blob: fa00eb4fb3e9f31cc893fa080a58295ab3d77b3e [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
Raymond Hettingerd2b03e12011-01-12 23:52:40 +000038def open(file, mode = "r", buffering = None,
39 encoding = None, errors = None,
40 newline = None, closefd = 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 Pitrou45a43722009-12-19 21:09:58 +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
Raymond Hettingerd2b03e12011-01-12 23:52:40 +0000287 def _unsupported(self, name):
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000288 """Internal: raise an exception for unsupported operations."""
289 raise UnsupportedOperation("%s.%s() not supported" %
290 (self.__class__.__name__, name))
291
292 ### Positioning ###
293
Raymond Hettingerd2b03e12011-01-12 23:52:40 +0000294 def seek(self, pos, whence = 0):
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000295 """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
Raymond Hettingerd2b03e12011-01-12 23:52:40 +0000309 def tell(self):
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000310 """Return current stream position."""
311 return self.seek(0, 1)
312
Raymond Hettingerd2b03e12011-01-12 23:52:40 +0000313 def truncate(self, pos = None):
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000314 """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
Raymond Hettingerd2b03e12011-01-12 23:52:40 +0000323 def flush(self):
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000324 """Flush write buffers, if applicable.
325
326 This is not implemented for read-only and non-blocking streams.
327 """
Antoine Pitroufaf90072010-05-03 16:58:19 +0000328 self._checkClosed()
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000329 # XXX Should this return the number of bytes written???
330
331 __closed = False
332
Raymond Hettingerd2b03e12011-01-12 23:52:40 +0000333 def close(self):
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000334 """Flush and close the IO object.
335
336 This method has no effect if the file is already closed.
337 """
338 if not self.__closed:
Antoine Pitroufaf90072010-05-03 16:58:19 +0000339 self.flush()
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000340 self.__closed = True
341
Raymond Hettingerd2b03e12011-01-12 23:52:40 +0000342 def __del__(self):
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000343 """Destructor. Calls close()."""
344 # The try/except block is in case this is called at program
345 # exit time, when it's possible that globals have already been
346 # deleted, and then the close() call might fail. Since
347 # there's nothing we can do about such failures and they annoy
348 # the end users, we suppress the traceback.
349 try:
350 self.close()
351 except:
352 pass
353
354 ### Inquiries ###
355
Raymond Hettingerd2b03e12011-01-12 23:52:40 +0000356 def seekable(self):
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000357 """Return whether object supports random access.
358
359 If False, seek(), tell() and truncate() will raise IOError.
360 This method may need to do a test seek().
361 """
362 return False
363
364 def _checkSeekable(self, msg=None):
365 """Internal: raise an IOError if file is not seekable
366 """
367 if not self.seekable():
368 raise IOError("File or stream is not seekable."
369 if msg is None else msg)
370
371
Raymond Hettingerd2b03e12011-01-12 23:52:40 +0000372 def readable(self):
373
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000374 """Return whether object was opened for reading.
375
376 If False, read() will raise IOError.
377 """
378 return False
379
380 def _checkReadable(self, msg=None):
381 """Internal: raise an IOError if file is not readable
382 """
383 if not self.readable():
384 raise IOError("File or stream is not readable."
385 if msg is None else msg)
386
Raymond Hettingerd2b03e12011-01-12 23:52:40 +0000387 def writable(self):
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000388 """Return whether object was opened for writing.
389
390 If False, write() and truncate() will raise IOError.
391 """
392 return False
393
394 def _checkWritable(self, msg=None):
395 """Internal: raise an IOError if file is not writable
396 """
397 if not self.writable():
398 raise IOError("File or stream is not writable."
399 if msg is None else msg)
400
401 @property
402 def closed(self):
403 """closed: bool. True iff the file has been closed.
404
405 For backwards compatibility, this is a property, not a predicate.
406 """
407 return self.__closed
408
409 def _checkClosed(self, msg=None):
410 """Internal: raise an ValueError if file is closed
411 """
412 if self.closed:
413 raise ValueError("I/O operation on closed file."
414 if msg is None else msg)
415
416 ### Context manager ###
417
Raymond Hettingerd2b03e12011-01-12 23:52:40 +0000418 def __enter__(self): # That's a forward reference
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000419 """Context management protocol. Returns self."""
420 self._checkClosed()
421 return self
422
Raymond Hettingerd2b03e12011-01-12 23:52:40 +0000423 def __exit__(self, *args):
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000424 """Context management protocol. Calls close()"""
425 self.close()
426
427 ### Lower-level APIs ###
428
429 # XXX Should these be present even if unimplemented?
430
Raymond Hettingerd2b03e12011-01-12 23:52:40 +0000431 def fileno(self):
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000432 """Returns underlying file descriptor if one exists.
433
434 An IOError is raised if the IO object does not use a file descriptor.
435 """
436 self._unsupported("fileno")
437
Raymond Hettingerd2b03e12011-01-12 23:52:40 +0000438 def isatty(self):
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000439 """Return whether this is an 'interactive' stream.
440
441 Return False if it can't be determined.
442 """
443 self._checkClosed()
444 return False
445
446 ### Readline[s] and writelines ###
447
Raymond Hettingerd2b03e12011-01-12 23:52:40 +0000448 def readline(self, limit = -1):
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000449 r"""Read and return a line from the stream.
450
451 If limit is specified, at most limit bytes will be read.
452
453 The line terminator is always b'\n' for binary files; for text
454 files, the newlines argument to open can be used to select the line
455 terminator(s) recognized.
456 """
457 # For backwards compatibility, a (slowish) readline().
458 if hasattr(self, "peek"):
459 def nreadahead():
460 readahead = self.peek(1)
461 if not readahead:
462 return 1
463 n = (readahead.find(b"\n") + 1) or len(readahead)
464 if limit >= 0:
465 n = min(n, limit)
466 return n
467 else:
468 def nreadahead():
469 return 1
470 if limit is None:
471 limit = -1
Benjamin Petersonb01138a2009-04-24 22:59:52 +0000472 elif not isinstance(limit, int):
473 raise TypeError("limit must be an integer")
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000474 res = bytearray()
475 while limit < 0 or len(res) < limit:
476 b = self.read(nreadahead())
477 if not b:
478 break
479 res += b
480 if res.endswith(b"\n"):
481 break
482 return bytes(res)
483
484 def __iter__(self):
485 self._checkClosed()
486 return self
487
488 def __next__(self):
489 line = self.readline()
490 if not line:
491 raise StopIteration
492 return line
493
494 def readlines(self, hint=None):
495 """Return a list of lines from the stream.
496
497 hint can be specified to control the number of lines read: no more
498 lines will be read if the total size (in bytes/characters) of all
499 lines so far exceeds hint.
500 """
501 if hint is None or hint <= 0:
502 return list(self)
503 n = 0
504 lines = []
505 for line in self:
506 lines.append(line)
507 n += len(line)
508 if n >= hint:
509 break
510 return lines
511
512 def writelines(self, lines):
513 self._checkClosed()
514 for line in lines:
515 self.write(line)
516
517io.IOBase.register(IOBase)
518
519
520class RawIOBase(IOBase):
521
522 """Base class for raw binary I/O."""
523
524 # The read() method is implemented by calling readinto(); derived
525 # classes that want to support read() only need to implement
526 # readinto() as a primitive operation. In general, readinto() can be
527 # more efficient than read().
528
529 # (It would be tempting to also provide an implementation of
530 # readinto() in terms of read(), in case the latter is a more suitable
531 # primitive operation, but that would lead to nasty recursion in case
532 # a subclass doesn't implement either.)
533
Raymond Hettingerd2b03e12011-01-12 23:52:40 +0000534 def read(self, n = -1):
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000535 """Read and return up to n bytes.
536
537 Returns an empty bytes object on EOF, or None if the object is
538 set not to block and has no data to read.
539 """
540 if n is None:
541 n = -1
542 if n < 0:
543 return self.readall()
544 b = bytearray(n.__index__())
545 n = self.readinto(b)
Antoine Pitroue5e75c62010-09-14 18:53:07 +0000546 if n is None:
547 return None
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000548 del b[n:]
549 return bytes(b)
550
551 def readall(self):
552 """Read until EOF, using multiple read() call."""
553 res = bytearray()
554 while True:
555 data = self.read(DEFAULT_BUFFER_SIZE)
556 if not data:
557 break
558 res += data
559 return bytes(res)
560
Raymond Hettingerd2b03e12011-01-12 23:52:40 +0000561 def readinto(self, b):
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000562 """Read up to len(b) bytes into b.
563
564 Returns number of bytes read (0 for EOF), or None if the object
Antoine Pitroue5e75c62010-09-14 18:53:07 +0000565 is set not to block and has no data to read.
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000566 """
567 self._unsupported("readinto")
568
Raymond Hettingerd2b03e12011-01-12 23:52:40 +0000569 def write(self, b):
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000570 """Write the given buffer to the IO stream.
571
572 Returns the number of bytes written, which may be less than len(b).
573 """
574 self._unsupported("write")
575
576io.RawIOBase.register(RawIOBase)
577from _io import FileIO
578RawIOBase.register(FileIO)
579
580
581class BufferedIOBase(IOBase):
582
583 """Base class for buffered IO objects.
584
585 The main difference with RawIOBase is that the read() method
586 supports omitting the size argument, and does not have a default
587 implementation that defers to readinto().
588
589 In addition, read(), readinto() and write() may raise
590 BlockingIOError if the underlying raw stream is in non-blocking
591 mode and not ready; unlike their raw counterparts, they will never
592 return None.
593
594 A typical implementation should not inherit from a RawIOBase
595 implementation, but wrap one.
596 """
597
Raymond Hettingerd2b03e12011-01-12 23:52:40 +0000598 def read(self, n = None):
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000599 """Read and return up to n bytes.
600
601 If the argument is omitted, None, or negative, reads and
602 returns all data until EOF.
603
604 If the argument is positive, and the underlying raw stream is
605 not 'interactive', multiple raw reads may be issued to satisfy
606 the byte count (unless EOF is reached first). But for
607 interactive raw streams (XXX and for pipes?), at most one raw
608 read will be issued, and a short result does not imply that
609 EOF is imminent.
610
611 Returns an empty bytes array on EOF.
612
613 Raises BlockingIOError if the underlying raw stream has no
614 data at the moment.
615 """
616 self._unsupported("read")
617
Raymond Hettingerd2b03e12011-01-12 23:52:40 +0000618 def read1(self, n = None):
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000619 """Read up to n bytes with at most one read() system call."""
620 self._unsupported("read1")
621
Raymond Hettingerd2b03e12011-01-12 23:52:40 +0000622 def readinto(self, b):
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000623 """Read up to len(b) bytes into b.
624
625 Like read(), this may issue multiple reads to the underlying raw
626 stream, unless the latter is 'interactive'.
627
628 Returns the number of bytes read (0 for EOF).
629
630 Raises BlockingIOError if the underlying raw stream has no
631 data at the moment.
632 """
633 # XXX This ought to work with anything that supports the buffer API
634 data = self.read(len(b))
635 n = len(data)
636 try:
637 b[:n] = data
638 except TypeError as err:
639 import array
640 if not isinstance(b, array.array):
641 raise err
642 b[:n] = array.array('b', data)
643 return n
644
Raymond Hettingerd2b03e12011-01-12 23:52:40 +0000645 def write(self, b):
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000646 """Write the given buffer to the IO stream.
647
648 Return the number of bytes written, which is never less than
649 len(b).
650
651 Raises BlockingIOError if the buffer is full and the
652 underlying raw stream cannot accept more data at the moment.
653 """
654 self._unsupported("write")
655
Raymond Hettingerd2b03e12011-01-12 23:52:40 +0000656 def detach(self):
Benjamin Petersond2e0c792009-05-01 20:40:59 +0000657 """
658 Separate the underlying raw stream from the buffer and return it.
659
660 After the raw stream has been detached, the buffer is in an unusable
661 state.
662 """
663 self._unsupported("detach")
664
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000665io.BufferedIOBase.register(BufferedIOBase)
666
667
668class _BufferedIOMixin(BufferedIOBase):
669
670 """A mixin implementation of BufferedIOBase with an underlying raw stream.
671
672 This passes most requests on to the underlying raw stream. It
673 does *not* provide implementations of read(), readinto() or
674 write().
675 """
676
677 def __init__(self, raw):
Antoine Pitrou6cfc5122010-12-21 21:26:09 +0000678 self._raw = raw
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000679
680 ### Positioning ###
681
682 def seek(self, pos, whence=0):
683 new_position = self.raw.seek(pos, whence)
684 if new_position < 0:
685 raise IOError("seek() returned an invalid position")
686 return new_position
687
688 def tell(self):
689 pos = self.raw.tell()
690 if pos < 0:
691 raise IOError("tell() returned an invalid position")
692 return pos
693
694 def truncate(self, pos=None):
695 # Flush the stream. We're mixing buffered I/O with lower-level I/O,
696 # and a flush may be necessary to synch both views of the current
697 # file state.
698 self.flush()
699
700 if pos is None:
701 pos = self.tell()
702 # XXX: Should seek() be used, instead of passing the position
703 # XXX directly to truncate?
704 return self.raw.truncate(pos)
705
706 ### Flush and close ###
707
708 def flush(self):
Antoine Pitroufaf90072010-05-03 16:58:19 +0000709 if self.closed:
710 raise ValueError("flush of closed file")
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000711 self.raw.flush()
712
713 def close(self):
Antoine Pitroufaf90072010-05-03 16:58:19 +0000714 if self.raw is not None and not self.closed:
715 self.flush()
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000716 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()
Antoine Pitrou6cfc5122010-12-21 21:26:09 +0000722 raw = self._raw
723 self._raw = None
Benjamin Petersond2e0c792009-05-01 20:40:59 +0000724 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
Antoine Pitrou6cfc5122010-12-21 21:26:09 +0000738 def raw(self):
739 return self._raw
740
741 @property
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000742 def closed(self):
743 return self.raw.closed
744
745 @property
746 def name(self):
747 return self.raw.name
748
749 @property
750 def mode(self):
751 return self.raw.mode
752
Antoine Pitrou716c4442009-05-23 19:04:03 +0000753 def __repr__(self):
754 clsname = self.__class__.__name__
755 try:
756 name = self.name
757 except AttributeError:
758 return "<_pyio.{0}>".format(clsname)
759 else:
760 return "<_pyio.{0} name={1!r}>".format(clsname, name)
761
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000762 ### Lower-level APIs ###
763
764 def fileno(self):
765 return self.raw.fileno()
766
767 def isatty(self):
768 return self.raw.isatty()
769
770
771class BytesIO(BufferedIOBase):
772
773 """Buffered I/O implementation using an in-memory bytes buffer."""
774
775 def __init__(self, initial_bytes=None):
776 buf = bytearray()
777 if initial_bytes is not None:
778 buf += initial_bytes
779 self._buffer = buf
780 self._pos = 0
781
782 def getvalue(self):
783 """Return the bytes value (contents) of the buffer
784 """
785 if self.closed:
786 raise ValueError("getvalue on closed file")
787 return bytes(self._buffer)
788
789 def read(self, n=None):
790 if self.closed:
791 raise ValueError("read from closed file")
792 if n is None:
793 n = -1
794 if n < 0:
795 n = len(self._buffer)
796 if len(self._buffer) <= self._pos:
797 return b""
798 newpos = min(len(self._buffer), self._pos + n)
799 b = self._buffer[self._pos : newpos]
800 self._pos = newpos
801 return bytes(b)
802
803 def read1(self, n):
804 """This is the same as read.
805 """
806 return self.read(n)
807
808 def write(self, b):
809 if self.closed:
810 raise ValueError("write to closed file")
811 if isinstance(b, str):
812 raise TypeError("can't write str to binary stream")
813 n = len(b)
814 if n == 0:
815 return 0
816 pos = self._pos
817 if pos > len(self._buffer):
818 # Inserts null bytes between the current end of the file
819 # and the new write position.
820 padding = b'\x00' * (pos - len(self._buffer))
821 self._buffer += padding
822 self._buffer[pos:pos + n] = b
823 self._pos += n
824 return n
825
826 def seek(self, pos, whence=0):
827 if self.closed:
828 raise ValueError("seek on closed file")
829 try:
830 pos = pos.__index__()
831 except AttributeError as err:
832 raise TypeError("an integer is required") from err
833 if whence == 0:
834 if pos < 0:
835 raise ValueError("negative seek position %r" % (pos,))
836 self._pos = pos
837 elif whence == 1:
838 self._pos = max(0, self._pos + pos)
839 elif whence == 2:
840 self._pos = max(0, len(self._buffer) + pos)
841 else:
842 raise ValueError("invalid whence value")
843 return self._pos
844
845 def tell(self):
846 if self.closed:
847 raise ValueError("tell on closed file")
848 return self._pos
849
850 def truncate(self, pos=None):
851 if self.closed:
852 raise ValueError("truncate on closed file")
853 if pos is None:
854 pos = self._pos
855 elif pos < 0:
856 raise ValueError("negative truncate position %r" % (pos,))
857 del self._buffer[pos:]
Antoine Pitrou66f9fea2010-01-31 23:20:26 +0000858 return pos
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000859
860 def readable(self):
861 return True
862
863 def writable(self):
864 return True
865
866 def seekable(self):
867 return True
868
869
870class BufferedReader(_BufferedIOMixin):
871
872 """BufferedReader(raw[, buffer_size])
873
874 A buffer for a readable, sequential BaseRawIO object.
875
876 The constructor creates a BufferedReader for the given readable raw
877 stream and buffer_size. If buffer_size is omitted, DEFAULT_BUFFER_SIZE
878 is used.
879 """
880
881 def __init__(self, raw, buffer_size=DEFAULT_BUFFER_SIZE):
882 """Create a new buffered reader using the given readable raw IO object.
883 """
Antoine Pitroucf4c7492009-04-19 00:09:36 +0000884 if not raw.readable():
885 raise IOError('"raw" argument must be readable.')
886
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000887 _BufferedIOMixin.__init__(self, raw)
888 if buffer_size <= 0:
889 raise ValueError("invalid buffer size")
890 self.buffer_size = buffer_size
891 self._reset_read_buf()
892 self._read_lock = Lock()
893
894 def _reset_read_buf(self):
895 self._read_buf = b""
896 self._read_pos = 0
897
898 def read(self, n=None):
899 """Read n bytes.
900
901 Returns exactly n bytes of data unless the underlying raw IO
902 stream reaches EOF or if the call would block in non-blocking
903 mode. If n is negative, read until EOF or until read() would
904 block.
905 """
906 if n is not None and n < -1:
907 raise ValueError("invalid number of bytes to read")
908 with self._read_lock:
909 return self._read_unlocked(n)
910
911 def _read_unlocked(self, n=None):
912 nodata_val = b""
913 empty_values = (b"", None)
914 buf = self._read_buf
915 pos = self._read_pos
916
917 # Special case for when the number of bytes to read is unspecified.
918 if n is None or n == -1:
919 self._reset_read_buf()
920 chunks = [buf[pos:]] # Strip the consumed bytes.
921 current_size = 0
922 while True:
923 # Read until EOF or until read() would block.
924 chunk = self.raw.read()
925 if chunk in empty_values:
926 nodata_val = chunk
927 break
928 current_size += len(chunk)
929 chunks.append(chunk)
930 return b"".join(chunks) or nodata_val
931
932 # The number of bytes to read is specified, return at most n bytes.
933 avail = len(buf) - pos # Length of the available buffered data.
934 if n <= avail:
935 # Fast path: the data to read is fully buffered.
936 self._read_pos += n
937 return buf[pos:pos+n]
938 # Slow path: read from the stream until enough bytes are read,
939 # or until an EOF occurs or until read() would block.
940 chunks = [buf[pos:]]
941 wanted = max(self.buffer_size, n)
942 while avail < n:
943 chunk = self.raw.read(wanted)
944 if chunk in empty_values:
945 nodata_val = chunk
946 break
947 avail += len(chunk)
948 chunks.append(chunk)
949 # n is more then avail only when an EOF occurred or when
950 # read() would have blocked.
951 n = min(n, avail)
952 out = b"".join(chunks)
953 self._read_buf = out[n:] # Save the extra data in the buffer.
954 self._read_pos = 0
955 return out[:n] if out else nodata_val
956
957 def peek(self, n=0):
958 """Returns buffered bytes without advancing the position.
959
960 The argument indicates a desired minimal number of bytes; we
961 do at most one raw read to satisfy it. We never return more
962 than self.buffer_size.
963 """
964 with self._read_lock:
965 return self._peek_unlocked(n)
966
967 def _peek_unlocked(self, n=0):
968 want = min(n, self.buffer_size)
969 have = len(self._read_buf) - self._read_pos
970 if have < want or have <= 0:
971 to_read = self.buffer_size - have
972 current = self.raw.read(to_read)
973 if current:
974 self._read_buf = self._read_buf[self._read_pos:] + current
975 self._read_pos = 0
976 return self._read_buf[self._read_pos:]
977
978 def read1(self, n):
979 """Reads up to n bytes, with at most one read() system call."""
980 # Returns up to n bytes. If at least one byte is buffered, we
981 # only return buffered bytes. Otherwise, we do one raw read.
982 if n < 0:
983 raise ValueError("number of bytes to read must be positive")
984 if n == 0:
985 return b""
986 with self._read_lock:
987 self._peek_unlocked(1)
988 return self._read_unlocked(
989 min(n, len(self._read_buf) - self._read_pos))
990
991 def tell(self):
992 return _BufferedIOMixin.tell(self) - len(self._read_buf) + self._read_pos
993
994 def seek(self, pos, whence=0):
995 if not (0 <= whence <= 2):
996 raise ValueError("invalid whence value")
997 with self._read_lock:
998 if whence == 1:
999 pos -= len(self._read_buf) - self._read_pos
1000 pos = _BufferedIOMixin.seek(self, pos, whence)
1001 self._reset_read_buf()
1002 return pos
1003
1004class BufferedWriter(_BufferedIOMixin):
1005
1006 """A buffer for a writeable sequential RawIO object.
1007
1008 The constructor creates a BufferedWriter for the given writeable raw
1009 stream. If the buffer_size is not given, it defaults to
Benjamin Peterson59406a92009-03-26 17:10:29 +00001010 DEFAULT_BUFFER_SIZE.
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001011 """
1012
Benjamin Peterson59406a92009-03-26 17:10:29 +00001013 _warning_stack_offset = 2
1014
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001015 def __init__(self, raw,
1016 buffer_size=DEFAULT_BUFFER_SIZE, max_buffer_size=None):
Antoine Pitroucf4c7492009-04-19 00:09:36 +00001017 if not raw.writable():
1018 raise IOError('"raw" argument must be writable.')
1019
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001020 _BufferedIOMixin.__init__(self, raw)
1021 if buffer_size <= 0:
1022 raise ValueError("invalid buffer size")
Benjamin Peterson59406a92009-03-26 17:10:29 +00001023 if max_buffer_size is not None:
1024 warnings.warn("max_buffer_size is deprecated", DeprecationWarning,
1025 self._warning_stack_offset)
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001026 self.buffer_size = buffer_size
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001027 self._write_buf = bytearray()
1028 self._write_lock = Lock()
1029
1030 def write(self, b):
1031 if self.closed:
1032 raise ValueError("write to closed file")
1033 if isinstance(b, str):
1034 raise TypeError("can't write str to binary stream")
1035 with self._write_lock:
1036 # XXX we can implement some more tricks to try and avoid
1037 # partial writes
1038 if len(self._write_buf) > self.buffer_size:
1039 # We're full, so let's pre-flush the buffer
1040 try:
1041 self._flush_unlocked()
1042 except BlockingIOError as e:
1043 # We can't accept anything else.
1044 # XXX Why not just let the exception pass through?
1045 raise BlockingIOError(e.errno, e.strerror, 0)
1046 before = len(self._write_buf)
1047 self._write_buf.extend(b)
1048 written = len(self._write_buf) - before
1049 if len(self._write_buf) > self.buffer_size:
1050 try:
1051 self._flush_unlocked()
1052 except BlockingIOError as e:
Benjamin Peterson394ee002009-03-05 22:33:59 +00001053 if len(self._write_buf) > self.buffer_size:
1054 # We've hit the buffer_size. We have to accept a partial
1055 # write and cut back our buffer.
1056 overage = len(self._write_buf) - self.buffer_size
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001057 written -= overage
Benjamin Peterson394ee002009-03-05 22:33:59 +00001058 self._write_buf = self._write_buf[:self.buffer_size]
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001059 raise BlockingIOError(e.errno, e.strerror, written)
1060 return written
1061
1062 def truncate(self, pos=None):
1063 with self._write_lock:
1064 self._flush_unlocked()
1065 if pos is None:
1066 pos = self.raw.tell()
1067 return self.raw.truncate(pos)
1068
1069 def flush(self):
1070 with self._write_lock:
1071 self._flush_unlocked()
1072
1073 def _flush_unlocked(self):
1074 if self.closed:
1075 raise ValueError("flush of closed file")
1076 written = 0
1077 try:
1078 while self._write_buf:
1079 n = self.raw.write(self._write_buf)
1080 if n > len(self._write_buf) or n < 0:
1081 raise IOError("write() returned incorrect number of bytes")
1082 del self._write_buf[:n]
1083 written += n
1084 except BlockingIOError as e:
1085 n = e.characters_written
1086 del self._write_buf[:n]
1087 written += n
1088 raise BlockingIOError(e.errno, e.strerror, written)
1089
1090 def tell(self):
1091 return _BufferedIOMixin.tell(self) + len(self._write_buf)
1092
1093 def seek(self, pos, whence=0):
1094 if not (0 <= whence <= 2):
1095 raise ValueError("invalid whence")
1096 with self._write_lock:
1097 self._flush_unlocked()
1098 return _BufferedIOMixin.seek(self, pos, whence)
1099
1100
1101class BufferedRWPair(BufferedIOBase):
1102
1103 """A buffered reader and writer object together.
1104
1105 A buffered reader object and buffered writer object put together to
1106 form a sequential IO object that can read and write. This is typically
1107 used with a socket or two-way pipe.
1108
1109 reader and writer are RawIOBase objects that are readable and
1110 writeable respectively. If the buffer_size is omitted it defaults to
Benjamin Peterson59406a92009-03-26 17:10:29 +00001111 DEFAULT_BUFFER_SIZE.
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001112 """
1113
1114 # XXX The usefulness of this (compared to having two separate IO
1115 # objects) is questionable.
1116
1117 def __init__(self, reader, writer,
1118 buffer_size=DEFAULT_BUFFER_SIZE, max_buffer_size=None):
1119 """Constructor.
1120
1121 The arguments are two RawIO instances.
1122 """
Benjamin Peterson59406a92009-03-26 17:10:29 +00001123 if max_buffer_size is not None:
1124 warnings.warn("max_buffer_size is deprecated", DeprecationWarning, 2)
Antoine Pitroucf4c7492009-04-19 00:09:36 +00001125
1126 if not reader.readable():
1127 raise IOError('"reader" argument must be readable.')
1128
1129 if not writer.writable():
1130 raise IOError('"writer" argument must be writable.')
1131
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001132 self.reader = BufferedReader(reader, buffer_size)
Benjamin Peterson59406a92009-03-26 17:10:29 +00001133 self.writer = BufferedWriter(writer, buffer_size)
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001134
1135 def read(self, n=None):
1136 if n is None:
1137 n = -1
1138 return self.reader.read(n)
1139
1140 def readinto(self, b):
1141 return self.reader.readinto(b)
1142
1143 def write(self, b):
1144 return self.writer.write(b)
1145
1146 def peek(self, n=0):
1147 return self.reader.peek(n)
1148
1149 def read1(self, n):
1150 return self.reader.read1(n)
1151
1152 def readable(self):
1153 return self.reader.readable()
1154
1155 def writable(self):
1156 return self.writer.writable()
1157
1158 def flush(self):
1159 return self.writer.flush()
1160
1161 def close(self):
1162 self.writer.close()
1163 self.reader.close()
1164
1165 def isatty(self):
1166 return self.reader.isatty() or self.writer.isatty()
1167
1168 @property
1169 def closed(self):
1170 return self.writer.closed
1171
1172
1173class BufferedRandom(BufferedWriter, BufferedReader):
1174
1175 """A buffered interface to random access streams.
1176
1177 The constructor creates a reader and writer for a seekable stream,
1178 raw, given in the first argument. If the buffer_size is omitted it
Benjamin Peterson59406a92009-03-26 17:10:29 +00001179 defaults to DEFAULT_BUFFER_SIZE.
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001180 """
1181
Benjamin Peterson59406a92009-03-26 17:10:29 +00001182 _warning_stack_offset = 3
1183
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001184 def __init__(self, raw,
1185 buffer_size=DEFAULT_BUFFER_SIZE, max_buffer_size=None):
1186 raw._checkSeekable()
1187 BufferedReader.__init__(self, raw, buffer_size)
1188 BufferedWriter.__init__(self, raw, buffer_size, max_buffer_size)
1189
1190 def seek(self, pos, whence=0):
1191 if not (0 <= whence <= 2):
1192 raise ValueError("invalid whence")
1193 self.flush()
1194 if self._read_buf:
1195 # Undo read ahead.
1196 with self._read_lock:
1197 self.raw.seek(self._read_pos - len(self._read_buf), 1)
1198 # First do the raw seek, then empty the read buffer, so that
1199 # if the raw seek fails, we don't lose buffered data forever.
1200 pos = self.raw.seek(pos, whence)
1201 with self._read_lock:
1202 self._reset_read_buf()
1203 if pos < 0:
1204 raise IOError("seek() returned invalid position")
1205 return pos
1206
1207 def tell(self):
1208 if self._write_buf:
1209 return BufferedWriter.tell(self)
1210 else:
1211 return BufferedReader.tell(self)
1212
1213 def truncate(self, pos=None):
1214 if pos is None:
1215 pos = self.tell()
1216 # Use seek to flush the read buffer.
Antoine Pitrou66f9fea2010-01-31 23:20:26 +00001217 return BufferedWriter.truncate(self, pos)
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001218
1219 def read(self, n=None):
1220 if n is None:
1221 n = -1
1222 self.flush()
1223 return BufferedReader.read(self, n)
1224
1225 def readinto(self, b):
1226 self.flush()
1227 return BufferedReader.readinto(self, b)
1228
1229 def peek(self, n=0):
1230 self.flush()
1231 return BufferedReader.peek(self, n)
1232
1233 def read1(self, n):
1234 self.flush()
1235 return BufferedReader.read1(self, n)
1236
1237 def write(self, b):
1238 if self._read_buf:
1239 # Undo readahead
1240 with self._read_lock:
1241 self.raw.seek(self._read_pos - len(self._read_buf), 1)
1242 self._reset_read_buf()
1243 return BufferedWriter.write(self, b)
1244
1245
1246class TextIOBase(IOBase):
1247
1248 """Base class for text I/O.
1249
1250 This class provides a character and line based interface to stream
1251 I/O. There is no readinto method because Python's character strings
1252 are immutable. There is no public constructor.
1253 """
1254
Raymond Hettingerd2b03e12011-01-12 23:52:40 +00001255 def read(self, n = -1):
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001256 """Read at most n characters from stream.
1257
1258 Read from underlying buffer until we have n characters or we hit EOF.
1259 If n is negative or omitted, read until EOF.
1260 """
1261 self._unsupported("read")
1262
Raymond Hettingerd2b03e12011-01-12 23:52:40 +00001263 def write(self, s):
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001264 """Write string s to stream."""
1265 self._unsupported("write")
1266
Raymond Hettingerd2b03e12011-01-12 23:52:40 +00001267 def truncate(self, pos = None):
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001268 """Truncate size to pos."""
1269 self._unsupported("truncate")
1270
Raymond Hettingerd2b03e12011-01-12 23:52:40 +00001271 def readline(self):
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001272 """Read until newline or EOF.
1273
1274 Returns an empty string if EOF is hit immediately.
1275 """
1276 self._unsupported("readline")
1277
Raymond Hettingerd2b03e12011-01-12 23:52:40 +00001278 def detach(self):
Benjamin Petersond2e0c792009-05-01 20:40:59 +00001279 """
1280 Separate the underlying buffer from the TextIOBase and return it.
1281
1282 After the underlying buffer has been detached, the TextIO is in an
1283 unusable state.
1284 """
1285 self._unsupported("detach")
1286
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001287 @property
1288 def encoding(self):
1289 """Subclasses should override."""
1290 return None
1291
1292 @property
1293 def newlines(self):
1294 """Line endings translated so far.
1295
1296 Only line endings translated during reading are considered.
1297
1298 Subclasses should override.
1299 """
1300 return None
1301
Benjamin Peterson0926ad12009-06-06 18:02:12 +00001302 @property
1303 def errors(self):
1304 """Error setting of the decoder or encoder.
1305
1306 Subclasses should override."""
1307 return None
1308
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001309io.TextIOBase.register(TextIOBase)
1310
1311
1312class IncrementalNewlineDecoder(codecs.IncrementalDecoder):
1313 r"""Codec used when reading a file in universal newlines mode. It wraps
1314 another incremental decoder, translating \r\n and \r into \n. It also
1315 records the types of newlines encountered. When used with
1316 translate=False, it ensures that the newline sequence is returned in
1317 one piece.
1318 """
1319 def __init__(self, decoder, translate, errors='strict'):
1320 codecs.IncrementalDecoder.__init__(self, errors=errors)
1321 self.translate = translate
1322 self.decoder = decoder
1323 self.seennl = 0
1324 self.pendingcr = False
1325
1326 def decode(self, input, final=False):
1327 # decode input (with the eventual \r from a previous pass)
1328 if self.decoder is None:
1329 output = input
1330 else:
1331 output = self.decoder.decode(input, final=final)
1332 if self.pendingcr and (output or final):
1333 output = "\r" + output
1334 self.pendingcr = False
1335
1336 # retain last \r even when not translating data:
1337 # then readline() is sure to get \r\n in one pass
1338 if output.endswith("\r") and not final:
1339 output = output[:-1]
1340 self.pendingcr = True
1341
1342 # Record which newlines are read
1343 crlf = output.count('\r\n')
1344 cr = output.count('\r') - crlf
1345 lf = output.count('\n') - crlf
1346 self.seennl |= (lf and self._LF) | (cr and self._CR) \
1347 | (crlf and self._CRLF)
1348
1349 if self.translate:
1350 if crlf:
1351 output = output.replace("\r\n", "\n")
1352 if cr:
1353 output = output.replace("\r", "\n")
1354
1355 return output
1356
1357 def getstate(self):
1358 if self.decoder is None:
1359 buf = b""
1360 flag = 0
1361 else:
1362 buf, flag = self.decoder.getstate()
1363 flag <<= 1
1364 if self.pendingcr:
1365 flag |= 1
1366 return buf, flag
1367
1368 def setstate(self, state):
1369 buf, flag = state
1370 self.pendingcr = bool(flag & 1)
1371 if self.decoder is not None:
1372 self.decoder.setstate((buf, flag >> 1))
1373
1374 def reset(self):
1375 self.seennl = 0
1376 self.pendingcr = False
1377 if self.decoder is not None:
1378 self.decoder.reset()
1379
1380 _LF = 1
1381 _CR = 2
1382 _CRLF = 4
1383
1384 @property
1385 def newlines(self):
1386 return (None,
1387 "\n",
1388 "\r",
1389 ("\r", "\n"),
1390 "\r\n",
1391 ("\n", "\r\n"),
1392 ("\r", "\r\n"),
1393 ("\r", "\n", "\r\n")
1394 )[self.seennl]
1395
1396
1397class TextIOWrapper(TextIOBase):
1398
1399 r"""Character and line based layer over a BufferedIOBase object, buffer.
1400
1401 encoding gives the name of the encoding that the stream will be
1402 decoded or encoded with. It defaults to locale.getpreferredencoding.
1403
1404 errors determines the strictness of encoding and decoding (see the
1405 codecs.register) and defaults to "strict".
1406
1407 newline can be None, '', '\n', '\r', or '\r\n'. It controls the
1408 handling of line endings. If it is None, universal newlines is
1409 enabled. With this enabled, on input, the lines endings '\n', '\r',
1410 or '\r\n' are translated to '\n' before being returned to the
1411 caller. Conversely, on output, '\n' is translated to the system
1412 default line seperator, os.linesep. If newline is any other of its
1413 legal values, that newline becomes the newline when the file is read
1414 and it is returned untranslated. On output, '\n' is converted to the
1415 newline.
1416
1417 If line_buffering is True, a call to flush is implied when a call to
1418 write contains a newline character.
1419 """
1420
1421 _CHUNK_SIZE = 2048
1422
1423 def __init__(self, buffer, encoding=None, errors=None, newline=None,
1424 line_buffering=False):
1425 if newline is not None and not isinstance(newline, str):
1426 raise TypeError("illegal newline type: %r" % (type(newline),))
1427 if newline not in (None, "", "\n", "\r", "\r\n"):
1428 raise ValueError("illegal newline value: %r" % (newline,))
1429 if encoding is None:
1430 try:
1431 encoding = os.device_encoding(buffer.fileno())
1432 except (AttributeError, UnsupportedOperation):
1433 pass
1434 if encoding is None:
1435 try:
1436 import locale
1437 except ImportError:
1438 # Importing locale may fail if Python is being built
1439 encoding = "ascii"
1440 else:
1441 encoding = locale.getpreferredencoding()
1442
1443 if not isinstance(encoding, str):
1444 raise ValueError("invalid encoding: %r" % encoding)
1445
1446 if errors is None:
1447 errors = "strict"
1448 else:
1449 if not isinstance(errors, str):
1450 raise ValueError("invalid errors: %r" % errors)
1451
Antoine Pitrou6cfc5122010-12-21 21:26:09 +00001452 self._buffer = buffer
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001453 self._line_buffering = line_buffering
1454 self._encoding = encoding
1455 self._errors = errors
1456 self._readuniversal = not newline
1457 self._readtranslate = newline is None
1458 self._readnl = newline
1459 self._writetranslate = newline != ''
1460 self._writenl = newline or os.linesep
1461 self._encoder = None
1462 self._decoder = None
1463 self._decoded_chars = '' # buffer for text returned from decoder
1464 self._decoded_chars_used = 0 # offset into _decoded_chars for read()
1465 self._snapshot = None # info for reconstructing decoder state
1466 self._seekable = self._telling = self.buffer.seekable()
1467
Antoine Pitroue4501852009-05-14 18:55:55 +00001468 if self._seekable and self.writable():
1469 position = self.buffer.tell()
1470 if position != 0:
1471 try:
1472 self._get_encoder().setstate(0)
1473 except LookupError:
1474 # Sometimes the encoder doesn't exist
1475 pass
1476
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001477 # self._snapshot is either None, or a tuple (dec_flags, next_input)
1478 # where dec_flags is the second (integer) item of the decoder state
1479 # and next_input is the chunk of input bytes that comes next after the
1480 # snapshot point. We use this to reconstruct decoder states in tell().
1481
1482 # Naming convention:
1483 # - "bytes_..." for integer variables that count input bytes
1484 # - "chars_..." for integer variables that count decoded characters
1485
Benjamin Petersonc4c0eae2009-03-09 00:07:03 +00001486 def __repr__(self):
Antoine Pitrou716c4442009-05-23 19:04:03 +00001487 try:
1488 name = self.name
1489 except AttributeError:
1490 return "<_pyio.TextIOWrapper encoding={0!r}>".format(self.encoding)
1491 else:
1492 return "<_pyio.TextIOWrapper name={0!r} encoding={1!r}>".format(
1493 name, self.encoding)
Benjamin Petersonc4c0eae2009-03-09 00:07:03 +00001494
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001495 @property
1496 def encoding(self):
1497 return self._encoding
1498
1499 @property
1500 def errors(self):
1501 return self._errors
1502
1503 @property
1504 def line_buffering(self):
1505 return self._line_buffering
1506
Antoine Pitrou6cfc5122010-12-21 21:26:09 +00001507 @property
1508 def buffer(self):
1509 return self._buffer
1510
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001511 def seekable(self):
1512 return self._seekable
1513
1514 def readable(self):
1515 return self.buffer.readable()
1516
1517 def writable(self):
1518 return self.buffer.writable()
1519
1520 def flush(self):
1521 self.buffer.flush()
1522 self._telling = self._seekable
1523
1524 def close(self):
Antoine Pitroufaf90072010-05-03 16:58:19 +00001525 if self.buffer is not None and not self.closed:
1526 self.flush()
Benjamin Petersond2e0c792009-05-01 20:40:59 +00001527 self.buffer.close()
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001528
1529 @property
1530 def closed(self):
1531 return self.buffer.closed
1532
1533 @property
1534 def name(self):
1535 return self.buffer.name
1536
1537 def fileno(self):
1538 return self.buffer.fileno()
1539
1540 def isatty(self):
1541 return self.buffer.isatty()
1542
1543 def write(self, s: str):
1544 if self.closed:
1545 raise ValueError("write to closed file")
1546 if not isinstance(s, str):
1547 raise TypeError("can't write %s to text stream" %
1548 s.__class__.__name__)
1549 length = len(s)
1550 haslf = (self._writetranslate or self._line_buffering) and "\n" in s
1551 if haslf and self._writetranslate and self._writenl != "\n":
1552 s = s.replace("\n", self._writenl)
1553 encoder = self._encoder or self._get_encoder()
1554 # XXX What if we were just reading?
1555 b = encoder.encode(s)
1556 self.buffer.write(b)
1557 if self._line_buffering and (haslf or "\r" in s):
1558 self.flush()
1559 self._snapshot = None
1560 if self._decoder:
1561 self._decoder.reset()
1562 return length
1563
1564 def _get_encoder(self):
1565 make_encoder = codecs.getincrementalencoder(self._encoding)
1566 self._encoder = make_encoder(self._errors)
1567 return self._encoder
1568
1569 def _get_decoder(self):
1570 make_decoder = codecs.getincrementaldecoder(self._encoding)
1571 decoder = make_decoder(self._errors)
1572 if self._readuniversal:
1573 decoder = IncrementalNewlineDecoder(decoder, self._readtranslate)
1574 self._decoder = decoder
1575 return decoder
1576
1577 # The following three methods implement an ADT for _decoded_chars.
1578 # Text returned from the decoder is buffered here until the client
1579 # requests it by calling our read() or readline() method.
1580 def _set_decoded_chars(self, chars):
1581 """Set the _decoded_chars buffer."""
1582 self._decoded_chars = chars
1583 self._decoded_chars_used = 0
1584
1585 def _get_decoded_chars(self, n=None):
1586 """Advance into the _decoded_chars buffer."""
1587 offset = self._decoded_chars_used
1588 if n is None:
1589 chars = self._decoded_chars[offset:]
1590 else:
1591 chars = self._decoded_chars[offset:offset + n]
1592 self._decoded_chars_used += len(chars)
1593 return chars
1594
1595 def _rewind_decoded_chars(self, n):
1596 """Rewind the _decoded_chars buffer."""
1597 if self._decoded_chars_used < n:
1598 raise AssertionError("rewind decoded_chars out of bounds")
1599 self._decoded_chars_used -= n
1600
1601 def _read_chunk(self):
1602 """
1603 Read and decode the next chunk of data from the BufferedReader.
1604 """
1605
1606 # The return value is True unless EOF was reached. The decoded
1607 # string is placed in self._decoded_chars (replacing its previous
1608 # value). The entire input chunk is sent to the decoder, though
1609 # some of it may remain buffered in the decoder, yet to be
1610 # converted.
1611
1612 if self._decoder is None:
1613 raise ValueError("no decoder")
1614
1615 if self._telling:
1616 # To prepare for tell(), we need to snapshot a point in the
1617 # file where the decoder's input buffer is empty.
1618
1619 dec_buffer, dec_flags = self._decoder.getstate()
1620 # Given this, we know there was a valid snapshot point
1621 # len(dec_buffer) bytes ago with decoder state (b'', dec_flags).
1622
1623 # Read a chunk, decode it, and put the result in self._decoded_chars.
1624 input_chunk = self.buffer.read1(self._CHUNK_SIZE)
1625 eof = not input_chunk
1626 self._set_decoded_chars(self._decoder.decode(input_chunk, eof))
1627
1628 if self._telling:
1629 # At the snapshot point, len(dec_buffer) bytes before the read,
1630 # the next input to be decoded is dec_buffer + input_chunk.
1631 self._snapshot = (dec_flags, dec_buffer + input_chunk)
1632
1633 return not eof
1634
1635 def _pack_cookie(self, position, dec_flags=0,
1636 bytes_to_feed=0, need_eof=0, chars_to_skip=0):
1637 # The meaning of a tell() cookie is: seek to position, set the
1638 # decoder flags to dec_flags, read bytes_to_feed bytes, feed them
1639 # into the decoder with need_eof as the EOF flag, then skip
1640 # chars_to_skip characters of the decoded result. For most simple
1641 # decoders, tell() will often just give a byte offset in the file.
1642 return (position | (dec_flags<<64) | (bytes_to_feed<<128) |
1643 (chars_to_skip<<192) | bool(need_eof)<<256)
1644
1645 def _unpack_cookie(self, bigint):
1646 rest, position = divmod(bigint, 1<<64)
1647 rest, dec_flags = divmod(rest, 1<<64)
1648 rest, bytes_to_feed = divmod(rest, 1<<64)
1649 need_eof, chars_to_skip = divmod(rest, 1<<64)
1650 return position, dec_flags, bytes_to_feed, need_eof, chars_to_skip
1651
1652 def tell(self):
1653 if not self._seekable:
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 Pitrou66f9fea2010-01-31 23:20:26 +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()
Antoine Pitrou6cfc5122010-12-21 21:26:09 +00001725 buffer = self._buffer
1726 self._buffer = None
Benjamin Petersond2e0c792009-05-01 20:40:59 +00001727 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
Georg Brandl194da4a2009-08-13 09:34:05 +00001940 if initial_value is not None:
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001941 if not isinstance(initial_value, str):
Georg Brandl194da4a2009-08-13 09:34:05 +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")