1200 # if the raw seek fails, we don't lose buffered data forever.
1201 pos = self.raw.seek(pos, whence)
1202 with self._read_lock:
1203 self._reset_read_buf()
1204 if pos < 0:
1205 raise IOError("seek() returned invalid position")
1206 return pos
1207
1208 def tell(self):
1209 if self._write_buf:
1210 return BufferedWriter.tell(self)
1211 else:
1212 return BufferedReader.tell(self)
1213
1214 def truncate(self, pos=None):
1215 if pos is None:
1216 pos = self.tell()
1217 # Use seek to flush the read buffer.
Antoine Pitrou905a2ff2010-01-31 22:47:27 +00001218 return BufferedWriter.truncate(self, pos)
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001219
1220 def read(self, n=None):
1221 if n is None:
1222 n = -1
1223 self.flush()
1224 return BufferedReader.read(self, n)
1225
1226 def readinto(self, b):
1227 self.flush()
1228 return BufferedReader.readinto(self, b)
1229
1230 def peek(self, n=0):
1231 self.flush()
1232 return BufferedReader.peek(self, n)
1233
1234 def read1(self, n):
1235 self.flush()
1236 return BufferedReader.read1(self, n)
1237
1238 def write(self, b):
1239 if self._read_buf:
1240 # Undo readahead
1241 with self._read_lock:
1242 self.raw.seek(self._read_pos - len(self._read_buf), 1)
1243 self._reset_read_buf()
1244 return BufferedWriter.write(self, b)
1245
1246
1247class TextIOBase(IOBase):
1248
1249 """Base class for text I/O.
1250
1251 This class provides a character and line based interface to stream
1252 I/O. There is no readinto method because Python's character strings
1253 are immutable. There is no public constructor.
1254 """
1255
1256 def read(self, n: int = -1) -> str:
1257 """Read at most n characters from stream.
1258
1259 Read from underlying buffer until we have n characters or we hit EOF.
1260 If n is negative or omitted, read until EOF.
1261 """
1262 self._unsupported("read")
1263
1264 def write(self, s: str) -> int:
1265 """Write string s to stream."""
1266 self._unsupported("write")
1267
1268 def truncate(self, pos: int = None) -> int:
1269 """Truncate size to pos."""
1270 self._unsupported("truncate")
1271
1272 def readline(self) -> str:
1273 """Read until newline or EOF.
1274
1275 Returns an empty string if EOF is hit immediately.
1276 """
1277 self._unsupported("readline")
1278
Benjamin Petersond2e0c792009-05-01 20:40:59 +00001279 def detach(self) -> None:
1280 """
1281 Separate the underlying buffer from the TextIOBase and return it.
1282
1283 After the underlying buffer has been detached, the TextIO is in an
1284 unusable state.
1285 """
1286 self._unsupported("detach")
1287
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001288 @property
1289 def encoding(self):
1290 """Subclasses should override."""
1291 return None
1292
1293 @property
1294 def newlines(self):
1295 """Line endings translated so far.
1296
1297 Only line endings translated during reading are considered.
1298
1299 Subclasses should override.
1300 """
1301 return None
1302
Benjamin Peterson0926ad12009-06-06 18:02:12 +00001303 @property
1304 def errors(self):
1305 """Error setting of the decoder or encoder.
1306
1307 Subclasses should override."""
1308 return None
1309
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001310io.TextIOBase.register(TextIOBase)
1311
1312
1313class IncrementalNewlineDecoder(codecs.IncrementalDecoder):
1314 r"""Codec used when reading a file in universal newlines mode. It wraps
1315 another incremental decoder, translating \r\n and \r into \n. It also
1316 records the types of newlines encountered. When used with
1317 translate=False, it ensures that the newline sequence is returned in
1318 one piece.
1319 """
1320 def __init__(self, decoder, translate, errors='strict'):
1321 codecs.IncrementalDecoder.__init__(self, errors=errors)
1322 self.translate = translate
1323 self.decoder = decoder
1324 self.seennl = 0
1325 self.pendingcr = False
1326
1327 def decode(self, input, final=False):
1328 # decode input (with the eventual \r from a previous pass)
1329 if self.decoder is None:
1330 output = input
1331 else:
1332 output = self.decoder.decode(input, final=final)
1333 if self.pendingcr and (output or final):
1334 output = "\r" + output
1335 self.pendingcr = False
1336
1337 # retain last \r even when not translating data:
1338 # then readline() is sure to get \r\n in one pass
1339 if output.endswith("\r") and not final:
1340 output = output[:-1]
1341 self.pendingcr = True
1342
1343 # Record which newlines are read
1344 crlf = output.count('\r\n')
1345 cr = output.count('\r') - crlf
1346 lf = output.count('\n') - crlf
1347 self.seennl |= (lf and self._LF) | (cr and self._CR) \
1348 | (crlf and self._CRLF)
1349
1350 if self.translate:
1351 if crlf:
1352 output = output.replace("\r\n", "\n")
1353 if cr:
1354 output = output.replace("\r", "\n")
1355
1356 return output
1357
1358 def getstate(self):
1359 if self.decoder is None:
1360 buf = b""
1361 flag = 0
1362 else:
1363 buf, flag = self.decoder.getstate()
1364 flag <<= 1
1365 if self.pendingcr:
1366 flag |= 1
1367 return buf, flag
1368
1369 def setstate(self, state):
1370 buf, flag = state
1371 self.pendingcr = bool(flag & 1)
1372 if self.decoder is not None:
1373 self.decoder.setstate((buf, flag >> 1))
1374
1375 def reset(self):
1376 self.seennl = 0
1377 self.pendingcr = False
1378 if self.decoder is not None:
1379 self.decoder.reset()
1380
1381 _LF = 1
1382 _CR = 2
1383 _CRLF = 4
1384
1385 @property
1386 def newlines(self):
1387 return (None,
1388 "\n",
1389 "\r",
1390 ("\r", "\n"),
1391 "\r\n",
1392 ("\n", "\r\n"),
1393 ("\r", "\r\n"),
1394 ("\r", "\n", "\r\n")
1395 )[self.seennl]
1396
1397
1398class TextIOWrapper(TextIOBase):
1399
1400 r"""Character and line based layer over a BufferedIOBase object, buffer.
1401
1402 encoding gives the name of the encoding that the stream will be
1403 decoded or encoded with. It defaults to locale.getpreferredencoding.
1404
1405 errors determines the strictness of encoding and decoding (see the
1406 codecs.register) and defaults to "strict".
1407
1408 newline can be None, '', '\n', '\r', or '\r\n'. It controls the
1409 handling of line endings. If it is None, universal newlines is
1410 enabled. With this enabled, on input, the lines endings '\n', '\r',
1411 or '\r\n' are translated to '\n' before being returned to the
1412 caller. Conversely, on output, '\n' is translated to the system
1413 default line seperator, os.linesep. If newline is any other of its
1414 legal values, that newline becomes the newline when the file is read
1415 and it is returned untranslated. On output, '\n' is converted to the
1416 newline.
1417
1418 If line_buffering is True, a call to flush is implied when a call to
1419 write contains a newline character.
1420 """
1421
1422 _CHUNK_SIZE = 2048
1423
1424 def __init__(self, buffer, encoding=None, errors=None, newline=None,
1425 line_buffering=False):
1426 if newline is not None and not isinstance(newline, str):
1427 raise TypeError("illegal newline type: %r" % (type(newline),))
1428 if newline not in (None, "", "\n", "\r", "\r\n"):
1429 raise ValueError("illegal newline value: %r" % (newline,))
1430 if encoding is None:
1431 try:
1432 encoding = os.device_encoding(buffer.fileno())
1433 except (AttributeError, UnsupportedOperation):
1434 pass
1435 if encoding is None:
1436 try:
1437 import locale
1438 except ImportError:
1439 # Importing locale may fail if Python is being built
1440 encoding = "ascii"
1441 else:
1442 encoding = locale.getpreferredencoding()
1443
1444 if not isinstance(encoding, str):
1445 raise ValueError("invalid encoding: %r" % encoding)
1446
1447 if errors is None:
1448 errors = "strict"
1449 else:
1450 if not isinstance(errors, str):
1451 raise ValueError("invalid errors: %r" % errors)
1452
1453 self.buffer = buffer
1454 self._line_buffering = line_buffering
1455 self._encoding = encoding
1456 self._errors = errors
1457 self._readuniversal = not newline
1458 self._readtranslate = newline is None
1459 self._readnl = newline
1460 self._writetranslate = newline != ''
1461 self._writenl = newline or os.linesep
1462 self._encoder = None
1463 self._decoder = None
1464 self._decoded_chars = '' # buffer for text returned from decoder
1465 self._decoded_chars_used = 0 # offset into _decoded_chars for read()
1466 self._snapshot = None # info for reconstructing decoder state
1467 self._seekable = self._telling = self.buffer.seekable()
1468
Antoine Pitroue4501852009-05-14 18:55:55 +00001469 if self._seekable and self.writable():
1470 position = self.buffer.tell()
1471 if position != 0:
1472 try:
1473 self._get_encoder().setstate(0)
1474 except LookupError:
1475 # Sometimes the encoder doesn't exist
1476 pass
1477
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001478 # self._snapshot is either None, or a tuple (dec_flags, next_input)
1479 # where dec_flags is the second (integer) item of the decoder state
1480 # and next_input is the chunk of input bytes that comes next after the
1481 # snapshot point. We use this to reconstruct decoder states in tell().
1482
1483 # Naming convention:
1484 # - "bytes_..." for integer variables that count input bytes
1485 # - "chars_..." for integer variables that count decoded characters
1486
Benjamin Petersonc4c0eae2009-03-09 00:07:03 +00001487 def __repr__(self):
Antoine Pitrou716c4442009-05-23 19:04:03 +00001488 try:
1489 name = self.name
1490 except AttributeError:
1491 return "<_pyio.TextIOWrapper encoding={0!r}>".format(self.encoding)
1492 else:
1493 return "<_pyio.TextIOWrapper name={0!r} encoding={1!r}>".format(
1494 name, self.encoding)
Benjamin Petersonc4c0eae2009-03-09 00:07:03 +00001495
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001496 @property
1497 def encoding(self):
1498 return self._encoding
1499
1500 @property
1501 def errors(self):
1502 return self._errors
1503
1504 @property
1505 def line_buffering(self):
1506 return self._line_buffering
1507
1508 def seekable(self):
1509 return self._seekable
1510
1511 def readable(self):
1512 return self.buffer.readable()
1513
1514 def writable(self):
1515 return self.buffer.writable()
1516
1517 def flush(self):
1518 self.buffer.flush()
1519 self._telling = self._seekable
1520
1521 def close(self):
Benjamin Petersond2e0c792009-05-01 20:40:59 +00001522 if self.buffer is not None:
1523 try:
1524 self.flush()
1525 except IOError:
1526 pass # If flush() fails, just give up
1527 self.buffer.close()
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001528
1529 @property
1530 def closed(self):
1531 return self.buffer.closed
1532
1533 @property
1534 def name(self):
1535 return self.buffer.name
1536
1537 def fileno(self):
1538 return self.buffer.fileno()
1539
1540 def isatty(self):
1541 return self.buffer.isatty()
1542
1543 def write(self, s: str):
1544 if self.closed:
1545 raise ValueError("write to closed file")
1546 if not isinstance(s, str):
1547 raise TypeError("can't write %s to text stream" %
1548 s.__class__.__name__)
1549 length = len(s)
1550 haslf = (self._writetranslate or self._line_buffering) and "\n" in s
1551 if haslf and self._writetranslate and self._writenl != "\n":
1552 s = s.replace("\n", self._writenl)
1553 encoder = self._encoder or self._get_encoder()
1554 # XXX What if we were just reading?
1555 b = encoder.encode(s)
1556 self.buffer.write(b)
1557 if self._line_buffering and (haslf or "\r" in s):
1558 self.flush()
1559 self._snapshot = None
1560 if self._decoder:
1561 self._decoder.reset()
1562 return length
1563
1564 def _get_encoder(self):
1565 make_encoder = codecs.getincrementalencoder(self._encoding)
1566 self._encoder = make_encoder(self._errors)
1567 return self._encoder
1568
1569 def _get_decoder(self):
1570 make_decoder = codecs.getincrementaldecoder(self._encoding)
1571 decoder = make_decoder(self._errors)
1572 if self._readuniversal:
1573 decoder = IncrementalNewlineDecoder(decoder, self._readtranslate)
1574 self._decoder = decoder
1575 return decoder
1576
1577 # The following three methods implement an ADT for _decoded_chars.
1578 # Text returned from the decoder is buffered here until the client
1579 # requests it by calling our read() or readline() method.
1580 def _set_decoded_chars(self, chars):
1581 """Set the _decoded_chars buffer."""
1582 self._decoded_chars = chars
1583 self._decoded_chars_used = 0
1584
1585 def _get_decoded_chars(self, n=None):
1586 """Advance into the _decoded_chars buffer."""
1587 offset = self._decoded_chars_used
1588 if n is None:
1589 chars = self._decoded_chars[offset:]
1590 else:
1591 chars = self._decoded_chars[offset:offset + n]
1592 self._decoded_chars_used += len(chars)
1593 return chars
1594
1595 def _rewind_decoded_chars(self, n):
1596 """Rewind the _decoded_chars buffer."""
1597 if self._decoded_chars_used < n:
1598 raise AssertionError("rewind decoded_chars out of bounds")
1599 self._decoded_chars_used -= n
1600
1601 def _read_chunk(self):
1602 """
1603 Read and decode the next chunk of data from the BufferedReader.
1604 """
1605
1606 # The return value is True unless EOF was reached. The decoded
1607 # string is placed in self._decoded_chars (replacing its previous
1608 # value). The entire input chunk is sent to the decoder, though
1609 # some of it may remain buffered in the decoder, yet to be
1610 # converted.
1611
1612 if self._decoder is None:
1613 raise ValueError("no decoder")
1614
1615 if self._telling:
1616 # To prepare for tell(), we need to snapshot a point in the
1617 # file where the decoder's input buffer is empty.
1618
1619 dec_buffer, dec_flags = self._decoder.getstate()
1620 # Given this, we know there was a valid snapshot point
1621 # len(dec_buffer) bytes ago with decoder state (b'', dec_flags).
1622
1623 # Read a chunk, decode it, and put the result in self._decoded_chars.
1624 input_chunk = self.buffer.read1(self._CHUNK_SIZE)
1625 eof = not input_chunk
1626 self._set_decoded_chars(self._decoder.decode(input_chunk, eof))
1627
1628 if self._telling:
1629 # At the snapshot point, len(dec_buffer) bytes before the read,
1630 # the next input to be decoded is dec_buffer + input_chunk.
1631 self._snapshot = (dec_flags, dec_buffer + input_chunk)
1632
1633 return not eof
1634
1635 def _pack_cookie(self, position, dec_flags=0,
1636 bytes_to_feed=0, need_eof=0, chars_to_skip=0):
1637 # The meaning of a tell() cookie is: seek to position, set the
1638 # decoder flags to dec_flags, read bytes_to_feed bytes, feed them
1639 # into the decoder with need_eof as the EOF flag, then skip
1640 # chars_to_skip characters of the decoded result. For most simple
1641 # decoders, tell() will often just give a byte offset in the file.
1642 return (position | (dec_flags<<64) | (bytes_to_feed<<128) |
1643 (chars_to_skip<<192) | bool(need_eof)<<256)
1644
1645 def _unpack_cookie(self, bigint):
1646 rest, position = divmod(bigint, 1<<64)
1647 rest, dec_flags = divmod(rest, 1<<64)
1648 rest, bytes_to_feed = divmod(rest, 1<<64)
1649 need_eof, chars_to_skip = divmod(rest, 1<<64)
1650 return position, dec_flags, bytes_to_feed, need_eof, chars_to_skip
1651
1652 def tell(self):
1653 if not self._seekable:
1654 raise IOError("underlying stream is not seekable")
1655 if not self._telling:
1656 raise IOError("telling position disabled by next() call")
1657 self.flush()
1658 position = self.buffer.tell()
1659 decoder = self._decoder
1660 if decoder is None or self._snapshot is None:
1661 if self._decoded_chars:
1662 # This should never happen.
1663 raise AssertionError("pending decoded text")
1664 return position
1665
1666 # Skip backward to the snapshot point (see _read_chunk).
1667 dec_flags, next_input = self._snapshot
1668 position -= len(next_input)
1669
1670 # How many decoded characters have been used up since the snapshot?
1671 chars_to_skip = self._decoded_chars_used
1672 if chars_to_skip == 0:
1673 # We haven't moved from the snapshot point.
1674 return self._pack_cookie(position, dec_flags)
1675
1676 # Starting from the snapshot position, we will walk the decoder
1677 # forward until it gives us enough decoded characters.
1678 saved_state = decoder.getstate()
1679 try:
1680 # Note our initial start point.
1681 decoder.setstate((b'', dec_flags))
1682 start_pos = position
1683 start_flags, bytes_fed, chars_decoded = dec_flags, 0, 0
1684 need_eof = 0
1685
1686 # Feed the decoder one byte at a time. As we go, note the
1687 # nearest "safe start point" before the current location
1688 # (a point where the decoder has nothing buffered, so seek()
1689 # can safely start from there and advance to this location).
1690 next_byte = bytearray(1)
1691 for next_byte[0] in next_input:
1692 bytes_fed += 1
1693 chars_decoded += len(decoder.decode(next_byte))
1694 dec_buffer, dec_flags = decoder.getstate()
1695 if not dec_buffer and chars_decoded <= chars_to_skip:
1696 # Decoder buffer is empty, so this is a safe start point.
1697 start_pos += bytes_fed
1698 chars_to_skip -= chars_decoded
1699 start_flags, bytes_fed, chars_decoded = dec_flags, 0, 0
1700 if chars_decoded >= chars_to_skip:
1701 break
1702 else:
1703 # We didn't get enough decoded data; signal EOF to get more.
1704 chars_decoded += len(decoder.decode(b'', final=True))
1705 need_eof = 1
1706 if chars_decoded < chars_to_skip:
1707 raise IOError("can't reconstruct logical file position")
1708
1709 # The returned cookie corresponds to the last safe start point.
1710 return self._pack_cookie(
1711 start_pos, start_flags, bytes_fed, need_eof, chars_to_skip)
1712 finally:
1713 decoder.setstate(saved_state)
1714
1715 def truncate(self, pos=None):
1716 self.flush()
1717 if pos is None:
1718 pos = self.tell()
Antoine Pitrou905a2ff2010-01-31 22:47:27 +00001719 return self.buffer.truncate(pos)
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001720
Benjamin Petersond2e0c792009-05-01 20:40:59 +00001721 def detach(self):
1722 if self.buffer is None:
1723 raise ValueError("buffer is already detached")
1724 self.flush()
1725 buffer = self.buffer
1726 self.buffer = None
1727 return buffer
1728
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001729 def seek(self, cookie, whence=0):
1730 if self.closed:
1731 raise ValueError("tell on closed file")
1732 if not self._seekable:
1733 raise IOError("underlying stream is not seekable")
1734 if whence == 1: # seek relative to current position
1735 if cookie != 0:
1736 raise IOError("can't do nonzero cur-relative seeks")
1737 # Seeking to the current position should attempt to
1738 # sync the underlying buffer with the current position.
1739 whence = 0
1740 cookie = self.tell()
1741 if whence == 2: # seek relative to end of file
1742 if cookie != 0:
1743 raise IOError("can't do nonzero end-relative seeks")
1744 self.flush()
1745 position = self.buffer.seek(0, 2)
1746 self._set_decoded_chars('')
1747 self._snapshot = None
1748 if self._decoder:
1749 self._decoder.reset()
1750 return position
1751 if whence != 0:
1752 raise ValueError("invalid whence (%r, should be 0, 1 or 2)" %
1753 (whence,))
1754 if cookie < 0:
1755 raise ValueError("negative seek position %r" % (cookie,))
1756 self.flush()
1757
1758 # The strategy of seek() is to go back to the safe start point
1759 # and replay the effect of read(chars_to_skip) from there.
1760 start_pos, dec_flags, bytes_to_feed, need_eof, chars_to_skip = \
1761 self._unpack_cookie(cookie)
1762
1763 # Seek back to the safe start point.
1764 self.buffer.seek(start_pos)
1765 self._set_decoded_chars('')
1766 self._snapshot = None
1767
1768 # Restore the decoder to its state from the safe start point.
Benjamin Peterson9363a652009-03-05 00:42:09 +00001769 if cookie == 0 and self._decoder:
1770 self._decoder.reset()
1771 elif self._decoder or dec_flags or chars_to_skip:
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001772 self._decoder = self._decoder or self._get_decoder()
1773 self._decoder.setstate((b'', dec_flags))
1774 self._snapshot = (dec_flags, b'')
1775
1776 if chars_to_skip:
1777 # Just like _read_chunk, feed the decoder and save a snapshot.
1778 input_chunk = self.buffer.read(bytes_to_feed)
1779 self._set_decoded_chars(
1780 self._decoder.decode(input_chunk, need_eof))
1781 self._snapshot = (dec_flags, input_chunk)
1782
1783 # Skip chars_to_skip of the decoded characters.
1784 if len(self._decoded_chars) < chars_to_skip:
1785 raise IOError("can't restore logical file position")
1786 self._decoded_chars_used = chars_to_skip
1787
Antoine Pitroue4501852009-05-14 18:55:55 +00001788 # Finally, reset the encoder (merely useful for proper BOM handling)
1789 try:
1790 encoder = self._encoder or self._get_encoder()
1791 except LookupError:
1792 # Sometimes the encoder doesn't exist
1793 pass
1794 else:
1795 if cookie != 0:
1796 encoder.setstate(0)
1797 else:
1798 encoder.reset()
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001799 return cookie
1800
1801 def read(self, n=None):
Benjamin Petersona1b49012009-03-31 23:11:32 +00001802 self._checkReadable()
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001803 if n is None:
1804 n = -1
1805 decoder = self._decoder or self._get_decoder()
1806 if n < 0:
1807 # Read everything.
1808 result = (self._get_decoded_chars() +
1809 decoder.decode(self.buffer.read(), final=True))
1810 self._set_decoded_chars('')
1811 self._snapshot = None
1812 return result
1813 else:
1814 # Keep reading chunks until we have n characters to return.
1815 eof = False
1816 result = self._get_decoded_chars(n)
1817 while len(result) < n and not eof:
1818 eof = not self._read_chunk()
1819 result += self._get_decoded_chars(n - len(result))
1820 return result
1821
1822 def __next__(self):
1823 self._telling = False
1824 line = self.readline()
1825 if not line:
1826 self._snapshot = None
1827 self._telling = self._seekable
1828 raise StopIteration
1829 return line
1830
1831 def readline(self, limit=None):
1832 if self.closed:
1833 raise ValueError("read from closed file")
1834 if limit is None:
1835 limit = -1
Benjamin Petersonb01138a2009-04-24 22:59:52 +00001836 elif not isinstance(limit, int):
1837 raise TypeError("limit must be an integer")
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001838
1839 # Grab all the decoded text (we will rewind any extra bits later).
1840 line = self._get_decoded_chars()
1841
1842 start = 0
1843 # Make the decoder if it doesn't already exist.
1844 if not self._decoder:
1845 self._get_decoder()
1846
1847 pos = endpos = None
1848 while True:
1849 if self._readtranslate:
1850 # Newlines are already translated, only search for \n
1851 pos = line.find('\n', start)
1852 if pos >= 0:
1853 endpos = pos + 1
1854 break
1855 else:
1856 start = len(line)
1857
1858 elif self._readuniversal:
1859 # Universal newline search. Find any of \r, \r\n, \n
1860 # The decoder ensures that \r\n are not split in two pieces
1861
1862 # In C we'd look for these in parallel of course.
1863 nlpos = line.find("\n", start)
1864 crpos = line.find("\r", start)
1865 if crpos == -1:
1866 if nlpos == -1:
1867 # Nothing found
1868 start = len(line)
1869 else:
1870 # Found \n
1871 endpos = nlpos + 1
1872 break
1873 elif nlpos == -1:
1874 # Found lone \r
1875 endpos = crpos + 1
1876 break
1877 elif nlpos < crpos:
1878 # Found \n
1879 endpos = nlpos + 1
1880 break
1881 elif nlpos == crpos + 1:
1882 # Found \r\n
1883 endpos = crpos + 2
1884 break
1885 else:
1886 # Found \r
1887 endpos = crpos + 1
1888 break
1889 else:
1890 # non-universal
1891 pos = line.find(self._readnl)
1892 if pos >= 0:
1893 endpos = pos + len(self._readnl)
1894 break
1895
1896 if limit >= 0 and len(line) >= limit:
1897 endpos = limit # reached length limit
1898 break
1899
1900 # No line ending seen yet - get more data'
1901 while self._read_chunk():
1902 if self._decoded_chars:
1903 break
1904 if self._decoded_chars:
1905 line += self._get_decoded_chars()
1906 else:
1907 # end of file
1908 self._set_decoded_chars('')
1909 self._snapshot = None
1910 return line
1911
1912 if limit >= 0 and endpos > limit:
1913 endpos = limit # don't exceed limit
1914
1915 # Rewind _decoded_chars to just after the line ending we found.
1916 self._rewind_decoded_chars(len(line) - endpos)
1917 return line[:endpos]
1918
1919 @property
1920 def newlines(self):
1921 return self._decoder.newlines if self._decoder else None
1922
1923
1924class StringIO(TextIOWrapper):
1925 """Text I/O implementation using an in-memory buffer.
1926
1927 The initial_value argument sets the value of object. The newline
1928 argument is like the one of TextIOWrapper's constructor.
1929 """
1930
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001931 def __init__(self, initial_value="", newline="\n"):
1932 super(StringIO, self).__init__(BytesIO(),
1933 encoding="utf-8",
1934 errors="strict",
1935 newline=newline)
Antoine Pitrou11446482009-04-04 14:09:30 +00001936 # Issue #5645: make universal newlines semantics the same as in the
1937 # C version, even under Windows.
1938 if newline is None:
1939 self._writetranslate = False
Alexandre Vassalottid2bb18b2009-07-22 03:07:33 +00001940 if initial_value is not None:
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001941 if not isinstance(initial_value, str):
Alexandre Vassalottid2bb18b2009-07-22 03:07:33 +00001942 raise TypeError("initial_value must be str or None, not {0}"
1943 .format(type(initial_value).__name__))
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001944 initial_value = str(initial_value)
1945 self.write(initial_value)
1946 self.seek(0)
1947
1948 def getvalue(self):
1949 self.flush()
1950 return self.buffer.getvalue().decode(self._encoding, self._errors)
Benjamin Peterson9fd459a2009-03-09 00:09:44 +00001951
1952 def __repr__(self):
1953 # TextIOWrapper tells the encoding in its repr. In StringIO,
1954 # that's a implementation detail.
1955 return object.__repr__(self)
Benjamin Petersonb487e632009-03-21 03:08:31 +00001956
1957 @property
Benjamin Peterson0926ad12009-06-06 18:02:12 +00001958 def errors(self):
1959 return None
1960
1961 @property
Benjamin Petersonb487e632009-03-21 03:08:31 +00001962 def encoding(self):
1963 return None
Benjamin Petersond2e0c792009-05-01 20:40:59 +00001964
1965 def detach(self):
1966 # This doesn't make sense on StringIO.
1967 self._unsupported("detach")