blob: c45062da9979fdcb7e8a10e2729f4bec3dbe9f73 [file] [log] [blame]
Christian Heimes1a6387e2008-03-26 12:49:49 +00001"""
Benjamin Peterson7bb4d2d2008-04-13 02:01:27 +00002The io module provides the Python interfaces to stream handling. The
3builtin open function is defined in this module.
4
5At the top of the I/O hierarchy is the abstract base class IOBase. It
6defines the basic interface to a stream. Note, however, that there is no
7seperation between reading and writing to streams; implementations are
8allowed to throw an IOError if they do not support a given operation.
9
10Extending IOBase is RawIOBase which deals simply with the reading and
11writing of raw bytes to a stream. FileIO subclasses RawIOBase to provide
12an interface to OS files.
13
14BufferedIOBase deals with buffering on a raw byte stream (RawIOBase). Its
15subclasses, BufferedWriter, BufferedReader, and BufferedRWPair buffer
16streams that are readable, writable, and both respectively.
17BufferedRandom provides a buffered interface to random access
18streams. BytesIO is a simple stream of in-memory bytes.
19
20Another IOBase subclass, TextIOBase, deals with the encoding and decoding
21of streams into text. TextIOWrapper, which extends it, is a buffered text
22interface to a buffered raw stream (`BufferedIOBase`). Finally, StringIO
23is a in-memory stream for text.
24
25Argument names are not part of the specification, and only the arguments
26of open() are intended to be used as keyword arguments.
27
28data:
29
30DEFAULT_BUFFER_SIZE
31
32 An int containing the default buffer size used by the module's buffered
33 I/O classes. open() uses the file's blksize (as obtained by os.stat) if
34 possible.
35"""
36# New I/O library conforming to PEP 3116.
37
38# This is a prototype; hopefully eventually some of this will be
39# reimplemented in C.
40
41# XXX edge cases when switching between reading/writing
42# XXX need to support 1 meaning line-buffered
43# XXX whenever an argument is None, use the default value
44# XXX read/write ops should check readable/writable
45# XXX buffered readinto should work with arbitrary buffer objects
46# XXX use incremental encoder for text output, at least for UTF-16 and UTF-8-SIG
47# XXX check writable, readable and seekable in appropriate places
Christian Heimes3784c6b2008-03-26 23:13:59 +000048from __future__ import print_function
49from __future__ import unicode_literals
Christian Heimes1a6387e2008-03-26 12:49:49 +000050
51__author__ = ("Guido van Rossum <guido@python.org>, "
52 "Mike Verdone <mike.verdone@gmail.com>, "
53 "Mark Russell <mark.russell@zen.co.uk>")
54
55__all__ = ["BlockingIOError", "open", "IOBase", "RawIOBase", "FileIO",
56 "BytesIO", "StringIO", "BufferedIOBase",
57 "BufferedReader", "BufferedWriter", "BufferedRWPair",
58 "BufferedRandom", "TextIOBase", "TextIOWrapper"]
59
60import os
61import abc
62import sys
63import codecs
64import _fileio
65import warnings
66
67# open() uses st_blksize whenever we can
68DEFAULT_BUFFER_SIZE = 8 * 1024 # bytes
69
70# py3k has only new style classes
71__metaclass__ = type
72
73class BlockingIOError(IOError):
74
75 """Exception raised when I/O would block on a non-blocking I/O stream."""
76
77 def __init__(self, errno, strerror, characters_written=0):
78 IOError.__init__(self, errno, strerror)
79 self.characters_written = characters_written
80
81
82def open(file, mode="r", buffering=None, encoding=None, errors=None,
83 newline=None, closefd=True):
Benjamin Peterson9ae080e2008-05-04 22:39:33 +000084 r"""Open file and return a stream. If the file cannot be opened, an IOError is
85 raised.
Christian Heimes1a6387e2008-03-26 12:49:49 +000086
Benjamin Peterson7bb4d2d2008-04-13 02:01:27 +000087 file is either a string giving the name (and the path if the file
88 isn't in the current working directory) of the file to be opened or an
89 integer file descriptor of the file to be wrapped. (If a file
90 descriptor is given, it is closed when the returned I/O object is
91 closed, unless closefd is set to False.)
Christian Heimes1a6387e2008-03-26 12:49:49 +000092
Benjamin Peterson7bb4d2d2008-04-13 02:01:27 +000093 mode is an optional string that specifies the mode in which the file
94 is opened. It defaults to 'r' which means open for reading in text
95 mode. Other common values are 'w' for writing (truncating the file if
96 it already exists), and 'a' for appending (which on some Unix systems,
97 means that all writes append to the end of the file regardless of the
98 current seek position). In text mode, if encoding is not specified the
99 encoding used is platform dependent. (For reading and writing raw
100 bytes use binary mode and leave encoding unspecified.) The available
101 modes are:
Christian Heimes1a6387e2008-03-26 12:49:49 +0000102
Benjamin Peterson7bb4d2d2008-04-13 02:01:27 +0000103 ========= ===============================================================
104 Character Meaning
105 --------- ---------------------------------------------------------------
106 'r' open for reading (default)
107 'w' open for writing, truncating the file first
108 'a' open for writing, appending to the end of the file if it exists
109 'b' binary mode
110 't' text mode (default)
111 '+' open a disk file for updating (reading and writing)
112 'U' universal newline mode (for backwards compatibility; unneeded
113 for new code)
114 ========= ===============================================================
Christian Heimes1a6387e2008-03-26 12:49:49 +0000115
Benjamin Peterson7bb4d2d2008-04-13 02:01:27 +0000116 The default mode is 'rt' (open for reading text). For binary random
117 access, the mode 'w+b' opens and truncates the file to 0 bytes, while
118 'r+b' opens the file without truncation.
Christian Heimes1a6387e2008-03-26 12:49:49 +0000119
Benjamin Peterson7bb4d2d2008-04-13 02:01:27 +0000120 Python distinguishes between files opened in binary and text modes,
121 even when the underlying operating system doesn't. Files opened in
122 binary mode (appending 'b' to the mode argument) return contents as
123 bytes objects without any decoding. In text mode (the default, or when
124 't' is appended to the mode argument), the contents of the file are
125 returned as strings, the bytes having been first decoded using a
126 platform-dependent encoding or using the specified encoding if given.
Christian Heimes1a6387e2008-03-26 12:49:49 +0000127
Benjamin Peterson7bb4d2d2008-04-13 02:01:27 +0000128 buffering is an optional integer used to set the buffering policy. By
129 default full buffering is on. Pass 0 to switch buffering off (only
130 allowed in binary mode), 1 to set line buffering, and an integer > 1
131 for full buffering.
Christian Heimes1a6387e2008-03-26 12:49:49 +0000132
Benjamin Peterson7bb4d2d2008-04-13 02:01:27 +0000133 encoding is the name of the encoding used to decode or encode the
134 file. This should only be used in text mode. The default encoding is
135 platform dependent, but any encoding supported by Python can be
136 passed. See the codecs module for the list of supported encodings.
Christian Heimes1a6387e2008-03-26 12:49:49 +0000137
Benjamin Peterson7bb4d2d2008-04-13 02:01:27 +0000138 errors is an optional string that specifies how encoding errors are to
139 be handled---this argument should not be used in binary mode. Pass
140 'strict' to raise a ValueError exception if there is an encoding error
141 (the default of None has the same effect), or pass 'ignore' to ignore
142 errors. (Note that ignoring encoding errors can lead to data loss.)
143 See the documentation for codecs.register for a list of the permitted
144 encoding error strings.
145
146 newline controls how universal newlines works (it only applies to text
147 mode). It can be None, '', '\n', '\r', and '\r\n'. It works as
148 follows:
149
150 * On input, if newline is None, universal newlines mode is
151 enabled. Lines in the input can end in '\n', '\r', or '\r\n', and
152 these are translated into '\n' before being returned to the
153 caller. If it is '', universal newline mode is enabled, but line
154 endings are returned to the caller untranslated. If it has any of
155 the other legal values, input lines are only terminated by the given
156 string, and the line ending is returned to the caller untranslated.
157
158 * On output, if newline is None, any '\n' characters written are
159 translated to the system default line separator, os.linesep. If
160 newline is '', no translation takes place. If newline is any of the
161 other legal values, any '\n' characters written are translated to
162 the given string.
163
164 If closefd is False, the underlying file descriptor will be kept open
165 when the file is closed. This does not work when a file name is given
166 and must be True in that case.
167
168 open() returns a file object whose type depends on the mode, and
169 through which the standard file operations such as reading and writing
170 are performed. When open() is used to open a file in a text mode ('w',
171 'r', 'wt', 'rt', etc.), it returns a TextIOWrapper. When used to open
172 a file in a binary mode, the returned class varies: in read binary
173 mode, it returns a BufferedReader; in write binary and append binary
174 modes, it returns a BufferedWriter, and in read/write mode, it returns
175 a BufferedRandom.
176
177 It is also possible to use a string or bytearray as a file for both
178 reading and writing. For strings StringIO can be used like a file
179 opened in a text mode, and for bytes a BytesIO can be used like a file
180 opened in a binary mode.
Christian Heimes1a6387e2008-03-26 12:49:49 +0000181 """
Christian Heimes3784c6b2008-03-26 23:13:59 +0000182 if not isinstance(file, (basestring, int)):
Christian Heimes1a6387e2008-03-26 12:49:49 +0000183 raise TypeError("invalid file: %r" % file)
Christian Heimes3784c6b2008-03-26 23:13:59 +0000184 if not isinstance(mode, basestring):
Christian Heimes1a6387e2008-03-26 12:49:49 +0000185 raise TypeError("invalid mode: %r" % mode)
186 if buffering is not None and not isinstance(buffering, int):
187 raise TypeError("invalid buffering: %r" % buffering)
Christian Heimes3784c6b2008-03-26 23:13:59 +0000188 if encoding is not None and not isinstance(encoding, basestring):
Christian Heimes1a6387e2008-03-26 12:49:49 +0000189 raise TypeError("invalid encoding: %r" % encoding)
Christian Heimes3784c6b2008-03-26 23:13:59 +0000190 if errors is not None and not isinstance(errors, basestring):
Christian Heimes1a6387e2008-03-26 12:49:49 +0000191 raise TypeError("invalid errors: %r" % errors)
192 modes = set(mode)
193 if modes - set("arwb+tU") or len(mode) > len(modes):
194 raise ValueError("invalid mode: %r" % mode)
195 reading = "r" in modes
196 writing = "w" in modes
197 appending = "a" in modes
198 updating = "+" in modes
199 text = "t" in modes
200 binary = "b" in modes
201 if "U" in modes:
202 if writing or appending:
203 raise ValueError("can't use U and writing mode at once")
204 reading = True
205 if text and binary:
206 raise ValueError("can't have text and binary mode at once")
207 if reading + writing + appending > 1:
208 raise ValueError("can't have read/write/append mode at once")
209 if not (reading or writing or appending):
210 raise ValueError("must have exactly one of read/write/append mode")
211 if binary and encoding is not None:
212 raise ValueError("binary mode doesn't take an encoding argument")
213 if binary and errors is not None:
214 raise ValueError("binary mode doesn't take an errors argument")
215 if binary and newline is not None:
216 raise ValueError("binary mode doesn't take a newline argument")
217 raw = FileIO(file,
218 (reading and "r" or "") +
219 (writing and "w" or "") +
220 (appending and "a" or "") +
221 (updating and "+" or ""),
222 closefd)
223 if buffering is None:
224 buffering = -1
225 line_buffering = False
226 if buffering == 1 or buffering < 0 and raw.isatty():
227 buffering = -1
228 line_buffering = True
229 if buffering < 0:
230 buffering = DEFAULT_BUFFER_SIZE
231 try:
232 bs = os.fstat(raw.fileno()).st_blksize
233 except (os.error, AttributeError):
234 pass
235 else:
236 if bs > 1:
237 buffering = bs
238 if buffering < 0:
239 raise ValueError("invalid buffering size")
240 if buffering == 0:
241 if binary:
242 raw._name = file
243 raw._mode = mode
244 return raw
245 raise ValueError("can't have unbuffered text I/O")
246 if updating:
247 buffer = BufferedRandom(raw, buffering)
248 elif writing or appending:
249 buffer = BufferedWriter(raw, buffering)
250 elif reading:
251 buffer = BufferedReader(raw, buffering)
252 else:
253 raise ValueError("unknown mode: %r" % mode)
254 if binary:
255 buffer.name = file
256 buffer.mode = mode
257 return buffer
258 text = TextIOWrapper(buffer, encoding, errors, newline, line_buffering)
259 text.name = file
260 text.mode = mode
261 return text
262
263class _DocDescriptor:
264 """Helper for builtins.open.__doc__
265 """
266 def __get__(self, obj, typ):
267 return (
268 "open(file, mode='r', buffering=None, encoding=None, "
269 "errors=None, newline=None, closefd=True)\n\n" +
270 open.__doc__)
271
272class OpenWrapper:
273 """Wrapper for builtins.open
274
275 Trick so that open won't become a bound method when stored
276 as a class variable (as dumbdbm does).
277
278 See initstdio() in Python/pythonrun.c.
279 """
280 __doc__ = _DocDescriptor()
281
282 def __new__(cls, *args, **kwargs):
283 return open(*args, **kwargs)
284
285
286class UnsupportedOperation(ValueError, IOError):
287 pass
288
289
290class IOBase(object):
291
Benjamin Peterson9ae080e2008-05-04 22:39:33 +0000292 """The abstract base class for all I/O classes, acting on streams of
Benjamin Peterson7bb4d2d2008-04-13 02:01:27 +0000293 bytes. There is no public constructor.
Christian Heimes1a6387e2008-03-26 12:49:49 +0000294
295 This class provides dummy implementations for many methods that
Benjamin Peterson7bb4d2d2008-04-13 02:01:27 +0000296 derived classes can override selectively; the default implementations
297 represent a file that cannot be read, written or seeked.
Christian Heimes1a6387e2008-03-26 12:49:49 +0000298
Benjamin Peterson7bb4d2d2008-04-13 02:01:27 +0000299 Even though IOBase does not declare read, readinto, or write because
300 their signatures will vary, implementations and clients should
301 consider those methods part of the interface. Also, implementations
302 may raise a IOError when operations they do not support are called.
Christian Heimes1a6387e2008-03-26 12:49:49 +0000303
Benjamin Peterson7bb4d2d2008-04-13 02:01:27 +0000304 The basic type used for binary data read from or written to a file is
305 bytes. bytearrays are accepted too, and in some cases (such as
306 readinto) needed. Text I/O classes work with str data.
307
308 Note that calling any method (even inquiries) on a closed stream is
309 undefined. Implementations may raise IOError in this case.
310
311 IOBase (and its subclasses) support the iterator protocol, meaning
312 that an IOBase object can be iterated over yielding the lines in a
313 stream.
314
315 IOBase also supports the :keyword:`with` statement. In this example,
316 fp is closed after the suite of the with statment is complete:
317
318 with open('spam.txt', 'r') as fp:
319 fp.write('Spam and eggs!')
Christian Heimes1a6387e2008-03-26 12:49:49 +0000320 """
321
322 __metaclass__ = abc.ABCMeta
323
324 ### Internal ###
325
326 def _unsupported(self, name):
327 """Internal: raise an exception for unsupported operations."""
328 raise UnsupportedOperation("%s.%s() not supported" %
329 (self.__class__.__name__, name))
330
331 ### Positioning ###
332
333 def seek(self, pos, whence = 0):
Benjamin Peterson9ae080e2008-05-04 22:39:33 +0000334 """Change stream position.
Christian Heimes1a6387e2008-03-26 12:49:49 +0000335
Benjamin Peterson7bb4d2d2008-04-13 02:01:27 +0000336 Change the stream position to byte offset offset. offset is
337 interpreted relative to the position indicated by whence. Values
338 for whence are:
339
340 * 0 -- start of stream (the default); offset should be zero or positive
341 * 1 -- current stream position; offset may be negative
342 * 2 -- end of stream; offset is usually negative
343
344 Return the new absolute position.
Christian Heimes1a6387e2008-03-26 12:49:49 +0000345 """
346 self._unsupported("seek")
347
348 def tell(self):
Benjamin Peterson9ae080e2008-05-04 22:39:33 +0000349 """Return current stream position."""
Christian Heimes1a6387e2008-03-26 12:49:49 +0000350 return self.seek(0, 1)
351
352 def truncate(self, pos = None):
Benjamin Peterson9ae080e2008-05-04 22:39:33 +0000353 """Truncate file to size bytes.
Christian Heimes1a6387e2008-03-26 12:49:49 +0000354
Benjamin Peterson9ae080e2008-05-04 22:39:33 +0000355 Size defaults to the current IO position as reported by tell(). Return
356 the new size.
Christian Heimes1a6387e2008-03-26 12:49:49 +0000357 """
358 self._unsupported("truncate")
359
360 ### Flush and close ###
361
362 def flush(self):
Benjamin Peterson9ae080e2008-05-04 22:39:33 +0000363 """Flush write buffers, if applicable.
Christian Heimes1a6387e2008-03-26 12:49:49 +0000364
Benjamin Peterson7bb4d2d2008-04-13 02:01:27 +0000365 This is not implemented for read-only and non-blocking streams.
Christian Heimes1a6387e2008-03-26 12:49:49 +0000366 """
367 # XXX Should this return the number of bytes written???
368
369 __closed = False
370
371 def close(self):
Benjamin Peterson9ae080e2008-05-04 22:39:33 +0000372 """Flush and close the IO object.
Christian Heimes1a6387e2008-03-26 12:49:49 +0000373
Benjamin Peterson7bb4d2d2008-04-13 02:01:27 +0000374 This method has no effect if the file is already closed.
Christian Heimes1a6387e2008-03-26 12:49:49 +0000375 """
376 if not self.__closed:
377 try:
378 self.flush()
379 except IOError:
380 pass # If flush() fails, just give up
381 self.__closed = True
382
383 def __del__(self):
384 """Destructor. Calls close()."""
385 # The try/except block is in case this is called at program
386 # exit time, when it's possible that globals have already been
387 # deleted, and then the close() call might fail. Since
388 # there's nothing we can do about such failures and they annoy
389 # the end users, we suppress the traceback.
390 try:
391 self.close()
392 except:
393 pass
394
395 ### Inquiries ###
396
397 def seekable(self):
Benjamin Peterson9ae080e2008-05-04 22:39:33 +0000398 """Return whether object supports random access.
Christian Heimes1a6387e2008-03-26 12:49:49 +0000399
400 If False, seek(), tell() and truncate() will raise IOError.
401 This method may need to do a test seek().
402 """
403 return False
404
405 def _checkSeekable(self, msg=None):
406 """Internal: raise an IOError if file is not seekable
407 """
408 if not self.seekable():
409 raise IOError("File or stream is not seekable."
410 if msg is None else msg)
411
412
413 def readable(self):
Benjamin Peterson9ae080e2008-05-04 22:39:33 +0000414 """Return whether object was opened for reading.
Christian Heimes1a6387e2008-03-26 12:49:49 +0000415
416 If False, read() will raise IOError.
417 """
418 return False
419
420 def _checkReadable(self, msg=None):
421 """Internal: raise an IOError if file is not readable
422 """
423 if not self.readable():
424 raise IOError("File or stream is not readable."
425 if msg is None else msg)
426
427 def writable(self):
Benjamin Peterson9ae080e2008-05-04 22:39:33 +0000428 """Return whether object was opened for writing.
Christian Heimes1a6387e2008-03-26 12:49:49 +0000429
430 If False, write() and truncate() will raise IOError.
431 """
432 return False
433
434 def _checkWritable(self, msg=None):
435 """Internal: raise an IOError if file is not writable
436 """
437 if not self.writable():
438 raise IOError("File or stream is not writable."
439 if msg is None else msg)
440
441 @property
442 def closed(self):
443 """closed: bool. True iff the file has been closed.
444
445 For backwards compatibility, this is a property, not a predicate.
446 """
447 return self.__closed
448
449 def _checkClosed(self, msg=None):
450 """Internal: raise an ValueError if file is closed
451 """
452 if self.closed:
453 raise ValueError("I/O operation on closed file."
454 if msg is None else msg)
455
456 ### Context manager ###
457
458 def __enter__(self):
459 """Context management protocol. Returns self."""
460 self._checkClosed()
461 return self
462
463 def __exit__(self, *args):
464 """Context management protocol. Calls close()"""
465 self.close()
466
467 ### Lower-level APIs ###
468
469 # XXX Should these be present even if unimplemented?
470
471 def fileno(self):
Benjamin Peterson9ae080e2008-05-04 22:39:33 +0000472 """Returns underlying file descriptor if one exists.
Christian Heimes1a6387e2008-03-26 12:49:49 +0000473
Benjamin Peterson9ae080e2008-05-04 22:39:33 +0000474 An IOError is raised if the IO object does not use a file descriptor.
Christian Heimes1a6387e2008-03-26 12:49:49 +0000475 """
476 self._unsupported("fileno")
477
478 def isatty(self):
Benjamin Peterson9ae080e2008-05-04 22:39:33 +0000479 """Return whether this is an 'interactive' stream.
Christian Heimes1a6387e2008-03-26 12:49:49 +0000480
Benjamin Peterson9ae080e2008-05-04 22:39:33 +0000481 Return False if it can't be determined.
Christian Heimes1a6387e2008-03-26 12:49:49 +0000482 """
483 self._checkClosed()
484 return False
485
486 ### Readline[s] and writelines ###
487
488 def readline(self, limit = -1):
Benjamin Peterson9ae080e2008-05-04 22:39:33 +0000489 r"""Read and return a line from the stream.
Benjamin Peterson7bb4d2d2008-04-13 02:01:27 +0000490
491 If limit is specified, at most limit bytes will be read.
492
493 The line terminator is always b'\n' for binary files; for text
494 files, the newlines argument to open can be used to select the line
495 terminator(s) recognized.
496 """
Alexandre Vassalotti1aed6242008-05-09 21:49:43 +0000497 self._checkClosed()
Christian Heimes1a6387e2008-03-26 12:49:49 +0000498 if hasattr(self, "peek"):
499 def nreadahead():
500 readahead = self.peek(1)
501 if not readahead:
502 return 1
503 n = (readahead.find(b"\n") + 1) or len(readahead)
504 if limit >= 0:
505 n = min(n, limit)
506 return n
507 else:
508 def nreadahead():
509 return 1
510 if limit is None:
511 limit = -1
Alexandre Vassalotti1aed6242008-05-09 21:49:43 +0000512 if not isinstance(limit, (int, long)):
513 raise TypeError("limit must be an integer")
Christian Heimes1a6387e2008-03-26 12:49:49 +0000514 res = bytearray()
515 while limit < 0 or len(res) < limit:
516 b = self.read(nreadahead())
517 if not b:
518 break
519 res += b
520 if res.endswith(b"\n"):
521 break
522 return bytes(res)
523
524 def __iter__(self):
525 self._checkClosed()
526 return self
527
528 def next(self):
529 line = self.readline()
530 if not line:
531 raise StopIteration
532 return line
533
534 def readlines(self, hint=None):
Benjamin Peterson9ae080e2008-05-04 22:39:33 +0000535 """Return a list of lines from the stream.
Benjamin Peterson7bb4d2d2008-04-13 02:01:27 +0000536
537 hint can be specified to control the number of lines read: no more
538 lines will be read if the total size (in bytes/characters) of all
539 lines so far exceeds hint.
540 """
Christian Heimes1a6387e2008-03-26 12:49:49 +0000541 if hint is None:
Alexandre Vassalotti1aed6242008-05-09 21:49:43 +0000542 hint = -1
543 if not isinstance(hint, (int, long)):
544 raise TypeError("hint must be an integer")
545 if hint <= 0:
Christian Heimes1a6387e2008-03-26 12:49:49 +0000546 return list(self)
547 n = 0
548 lines = []
549 for line in self:
550 lines.append(line)
551 n += len(line)
552 if n >= hint:
553 break
554 return lines
555
556 def writelines(self, lines):
557 self._checkClosed()
558 for line in lines:
559 self.write(line)
560
561
562class RawIOBase(IOBase):
563
Benjamin Peterson7bb4d2d2008-04-13 02:01:27 +0000564 """Base class for raw binary I/O."""
Christian Heimes1a6387e2008-03-26 12:49:49 +0000565
Benjamin Peterson7bb4d2d2008-04-13 02:01:27 +0000566 # The read() method is implemented by calling readinto(); derived
567 # classes that want to support read() only need to implement
568 # readinto() as a primitive operation. In general, readinto() can be
569 # more efficient than read().
Christian Heimes1a6387e2008-03-26 12:49:49 +0000570
Benjamin Peterson7bb4d2d2008-04-13 02:01:27 +0000571 # (It would be tempting to also provide an implementation of
572 # readinto() in terms of read(), in case the latter is a more suitable
573 # primitive operation, but that would lead to nasty recursion in case
574 # a subclass doesn't implement either.)
Christian Heimes1a6387e2008-03-26 12:49:49 +0000575
576 def read(self, n = -1):
Benjamin Peterson9ae080e2008-05-04 22:39:33 +0000577 """Read and return up to n bytes.
Christian Heimes1a6387e2008-03-26 12:49:49 +0000578
579 Returns an empty bytes array on EOF, or None if the object is
580 set not to block and has no data to read.
581 """
582 if n is None:
583 n = -1
584 if n < 0:
585 return self.readall()
586 b = bytearray(n.__index__())
587 n = self.readinto(b)
588 del b[n:]
589 return bytes(b)
590
591 def readall(self):
Benjamin Peterson9ae080e2008-05-04 22:39:33 +0000592 """Read until EOF, using multiple read() call."""
Christian Heimes1a6387e2008-03-26 12:49:49 +0000593 res = bytearray()
594 while True:
595 data = self.read(DEFAULT_BUFFER_SIZE)
596 if not data:
597 break
598 res += data
599 return bytes(res)
600
601 def readinto(self, b):
Benjamin Peterson9ae080e2008-05-04 22:39:33 +0000602 """Read up to len(b) bytes into b.
Christian Heimes1a6387e2008-03-26 12:49:49 +0000603
604 Returns number of bytes read (0 for EOF), or None if the object
605 is set not to block as has no data to read.
606 """
607 self._unsupported("readinto")
608
609 def write(self, b):
Benjamin Peterson9ae080e2008-05-04 22:39:33 +0000610 """Write the given buffer to the IO stream.
Christian Heimes1a6387e2008-03-26 12:49:49 +0000611
612 Returns the number of bytes written, which may be less than len(b).
613 """
614 self._unsupported("write")
615
616
617class FileIO(_fileio._FileIO, RawIOBase):
618
Benjamin Peterson7bb4d2d2008-04-13 02:01:27 +0000619 """Raw I/O implementation for OS files."""
Christian Heimes1a6387e2008-03-26 12:49:49 +0000620
Benjamin Peterson7bb4d2d2008-04-13 02:01:27 +0000621 # This multiply inherits from _FileIO and RawIOBase to make
622 # isinstance(io.FileIO(), io.RawIOBase) return True without requiring
623 # that _fileio._FileIO inherits from io.RawIOBase (which would be hard
624 # to do since _fileio.c is written in C).
Christian Heimes1a6387e2008-03-26 12:49:49 +0000625
626 def close(self):
627 _fileio._FileIO.close(self)
628 RawIOBase.close(self)
629
630 @property
631 def name(self):
632 return self._name
633
634 @property
635 def mode(self):
636 return self._mode
637
638
639class BufferedIOBase(IOBase):
640
641 """Base class for buffered IO objects.
642
643 The main difference with RawIOBase is that the read() method
644 supports omitting the size argument, and does not have a default
645 implementation that defers to readinto().
646
647 In addition, read(), readinto() and write() may raise
648 BlockingIOError if the underlying raw stream is in non-blocking
649 mode and not ready; unlike their raw counterparts, they will never
650 return None.
651
652 A typical implementation should not inherit from a RawIOBase
653 implementation, but wrap one.
654 """
655
656 def read(self, n = None):
Benjamin Peterson9ae080e2008-05-04 22:39:33 +0000657 """Read and return up to n bytes.
Christian Heimes1a6387e2008-03-26 12:49:49 +0000658
659 If the argument is omitted, None, or negative, reads and
660 returns all data until EOF.
661
662 If the argument is positive, and the underlying raw stream is
663 not 'interactive', multiple raw reads may be issued to satisfy
664 the byte count (unless EOF is reached first). But for
665 interactive raw streams (XXX and for pipes?), at most one raw
666 read will be issued, and a short result does not imply that
667 EOF is imminent.
668
669 Returns an empty bytes array on EOF.
670
671 Raises BlockingIOError if the underlying raw stream has no
672 data at the moment.
673 """
674 self._unsupported("read")
675
676 def readinto(self, b):
Benjamin Peterson9ae080e2008-05-04 22:39:33 +0000677 """Read up to len(b) bytes into b.
Christian Heimes1a6387e2008-03-26 12:49:49 +0000678
Benjamin Peterson7bb4d2d2008-04-13 02:01:27 +0000679 Like read(), this may issue multiple reads to the underlying raw
680 stream, unless the latter is 'interactive'.
Christian Heimes1a6387e2008-03-26 12:49:49 +0000681
682 Returns the number of bytes read (0 for EOF).
683
684 Raises BlockingIOError if the underlying raw stream has no
685 data at the moment.
686 """
687 # XXX This ought to work with anything that supports the buffer API
688 data = self.read(len(b))
689 n = len(data)
690 try:
691 b[:n] = data
692 except TypeError as err:
693 import array
694 if not isinstance(b, array.array):
695 raise err
Alexandre Vassalotti1aed6242008-05-09 21:49:43 +0000696 b[:n] = array.array(b'b', data)
Christian Heimes1a6387e2008-03-26 12:49:49 +0000697 return n
698
699 def write(self, b):
Benjamin Peterson9ae080e2008-05-04 22:39:33 +0000700 """Write the given buffer to the IO stream.
Christian Heimes1a6387e2008-03-26 12:49:49 +0000701
Benjamin Peterson9ae080e2008-05-04 22:39:33 +0000702 Return the number of bytes written, which is never less than
Christian Heimes1a6387e2008-03-26 12:49:49 +0000703 len(b).
704
705 Raises BlockingIOError if the buffer is full and the
706 underlying raw stream cannot accept more data at the moment.
707 """
708 self._unsupported("write")
709
710
711class _BufferedIOMixin(BufferedIOBase):
712
713 """A mixin implementation of BufferedIOBase with an underlying raw stream.
714
715 This passes most requests on to the underlying raw stream. It
716 does *not* provide implementations of read(), readinto() or
717 write().
718 """
719
720 def __init__(self, raw):
721 self.raw = raw
722
723 ### Positioning ###
724
725 def seek(self, pos, whence=0):
726 return self.raw.seek(pos, whence)
727
728 def tell(self):
729 return self.raw.tell()
730
731 def truncate(self, pos=None):
732 # Flush the stream. We're mixing buffered I/O with lower-level I/O,
733 # and a flush may be necessary to synch both views of the current
734 # file state.
735 self.flush()
736
737 if pos is None:
738 pos = self.tell()
Alexandre Vassalotti1aed6242008-05-09 21:49:43 +0000739 # XXX: Should seek() be used, instead of passing the position
740 # XXX directly to truncate?
Christian Heimes1a6387e2008-03-26 12:49:49 +0000741 return self.raw.truncate(pos)
742
743 ### Flush and close ###
744
745 def flush(self):
746 self.raw.flush()
747
748 def close(self):
749 if not self.closed:
750 try:
751 self.flush()
752 except IOError:
753 pass # If flush() fails, just give up
754 self.raw.close()
755
756 ### Inquiries ###
757
758 def seekable(self):
759 return self.raw.seekable()
760
761 def readable(self):
762 return self.raw.readable()
763
764 def writable(self):
765 return self.raw.writable()
766
767 @property
768 def closed(self):
769 return self.raw.closed
770
771 ### Lower-level APIs ###
772
773 def fileno(self):
774 return self.raw.fileno()
775
776 def isatty(self):
777 return self.raw.isatty()
778
779
Alexandre Vassalotti1aed6242008-05-09 21:49:43 +0000780class _BytesIO(BufferedIOBase):
Christian Heimes1a6387e2008-03-26 12:49:49 +0000781
782 """Buffered I/O implementation using an in-memory bytes buffer."""
783
784 # XXX More docs
785
786 def __init__(self, initial_bytes=None):
787 buf = bytearray()
788 if initial_bytes is not None:
Alexandre Vassalotti1aed6242008-05-09 21:49:43 +0000789 buf += bytearray(initial_bytes)
Christian Heimes1a6387e2008-03-26 12:49:49 +0000790 self._buffer = buf
791 self._pos = 0
792
793 def getvalue(self):
Benjamin Peterson9ae080e2008-05-04 22:39:33 +0000794 """Return the bytes value (contents) of the buffer
Benjamin Peterson7bb4d2d2008-04-13 02:01:27 +0000795 """
Alexandre Vassalotti1aed6242008-05-09 21:49:43 +0000796 if self.closed:
797 raise ValueError("getvalue on closed file")
Christian Heimes1a6387e2008-03-26 12:49:49 +0000798 return bytes(self._buffer)
799
800 def read(self, n=None):
Alexandre Vassalotti1aed6242008-05-09 21:49:43 +0000801 if self.closed:
802 raise ValueError("read from closed file")
Christian Heimes1a6387e2008-03-26 12:49:49 +0000803 if n is None:
804 n = -1
Alexandre Vassalotti1aed6242008-05-09 21:49:43 +0000805 if not isinstance(n, (int, long)):
806 raise TypeError("argument must be an integer")
Christian Heimes1a6387e2008-03-26 12:49:49 +0000807 if n < 0:
808 n = len(self._buffer)
Alexandre Vassalotti1aed6242008-05-09 21:49:43 +0000809 if len(self._buffer) <= self._pos:
810 return b""
Christian Heimes1a6387e2008-03-26 12:49:49 +0000811 newpos = min(len(self._buffer), self._pos + n)
812 b = self._buffer[self._pos : newpos]
813 self._pos = newpos
814 return bytes(b)
815
816 def read1(self, n):
Benjamin Peterson9ae080e2008-05-04 22:39:33 +0000817 """this is the same as read.
Benjamin Peterson7bb4d2d2008-04-13 02:01:27 +0000818 """
Christian Heimes1a6387e2008-03-26 12:49:49 +0000819 return self.read(n)
820
821 def write(self, b):
822 if self.closed:
823 raise ValueError("write to closed file")
824 if isinstance(b, unicode):
825 raise TypeError("can't write unicode to binary stream")
826 n = len(b)
Alexandre Vassalotti1aed6242008-05-09 21:49:43 +0000827 if n == 0:
828 return 0
Alexandre Vassalotti844f7572008-05-10 19:59:16 +0000829 pos = self._pos
830 if pos > len(self._buffer):
Christian Heimes1a6387e2008-03-26 12:49:49 +0000831 # Inserts null bytes between the current end of the file
832 # and the new write position.
Alexandre Vassalotti844f7572008-05-10 19:59:16 +0000833 padding = b'\x00' * (pos - len(self._buffer))
834 self._buffer += padding
835 self._buffer[pos:pos + n] = b
836 self._pos += n
Christian Heimes1a6387e2008-03-26 12:49:49 +0000837 return n
838
839 def seek(self, pos, whence=0):
Alexandre Vassalotti1aed6242008-05-09 21:49:43 +0000840 if self.closed:
841 raise ValueError("seek on closed file")
Christian Heimes1a6387e2008-03-26 12:49:49 +0000842 try:
843 pos = pos.__index__()
844 except AttributeError as err:
845 raise TypeError("an integer is required") # from err
846 if whence == 0:
Alexandre Vassalotti1aed6242008-05-09 21:49:43 +0000847 if pos < 0:
848 raise ValueError("negative seek position %r" % (pos,))
849 self._pos = pos
Christian Heimes1a6387e2008-03-26 12:49:49 +0000850 elif whence == 1:
851 self._pos = max(0, self._pos + pos)
852 elif whence == 2:
853 self._pos = max(0, len(self._buffer) + pos)
854 else:
Alexandre Vassalotti1aed6242008-05-09 21:49:43 +0000855 raise ValueError("invalid whence value")
Christian Heimes1a6387e2008-03-26 12:49:49 +0000856 return self._pos
857
858 def tell(self):
Alexandre Vassalotti1aed6242008-05-09 21:49:43 +0000859 if self.closed:
860 raise ValueError("tell on closed file")
Christian Heimes1a6387e2008-03-26 12:49:49 +0000861 return self._pos
862
863 def truncate(self, pos=None):
Alexandre Vassalotti1aed6242008-05-09 21:49:43 +0000864 if self.closed:
865 raise ValueError("truncate on closed file")
Christian Heimes1a6387e2008-03-26 12:49:49 +0000866 if pos is None:
867 pos = self._pos
Alexandre Vassalotti1aed6242008-05-09 21:49:43 +0000868 elif pos < 0:
869 raise ValueError("negative truncate position %r" % (pos,))
Christian Heimes1a6387e2008-03-26 12:49:49 +0000870 del self._buffer[pos:]
Alexandre Vassalotti1aed6242008-05-09 21:49:43 +0000871 return self.seek(pos)
Christian Heimes1a6387e2008-03-26 12:49:49 +0000872
873 def readable(self):
874 return True
875
876 def writable(self):
877 return True
878
879 def seekable(self):
880 return True
881
Alexandre Vassalotti1aed6242008-05-09 21:49:43 +0000882# Use the faster implementation of BytesIO if available
883try:
884 import _bytesio
885
886 class BytesIO(_bytesio._BytesIO, BufferedIOBase):
887 __doc__ = _bytesio._BytesIO.__doc__
888
889except ImportError:
890 BytesIO = _BytesIO
891
Christian Heimes1a6387e2008-03-26 12:49:49 +0000892
893class BufferedReader(_BufferedIOMixin):
894
Benjamin Peterson7bb4d2d2008-04-13 02:01:27 +0000895 """BufferedReader(raw[, buffer_size])
896
897 A buffer for a readable, sequential BaseRawIO object.
898
899 The constructor creates a BufferedReader for the given readable raw
900 stream and buffer_size. If buffer_size is omitted, DEFAULT_BUFFER_SIZE
901 is used.
902 """
Christian Heimes1a6387e2008-03-26 12:49:49 +0000903
904 def __init__(self, raw, buffer_size=DEFAULT_BUFFER_SIZE):
905 """Create a new buffered reader using the given readable raw IO object.
906 """
907 raw._checkReadable()
908 _BufferedIOMixin.__init__(self, raw)
909 self._read_buf = b""
910 self.buffer_size = buffer_size
911
912 def read(self, n=None):
913 """Read n bytes.
914
915 Returns exactly n bytes of data unless the underlying raw IO
916 stream reaches EOF or if the call would block in non-blocking
917 mode. If n is negative, read until EOF or until read() would
918 block.
919 """
920 if n is None:
921 n = -1
922 nodata_val = b""
923 while n < 0 or len(self._read_buf) < n:
924 to_read = max(self.buffer_size,
925 n if n is not None else 2*len(self._read_buf))
926 current = self.raw.read(to_read)
927 if current in (b"", None):
928 nodata_val = current
929 break
930 self._read_buf += current
931 if self._read_buf:
932 if n < 0:
933 n = len(self._read_buf)
934 out = self._read_buf[:n]
935 self._read_buf = self._read_buf[n:]
936 else:
937 out = nodata_val
938 return out
939
940 def peek(self, n=0):
941 """Returns buffered bytes without advancing the position.
942
943 The argument indicates a desired minimal number of bytes; we
944 do at most one raw read to satisfy it. We never return more
945 than self.buffer_size.
946 """
947 want = min(n, self.buffer_size)
948 have = len(self._read_buf)
949 if have < want:
950 to_read = self.buffer_size - have
951 current = self.raw.read(to_read)
952 if current:
953 self._read_buf += current
954 return self._read_buf
955
956 def read1(self, n):
Benjamin Peterson7bb4d2d2008-04-13 02:01:27 +0000957 """Reads up to n bytes, with at most one read() system call."""
958 # Returns up to n bytes. If at least one byte is buffered, we
959 # only return buffered bytes. Otherwise, we do one raw read.
Christian Heimes1a6387e2008-03-26 12:49:49 +0000960 if n <= 0:
961 return b""
962 self.peek(1)
963 return self.read(min(n, len(self._read_buf)))
964
965 def tell(self):
966 return self.raw.tell() - len(self._read_buf)
967
968 def seek(self, pos, whence=0):
969 if whence == 1:
970 pos -= len(self._read_buf)
971 pos = self.raw.seek(pos, whence)
972 self._read_buf = b""
973 return pos
974
975
976class BufferedWriter(_BufferedIOMixin):
977
Benjamin Peterson9ae080e2008-05-04 22:39:33 +0000978 """A buffer for a writeable sequential RawIO object.
Benjamin Peterson7bb4d2d2008-04-13 02:01:27 +0000979
980 The constructor creates a BufferedWriter for the given writeable raw
981 stream. If the buffer_size is not given, it defaults to
982 DEAFULT_BUFFER_SIZE. If max_buffer_size is omitted, it defaults to
983 twice the buffer size.
984 """
Christian Heimes1a6387e2008-03-26 12:49:49 +0000985
986 def __init__(self, raw,
987 buffer_size=DEFAULT_BUFFER_SIZE, max_buffer_size=None):
988 raw._checkWritable()
989 _BufferedIOMixin.__init__(self, raw)
990 self.buffer_size = buffer_size
991 self.max_buffer_size = (2*buffer_size
992 if max_buffer_size is None
993 else max_buffer_size)
994 self._write_buf = bytearray()
995
996 def write(self, b):
997 if self.closed:
998 raise ValueError("write to closed file")
999 if isinstance(b, unicode):
1000 raise TypeError("can't write unicode to binary stream")
1001 # XXX we can implement some more tricks to try and avoid partial writes
1002 if len(self._write_buf) > self.buffer_size:
1003 # We're full, so let's pre-flush the buffer
1004 try:
1005 self.flush()
1006 except BlockingIOError as e:
1007 # We can't accept anything else.
1008 # XXX Why not just let the exception pass through?
1009 raise BlockingIOError(e.errno, e.strerror, 0)
1010 before = len(self._write_buf)
1011 self._write_buf.extend(b)
1012 written = len(self._write_buf) - before
1013 if len(self._write_buf) > self.buffer_size:
1014 try:
1015 self.flush()
1016 except BlockingIOError as e:
1017 if (len(self._write_buf) > self.max_buffer_size):
1018 # We've hit max_buffer_size. We have to accept a partial
1019 # write and cut back our buffer.
1020 overage = len(self._write_buf) - self.max_buffer_size
1021 self._write_buf = self._write_buf[:self.max_buffer_size]
1022 raise BlockingIOError(e.errno, e.strerror, overage)
1023 return written
1024
Alexandre Vassalotti1aed6242008-05-09 21:49:43 +00001025 def truncate(self, pos=None):
1026 self.flush()
1027 if pos is None:
1028 pos = self.raw.tell()
1029 return self.raw.truncate(pos)
1030
Christian Heimes1a6387e2008-03-26 12:49:49 +00001031 def flush(self):
1032 if self.closed:
1033 raise ValueError("flush of closed file")
1034 written = 0
1035 try:
1036 while self._write_buf:
1037 n = self.raw.write(self._write_buf)
1038 del self._write_buf[:n]
1039 written += n
1040 except BlockingIOError as e:
1041 n = e.characters_written
1042 del self._write_buf[:n]
1043 written += n
1044 raise BlockingIOError(e.errno, e.strerror, written)
1045
1046 def tell(self):
1047 return self.raw.tell() + len(self._write_buf)
1048
1049 def seek(self, pos, whence=0):
1050 self.flush()
1051 return self.raw.seek(pos, whence)
1052
1053
1054class BufferedRWPair(BufferedIOBase):
1055
1056 """A buffered reader and writer object together.
1057
Benjamin Peterson7bb4d2d2008-04-13 02:01:27 +00001058 A buffered reader object and buffered writer object put together to
1059 form a sequential IO object that can read and write. This is typically
1060 used with a socket or two-way pipe.
Christian Heimes1a6387e2008-03-26 12:49:49 +00001061
Benjamin Peterson7bb4d2d2008-04-13 02:01:27 +00001062 reader and writer are RawIOBase objects that are readable and
1063 writeable respectively. If the buffer_size is omitted it defaults to
1064 DEFAULT_BUFFER_SIZE. The max_buffer_size (for the buffered writer)
1065 defaults to twice the buffer size.
Christian Heimes1a6387e2008-03-26 12:49:49 +00001066 """
1067
Benjamin Peterson7bb4d2d2008-04-13 02:01:27 +00001068 # XXX The usefulness of this (compared to having two separate IO
1069 # objects) is questionable.
1070
Christian Heimes1a6387e2008-03-26 12:49:49 +00001071 def __init__(self, reader, writer,
1072 buffer_size=DEFAULT_BUFFER_SIZE, max_buffer_size=None):
1073 """Constructor.
1074
1075 The arguments are two RawIO instances.
1076 """
1077 reader._checkReadable()
1078 writer._checkWritable()
1079 self.reader = BufferedReader(reader, buffer_size)
1080 self.writer = BufferedWriter(writer, buffer_size, max_buffer_size)
1081
1082 def read(self, n=None):
1083 if n is None:
1084 n = -1
1085 return self.reader.read(n)
1086
1087 def readinto(self, b):
1088 return self.reader.readinto(b)
1089
1090 def write(self, b):
1091 return self.writer.write(b)
1092
1093 def peek(self, n=0):
1094 return self.reader.peek(n)
1095
1096 def read1(self, n):
1097 return self.reader.read1(n)
1098
1099 def readable(self):
1100 return self.reader.readable()
1101
1102 def writable(self):
1103 return self.writer.writable()
1104
1105 def flush(self):
1106 return self.writer.flush()
1107
1108 def close(self):
1109 self.writer.close()
1110 self.reader.close()
1111
1112 def isatty(self):
1113 return self.reader.isatty() or self.writer.isatty()
1114
1115 @property
1116 def closed(self):
1117 return self.writer.closed()
1118
1119
1120class BufferedRandom(BufferedWriter, BufferedReader):
1121
Benjamin Peterson9ae080e2008-05-04 22:39:33 +00001122 """A buffered interface to random access streams.
Benjamin Peterson7bb4d2d2008-04-13 02:01:27 +00001123
1124 The constructor creates a reader and writer for a seekable stream,
1125 raw, given in the first argument. If the buffer_size is omitted it
1126 defaults to DEFAULT_BUFFER_SIZE. The max_buffer_size (for the buffered
1127 writer) defaults to twice the buffer size.
1128 """
Christian Heimes1a6387e2008-03-26 12:49:49 +00001129
1130 def __init__(self, raw,
1131 buffer_size=DEFAULT_BUFFER_SIZE, max_buffer_size=None):
1132 raw._checkSeekable()
1133 BufferedReader.__init__(self, raw, buffer_size)
1134 BufferedWriter.__init__(self, raw, buffer_size, max_buffer_size)
1135
1136 def seek(self, pos, whence=0):
1137 self.flush()
1138 # First do the raw seek, then empty the read buffer, so that
1139 # if the raw seek fails, we don't lose buffered data forever.
1140 pos = self.raw.seek(pos, whence)
1141 self._read_buf = b""
1142 return pos
1143
1144 def tell(self):
1145 if (self._write_buf):
1146 return self.raw.tell() + len(self._write_buf)
1147 else:
1148 return self.raw.tell() - len(self._read_buf)
1149
Alexandre Vassalotti1aed6242008-05-09 21:49:43 +00001150 def truncate(self, pos=None):
1151 if pos is None:
1152 pos = self.tell()
1153 # Use seek to flush the read buffer.
1154 self.seek(pos)
1155 return BufferedWriter.truncate(self)
1156
Christian Heimes1a6387e2008-03-26 12:49:49 +00001157 def read(self, n=None):
1158 if n is None:
1159 n = -1
1160 self.flush()
1161 return BufferedReader.read(self, n)
1162
1163 def readinto(self, b):
1164 self.flush()
1165 return BufferedReader.readinto(self, b)
1166
1167 def peek(self, n=0):
1168 self.flush()
1169 return BufferedReader.peek(self, n)
1170
1171 def read1(self, n):
1172 self.flush()
1173 return BufferedReader.read1(self, n)
1174
1175 def write(self, b):
1176 if self._read_buf:
1177 self.raw.seek(-len(self._read_buf), 1) # Undo readahead
1178 self._read_buf = b""
1179 return BufferedWriter.write(self, b)
1180
1181
1182class TextIOBase(IOBase):
1183
1184 """Base class for text I/O.
1185
Benjamin Peterson7bb4d2d2008-04-13 02:01:27 +00001186 This class provides a character and line based interface to stream
1187 I/O. There is no readinto method because Python's character strings
1188 are immutable. There is no public constructor.
Christian Heimes1a6387e2008-03-26 12:49:49 +00001189 """
1190
1191 def read(self, n = -1):
Benjamin Peterson9ae080e2008-05-04 22:39:33 +00001192 """Read at most n characters from stream.
Christian Heimes1a6387e2008-03-26 12:49:49 +00001193
1194 Read from underlying buffer until we have n characters or we hit EOF.
1195 If n is negative or omitted, read until EOF.
1196 """
1197 self._unsupported("read")
1198
1199 def write(self, s):
Benjamin Peterson9ae080e2008-05-04 22:39:33 +00001200 """Write string s to stream."""
Christian Heimes1a6387e2008-03-26 12:49:49 +00001201 self._unsupported("write")
1202
1203 def truncate(self, pos = None):
Benjamin Peterson9ae080e2008-05-04 22:39:33 +00001204 """Truncate size to pos."""
Alexandre Vassalotti1aed6242008-05-09 21:49:43 +00001205 self._unsupported("truncate")
Christian Heimes1a6387e2008-03-26 12:49:49 +00001206
1207 def readline(self):
Benjamin Peterson9ae080e2008-05-04 22:39:33 +00001208 """Read until newline or EOF.
Christian Heimes1a6387e2008-03-26 12:49:49 +00001209
1210 Returns an empty string if EOF is hit immediately.
1211 """
1212 self._unsupported("readline")
1213
1214 @property
1215 def encoding(self):
1216 """Subclasses should override."""
1217 return None
1218
1219 @property
1220 def newlines(self):
Benjamin Peterson9ae080e2008-05-04 22:39:33 +00001221 """Line endings translated so far.
Christian Heimes1a6387e2008-03-26 12:49:49 +00001222
1223 Only line endings translated during reading are considered.
1224
1225 Subclasses should override.
1226 """
1227 return None
1228
1229
1230class IncrementalNewlineDecoder(codecs.IncrementalDecoder):
1231 """Codec used when reading a file in universal newlines mode.
1232 It wraps another incremental decoder, translating \\r\\n and \\r into \\n.
1233 It also records the types of newlines encountered.
1234 When used with translate=False, it ensures that the newline sequence is
1235 returned in one piece.
1236 """
1237 def __init__(self, decoder, translate, errors='strict'):
1238 codecs.IncrementalDecoder.__init__(self, errors=errors)
1239 self.buffer = b''
1240 self.translate = translate
1241 self.decoder = decoder
1242 self.seennl = 0
1243
1244 def decode(self, input, final=False):
1245 # decode input (with the eventual \r from a previous pass)
1246 if self.buffer:
1247 input = self.buffer + input
1248
1249 output = self.decoder.decode(input, final=final)
1250
1251 # retain last \r even when not translating data:
1252 # then readline() is sure to get \r\n in one pass
1253 if output.endswith("\r") and not final:
1254 output = output[:-1]
1255 self.buffer = b'\r'
1256 else:
1257 self.buffer = b''
1258
1259 # Record which newlines are read
1260 crlf = output.count('\r\n')
1261 cr = output.count('\r') - crlf
1262 lf = output.count('\n') - crlf
1263 self.seennl |= (lf and self._LF) | (cr and self._CR) \
1264 | (crlf and self._CRLF)
1265
1266 if self.translate:
1267 if crlf:
1268 output = output.replace("\r\n", "\n")
1269 if cr:
1270 output = output.replace("\r", "\n")
1271
1272 return output
1273
1274 def getstate(self):
1275 buf, flag = self.decoder.getstate()
1276 return buf + self.buffer, flag
1277
1278 def setstate(self, state):
1279 buf, flag = state
1280 if buf.endswith(b'\r'):
1281 self.buffer = b'\r'
1282 buf = buf[:-1]
1283 else:
1284 self.buffer = b''
1285 self.decoder.setstate((buf, flag))
1286
1287 def reset(self):
1288 self.seennl = 0
1289 self.buffer = b''
1290 self.decoder.reset()
1291
1292 _LF = 1
1293 _CR = 2
1294 _CRLF = 4
1295
1296 @property
1297 def newlines(self):
1298 return (None,
1299 "\n",
1300 "\r",
1301 ("\r", "\n"),
1302 "\r\n",
1303 ("\n", "\r\n"),
1304 ("\r", "\r\n"),
1305 ("\r", "\n", "\r\n")
1306 )[self.seennl]
1307
1308
1309class TextIOWrapper(TextIOBase):
1310
Benjamin Peterson9ae080e2008-05-04 22:39:33 +00001311 r"""Character and line based layer over a BufferedIOBase object, buffer.
Benjamin Peterson7bb4d2d2008-04-13 02:01:27 +00001312
1313 encoding gives the name of the encoding that the stream will be
1314 decoded or encoded with. It defaults to locale.getpreferredencoding.
1315
1316 errors determines the strictness of encoding and decoding (see the
1317 codecs.register) and defaults to "strict".
1318
1319 newline can be None, '', '\n', '\r', or '\r\n'. It controls the
1320 handling of line endings. If it is None, universal newlines is
1321 enabled. With this enabled, on input, the lines endings '\n', '\r',
1322 or '\r\n' are translated to '\n' before being returned to the
1323 caller. Conversely, on output, '\n' is translated to the system
1324 default line seperator, os.linesep. If newline is any other of its
1325 legal values, that newline becomes the newline when the file is read
1326 and it is returned untranslated. On output, '\n' is converted to the
1327 newline.
1328
1329 If line_buffering is True, a call to flush is implied when a call to
1330 write contains a newline character.
Christian Heimes1a6387e2008-03-26 12:49:49 +00001331 """
1332
1333 _CHUNK_SIZE = 128
1334
1335 def __init__(self, buffer, encoding=None, errors=None, newline=None,
1336 line_buffering=False):
1337 if newline not in (None, "", "\n", "\r", "\r\n"):
1338 raise ValueError("illegal newline value: %r" % (newline,))
1339 if encoding is None:
1340 try:
1341 encoding = os.device_encoding(buffer.fileno())
1342 except (AttributeError, UnsupportedOperation):
1343 pass
1344 if encoding is None:
1345 try:
1346 import locale
1347 except ImportError:
1348 # Importing locale may fail if Python is being built
1349 encoding = "ascii"
1350 else:
1351 encoding = locale.getpreferredencoding()
1352
Christian Heimes3784c6b2008-03-26 23:13:59 +00001353 if not isinstance(encoding, basestring):
Christian Heimes1a6387e2008-03-26 12:49:49 +00001354 raise ValueError("invalid encoding: %r" % encoding)
1355
1356 if errors is None:
1357 errors = "strict"
1358 else:
Christian Heimes3784c6b2008-03-26 23:13:59 +00001359 if not isinstance(errors, basestring):
Christian Heimes1a6387e2008-03-26 12:49:49 +00001360 raise ValueError("invalid errors: %r" % errors)
1361
1362 self.buffer = buffer
1363 self._line_buffering = line_buffering
1364 self._encoding = encoding
1365 self._errors = errors
1366 self._readuniversal = not newline
1367 self._readtranslate = newline is None
1368 self._readnl = newline
1369 self._writetranslate = newline != ''
1370 self._writenl = newline or os.linesep
1371 self._encoder = None
1372 self._decoder = None
1373 self._decoded_chars = '' # buffer for text returned from decoder
1374 self._decoded_chars_used = 0 # offset into _decoded_chars for read()
1375 self._snapshot = None # info for reconstructing decoder state
1376 self._seekable = self._telling = self.buffer.seekable()
1377
1378 # self._snapshot is either None, or a tuple (dec_flags, next_input)
1379 # where dec_flags is the second (integer) item of the decoder state
1380 # and next_input is the chunk of input bytes that comes next after the
1381 # snapshot point. We use this to reconstruct decoder states in tell().
1382
1383 # Naming convention:
1384 # - "bytes_..." for integer variables that count input bytes
1385 # - "chars_..." for integer variables that count decoded characters
1386
Christian Heimes1a6387e2008-03-26 12:49:49 +00001387 @property
1388 def encoding(self):
1389 return self._encoding
1390
1391 @property
1392 def errors(self):
1393 return self._errors
1394
1395 @property
1396 def line_buffering(self):
1397 return self._line_buffering
1398
1399 def seekable(self):
1400 return self._seekable
1401
Alexandre Vassalotti1aed6242008-05-09 21:49:43 +00001402 def readable(self):
1403 return self.buffer.readable()
1404
1405 def writable(self):
1406 return self.buffer.writable()
1407
Christian Heimes1a6387e2008-03-26 12:49:49 +00001408 def flush(self):
1409 self.buffer.flush()
1410 self._telling = self._seekable
1411
1412 def close(self):
1413 try:
1414 self.flush()
1415 except:
1416 pass # If flush() fails, just give up
1417 self.buffer.close()
1418
1419 @property
1420 def closed(self):
1421 return self.buffer.closed
1422
1423 def fileno(self):
1424 return self.buffer.fileno()
1425
1426 def isatty(self):
1427 return self.buffer.isatty()
1428
1429 def write(self, s):
1430 if self.closed:
1431 raise ValueError("write to closed file")
1432 if not isinstance(s, unicode):
1433 raise TypeError("can't write %s to text stream" %
1434 s.__class__.__name__)
1435 length = len(s)
1436 haslf = (self._writetranslate or self._line_buffering) and "\n" in s
1437 if haslf and self._writetranslate and self._writenl != "\n":
1438 s = s.replace("\n", self._writenl)
1439 encoder = self._encoder or self._get_encoder()
1440 # XXX What if we were just reading?
1441 b = encoder.encode(s)
1442 self.buffer.write(b)
1443 if self._line_buffering and (haslf or "\r" in s):
1444 self.flush()
1445 self._snapshot = None
1446 if self._decoder:
1447 self._decoder.reset()
1448 return length
1449
1450 def _get_encoder(self):
1451 make_encoder = codecs.getincrementalencoder(self._encoding)
1452 self._encoder = make_encoder(self._errors)
1453 return self._encoder
1454
1455 def _get_decoder(self):
1456 make_decoder = codecs.getincrementaldecoder(self._encoding)
1457 decoder = make_decoder(self._errors)
1458 if self._readuniversal:
1459 decoder = IncrementalNewlineDecoder(decoder, self._readtranslate)
1460 self._decoder = decoder
1461 return decoder
1462
1463 # The following three methods implement an ADT for _decoded_chars.
1464 # Text returned from the decoder is buffered here until the client
1465 # requests it by calling our read() or readline() method.
1466 def _set_decoded_chars(self, chars):
1467 """Set the _decoded_chars buffer."""
1468 self._decoded_chars = chars
1469 self._decoded_chars_used = 0
1470
1471 def _get_decoded_chars(self, n=None):
1472 """Advance into the _decoded_chars buffer."""
1473 offset = self._decoded_chars_used
1474 if n is None:
1475 chars = self._decoded_chars[offset:]
1476 else:
1477 chars = self._decoded_chars[offset:offset + n]
1478 self._decoded_chars_used += len(chars)
1479 return chars
1480
1481 def _rewind_decoded_chars(self, n):
1482 """Rewind the _decoded_chars buffer."""
1483 if self._decoded_chars_used < n:
1484 raise AssertionError("rewind decoded_chars out of bounds")
1485 self._decoded_chars_used -= n
1486
1487 def _read_chunk(self):
1488 """
1489 Read and decode the next chunk of data from the BufferedReader.
1490
1491 The return value is True unless EOF was reached. The decoded string
1492 is placed in self._decoded_chars (replacing its previous value).
1493 The entire input chunk is sent to the decoder, though some of it
1494 may remain buffered in the decoder, yet to be converted.
1495 """
1496
1497 if self._decoder is None:
1498 raise ValueError("no decoder")
1499
1500 if self._telling:
1501 # To prepare for tell(), we need to snapshot a point in the
1502 # file where the decoder's input buffer is empty.
1503
1504 dec_buffer, dec_flags = self._decoder.getstate()
1505 # Given this, we know there was a valid snapshot point
1506 # len(dec_buffer) bytes ago with decoder state (b'', dec_flags).
1507
1508 # Read a chunk, decode it, and put the result in self._decoded_chars.
1509 input_chunk = self.buffer.read1(self._CHUNK_SIZE)
1510 eof = not input_chunk
1511 self._set_decoded_chars(self._decoder.decode(input_chunk, eof))
1512
1513 if self._telling:
1514 # At the snapshot point, len(dec_buffer) bytes before the read,
1515 # the next input to be decoded is dec_buffer + input_chunk.
1516 self._snapshot = (dec_flags, dec_buffer + input_chunk)
1517
1518 return not eof
1519
1520 def _pack_cookie(self, position, dec_flags=0,
1521 bytes_to_feed=0, need_eof=0, chars_to_skip=0):
1522 # The meaning of a tell() cookie is: seek to position, set the
1523 # decoder flags to dec_flags, read bytes_to_feed bytes, feed them
1524 # into the decoder with need_eof as the EOF flag, then skip
1525 # chars_to_skip characters of the decoded result. For most simple
1526 # decoders, tell() will often just give a byte offset in the file.
1527 return (position | (dec_flags<<64) | (bytes_to_feed<<128) |
1528 (chars_to_skip<<192) | bool(need_eof)<<256)
1529
1530 def _unpack_cookie(self, bigint):
1531 rest, position = divmod(bigint, 1<<64)
1532 rest, dec_flags = divmod(rest, 1<<64)
1533 rest, bytes_to_feed = divmod(rest, 1<<64)
1534 need_eof, chars_to_skip = divmod(rest, 1<<64)
1535 return position, dec_flags, bytes_to_feed, need_eof, chars_to_skip
1536
1537 def tell(self):
1538 if not self._seekable:
1539 raise IOError("underlying stream is not seekable")
1540 if not self._telling:
1541 raise IOError("telling position disabled by next() call")
1542 self.flush()
1543 position = self.buffer.tell()
1544 decoder = self._decoder
1545 if decoder is None or self._snapshot is None:
1546 if self._decoded_chars:
1547 # This should never happen.
1548 raise AssertionError("pending decoded text")
1549 return position
1550
1551 # Skip backward to the snapshot point (see _read_chunk).
1552 dec_flags, next_input = self._snapshot
1553 position -= len(next_input)
1554
1555 # How many decoded characters have been used up since the snapshot?
1556 chars_to_skip = self._decoded_chars_used
1557 if chars_to_skip == 0:
1558 # We haven't moved from the snapshot point.
1559 return self._pack_cookie(position, dec_flags)
1560
1561 # Starting from the snapshot position, we will walk the decoder
1562 # forward until it gives us enough decoded characters.
1563 saved_state = decoder.getstate()
1564 try:
1565 # Note our initial start point.
1566 decoder.setstate((b'', dec_flags))
1567 start_pos = position
1568 start_flags, bytes_fed, chars_decoded = dec_flags, 0, 0
1569 need_eof = 0
1570
1571 # Feed the decoder one byte at a time. As we go, note the
1572 # nearest "safe start point" before the current location
1573 # (a point where the decoder has nothing buffered, so seek()
1574 # can safely start from there and advance to this location).
Amaury Forgeot d'Arc7684f852008-05-03 12:21:13 +00001575 for next_byte in next_input:
Christian Heimes1a6387e2008-03-26 12:49:49 +00001576 bytes_fed += 1
1577 chars_decoded += len(decoder.decode(next_byte))
1578 dec_buffer, dec_flags = decoder.getstate()
1579 if not dec_buffer and chars_decoded <= chars_to_skip:
1580 # Decoder buffer is empty, so this is a safe start point.
1581 start_pos += bytes_fed
1582 chars_to_skip -= chars_decoded
1583 start_flags, bytes_fed, chars_decoded = dec_flags, 0, 0
1584 if chars_decoded >= chars_to_skip:
1585 break
1586 else:
1587 # We didn't get enough decoded data; signal EOF to get more.
1588 chars_decoded += len(decoder.decode(b'', final=True))
1589 need_eof = 1
1590 if chars_decoded < chars_to_skip:
1591 raise IOError("can't reconstruct logical file position")
1592
1593 # The returned cookie corresponds to the last safe start point.
1594 return self._pack_cookie(
1595 start_pos, start_flags, bytes_fed, need_eof, chars_to_skip)
1596 finally:
1597 decoder.setstate(saved_state)
1598
Alexandre Vassalotti1aed6242008-05-09 21:49:43 +00001599 def truncate(self, pos=None):
1600 self.flush()
1601 if pos is None:
1602 pos = self.tell()
1603 self.seek(pos)
1604 return self.buffer.truncate()
1605
Christian Heimes1a6387e2008-03-26 12:49:49 +00001606 def seek(self, cookie, whence=0):
Alexandre Vassalotti1aed6242008-05-09 21:49:43 +00001607 if self.closed:
1608 raise ValueError("tell on closed file")
Christian Heimes1a6387e2008-03-26 12:49:49 +00001609 if not self._seekable:
1610 raise IOError("underlying stream is not seekable")
1611 if whence == 1: # seek relative to current position
1612 if cookie != 0:
1613 raise IOError("can't do nonzero cur-relative seeks")
1614 # Seeking to the current position should attempt to
1615 # sync the underlying buffer with the current position.
1616 whence = 0
1617 cookie = self.tell()
1618 if whence == 2: # seek relative to end of file
1619 if cookie != 0:
1620 raise IOError("can't do nonzero end-relative seeks")
1621 self.flush()
1622 position = self.buffer.seek(0, 2)
1623 self._set_decoded_chars('')
1624 self._snapshot = None
1625 if self._decoder:
1626 self._decoder.reset()
1627 return position
1628 if whence != 0:
1629 raise ValueError("invalid whence (%r, should be 0, 1 or 2)" %
1630 (whence,))
1631 if cookie < 0:
1632 raise ValueError("negative seek position %r" % (cookie,))
1633 self.flush()
1634
1635 # The strategy of seek() is to go back to the safe start point
1636 # and replay the effect of read(chars_to_skip) from there.
1637 start_pos, dec_flags, bytes_to_feed, need_eof, chars_to_skip = \
1638 self._unpack_cookie(cookie)
1639
1640 # Seek back to the safe start point.
1641 self.buffer.seek(start_pos)
1642 self._set_decoded_chars('')
1643 self._snapshot = None
1644
1645 # Restore the decoder to its state from the safe start point.
1646 if self._decoder or dec_flags or chars_to_skip:
1647 self._decoder = self._decoder or self._get_decoder()
1648 self._decoder.setstate((b'', dec_flags))
1649 self._snapshot = (dec_flags, b'')
1650
1651 if chars_to_skip:
1652 # Just like _read_chunk, feed the decoder and save a snapshot.
1653 input_chunk = self.buffer.read(bytes_to_feed)
1654 self._set_decoded_chars(
1655 self._decoder.decode(input_chunk, need_eof))
1656 self._snapshot = (dec_flags, input_chunk)
1657
1658 # Skip chars_to_skip of the decoded characters.
1659 if len(self._decoded_chars) < chars_to_skip:
1660 raise IOError("can't restore logical file position")
1661 self._decoded_chars_used = chars_to_skip
1662
1663 return cookie
1664
1665 def read(self, n=None):
1666 if n is None:
1667 n = -1
1668 decoder = self._decoder or self._get_decoder()
1669 if n < 0:
1670 # Read everything.
1671 result = (self._get_decoded_chars() +
1672 decoder.decode(self.buffer.read(), final=True))
1673 self._set_decoded_chars('')
1674 self._snapshot = None
1675 return result
1676 else:
1677 # Keep reading chunks until we have n characters to return.
1678 eof = False
1679 result = self._get_decoded_chars(n)
1680 while len(result) < n and not eof:
1681 eof = not self._read_chunk()
1682 result += self._get_decoded_chars(n - len(result))
1683 return result
1684
1685 def next(self):
1686 self._telling = False
1687 line = self.readline()
1688 if not line:
1689 self._snapshot = None
1690 self._telling = self._seekable
1691 raise StopIteration
1692 return line
1693
1694 def readline(self, limit=None):
Alexandre Vassalotti1aed6242008-05-09 21:49:43 +00001695 if self.closed:
1696 raise ValueError("read from closed file")
Christian Heimes1a6387e2008-03-26 12:49:49 +00001697 if limit is None:
1698 limit = -1
Alexandre Vassalotti1aed6242008-05-09 21:49:43 +00001699 if not isinstance(limit, (int, long)):
1700 raise TypeError("limit must be an integer")
Christian Heimes1a6387e2008-03-26 12:49:49 +00001701
1702 # Grab all the decoded text (we will rewind any extra bits later).
1703 line = self._get_decoded_chars()
1704
1705 start = 0
1706 decoder = self._decoder or self._get_decoder()
1707
1708 pos = endpos = None
1709 while True:
1710 if self._readtranslate:
1711 # Newlines are already translated, only search for \n
1712 pos = line.find('\n', start)
1713 if pos >= 0:
1714 endpos = pos + 1
1715 break
1716 else:
1717 start = len(line)
1718
1719 elif self._readuniversal:
1720 # Universal newline search. Find any of \r, \r\n, \n
1721 # The decoder ensures that \r\n are not split in two pieces
1722
1723 # In C we'd look for these in parallel of course.
1724 nlpos = line.find("\n", start)
1725 crpos = line.find("\r", start)
1726 if crpos == -1:
1727 if nlpos == -1:
1728 # Nothing found
1729 start = len(line)
1730 else:
1731 # Found \n
1732 endpos = nlpos + 1
1733 break
1734 elif nlpos == -1:
1735 # Found lone \r
1736 endpos = crpos + 1
1737 break
1738 elif nlpos < crpos:
1739 # Found \n
1740 endpos = nlpos + 1
1741 break
1742 elif nlpos == crpos + 1:
1743 # Found \r\n
1744 endpos = crpos + 2
1745 break
1746 else:
1747 # Found \r
1748 endpos = crpos + 1
1749 break
1750 else:
1751 # non-universal
1752 pos = line.find(self._readnl)
1753 if pos >= 0:
1754 endpos = pos + len(self._readnl)
1755 break
1756
1757 if limit >= 0 and len(line) >= limit:
1758 endpos = limit # reached length limit
1759 break
1760
1761 # No line ending seen yet - get more data
1762 more_line = ''
1763 while self._read_chunk():
1764 if self._decoded_chars:
1765 break
1766 if self._decoded_chars:
1767 line += self._get_decoded_chars()
1768 else:
1769 # end of file
1770 self._set_decoded_chars('')
1771 self._snapshot = None
1772 return line
1773
1774 if limit >= 0 and endpos > limit:
1775 endpos = limit # don't exceed limit
1776
1777 # Rewind _decoded_chars to just after the line ending we found.
1778 self._rewind_decoded_chars(len(line) - endpos)
1779 return line[:endpos]
1780
1781 @property
1782 def newlines(self):
1783 return self._decoder.newlines if self._decoder else None
1784
1785class StringIO(TextIOWrapper):
1786
Benjamin Peterson9ae080e2008-05-04 22:39:33 +00001787 """An in-memory stream for text. The initial_value argument sets the
Benjamin Peterson7bb4d2d2008-04-13 02:01:27 +00001788 value of object. The other arguments are like those of TextIOWrapper's
1789 constructor.
1790 """
Christian Heimes1a6387e2008-03-26 12:49:49 +00001791
1792 def __init__(self, initial_value="", encoding="utf-8",
1793 errors="strict", newline="\n"):
1794 super(StringIO, self).__init__(BytesIO(),
1795 encoding=encoding,
1796 errors=errors,
1797 newline=newline)
1798 if initial_value:
1799 if not isinstance(initial_value, unicode):
1800 initial_value = unicode(initial_value)
1801 self.write(initial_value)
1802 self.seek(0)
1803
1804 def getvalue(self):
1805 self.flush()
1806 return self.buffer.getvalue().decode(self._encoding, self._errors)