blob: 3fa932563b3ff2d47c48b22765de65a830b18148 [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
Benjamin Petersonc3be11a2010-04-27 21:24:03 +000016from io import (__all__, SEEK_SET, SEEK_CUR, SEEK_END)
Antoine Pitroud843c2d2011-02-25 21:34:39 +000017from errno import EINTR
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
Georg Brandl4d73b572011-01-13 07:13:06 +000038def open(file, mode="r", buffering=-1, encoding=None, errors=None,
39 newline=None, closefd=True):
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +000040
41 r"""Open file and return a stream. Raise IOError upon failure.
42
43 file is either a text or byte string giving the name (and the path
44 if the file isn't in the current working directory) of the file to
45 be opened or an integer file descriptor of the file to be
46 wrapped. (If a file descriptor is given, it is closed when the
47 returned I/O object is closed, unless closefd is set to False.)
48
49 mode is an optional string that specifies the mode in which the file
50 is opened. It defaults to 'r' which means open for reading in text
51 mode. Other common values are 'w' for writing (truncating the file if
52 it already exists), and 'a' for appending (which on some Unix systems,
53 means that all writes append to the end of the file regardless of the
54 current seek position). In text mode, if encoding is not specified the
55 encoding used is platform dependent. (For reading and writing raw
56 bytes use binary mode and leave encoding unspecified.) The available
57 modes are:
58
59 ========= ===============================================================
60 Character Meaning
61 --------- ---------------------------------------------------------------
62 'r' open for reading (default)
63 'w' open for writing, truncating the file first
64 'a' open for writing, appending to the end of the file if it exists
65 'b' binary mode
66 't' text mode (default)
67 '+' open a disk file for updating (reading and writing)
68 'U' universal newline mode (for backwards compatibility; unneeded
69 for new code)
70 ========= ===============================================================
71
72 The default mode is 'rt' (open for reading text). For binary random
73 access, the mode 'w+b' opens and truncates the file to 0 bytes, while
74 'r+b' opens the file without truncation.
75
76 Python distinguishes between files opened in binary and text modes,
77 even when the underlying operating system doesn't. Files opened in
78 binary mode (appending 'b' to the mode argument) return contents as
79 bytes objects without any decoding. In text mode (the default, or when
80 't' is appended to the mode argument), the contents of the file are
81 returned as strings, the bytes having been first decoded using a
82 platform-dependent encoding or using the specified encoding if given.
83
Antoine Pitroud5587bc2009-12-19 21:08:31 +000084 buffering is an optional integer used to set the buffering policy.
85 Pass 0 to switch buffering off (only allowed in binary mode), 1 to select
86 line buffering (only usable in text mode), and an integer > 1 to indicate
87 the size of a fixed-size chunk buffer. When no buffering argument is
88 given, the default buffering policy works as follows:
89
90 * Binary files are buffered in fixed-size chunks; the size of the buffer
91 is chosen using a heuristic trying to determine the underlying device's
92 "block size" and falling back on `io.DEFAULT_BUFFER_SIZE`.
93 On many systems, the buffer will typically be 4096 or 8192 bytes long.
94
95 * "Interactive" text files (files for which isatty() returns True)
96 use line buffering. Other text files use the policy described above
97 for binary files.
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +000098
Raymond Hettingercbb80892011-01-13 18:15:51 +000099 encoding is the str name of the encoding used to decode or encode the
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000100 file. This should only be used in text mode. The default encoding is
101 platform dependent, but any encoding supported by Python can be
102 passed. See the codecs module for the list of supported encodings.
103
104 errors is an optional string that specifies how encoding errors are to
105 be handled---this argument should not be used in binary mode. Pass
106 'strict' to raise a ValueError exception if there is an encoding error
107 (the default of None has the same effect), or pass 'ignore' to ignore
108 errors. (Note that ignoring encoding errors can lead to data loss.)
109 See the documentation for codecs.register for a list of the permitted
110 encoding error strings.
111
Raymond Hettingercbb80892011-01-13 18:15:51 +0000112 newline is a string controlling how universal newlines works (it only
113 applies to text mode). It can be None, '', '\n', '\r', and '\r\n'. It works
114 as follows:
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000115
116 * On input, if newline is None, universal newlines mode is
117 enabled. Lines in the input can end in '\n', '\r', or '\r\n', and
118 these are translated into '\n' before being returned to the
119 caller. If it is '', universal newline mode is enabled, but line
120 endings are returned to the caller untranslated. If it has any of
121 the other legal values, input lines are only terminated by the given
122 string, and the line ending is returned to the caller untranslated.
123
124 * On output, if newline is None, any '\n' characters written are
125 translated to the system default line separator, os.linesep. If
126 newline is '', no translation takes place. If newline is any of the
127 other legal values, any '\n' characters written are translated to
128 the given string.
129
Raymond Hettingercbb80892011-01-13 18:15:51 +0000130 closedfd is a bool. If closefd is False, the underlying file descriptor will
131 be kept open when the file is closed. This does not work when a file name is
132 given and must be True in that case.
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000133
134 open() returns a file object whose type depends on the mode, and
135 through which the standard file operations such as reading and writing
136 are performed. When open() is used to open a file in a text mode ('w',
137 'r', 'wt', 'rt', etc.), it returns a TextIOWrapper. When used to open
138 a file in a binary mode, the returned class varies: in read binary
139 mode, it returns a BufferedReader; in write binary and append binary
140 modes, it returns a BufferedWriter, and in read/write mode, it returns
141 a BufferedRandom.
142
143 It is also possible to use a string or bytearray as a file for both
144 reading and writing. For strings StringIO can be used like a file
145 opened in a text mode, and for bytes a BytesIO can be used like a file
146 opened in a binary mode.
147 """
148 if not isinstance(file, (str, bytes, int)):
149 raise TypeError("invalid file: %r" % file)
150 if not isinstance(mode, str):
151 raise TypeError("invalid mode: %r" % mode)
Benjamin Peterson95e392c2010-04-27 21:07:21 +0000152 if not isinstance(buffering, int):
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000153 raise TypeError("invalid buffering: %r" % buffering)
154 if encoding is not None and not isinstance(encoding, str):
155 raise TypeError("invalid encoding: %r" % encoding)
156 if errors is not None and not isinstance(errors, str):
157 raise TypeError("invalid errors: %r" % errors)
158 modes = set(mode)
159 if modes - set("arwb+tU") or len(mode) > len(modes):
160 raise ValueError("invalid mode: %r" % mode)
161 reading = "r" in modes
162 writing = "w" in modes
163 appending = "a" in modes
164 updating = "+" in modes
165 text = "t" in modes
166 binary = "b" in modes
167 if "U" in modes:
168 if writing or appending:
169 raise ValueError("can't use U and writing mode at once")
170 reading = True
171 if text and binary:
172 raise ValueError("can't have text and binary mode at once")
173 if reading + writing + appending > 1:
174 raise ValueError("can't have read/write/append mode at once")
175 if not (reading or writing or appending):
176 raise ValueError("must have exactly one of read/write/append mode")
177 if binary and encoding is not None:
178 raise ValueError("binary mode doesn't take an encoding argument")
179 if binary and errors is not None:
180 raise ValueError("binary mode doesn't take an errors argument")
181 if binary and newline is not None:
182 raise ValueError("binary mode doesn't take a newline argument")
183 raw = FileIO(file,
184 (reading and "r" or "") +
185 (writing and "w" or "") +
186 (appending and "a" or "") +
187 (updating and "+" or ""),
188 closefd)
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000189 line_buffering = False
190 if buffering == 1 or buffering < 0 and raw.isatty():
191 buffering = -1
192 line_buffering = True
193 if buffering < 0:
194 buffering = DEFAULT_BUFFER_SIZE
195 try:
196 bs = os.fstat(raw.fileno()).st_blksize
197 except (os.error, AttributeError):
198 pass
199 else:
200 if bs > 1:
201 buffering = bs
202 if buffering < 0:
203 raise ValueError("invalid buffering size")
204 if buffering == 0:
205 if binary:
206 return raw
207 raise ValueError("can't have unbuffered text I/O")
208 if updating:
209 buffer = BufferedRandom(raw, buffering)
210 elif writing or appending:
211 buffer = BufferedWriter(raw, buffering)
212 elif reading:
213 buffer = BufferedReader(raw, buffering)
214 else:
215 raise ValueError("unknown mode: %r" % mode)
216 if binary:
217 return buffer
218 text = TextIOWrapper(buffer, encoding, errors, newline, line_buffering)
219 text.mode = mode
220 return text
221
222
223class DocDescriptor:
224 """Helper for builtins.open.__doc__
225 """
226 def __get__(self, obj, typ):
227 return (
Benjamin Petersonc3be11a2010-04-27 21:24:03 +0000228 "open(file, mode='r', buffering=-1, encoding=None, "
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000229 "errors=None, newline=None, closefd=True)\n\n" +
230 open.__doc__)
231
232class OpenWrapper:
233 """Wrapper for builtins.open
234
235 Trick so that open won't become a bound method when stored
236 as a class variable (as dbm.dumb does).
237
238 See initstdio() in Python/pythonrun.c.
239 """
240 __doc__ = DocDescriptor()
241
242 def __new__(cls, *args, **kwargs):
243 return open(*args, **kwargs)
244
245
Antoine Pitrou0d739d72010-09-05 23:01:12 +0000246# In normal operation, both `UnsupportedOperation`s should be bound to the
247# same object.
248try:
249 UnsupportedOperation = io.UnsupportedOperation
250except AttributeError:
251 class UnsupportedOperation(ValueError, IOError):
252 pass
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000253
254
255class IOBase(metaclass=abc.ABCMeta):
256
257 """The abstract base class for all I/O classes, acting on streams of
258 bytes. There is no public constructor.
259
260 This class provides dummy implementations for many methods that
261 derived classes can override selectively; the default implementations
262 represent a file that cannot be read, written or seeked.
263
264 Even though IOBase does not declare read, readinto, or write because
265 their signatures will vary, implementations and clients should
266 consider those methods part of the interface. Also, implementations
Amaury Forgeot d'Arc616453c2010-09-06 22:31:52 +0000267 may raise UnsupportedOperation when operations they do not support are
268 called.
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000269
270 The basic type used for binary data read from or written to a file is
271 bytes. bytearrays are accepted too, and in some cases (such as
272 readinto) needed. Text I/O classes work with str data.
273
274 Note that calling any method (even inquiries) on a closed stream is
275 undefined. Implementations may raise IOError in this case.
276
277 IOBase (and its subclasses) support the iterator protocol, meaning
278 that an IOBase object can be iterated over yielding the lines in a
279 stream.
280
281 IOBase also supports the :keyword:`with` statement. In this example,
282 fp is closed after the suite of the with statement is complete:
283
284 with open('spam.txt', 'r') as fp:
285 fp.write('Spam and eggs!')
286 """
287
288 ### Internal ###
289
Raymond Hettinger3c940242011-01-12 23:39:31 +0000290 def _unsupported(self, name):
Raymond Hettingercbb80892011-01-13 18:15:51 +0000291 """Internal: raise an IOError exception for unsupported operations."""
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000292 raise UnsupportedOperation("%s.%s() not supported" %
293 (self.__class__.__name__, name))
294
295 ### Positioning ###
296
Georg Brandl4d73b572011-01-13 07:13:06 +0000297 def seek(self, pos, whence=0):
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000298 """Change stream position.
299
300 Change the stream position to byte offset offset. offset is
301 interpreted relative to the position indicated by whence. Values
Raymond Hettingercbb80892011-01-13 18:15:51 +0000302 for whence are ints:
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000303
304 * 0 -- start of stream (the default); offset should be zero or positive
305 * 1 -- current stream position; offset may be negative
306 * 2 -- end of stream; offset is usually negative
307
Raymond Hettingercbb80892011-01-13 18:15:51 +0000308 Return an int indicating the new absolute position.
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000309 """
310 self._unsupported("seek")
311
Raymond Hettinger3c940242011-01-12 23:39:31 +0000312 def tell(self):
Raymond Hettingercbb80892011-01-13 18:15:51 +0000313 """Return an int indicating the current stream position."""
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000314 return self.seek(0, 1)
315
Georg Brandl4d73b572011-01-13 07:13:06 +0000316 def truncate(self, pos=None):
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000317 """Truncate file to size bytes.
318
319 Size defaults to the current IO position as reported by tell(). Return
320 the new size.
321 """
322 self._unsupported("truncate")
323
324 ### Flush and close ###
325
Raymond Hettinger3c940242011-01-12 23:39:31 +0000326 def flush(self):
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000327 """Flush write buffers, if applicable.
328
329 This is not implemented for read-only and non-blocking streams.
330 """
Antoine Pitrou6be88762010-05-03 16:48:20 +0000331 self._checkClosed()
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000332 # XXX Should this return the number of bytes written???
333
334 __closed = False
335
Raymond Hettinger3c940242011-01-12 23:39:31 +0000336 def close(self):
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000337 """Flush and close the IO object.
338
339 This method has no effect if the file is already closed.
340 """
341 if not self.__closed:
Antoine Pitrou6be88762010-05-03 16:48:20 +0000342 self.flush()
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000343 self.__closed = True
344
Raymond Hettinger3c940242011-01-12 23:39:31 +0000345 def __del__(self):
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000346 """Destructor. Calls close()."""
347 # The try/except block is in case this is called at program
348 # exit time, when it's possible that globals have already been
349 # deleted, and then the close() call might fail. Since
350 # there's nothing we can do about such failures and they annoy
351 # the end users, we suppress the traceback.
352 try:
353 self.close()
354 except:
355 pass
356
357 ### Inquiries ###
358
Raymond Hettinger3c940242011-01-12 23:39:31 +0000359 def seekable(self):
Raymond Hettingercbb80892011-01-13 18:15:51 +0000360 """Return a bool indicating whether object supports random access.
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000361
Amaury Forgeot d'Arcada99482010-09-06 22:23:13 +0000362 If False, seek(), tell() and truncate() will raise UnsupportedOperation.
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000363 This method may need to do a test seek().
364 """
365 return False
366
367 def _checkSeekable(self, msg=None):
Amaury Forgeot d'Arcada99482010-09-06 22:23:13 +0000368 """Internal: raise UnsupportedOperation if file is not seekable
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000369 """
370 if not self.seekable():
Antoine Pitrou0d739d72010-09-05 23:01:12 +0000371 raise UnsupportedOperation("File or stream is not seekable."
372 if msg is None else msg)
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000373
Raymond Hettinger3c940242011-01-12 23:39:31 +0000374 def readable(self):
Raymond Hettingercbb80892011-01-13 18:15:51 +0000375 """Return a bool indicating whether object was opened for reading.
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000376
Amaury Forgeot d'Arcada99482010-09-06 22:23:13 +0000377 If False, read() will raise UnsupportedOperation.
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000378 """
379 return False
380
381 def _checkReadable(self, msg=None):
Amaury Forgeot d'Arcada99482010-09-06 22:23:13 +0000382 """Internal: raise UnsupportedOperation if file is not readable
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000383 """
384 if not self.readable():
Antoine Pitrou0d739d72010-09-05 23:01:12 +0000385 raise UnsupportedOperation("File or stream is not readable."
386 if msg is None else msg)
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000387
Raymond Hettinger3c940242011-01-12 23:39:31 +0000388 def writable(self):
Raymond Hettingercbb80892011-01-13 18:15:51 +0000389 """Return a bool indicating whether object was opened for writing.
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000390
Amaury Forgeot d'Arcada99482010-09-06 22:23:13 +0000391 If False, write() and truncate() will raise UnsupportedOperation.
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000392 """
393 return False
394
395 def _checkWritable(self, msg=None):
Amaury Forgeot d'Arcada99482010-09-06 22:23:13 +0000396 """Internal: raise UnsupportedOperation if file is not writable
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000397 """
398 if not self.writable():
Antoine Pitrou0d739d72010-09-05 23:01:12 +0000399 raise UnsupportedOperation("File or stream is not writable."
400 if msg is None else msg)
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000401
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
Raymond Hettinger3c940242011-01-12 23:39:31 +0000419 def __enter__(self): # That's a forward reference
Raymond Hettingercbb80892011-01-13 18:15:51 +0000420 """Context management protocol. Returns self (an instance of IOBase)."""
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000421 self._checkClosed()
422 return self
423
Raymond Hettinger3c940242011-01-12 23:39:31 +0000424 def __exit__(self, *args):
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000425 """Context management protocol. Calls close()"""
426 self.close()
427
428 ### Lower-level APIs ###
429
430 # XXX Should these be present even if unimplemented?
431
Raymond Hettinger3c940242011-01-12 23:39:31 +0000432 def fileno(self):
Raymond Hettingercbb80892011-01-13 18:15:51 +0000433 """Returns underlying file descriptor (an int) if one exists.
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000434
435 An IOError is raised if the IO object does not use a file descriptor.
436 """
437 self._unsupported("fileno")
438
Raymond Hettinger3c940242011-01-12 23:39:31 +0000439 def isatty(self):
Raymond Hettingercbb80892011-01-13 18:15:51 +0000440 """Return a bool indicating whether this is an 'interactive' stream.
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000441
442 Return False if it can't be determined.
443 """
444 self._checkClosed()
445 return False
446
447 ### Readline[s] and writelines ###
448
Georg Brandl4d73b572011-01-13 07:13:06 +0000449 def readline(self, limit=-1):
Raymond Hettingercbb80892011-01-13 18:15:51 +0000450 r"""Read and return a line of bytes from the stream.
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000451
452 If limit is specified, at most limit bytes will be read.
Raymond Hettingercbb80892011-01-13 18:15:51 +0000453 Limit should be an int.
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000454
455 The line terminator is always b'\n' for binary files; for text
456 files, the newlines argument to open can be used to select the line
457 terminator(s) recognized.
458 """
459 # For backwards compatibility, a (slowish) readline().
460 if hasattr(self, "peek"):
461 def nreadahead():
462 readahead = self.peek(1)
463 if not readahead:
464 return 1
465 n = (readahead.find(b"\n") + 1) or len(readahead)
466 if limit >= 0:
467 n = min(n, limit)
468 return n
469 else:
470 def nreadahead():
471 return 1
472 if limit is None:
473 limit = -1
Benjamin Petersonb01138a2009-04-24 22:59:52 +0000474 elif not isinstance(limit, int):
475 raise TypeError("limit must be an integer")
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000476 res = bytearray()
477 while limit < 0 or len(res) < limit:
478 b = self.read(nreadahead())
479 if not b:
480 break
481 res += b
482 if res.endswith(b"\n"):
483 break
484 return bytes(res)
485
486 def __iter__(self):
487 self._checkClosed()
488 return self
489
490 def __next__(self):
491 line = self.readline()
492 if not line:
493 raise StopIteration
494 return line
495
496 def readlines(self, hint=None):
497 """Return a list of lines from the stream.
498
499 hint can be specified to control the number of lines read: no more
500 lines will be read if the total size (in bytes/characters) of all
501 lines so far exceeds hint.
502 """
503 if hint is None or hint <= 0:
504 return list(self)
505 n = 0
506 lines = []
507 for line in self:
508 lines.append(line)
509 n += len(line)
510 if n >= hint:
511 break
512 return lines
513
514 def writelines(self, lines):
515 self._checkClosed()
516 for line in lines:
517 self.write(line)
518
519io.IOBase.register(IOBase)
520
521
522class RawIOBase(IOBase):
523
524 """Base class for raw binary I/O."""
525
526 # The read() method is implemented by calling readinto(); derived
527 # classes that want to support read() only need to implement
528 # readinto() as a primitive operation. In general, readinto() can be
529 # more efficient than read().
530
531 # (It would be tempting to also provide an implementation of
532 # readinto() in terms of read(), in case the latter is a more suitable
533 # primitive operation, but that would lead to nasty recursion in case
534 # a subclass doesn't implement either.)
535
Georg Brandl4d73b572011-01-13 07:13:06 +0000536 def read(self, n=-1):
Raymond Hettingercbb80892011-01-13 18:15:51 +0000537 """Read and return up to n bytes, where n is an int.
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000538
539 Returns an empty bytes object on EOF, or None if the object is
540 set not to block and has no data to read.
541 """
542 if n is None:
543 n = -1
544 if n < 0:
545 return self.readall()
546 b = bytearray(n.__index__())
547 n = self.readinto(b)
Antoine Pitrou328ec742010-09-14 18:37:24 +0000548 if n is None:
549 return None
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000550 del b[n:]
551 return bytes(b)
552
553 def readall(self):
554 """Read until EOF, using multiple read() call."""
555 res = bytearray()
556 while True:
557 data = self.read(DEFAULT_BUFFER_SIZE)
558 if not data:
559 break
560 res += data
Victor Stinnera80987f2011-05-25 22:47:16 +0200561 if res:
562 return bytes(res)
563 else:
564 # b'' or None
565 return data
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000566
Raymond Hettinger3c940242011-01-12 23:39:31 +0000567 def readinto(self, b):
Raymond Hettingercbb80892011-01-13 18:15:51 +0000568 """Read up to len(b) bytes into bytearray b.
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000569
Raymond Hettingercbb80892011-01-13 18:15:51 +0000570 Returns an int representing the number of bytes read (0 for EOF), or
571 None if the object is set not to block and has no data to read.
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000572 """
573 self._unsupported("readinto")
574
Raymond Hettinger3c940242011-01-12 23:39:31 +0000575 def write(self, b):
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000576 """Write the given buffer to the IO stream.
577
578 Returns the number of bytes written, which may be less than len(b).
579 """
580 self._unsupported("write")
581
582io.RawIOBase.register(RawIOBase)
583from _io import FileIO
584RawIOBase.register(FileIO)
585
586
587class BufferedIOBase(IOBase):
588
589 """Base class for buffered IO objects.
590
591 The main difference with RawIOBase is that the read() method
592 supports omitting the size argument, and does not have a default
593 implementation that defers to readinto().
594
595 In addition, read(), readinto() and write() may raise
596 BlockingIOError if the underlying raw stream is in non-blocking
597 mode and not ready; unlike their raw counterparts, they will never
598 return None.
599
600 A typical implementation should not inherit from a RawIOBase
601 implementation, but wrap one.
602 """
603
Georg Brandl4d73b572011-01-13 07:13:06 +0000604 def read(self, n=None):
Raymond Hettingercbb80892011-01-13 18:15:51 +0000605 """Read and return up to n bytes, where n is an int.
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000606
607 If the argument is omitted, None, or negative, reads and
608 returns all data until EOF.
609
610 If the argument is positive, and the underlying raw stream is
611 not 'interactive', multiple raw reads may be issued to satisfy
612 the byte count (unless EOF is reached first). But for
613 interactive raw streams (XXX and for pipes?), at most one raw
614 read will be issued, and a short result does not imply that
615 EOF is imminent.
616
617 Returns an empty bytes array on EOF.
618
619 Raises BlockingIOError if the underlying raw stream has no
620 data at the moment.
621 """
622 self._unsupported("read")
623
Georg Brandl4d73b572011-01-13 07:13:06 +0000624 def read1(self, n=None):
Raymond Hettingercbb80892011-01-13 18:15:51 +0000625 """Read up to n bytes with at most one read() system call,
626 where n is an int.
627 """
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000628 self._unsupported("read1")
629
Raymond Hettinger3c940242011-01-12 23:39:31 +0000630 def readinto(self, b):
Raymond Hettingercbb80892011-01-13 18:15:51 +0000631 """Read up to len(b) bytes into bytearray b.
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000632
633 Like read(), this may issue multiple reads to the underlying raw
634 stream, unless the latter is 'interactive'.
635
Raymond Hettingercbb80892011-01-13 18:15:51 +0000636 Returns an int representing the number of bytes read (0 for EOF).
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000637
638 Raises BlockingIOError if the underlying raw stream has no
639 data at the moment.
640 """
641 # XXX This ought to work with anything that supports the buffer API
642 data = self.read(len(b))
643 n = len(data)
644 try:
645 b[:n] = data
646 except TypeError as err:
647 import array
648 if not isinstance(b, array.array):
649 raise err
650 b[:n] = array.array('b', data)
651 return n
652
Raymond Hettinger3c940242011-01-12 23:39:31 +0000653 def write(self, b):
Raymond Hettingercbb80892011-01-13 18:15:51 +0000654 """Write the given bytes buffer to the IO stream.
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000655
656 Return the number of bytes written, which is never less than
657 len(b).
658
659 Raises BlockingIOError if the buffer is full and the
660 underlying raw stream cannot accept more data at the moment.
661 """
662 self._unsupported("write")
663
Raymond Hettinger3c940242011-01-12 23:39:31 +0000664 def detach(self):
Benjamin Petersond2e0c792009-05-01 20:40:59 +0000665 """
666 Separate the underlying raw stream from the buffer and return it.
667
668 After the raw stream has been detached, the buffer is in an unusable
669 state.
670 """
671 self._unsupported("detach")
672
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000673io.BufferedIOBase.register(BufferedIOBase)
674
675
676class _BufferedIOMixin(BufferedIOBase):
677
678 """A mixin implementation of BufferedIOBase with an underlying raw stream.
679
680 This passes most requests on to the underlying raw stream. It
681 does *not* provide implementations of read(), readinto() or
682 write().
683 """
684
685 def __init__(self, raw):
Antoine Pitrou7f8f4182010-12-21 21:20:59 +0000686 self._raw = raw
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000687
688 ### Positioning ###
689
690 def seek(self, pos, whence=0):
691 new_position = self.raw.seek(pos, whence)
692 if new_position < 0:
693 raise IOError("seek() returned an invalid position")
694 return new_position
695
696 def tell(self):
697 pos = self.raw.tell()
698 if pos < 0:
699 raise IOError("tell() returned an invalid position")
700 return pos
701
702 def truncate(self, pos=None):
703 # Flush the stream. We're mixing buffered I/O with lower-level I/O,
704 # and a flush may be necessary to synch both views of the current
705 # file state.
706 self.flush()
707
708 if pos is None:
709 pos = self.tell()
710 # XXX: Should seek() be used, instead of passing the position
711 # XXX directly to truncate?
712 return self.raw.truncate(pos)
713
714 ### Flush and close ###
715
716 def flush(self):
Antoine Pitrou6be88762010-05-03 16:48:20 +0000717 if self.closed:
718 raise ValueError("flush of closed file")
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000719 self.raw.flush()
720
721 def close(self):
Antoine Pitrou6be88762010-05-03 16:48:20 +0000722 if self.raw is not None and not self.closed:
723 self.flush()
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000724 self.raw.close()
725
Benjamin Petersond2e0c792009-05-01 20:40:59 +0000726 def detach(self):
727 if self.raw is None:
728 raise ValueError("raw stream already detached")
729 self.flush()
Antoine Pitrou7f8f4182010-12-21 21:20:59 +0000730 raw = self._raw
731 self._raw = None
Benjamin Petersond2e0c792009-05-01 20:40:59 +0000732 return raw
733
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000734 ### Inquiries ###
735
736 def seekable(self):
737 return self.raw.seekable()
738
739 def readable(self):
740 return self.raw.readable()
741
742 def writable(self):
743 return self.raw.writable()
744
745 @property
Antoine Pitrou7f8f4182010-12-21 21:20:59 +0000746 def raw(self):
747 return self._raw
748
749 @property
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000750 def closed(self):
751 return self.raw.closed
752
753 @property
754 def name(self):
755 return self.raw.name
756
757 @property
758 def mode(self):
759 return self.raw.mode
760
Antoine Pitrou243757e2010-11-05 21:15:39 +0000761 def __getstate__(self):
762 raise TypeError("can not serialize a '{0}' object"
763 .format(self.__class__.__name__))
764
Antoine Pitrou716c4442009-05-23 19:04:03 +0000765 def __repr__(self):
766 clsname = self.__class__.__name__
767 try:
768 name = self.name
769 except AttributeError:
770 return "<_pyio.{0}>".format(clsname)
771 else:
772 return "<_pyio.{0} name={1!r}>".format(clsname, name)
773
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000774 ### Lower-level APIs ###
775
776 def fileno(self):
777 return self.raw.fileno()
778
779 def isatty(self):
780 return self.raw.isatty()
781
782
783class BytesIO(BufferedIOBase):
784
785 """Buffered I/O implementation using an in-memory bytes buffer."""
786
787 def __init__(self, initial_bytes=None):
788 buf = bytearray()
789 if initial_bytes is not None:
790 buf += initial_bytes
791 self._buffer = buf
792 self._pos = 0
793
Alexandre Vassalotticf76e1a2009-07-22 03:24:36 +0000794 def __getstate__(self):
795 if self.closed:
796 raise ValueError("__getstate__ on closed file")
797 return self.__dict__.copy()
798
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000799 def getvalue(self):
800 """Return the bytes value (contents) of the buffer
801 """
802 if self.closed:
803 raise ValueError("getvalue on closed file")
804 return bytes(self._buffer)
805
Antoine Pitrou972ee132010-09-06 18:48:21 +0000806 def getbuffer(self):
807 """Return a readable and writable view of the buffer.
808 """
809 return memoryview(self._buffer)
810
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000811 def read(self, n=None):
812 if self.closed:
813 raise ValueError("read from closed file")
814 if n is None:
815 n = -1
816 if n < 0:
817 n = len(self._buffer)
818 if len(self._buffer) <= self._pos:
819 return b""
820 newpos = min(len(self._buffer), self._pos + n)
821 b = self._buffer[self._pos : newpos]
822 self._pos = newpos
823 return bytes(b)
824
825 def read1(self, n):
826 """This is the same as read.
827 """
828 return self.read(n)
829
830 def write(self, b):
831 if self.closed:
832 raise ValueError("write to closed file")
833 if isinstance(b, str):
834 raise TypeError("can't write str to binary stream")
835 n = len(b)
836 if n == 0:
837 return 0
838 pos = self._pos
839 if pos > len(self._buffer):
840 # Inserts null bytes between the current end of the file
841 # and the new write position.
842 padding = b'\x00' * (pos - len(self._buffer))
843 self._buffer += padding
844 self._buffer[pos:pos + n] = b
845 self._pos += n
846 return n
847
848 def seek(self, pos, whence=0):
849 if self.closed:
850 raise ValueError("seek on closed file")
851 try:
Florent Xiclunab14930c2010-03-13 15:26:44 +0000852 pos.__index__
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000853 except AttributeError as err:
854 raise TypeError("an integer is required") from err
855 if whence == 0:
856 if pos < 0:
857 raise ValueError("negative seek position %r" % (pos,))
858 self._pos = pos
859 elif whence == 1:
860 self._pos = max(0, self._pos + pos)
861 elif whence == 2:
862 self._pos = max(0, len(self._buffer) + pos)
863 else:
864 raise ValueError("invalid whence value")
865 return self._pos
866
867 def tell(self):
868 if self.closed:
869 raise ValueError("tell on closed file")
870 return self._pos
871
872 def truncate(self, pos=None):
873 if self.closed:
874 raise ValueError("truncate on closed file")
875 if pos is None:
876 pos = self._pos
Florent Xiclunab14930c2010-03-13 15:26:44 +0000877 else:
878 try:
879 pos.__index__
880 except AttributeError as err:
881 raise TypeError("an integer is required") from err
882 if pos < 0:
883 raise ValueError("negative truncate position %r" % (pos,))
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000884 del self._buffer[pos:]
Antoine Pitrou905a2ff2010-01-31 22:47:27 +0000885 return pos
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000886
887 def readable(self):
888 return True
889
890 def writable(self):
891 return True
892
893 def seekable(self):
894 return True
895
896
897class BufferedReader(_BufferedIOMixin):
898
899 """BufferedReader(raw[, buffer_size])
900
901 A buffer for a readable, sequential BaseRawIO object.
902
903 The constructor creates a BufferedReader for the given readable raw
904 stream and buffer_size. If buffer_size is omitted, DEFAULT_BUFFER_SIZE
905 is used.
906 """
907
908 def __init__(self, raw, buffer_size=DEFAULT_BUFFER_SIZE):
909 """Create a new buffered reader using the given readable raw IO object.
910 """
Antoine Pitroucf4c7492009-04-19 00:09:36 +0000911 if not raw.readable():
912 raise IOError('"raw" argument must be readable.')
913
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000914 _BufferedIOMixin.__init__(self, raw)
915 if buffer_size <= 0:
916 raise ValueError("invalid buffer size")
917 self.buffer_size = buffer_size
918 self._reset_read_buf()
919 self._read_lock = Lock()
920
921 def _reset_read_buf(self):
922 self._read_buf = b""
923 self._read_pos = 0
924
925 def read(self, n=None):
926 """Read n bytes.
927
928 Returns exactly n bytes of data unless the underlying raw IO
929 stream reaches EOF or if the call would block in non-blocking
930 mode. If n is negative, read until EOF or until read() would
931 block.
932 """
933 if n is not None and n < -1:
934 raise ValueError("invalid number of bytes to read")
935 with self._read_lock:
936 return self._read_unlocked(n)
937
938 def _read_unlocked(self, n=None):
939 nodata_val = b""
940 empty_values = (b"", None)
941 buf = self._read_buf
942 pos = self._read_pos
943
944 # Special case for when the number of bytes to read is unspecified.
945 if n is None or n == -1:
946 self._reset_read_buf()
947 chunks = [buf[pos:]] # Strip the consumed bytes.
948 current_size = 0
949 while True:
950 # Read until EOF or until read() would block.
Antoine Pitroud843c2d2011-02-25 21:34:39 +0000951 try:
952 chunk = self.raw.read()
953 except IOError as e:
954 if e.errno != EINTR:
955 raise
956 continue
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000957 if chunk in empty_values:
958 nodata_val = chunk
959 break
960 current_size += len(chunk)
961 chunks.append(chunk)
962 return b"".join(chunks) or nodata_val
963
964 # The number of bytes to read is specified, return at most n bytes.
965 avail = len(buf) - pos # Length of the available buffered data.
966 if n <= avail:
967 # Fast path: the data to read is fully buffered.
968 self._read_pos += n
969 return buf[pos:pos+n]
970 # Slow path: read from the stream until enough bytes are read,
971 # or until an EOF occurs or until read() would block.
972 chunks = [buf[pos:]]
973 wanted = max(self.buffer_size, n)
974 while avail < n:
Antoine Pitroud843c2d2011-02-25 21:34:39 +0000975 try:
976 chunk = self.raw.read(wanted)
977 except IOError as e:
978 if e.errno != EINTR:
979 raise
980 continue
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +0000981 if chunk in empty_values:
982 nodata_val = chunk
983 break
984 avail += len(chunk)
985 chunks.append(chunk)
986 # n is more then avail only when an EOF occurred or when
987 # read() would have blocked.
988 n = min(n, avail)
989 out = b"".join(chunks)
990 self._read_buf = out[n:] # Save the extra data in the buffer.
991 self._read_pos = 0
992 return out[:n] if out else nodata_val
993
994 def peek(self, n=0):
995 """Returns buffered bytes without advancing the position.
996
997 The argument indicates a desired minimal number of bytes; we
998 do at most one raw read to satisfy it. We never return more
999 than self.buffer_size.
1000 """
1001 with self._read_lock:
1002 return self._peek_unlocked(n)
1003
1004 def _peek_unlocked(self, n=0):
1005 want = min(n, self.buffer_size)
1006 have = len(self._read_buf) - self._read_pos
1007 if have < want or have <= 0:
1008 to_read = self.buffer_size - have
Antoine Pitroud843c2d2011-02-25 21:34:39 +00001009 while True:
1010 try:
1011 current = self.raw.read(to_read)
1012 except IOError as e:
1013 if e.errno != EINTR:
1014 raise
1015 continue
1016 break
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001017 if current:
1018 self._read_buf = self._read_buf[self._read_pos:] + current
1019 self._read_pos = 0
1020 return self._read_buf[self._read_pos:]
1021
1022 def read1(self, n):
1023 """Reads up to n bytes, with at most one read() system call."""
1024 # Returns up to n bytes. If at least one byte is buffered, we
1025 # only return buffered bytes. Otherwise, we do one raw read.
1026 if n < 0:
1027 raise ValueError("number of bytes to read must be positive")
1028 if n == 0:
1029 return b""
1030 with self._read_lock:
1031 self._peek_unlocked(1)
1032 return self._read_unlocked(
1033 min(n, len(self._read_buf) - self._read_pos))
1034
1035 def tell(self):
1036 return _BufferedIOMixin.tell(self) - len(self._read_buf) + self._read_pos
1037
1038 def seek(self, pos, whence=0):
1039 if not (0 <= whence <= 2):
1040 raise ValueError("invalid whence value")
1041 with self._read_lock:
1042 if whence == 1:
1043 pos -= len(self._read_buf) - self._read_pos
1044 pos = _BufferedIOMixin.seek(self, pos, whence)
1045 self._reset_read_buf()
1046 return pos
1047
1048class BufferedWriter(_BufferedIOMixin):
1049
1050 """A buffer for a writeable sequential RawIO object.
1051
1052 The constructor creates a BufferedWriter for the given writeable raw
1053 stream. If the buffer_size is not given, it defaults to
Benjamin Peterson59406a92009-03-26 17:10:29 +00001054 DEFAULT_BUFFER_SIZE.
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001055 """
1056
Benjamin Peterson59406a92009-03-26 17:10:29 +00001057 _warning_stack_offset = 2
1058
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001059 def __init__(self, raw,
1060 buffer_size=DEFAULT_BUFFER_SIZE, max_buffer_size=None):
Antoine Pitroucf4c7492009-04-19 00:09:36 +00001061 if not raw.writable():
1062 raise IOError('"raw" argument must be writable.')
1063
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001064 _BufferedIOMixin.__init__(self, raw)
1065 if buffer_size <= 0:
1066 raise ValueError("invalid buffer size")
Benjamin Peterson59406a92009-03-26 17:10:29 +00001067 if max_buffer_size is not None:
1068 warnings.warn("max_buffer_size is deprecated", DeprecationWarning,
1069 self._warning_stack_offset)
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001070 self.buffer_size = buffer_size
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001071 self._write_buf = bytearray()
1072 self._write_lock = Lock()
1073
1074 def write(self, b):
1075 if self.closed:
1076 raise ValueError("write to closed file")
1077 if isinstance(b, str):
1078 raise TypeError("can't write str to binary stream")
1079 with self._write_lock:
1080 # XXX we can implement some more tricks to try and avoid
1081 # partial writes
1082 if len(self._write_buf) > self.buffer_size:
1083 # We're full, so let's pre-flush the buffer
1084 try:
1085 self._flush_unlocked()
1086 except BlockingIOError as e:
1087 # We can't accept anything else.
1088 # XXX Why not just let the exception pass through?
1089 raise BlockingIOError(e.errno, e.strerror, 0)
1090 before = len(self._write_buf)
1091 self._write_buf.extend(b)
1092 written = len(self._write_buf) - before
1093 if len(self._write_buf) > self.buffer_size:
1094 try:
1095 self._flush_unlocked()
1096 except BlockingIOError as e:
Benjamin Peterson394ee002009-03-05 22:33:59 +00001097 if len(self._write_buf) > self.buffer_size:
1098 # We've hit the buffer_size. We have to accept a partial
1099 # write and cut back our buffer.
1100 overage = len(self._write_buf) - self.buffer_size
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001101 written -= overage
Benjamin Peterson394ee002009-03-05 22:33:59 +00001102 self._write_buf = self._write_buf[:self.buffer_size]
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001103 raise BlockingIOError(e.errno, e.strerror, written)
1104 return written
1105
1106 def truncate(self, pos=None):
1107 with self._write_lock:
1108 self._flush_unlocked()
1109 if pos is None:
1110 pos = self.raw.tell()
1111 return self.raw.truncate(pos)
1112
1113 def flush(self):
1114 with self._write_lock:
1115 self._flush_unlocked()
1116
1117 def _flush_unlocked(self):
1118 if self.closed:
1119 raise ValueError("flush of closed file")
1120 written = 0
1121 try:
1122 while self._write_buf:
Antoine Pitroud843c2d2011-02-25 21:34:39 +00001123 try:
1124 n = self.raw.write(self._write_buf)
1125 except IOError as e:
1126 if e.errno != EINTR:
1127 raise
1128 continue
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001129 if n > len(self._write_buf) or n < 0:
1130 raise IOError("write() returned incorrect number of bytes")
1131 del self._write_buf[:n]
1132 written += n
1133 except BlockingIOError as e:
1134 n = e.characters_written
1135 del self._write_buf[:n]
1136 written += n
1137 raise BlockingIOError(e.errno, e.strerror, written)
1138
1139 def tell(self):
1140 return _BufferedIOMixin.tell(self) + len(self._write_buf)
1141
1142 def seek(self, pos, whence=0):
1143 if not (0 <= whence <= 2):
1144 raise ValueError("invalid whence")
1145 with self._write_lock:
1146 self._flush_unlocked()
1147 return _BufferedIOMixin.seek(self, pos, whence)
1148
1149
1150class BufferedRWPair(BufferedIOBase):
1151
1152 """A buffered reader and writer object together.
1153
1154 A buffered reader object and buffered writer object put together to
1155 form a sequential IO object that can read and write. This is typically
1156 used with a socket or two-way pipe.
1157
1158 reader and writer are RawIOBase objects that are readable and
1159 writeable respectively. If the buffer_size is omitted it defaults to
Benjamin Peterson59406a92009-03-26 17:10:29 +00001160 DEFAULT_BUFFER_SIZE.
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001161 """
1162
1163 # XXX The usefulness of this (compared to having two separate IO
1164 # objects) is questionable.
1165
1166 def __init__(self, reader, writer,
1167 buffer_size=DEFAULT_BUFFER_SIZE, max_buffer_size=None):
1168 """Constructor.
1169
1170 The arguments are two RawIO instances.
1171 """
Benjamin Peterson59406a92009-03-26 17:10:29 +00001172 if max_buffer_size is not None:
1173 warnings.warn("max_buffer_size is deprecated", DeprecationWarning, 2)
Antoine Pitroucf4c7492009-04-19 00:09:36 +00001174
1175 if not reader.readable():
1176 raise IOError('"reader" argument must be readable.')
1177
1178 if not writer.writable():
1179 raise IOError('"writer" argument must be writable.')
1180
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001181 self.reader = BufferedReader(reader, buffer_size)
Benjamin Peterson59406a92009-03-26 17:10:29 +00001182 self.writer = BufferedWriter(writer, buffer_size)
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001183
1184 def read(self, n=None):
1185 if n is None:
1186 n = -1
1187 return self.reader.read(n)
1188
1189 def readinto(self, b):
1190 return self.reader.readinto(b)
1191
1192 def write(self, b):
1193 return self.writer.write(b)
1194
1195 def peek(self, n=0):
1196 return self.reader.peek(n)
1197
1198 def read1(self, n):
1199 return self.reader.read1(n)
1200
1201 def readable(self):
1202 return self.reader.readable()
1203
1204 def writable(self):
1205 return self.writer.writable()
1206
1207 def flush(self):
1208 return self.writer.flush()
1209
1210 def close(self):
1211 self.writer.close()
1212 self.reader.close()
1213
1214 def isatty(self):
1215 return self.reader.isatty() or self.writer.isatty()
1216
1217 @property
1218 def closed(self):
1219 return self.writer.closed
1220
1221
1222class BufferedRandom(BufferedWriter, BufferedReader):
1223
1224 """A buffered interface to random access streams.
1225
1226 The constructor creates a reader and writer for a seekable stream,
1227 raw, given in the first argument. If the buffer_size is omitted it
Benjamin Peterson59406a92009-03-26 17:10:29 +00001228 defaults to DEFAULT_BUFFER_SIZE.
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001229 """
1230
Benjamin Peterson59406a92009-03-26 17:10:29 +00001231 _warning_stack_offset = 3
1232
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001233 def __init__(self, raw,
1234 buffer_size=DEFAULT_BUFFER_SIZE, max_buffer_size=None):
1235 raw._checkSeekable()
1236 BufferedReader.__init__(self, raw, buffer_size)
1237 BufferedWriter.__init__(self, raw, buffer_size, max_buffer_size)
1238
1239 def seek(self, pos, whence=0):
1240 if not (0 <= whence <= 2):
1241 raise ValueError("invalid whence")
1242 self.flush()
1243 if self._read_buf:
1244 # Undo read ahead.
1245 with self._read_lock:
1246 self.raw.seek(self._read_pos - len(self._read_buf), 1)
1247 # First do the raw seek, then empty the read buffer, so that
1248 # if the raw seek fails, we don't lose buffered data forever.
1249 pos = self.raw.seek(pos, whence)
1250 with self._read_lock:
1251 self._reset_read_buf()
1252 if pos < 0:
1253 raise IOError("seek() returned invalid position")
1254 return pos
1255
1256 def tell(self):
1257 if self._write_buf:
1258 return BufferedWriter.tell(self)
1259 else:
1260 return BufferedReader.tell(self)
1261
1262 def truncate(self, pos=None):
1263 if pos is None:
1264 pos = self.tell()
1265 # Use seek to flush the read buffer.
Antoine Pitrou905a2ff2010-01-31 22:47:27 +00001266 return BufferedWriter.truncate(self, pos)
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001267
1268 def read(self, n=None):
1269 if n is None:
1270 n = -1
1271 self.flush()
1272 return BufferedReader.read(self, n)
1273
1274 def readinto(self, b):
1275 self.flush()
1276 return BufferedReader.readinto(self, b)
1277
1278 def peek(self, n=0):
1279 self.flush()
1280 return BufferedReader.peek(self, n)
1281
1282 def read1(self, n):
1283 self.flush()
1284 return BufferedReader.read1(self, n)
1285
1286 def write(self, b):
1287 if self._read_buf:
1288 # Undo readahead
1289 with self._read_lock:
1290 self.raw.seek(self._read_pos - len(self._read_buf), 1)
1291 self._reset_read_buf()
1292 return BufferedWriter.write(self, b)
1293
1294
1295class TextIOBase(IOBase):
1296
1297 """Base class for text I/O.
1298
1299 This class provides a character and line based interface to stream
1300 I/O. There is no readinto method because Python's character strings
1301 are immutable. There is no public constructor.
1302 """
1303
Georg Brandl4d73b572011-01-13 07:13:06 +00001304 def read(self, n=-1):
Raymond Hettingercbb80892011-01-13 18:15:51 +00001305 """Read at most n characters from stream, where n is an int.
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001306
1307 Read from underlying buffer until we have n characters or we hit EOF.
1308 If n is negative or omitted, read until EOF.
Raymond Hettingercbb80892011-01-13 18:15:51 +00001309
1310 Returns a string.
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001311 """
1312 self._unsupported("read")
1313
Raymond Hettinger3c940242011-01-12 23:39:31 +00001314 def write(self, s):
Raymond Hettingercbb80892011-01-13 18:15:51 +00001315 """Write string s to stream and returning an int."""
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001316 self._unsupported("write")
1317
Georg Brandl4d73b572011-01-13 07:13:06 +00001318 def truncate(self, pos=None):
Raymond Hettingercbb80892011-01-13 18:15:51 +00001319 """Truncate size to pos, where pos is an int."""
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001320 self._unsupported("truncate")
1321
Raymond Hettinger3c940242011-01-12 23:39:31 +00001322 def readline(self):
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001323 """Read until newline or EOF.
1324
1325 Returns an empty string if EOF is hit immediately.
1326 """
1327 self._unsupported("readline")
1328
Raymond Hettinger3c940242011-01-12 23:39:31 +00001329 def detach(self):
Benjamin Petersond2e0c792009-05-01 20:40:59 +00001330 """
1331 Separate the underlying buffer from the TextIOBase and return it.
1332
1333 After the underlying buffer has been detached, the TextIO is in an
1334 unusable state.
1335 """
1336 self._unsupported("detach")
1337
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001338 @property
1339 def encoding(self):
1340 """Subclasses should override."""
1341 return None
1342
1343 @property
1344 def newlines(self):
1345 """Line endings translated so far.
1346
1347 Only line endings translated during reading are considered.
1348
1349 Subclasses should override.
1350 """
1351 return None
1352
Benjamin Peterson0926ad12009-06-06 18:02:12 +00001353 @property
1354 def errors(self):
1355 """Error setting of the decoder or encoder.
1356
1357 Subclasses should override."""
1358 return None
1359
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001360io.TextIOBase.register(TextIOBase)
1361
1362
1363class IncrementalNewlineDecoder(codecs.IncrementalDecoder):
1364 r"""Codec used when reading a file in universal newlines mode. It wraps
1365 another incremental decoder, translating \r\n and \r into \n. It also
1366 records the types of newlines encountered. When used with
1367 translate=False, it ensures that the newline sequence is returned in
1368 one piece.
1369 """
1370 def __init__(self, decoder, translate, errors='strict'):
1371 codecs.IncrementalDecoder.__init__(self, errors=errors)
1372 self.translate = translate
1373 self.decoder = decoder
1374 self.seennl = 0
1375 self.pendingcr = False
1376
1377 def decode(self, input, final=False):
1378 # decode input (with the eventual \r from a previous pass)
1379 if self.decoder is None:
1380 output = input
1381 else:
1382 output = self.decoder.decode(input, final=final)
1383 if self.pendingcr and (output or final):
1384 output = "\r" + output
1385 self.pendingcr = False
1386
1387 # retain last \r even when not translating data:
1388 # then readline() is sure to get \r\n in one pass
1389 if output.endswith("\r") and not final:
1390 output = output[:-1]
1391 self.pendingcr = True
1392
1393 # Record which newlines are read
1394 crlf = output.count('\r\n')
1395 cr = output.count('\r') - crlf
1396 lf = output.count('\n') - crlf
1397 self.seennl |= (lf and self._LF) | (cr and self._CR) \
1398 | (crlf and self._CRLF)
1399
1400 if self.translate:
1401 if crlf:
1402 output = output.replace("\r\n", "\n")
1403 if cr:
1404 output = output.replace("\r", "\n")
1405
1406 return output
1407
1408 def getstate(self):
1409 if self.decoder is None:
1410 buf = b""
1411 flag = 0
1412 else:
1413 buf, flag = self.decoder.getstate()
1414 flag <<= 1
1415 if self.pendingcr:
1416 flag |= 1
1417 return buf, flag
1418
1419 def setstate(self, state):
1420 buf, flag = state
1421 self.pendingcr = bool(flag & 1)
1422 if self.decoder is not None:
1423 self.decoder.setstate((buf, flag >> 1))
1424
1425 def reset(self):
1426 self.seennl = 0
1427 self.pendingcr = False
1428 if self.decoder is not None:
1429 self.decoder.reset()
1430
1431 _LF = 1
1432 _CR = 2
1433 _CRLF = 4
1434
1435 @property
1436 def newlines(self):
1437 return (None,
1438 "\n",
1439 "\r",
1440 ("\r", "\n"),
1441 "\r\n",
1442 ("\n", "\r\n"),
1443 ("\r", "\r\n"),
1444 ("\r", "\n", "\r\n")
1445 )[self.seennl]
1446
1447
1448class TextIOWrapper(TextIOBase):
1449
1450 r"""Character and line based layer over a BufferedIOBase object, buffer.
1451
1452 encoding gives the name of the encoding that the stream will be
1453 decoded or encoded with. It defaults to locale.getpreferredencoding.
1454
1455 errors determines the strictness of encoding and decoding (see the
1456 codecs.register) and defaults to "strict".
1457
1458 newline can be None, '', '\n', '\r', or '\r\n'. It controls the
1459 handling of line endings. If it is None, universal newlines is
1460 enabled. With this enabled, on input, the lines endings '\n', '\r',
1461 or '\r\n' are translated to '\n' before being returned to the
1462 caller. Conversely, on output, '\n' is translated to the system
Éric Araujo39242302011-11-03 00:08:48 +01001463 default line separator, os.linesep. If newline is any other of its
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001464 legal values, that newline becomes the newline when the file is read
1465 and it is returned untranslated. On output, '\n' is converted to the
1466 newline.
1467
1468 If line_buffering is True, a call to flush is implied when a call to
1469 write contains a newline character.
1470 """
1471
1472 _CHUNK_SIZE = 2048
1473
1474 def __init__(self, buffer, encoding=None, errors=None, newline=None,
Antoine Pitroue96ec682011-07-23 21:46:35 +02001475 line_buffering=False, write_through=False):
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001476 if newline is not None and not isinstance(newline, str):
1477 raise TypeError("illegal newline type: %r" % (type(newline),))
1478 if newline not in (None, "", "\n", "\r", "\r\n"):
1479 raise ValueError("illegal newline value: %r" % (newline,))
1480 if encoding is None:
1481 try:
1482 encoding = os.device_encoding(buffer.fileno())
1483 except (AttributeError, UnsupportedOperation):
1484 pass
1485 if encoding is None:
1486 try:
1487 import locale
1488 except ImportError:
1489 # Importing locale may fail if Python is being built
1490 encoding = "ascii"
1491 else:
1492 encoding = locale.getpreferredencoding()
1493
1494 if not isinstance(encoding, str):
1495 raise ValueError("invalid encoding: %r" % encoding)
1496
1497 if errors is None:
1498 errors = "strict"
1499 else:
1500 if not isinstance(errors, str):
1501 raise ValueError("invalid errors: %r" % errors)
1502
Antoine Pitrou7f8f4182010-12-21 21:20:59 +00001503 self._buffer = buffer
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001504 self._line_buffering = line_buffering
1505 self._encoding = encoding
1506 self._errors = errors
1507 self._readuniversal = not newline
1508 self._readtranslate = newline is None
1509 self._readnl = newline
1510 self._writetranslate = newline != ''
1511 self._writenl = newline or os.linesep
1512 self._encoder = None
1513 self._decoder = None
1514 self._decoded_chars = '' # buffer for text returned from decoder
1515 self._decoded_chars_used = 0 # offset into _decoded_chars for read()
1516 self._snapshot = None # info for reconstructing decoder state
1517 self._seekable = self._telling = self.buffer.seekable()
Antoine Pitroue96ec682011-07-23 21:46:35 +02001518 self._has_read1 = hasattr(self.buffer, 'read1')
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001519
Antoine Pitroue4501852009-05-14 18:55:55 +00001520 if self._seekable and self.writable():
1521 position = self.buffer.tell()
1522 if position != 0:
1523 try:
1524 self._get_encoder().setstate(0)
1525 except LookupError:
1526 # Sometimes the encoder doesn't exist
1527 pass
1528
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001529 # self._snapshot is either None, or a tuple (dec_flags, next_input)
1530 # where dec_flags is the second (integer) item of the decoder state
1531 # and next_input is the chunk of input bytes that comes next after the
1532 # snapshot point. We use this to reconstruct decoder states in tell().
1533
1534 # Naming convention:
1535 # - "bytes_..." for integer variables that count input bytes
1536 # - "chars_..." for integer variables that count decoded characters
1537
Benjamin Petersonc4c0eae2009-03-09 00:07:03 +00001538 def __repr__(self):
Antoine Pitroua4815ca2011-01-09 20:38:15 +00001539 result = "<_pyio.TextIOWrapper"
Antoine Pitrou716c4442009-05-23 19:04:03 +00001540 try:
1541 name = self.name
1542 except AttributeError:
Antoine Pitroua4815ca2011-01-09 20:38:15 +00001543 pass
Antoine Pitrou716c4442009-05-23 19:04:03 +00001544 else:
Antoine Pitroua4815ca2011-01-09 20:38:15 +00001545 result += " name={0!r}".format(name)
1546 try:
1547 mode = self.mode
1548 except AttributeError:
1549 pass
1550 else:
1551 result += " mode={0!r}".format(mode)
1552 return result + " encoding={0!r}>".format(self.encoding)
Benjamin Petersonc4c0eae2009-03-09 00:07:03 +00001553
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001554 @property
1555 def encoding(self):
1556 return self._encoding
1557
1558 @property
1559 def errors(self):
1560 return self._errors
1561
1562 @property
1563 def line_buffering(self):
1564 return self._line_buffering
1565
Antoine Pitrou7f8f4182010-12-21 21:20:59 +00001566 @property
1567 def buffer(self):
1568 return self._buffer
1569
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001570 def seekable(self):
1571 return self._seekable
1572
1573 def readable(self):
1574 return self.buffer.readable()
1575
1576 def writable(self):
1577 return self.buffer.writable()
1578
1579 def flush(self):
1580 self.buffer.flush()
1581 self._telling = self._seekable
1582
1583 def close(self):
Antoine Pitrou6be88762010-05-03 16:48:20 +00001584 if self.buffer is not None and not self.closed:
1585 self.flush()
Benjamin Petersond2e0c792009-05-01 20:40:59 +00001586 self.buffer.close()
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001587
1588 @property
1589 def closed(self):
1590 return self.buffer.closed
1591
1592 @property
1593 def name(self):
1594 return self.buffer.name
1595
1596 def fileno(self):
1597 return self.buffer.fileno()
1598
1599 def isatty(self):
1600 return self.buffer.isatty()
1601
Raymond Hettinger00fa0392011-01-13 02:52:26 +00001602 def write(self, s):
Raymond Hettingercbb80892011-01-13 18:15:51 +00001603 'Write data, where s is a str'
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001604 if self.closed:
1605 raise ValueError("write to closed file")
1606 if not isinstance(s, str):
1607 raise TypeError("can't write %s to text stream" %
1608 s.__class__.__name__)
1609 length = len(s)
1610 haslf = (self._writetranslate or self._line_buffering) and "\n" in s
1611 if haslf and self._writetranslate and self._writenl != "\n":
1612 s = s.replace("\n", self._writenl)
1613 encoder = self._encoder or self._get_encoder()
1614 # XXX What if we were just reading?
1615 b = encoder.encode(s)
1616 self.buffer.write(b)
1617 if self._line_buffering and (haslf or "\r" in s):
1618 self.flush()
1619 self._snapshot = None
1620 if self._decoder:
1621 self._decoder.reset()
1622 return length
1623
1624 def _get_encoder(self):
1625 make_encoder = codecs.getincrementalencoder(self._encoding)
1626 self._encoder = make_encoder(self._errors)
1627 return self._encoder
1628
1629 def _get_decoder(self):
1630 make_decoder = codecs.getincrementaldecoder(self._encoding)
1631 decoder = make_decoder(self._errors)
1632 if self._readuniversal:
1633 decoder = IncrementalNewlineDecoder(decoder, self._readtranslate)
1634 self._decoder = decoder
1635 return decoder
1636
1637 # The following three methods implement an ADT for _decoded_chars.
1638 # Text returned from the decoder is buffered here until the client
1639 # requests it by calling our read() or readline() method.
1640 def _set_decoded_chars(self, chars):
1641 """Set the _decoded_chars buffer."""
1642 self._decoded_chars = chars
1643 self._decoded_chars_used = 0
1644
1645 def _get_decoded_chars(self, n=None):
1646 """Advance into the _decoded_chars buffer."""
1647 offset = self._decoded_chars_used
1648 if n is None:
1649 chars = self._decoded_chars[offset:]
1650 else:
1651 chars = self._decoded_chars[offset:offset + n]
1652 self._decoded_chars_used += len(chars)
1653 return chars
1654
1655 def _rewind_decoded_chars(self, n):
1656 """Rewind the _decoded_chars buffer."""
1657 if self._decoded_chars_used < n:
1658 raise AssertionError("rewind decoded_chars out of bounds")
1659 self._decoded_chars_used -= n
1660
1661 def _read_chunk(self):
1662 """
1663 Read and decode the next chunk of data from the BufferedReader.
1664 """
1665
1666 # The return value is True unless EOF was reached. The decoded
1667 # string is placed in self._decoded_chars (replacing its previous
1668 # value). The entire input chunk is sent to the decoder, though
1669 # some of it may remain buffered in the decoder, yet to be
1670 # converted.
1671
1672 if self._decoder is None:
1673 raise ValueError("no decoder")
1674
1675 if self._telling:
1676 # To prepare for tell(), we need to snapshot a point in the
1677 # file where the decoder's input buffer is empty.
1678
1679 dec_buffer, dec_flags = self._decoder.getstate()
1680 # Given this, we know there was a valid snapshot point
1681 # len(dec_buffer) bytes ago with decoder state (b'', dec_flags).
1682
1683 # Read a chunk, decode it, and put the result in self._decoded_chars.
Antoine Pitroue96ec682011-07-23 21:46:35 +02001684 if self._has_read1:
1685 input_chunk = self.buffer.read1(self._CHUNK_SIZE)
1686 else:
1687 input_chunk = self.buffer.read(self._CHUNK_SIZE)
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001688 eof = not input_chunk
1689 self._set_decoded_chars(self._decoder.decode(input_chunk, eof))
1690
1691 if self._telling:
1692 # At the snapshot point, len(dec_buffer) bytes before the read,
1693 # the next input to be decoded is dec_buffer + input_chunk.
1694 self._snapshot = (dec_flags, dec_buffer + input_chunk)
1695
1696 return not eof
1697
1698 def _pack_cookie(self, position, dec_flags=0,
1699 bytes_to_feed=0, need_eof=0, chars_to_skip=0):
1700 # The meaning of a tell() cookie is: seek to position, set the
1701 # decoder flags to dec_flags, read bytes_to_feed bytes, feed them
1702 # into the decoder with need_eof as the EOF flag, then skip
1703 # chars_to_skip characters of the decoded result. For most simple
1704 # decoders, tell() will often just give a byte offset in the file.
1705 return (position | (dec_flags<<64) | (bytes_to_feed<<128) |
1706 (chars_to_skip<<192) | bool(need_eof)<<256)
1707
1708 def _unpack_cookie(self, bigint):
1709 rest, position = divmod(bigint, 1<<64)
1710 rest, dec_flags = divmod(rest, 1<<64)
1711 rest, bytes_to_feed = divmod(rest, 1<<64)
1712 need_eof, chars_to_skip = divmod(rest, 1<<64)
1713 return position, dec_flags, bytes_to_feed, need_eof, chars_to_skip
1714
1715 def tell(self):
1716 if not self._seekable:
Antoine Pitrou0d739d72010-09-05 23:01:12 +00001717 raise UnsupportedOperation("underlying stream is not seekable")
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001718 if not self._telling:
1719 raise IOError("telling position disabled by next() call")
1720 self.flush()
1721 position = self.buffer.tell()
1722 decoder = self._decoder
1723 if decoder is None or self._snapshot is None:
1724 if self._decoded_chars:
1725 # This should never happen.
1726 raise AssertionError("pending decoded text")
1727 return position
1728
1729 # Skip backward to the snapshot point (see _read_chunk).
1730 dec_flags, next_input = self._snapshot
1731 position -= len(next_input)
1732
1733 # How many decoded characters have been used up since the snapshot?
1734 chars_to_skip = self._decoded_chars_used
1735 if chars_to_skip == 0:
1736 # We haven't moved from the snapshot point.
1737 return self._pack_cookie(position, dec_flags)
1738
1739 # Starting from the snapshot position, we will walk the decoder
1740 # forward until it gives us enough decoded characters.
1741 saved_state = decoder.getstate()
1742 try:
1743 # Note our initial start point.
1744 decoder.setstate((b'', dec_flags))
1745 start_pos = position
1746 start_flags, bytes_fed, chars_decoded = dec_flags, 0, 0
1747 need_eof = 0
1748
1749 # Feed the decoder one byte at a time. As we go, note the
1750 # nearest "safe start point" before the current location
1751 # (a point where the decoder has nothing buffered, so seek()
1752 # can safely start from there and advance to this location).
1753 next_byte = bytearray(1)
1754 for next_byte[0] in next_input:
1755 bytes_fed += 1
1756 chars_decoded += len(decoder.decode(next_byte))
1757 dec_buffer, dec_flags = decoder.getstate()
1758 if not dec_buffer and chars_decoded <= chars_to_skip:
1759 # Decoder buffer is empty, so this is a safe start point.
1760 start_pos += bytes_fed
1761 chars_to_skip -= chars_decoded
1762 start_flags, bytes_fed, chars_decoded = dec_flags, 0, 0
1763 if chars_decoded >= chars_to_skip:
1764 break
1765 else:
1766 # We didn't get enough decoded data; signal EOF to get more.
1767 chars_decoded += len(decoder.decode(b'', final=True))
1768 need_eof = 1
1769 if chars_decoded < chars_to_skip:
1770 raise IOError("can't reconstruct logical file position")
1771
1772 # The returned cookie corresponds to the last safe start point.
1773 return self._pack_cookie(
1774 start_pos, start_flags, bytes_fed, need_eof, chars_to_skip)
1775 finally:
1776 decoder.setstate(saved_state)
1777
1778 def truncate(self, pos=None):
1779 self.flush()
1780 if pos is None:
1781 pos = self.tell()
Antoine Pitrou905a2ff2010-01-31 22:47:27 +00001782 return self.buffer.truncate(pos)
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001783
Benjamin Petersond2e0c792009-05-01 20:40:59 +00001784 def detach(self):
1785 if self.buffer is None:
1786 raise ValueError("buffer is already detached")
1787 self.flush()
Antoine Pitrou7f8f4182010-12-21 21:20:59 +00001788 buffer = self._buffer
1789 self._buffer = None
Benjamin Petersond2e0c792009-05-01 20:40:59 +00001790 return buffer
1791
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001792 def seek(self, cookie, whence=0):
1793 if self.closed:
1794 raise ValueError("tell on closed file")
1795 if not self._seekable:
Antoine Pitrou0d739d72010-09-05 23:01:12 +00001796 raise UnsupportedOperation("underlying stream is not seekable")
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001797 if whence == 1: # seek relative to current position
1798 if cookie != 0:
Antoine Pitrou0d739d72010-09-05 23:01:12 +00001799 raise UnsupportedOperation("can't do nonzero cur-relative seeks")
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001800 # Seeking to the current position should attempt to
1801 # sync the underlying buffer with the current position.
1802 whence = 0
1803 cookie = self.tell()
1804 if whence == 2: # seek relative to end of file
1805 if cookie != 0:
Antoine Pitrou0d739d72010-09-05 23:01:12 +00001806 raise UnsupportedOperation("can't do nonzero end-relative seeks")
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001807 self.flush()
1808 position = self.buffer.seek(0, 2)
1809 self._set_decoded_chars('')
1810 self._snapshot = None
1811 if self._decoder:
1812 self._decoder.reset()
1813 return position
1814 if whence != 0:
1815 raise ValueError("invalid whence (%r, should be 0, 1 or 2)" %
1816 (whence,))
1817 if cookie < 0:
1818 raise ValueError("negative seek position %r" % (cookie,))
1819 self.flush()
1820
1821 # The strategy of seek() is to go back to the safe start point
1822 # and replay the effect of read(chars_to_skip) from there.
1823 start_pos, dec_flags, bytes_to_feed, need_eof, chars_to_skip = \
1824 self._unpack_cookie(cookie)
1825
1826 # Seek back to the safe start point.
1827 self.buffer.seek(start_pos)
1828 self._set_decoded_chars('')
1829 self._snapshot = None
1830
1831 # Restore the decoder to its state from the safe start point.
Benjamin Peterson9363a652009-03-05 00:42:09 +00001832 if cookie == 0 and self._decoder:
1833 self._decoder.reset()
1834 elif self._decoder or dec_flags or chars_to_skip:
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001835 self._decoder = self._decoder or self._get_decoder()
1836 self._decoder.setstate((b'', dec_flags))
1837 self._snapshot = (dec_flags, b'')
1838
1839 if chars_to_skip:
1840 # Just like _read_chunk, feed the decoder and save a snapshot.
1841 input_chunk = self.buffer.read(bytes_to_feed)
1842 self._set_decoded_chars(
1843 self._decoder.decode(input_chunk, need_eof))
1844 self._snapshot = (dec_flags, input_chunk)
1845
1846 # Skip chars_to_skip of the decoded characters.
1847 if len(self._decoded_chars) < chars_to_skip:
1848 raise IOError("can't restore logical file position")
1849 self._decoded_chars_used = chars_to_skip
1850
Antoine Pitroue4501852009-05-14 18:55:55 +00001851 # Finally, reset the encoder (merely useful for proper BOM handling)
1852 try:
1853 encoder = self._encoder or self._get_encoder()
1854 except LookupError:
1855 # Sometimes the encoder doesn't exist
1856 pass
1857 else:
1858 if cookie != 0:
1859 encoder.setstate(0)
1860 else:
1861 encoder.reset()
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001862 return cookie
1863
1864 def read(self, n=None):
Benjamin Petersona1b49012009-03-31 23:11:32 +00001865 self._checkReadable()
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001866 if n is None:
1867 n = -1
1868 decoder = self._decoder or self._get_decoder()
Florent Xiclunab14930c2010-03-13 15:26:44 +00001869 try:
1870 n.__index__
1871 except AttributeError as err:
1872 raise TypeError("an integer is required") from err
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001873 if n < 0:
1874 # Read everything.
1875 result = (self._get_decoded_chars() +
1876 decoder.decode(self.buffer.read(), final=True))
1877 self._set_decoded_chars('')
1878 self._snapshot = None
1879 return result
1880 else:
1881 # Keep reading chunks until we have n characters to return.
1882 eof = False
1883 result = self._get_decoded_chars(n)
1884 while len(result) < n and not eof:
1885 eof = not self._read_chunk()
1886 result += self._get_decoded_chars(n - len(result))
1887 return result
1888
1889 def __next__(self):
1890 self._telling = False
1891 line = self.readline()
1892 if not line:
1893 self._snapshot = None
1894 self._telling = self._seekable
1895 raise StopIteration
1896 return line
1897
1898 def readline(self, limit=None):
1899 if self.closed:
1900 raise ValueError("read from closed file")
1901 if limit is None:
1902 limit = -1
Benjamin Petersonb01138a2009-04-24 22:59:52 +00001903 elif not isinstance(limit, int):
1904 raise TypeError("limit must be an integer")
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001905
1906 # Grab all the decoded text (we will rewind any extra bits later).
1907 line = self._get_decoded_chars()
1908
1909 start = 0
1910 # Make the decoder if it doesn't already exist.
1911 if not self._decoder:
1912 self._get_decoder()
1913
1914 pos = endpos = None
1915 while True:
1916 if self._readtranslate:
1917 # Newlines are already translated, only search for \n
1918 pos = line.find('\n', start)
1919 if pos >= 0:
1920 endpos = pos + 1
1921 break
1922 else:
1923 start = len(line)
1924
1925 elif self._readuniversal:
1926 # Universal newline search. Find any of \r, \r\n, \n
1927 # The decoder ensures that \r\n are not split in two pieces
1928
1929 # In C we'd look for these in parallel of course.
1930 nlpos = line.find("\n", start)
1931 crpos = line.find("\r", start)
1932 if crpos == -1:
1933 if nlpos == -1:
1934 # Nothing found
1935 start = len(line)
1936 else:
1937 # Found \n
1938 endpos = nlpos + 1
1939 break
1940 elif nlpos == -1:
1941 # Found lone \r
1942 endpos = crpos + 1
1943 break
1944 elif nlpos < crpos:
1945 # Found \n
1946 endpos = nlpos + 1
1947 break
1948 elif nlpos == crpos + 1:
1949 # Found \r\n
1950 endpos = crpos + 2
1951 break
1952 else:
1953 # Found \r
1954 endpos = crpos + 1
1955 break
1956 else:
1957 # non-universal
1958 pos = line.find(self._readnl)
1959 if pos >= 0:
1960 endpos = pos + len(self._readnl)
1961 break
1962
1963 if limit >= 0 and len(line) >= limit:
1964 endpos = limit # reached length limit
1965 break
1966
1967 # No line ending seen yet - get more data'
1968 while self._read_chunk():
1969 if self._decoded_chars:
1970 break
1971 if self._decoded_chars:
1972 line += self._get_decoded_chars()
1973 else:
1974 # end of file
1975 self._set_decoded_chars('')
1976 self._snapshot = None
1977 return line
1978
1979 if limit >= 0 and endpos > limit:
1980 endpos = limit # don't exceed limit
1981
1982 # Rewind _decoded_chars to just after the line ending we found.
1983 self._rewind_decoded_chars(len(line) - endpos)
1984 return line[:endpos]
1985
1986 @property
1987 def newlines(self):
1988 return self._decoder.newlines if self._decoder else None
1989
1990
1991class StringIO(TextIOWrapper):
1992 """Text I/O implementation using an in-memory buffer.
1993
1994 The initial_value argument sets the value of object. The newline
1995 argument is like the one of TextIOWrapper's constructor.
1996 """
1997
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00001998 def __init__(self, initial_value="", newline="\n"):
1999 super(StringIO, self).__init__(BytesIO(),
2000 encoding="utf-8",
2001 errors="strict",
2002 newline=newline)
Antoine Pitrou11446482009-04-04 14:09:30 +00002003 # Issue #5645: make universal newlines semantics the same as in the
2004 # C version, even under Windows.
2005 if newline is None:
2006 self._writetranslate = False
Alexandre Vassalottid2bb18b2009-07-22 03:07:33 +00002007 if initial_value is not None:
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00002008 if not isinstance(initial_value, str):
Alexandre Vassalottid2bb18b2009-07-22 03:07:33 +00002009 raise TypeError("initial_value must be str or None, not {0}"
2010 .format(type(initial_value).__name__))
Benjamin Peterson4fa88fa2009-03-04 00:14:51 +00002011 initial_value = str(initial_value)
2012 self.write(initial_value)
2013 self.seek(0)
2014
2015 def getvalue(self):
2016 self.flush()
2017 return self.buffer.getvalue().decode(self._encoding, self._errors)
Benjamin Peterson9fd459a2009-03-09 00:09:44 +00002018
2019 def __repr__(self):
2020 # TextIOWrapper tells the encoding in its repr. In StringIO,
2021 # that's a implementation detail.
2022 return object.__repr__(self)
Benjamin Petersonb487e632009-03-21 03:08:31 +00002023
2024 @property
Benjamin Peterson0926ad12009-06-06 18:02:12 +00002025 def errors(self):
2026 return None
2027
2028 @property
Benjamin Petersonb487e632009-03-21 03:08:31 +00002029 def encoding(self):
2030 return None
Benjamin Petersond2e0c792009-05-01 20:40:59 +00002031
2032 def detach(self):
2033 # This doesn't make sense on StringIO.
2034 self._unsupported("detach")