blob: 70f2abaf57e2b8926aba8ad0d4e65f96c9133f02 [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
Jesus Cea585ad8a2009-07-02 15:37:21 +00007separation between reading and writing to streams; implementations are
Benjamin Peterson7bb4d2d2008-04-13 02:01:27 +00008allowed 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
Christian Heimes1a6387e2008-03-26 12:49:49 +000062import codecs
63import _fileio
Antoine Pitrou11ec65d2008-08-14 21:04:30 +000064import threading
Christian Heimes1a6387e2008-03-26 12:49:49 +000065
66# open() uses st_blksize whenever we can
67DEFAULT_BUFFER_SIZE = 8 * 1024 # bytes
68
69# py3k has only new style classes
70__metaclass__ = type
71
72class BlockingIOError(IOError):
73
74 """Exception raised when I/O would block on a non-blocking I/O stream."""
75
76 def __init__(self, errno, strerror, characters_written=0):
77 IOError.__init__(self, errno, strerror)
78 self.characters_written = characters_written
79
80
81def open(file, mode="r", buffering=None, encoding=None, errors=None,
82 newline=None, closefd=True):
Benjamin Peterson9ae080e2008-05-04 22:39:33 +000083 r"""Open file and return a stream. If the file cannot be opened, an IOError is
84 raised.
Christian Heimes1a6387e2008-03-26 12:49:49 +000085
Benjamin Peterson7bb4d2d2008-04-13 02:01:27 +000086 file is either a string giving the name (and the path if the file
87 isn't in the current working directory) of the file to be opened or an
88 integer file descriptor of the file to be wrapped. (If a file
89 descriptor is given, it is closed when the returned I/O object is
90 closed, unless closefd is set to False.)
Christian Heimes1a6387e2008-03-26 12:49:49 +000091
Benjamin Peterson7bb4d2d2008-04-13 02:01:27 +000092 mode is an optional string that specifies the mode in which the file
93 is opened. It defaults to 'r' which means open for reading in text
94 mode. Other common values are 'w' for writing (truncating the file if
95 it already exists), and 'a' for appending (which on some Unix systems,
96 means that all writes append to the end of the file regardless of the
97 current seek position). In text mode, if encoding is not specified the
98 encoding used is platform dependent. (For reading and writing raw
99 bytes use binary mode and leave encoding unspecified.) The available
100 modes are:
Christian Heimes1a6387e2008-03-26 12:49:49 +0000101
Benjamin Peterson7bb4d2d2008-04-13 02:01:27 +0000102 ========= ===============================================================
103 Character Meaning
104 --------- ---------------------------------------------------------------
105 'r' open for reading (default)
106 'w' open for writing, truncating the file first
107 'a' open for writing, appending to the end of the file if it exists
108 'b' binary mode
109 't' text mode (default)
110 '+' open a disk file for updating (reading and writing)
111 'U' universal newline mode (for backwards compatibility; unneeded
112 for new code)
113 ========= ===============================================================
Christian Heimes1a6387e2008-03-26 12:49:49 +0000114
Benjamin Peterson7bb4d2d2008-04-13 02:01:27 +0000115 The default mode is 'rt' (open for reading text). For binary random
116 access, the mode 'w+b' opens and truncates the file to 0 bytes, while
117 'r+b' opens the file without truncation.
Christian Heimes1a6387e2008-03-26 12:49:49 +0000118
Benjamin Peterson7bb4d2d2008-04-13 02:01:27 +0000119 Python distinguishes between files opened in binary and text modes,
120 even when the underlying operating system doesn't. Files opened in
121 binary mode (appending 'b' to the mode argument) return contents as
122 bytes objects without any decoding. In text mode (the default, or when
123 't' is appended to the mode argument), the contents of the file are
124 returned as strings, the bytes having been first decoded using a
125 platform-dependent encoding or using the specified encoding if given.
Christian Heimes1a6387e2008-03-26 12:49:49 +0000126
Antoine Pitroub9767262009-12-19 21:03:36 +0000127 buffering is an optional integer used to set the buffering policy.
128 Pass 0 to switch buffering off (only allowed in binary mode), 1 to select
129 line buffering (only usable in text mode), and an integer > 1 to indicate
130 the size of a fixed-size chunk buffer. When no buffering argument is
131 given, the default buffering policy works as follows:
132
133 * Binary files are buffered in fixed-size chunks; the size of the buffer
134 is chosen using a heuristic trying to determine the underlying device's
135 "block size" and falling back on `io.DEFAULT_BUFFER_SIZE`.
136 On many systems, the buffer will typically be 4096 or 8192 bytes long.
137
138 * "Interactive" text files (files for which isatty() returns True)
139 use line buffering. Other text files use the policy described above
140 for binary files.
Christian Heimes1a6387e2008-03-26 12:49:49 +0000141
Benjamin Peterson7bb4d2d2008-04-13 02:01:27 +0000142 encoding is the name of the encoding used to decode or encode the
143 file. This should only be used in text mode. The default encoding is
144 platform dependent, but any encoding supported by Python can be
145 passed. See the codecs module for the list of supported encodings.
Christian Heimes1a6387e2008-03-26 12:49:49 +0000146
Benjamin Peterson7bb4d2d2008-04-13 02:01:27 +0000147 errors is an optional string that specifies how encoding errors are to
148 be handled---this argument should not be used in binary mode. Pass
149 'strict' to raise a ValueError exception if there is an encoding error
150 (the default of None has the same effect), or pass 'ignore' to ignore
151 errors. (Note that ignoring encoding errors can lead to data loss.)
152 See the documentation for codecs.register for a list of the permitted
153 encoding error strings.
154
155 newline controls how universal newlines works (it only applies to text
156 mode). It can be None, '', '\n', '\r', and '\r\n'. It works as
157 follows:
158
159 * On input, if newline is None, universal newlines mode is
160 enabled. Lines in the input can end in '\n', '\r', or '\r\n', and
161 these are translated into '\n' before being returned to the
162 caller. If it is '', universal newline mode is enabled, but line
163 endings are returned to the caller untranslated. If it has any of
164 the other legal values, input lines are only terminated by the given
165 string, and the line ending is returned to the caller untranslated.
166
167 * On output, if newline is None, any '\n' characters written are
168 translated to the system default line separator, os.linesep. If
169 newline is '', no translation takes place. If newline is any of the
170 other legal values, any '\n' characters written are translated to
171 the given string.
172
173 If closefd is False, the underlying file descriptor will be kept open
174 when the file is closed. This does not work when a file name is given
175 and must be True in that case.
176
177 open() returns a file object whose type depends on the mode, and
178 through which the standard file operations such as reading and writing
179 are performed. When open() is used to open a file in a text mode ('w',
180 'r', 'wt', 'rt', etc.), it returns a TextIOWrapper. When used to open
181 a file in a binary mode, the returned class varies: in read binary
182 mode, it returns a BufferedReader; in write binary and append binary
183 modes, it returns a BufferedWriter, and in read/write mode, it returns
184 a BufferedRandom.
185
186 It is also possible to use a string or bytearray as a file for both
187 reading and writing. For strings StringIO can be used like a file
188 opened in a text mode, and for bytes a BytesIO can be used like a file
189 opened in a binary mode.
Christian Heimes1a6387e2008-03-26 12:49:49 +0000190 """
Christian Heimes3784c6b2008-03-26 23:13:59 +0000191 if not isinstance(file, (basestring, int)):
Christian Heimes1a6387e2008-03-26 12:49:49 +0000192 raise TypeError("invalid file: %r" % file)
Christian Heimes3784c6b2008-03-26 23:13:59 +0000193 if not isinstance(mode, basestring):
Christian Heimes1a6387e2008-03-26 12:49:49 +0000194 raise TypeError("invalid mode: %r" % mode)
195 if buffering is not None and not isinstance(buffering, int):
196 raise TypeError("invalid buffering: %r" % buffering)
Christian Heimes3784c6b2008-03-26 23:13:59 +0000197 if encoding is not None and not isinstance(encoding, basestring):
Christian Heimes1a6387e2008-03-26 12:49:49 +0000198 raise TypeError("invalid encoding: %r" % encoding)
Christian Heimes3784c6b2008-03-26 23:13:59 +0000199 if errors is not None and not isinstance(errors, basestring):
Christian Heimes1a6387e2008-03-26 12:49:49 +0000200 raise TypeError("invalid errors: %r" % errors)
201 modes = set(mode)
202 if modes - set("arwb+tU") or len(mode) > len(modes):
203 raise ValueError("invalid mode: %r" % mode)
204 reading = "r" in modes
205 writing = "w" in modes
206 appending = "a" in modes
207 updating = "+" in modes
208 text = "t" in modes
209 binary = "b" in modes
210 if "U" in modes:
211 if writing or appending:
212 raise ValueError("can't use U and writing mode at once")
213 reading = True
214 if text and binary:
215 raise ValueError("can't have text and binary mode at once")
216 if reading + writing + appending > 1:
217 raise ValueError("can't have read/write/append mode at once")
218 if not (reading or writing or appending):
219 raise ValueError("must have exactly one of read/write/append mode")
220 if binary and encoding is not None:
221 raise ValueError("binary mode doesn't take an encoding argument")
222 if binary and errors is not None:
223 raise ValueError("binary mode doesn't take an errors argument")
224 if binary and newline is not None:
225 raise ValueError("binary mode doesn't take a newline argument")
226 raw = FileIO(file,
227 (reading and "r" or "") +
228 (writing and "w" or "") +
229 (appending and "a" or "") +
230 (updating and "+" or ""),
231 closefd)
232 if buffering is None:
233 buffering = -1
234 line_buffering = False
235 if buffering == 1 or buffering < 0 and raw.isatty():
236 buffering = -1
237 line_buffering = True
238 if buffering < 0:
239 buffering = DEFAULT_BUFFER_SIZE
240 try:
241 bs = os.fstat(raw.fileno()).st_blksize
242 except (os.error, AttributeError):
243 pass
244 else:
245 if bs > 1:
246 buffering = bs
247 if buffering < 0:
248 raise ValueError("invalid buffering size")
249 if buffering == 0:
250 if binary:
Christian Heimes1a6387e2008-03-26 12:49:49 +0000251 return raw
252 raise ValueError("can't have unbuffered text I/O")
253 if updating:
254 buffer = BufferedRandom(raw, buffering)
255 elif writing or appending:
256 buffer = BufferedWriter(raw, buffering)
257 elif reading:
258 buffer = BufferedReader(raw, buffering)
259 else:
260 raise ValueError("unknown mode: %r" % mode)
261 if binary:
Christian Heimes1a6387e2008-03-26 12:49:49 +0000262 return buffer
263 text = TextIOWrapper(buffer, encoding, errors, newline, line_buffering)
Christian Heimes1a6387e2008-03-26 12:49:49 +0000264 text.mode = mode
265 return text
266
267class _DocDescriptor:
268 """Helper for builtins.open.__doc__
269 """
270 def __get__(self, obj, typ):
271 return (
272 "open(file, mode='r', buffering=None, encoding=None, "
273 "errors=None, newline=None, closefd=True)\n\n" +
274 open.__doc__)
275
276class OpenWrapper:
277 """Wrapper for builtins.open
278
279 Trick so that open won't become a bound method when stored
280 as a class variable (as dumbdbm does).
281
282 See initstdio() in Python/pythonrun.c.
283 """
284 __doc__ = _DocDescriptor()
285
286 def __new__(cls, *args, **kwargs):
287 return open(*args, **kwargs)
288
289
290class UnsupportedOperation(ValueError, IOError):
291 pass
292
293
294class IOBase(object):
295
Benjamin Peterson9ae080e2008-05-04 22:39:33 +0000296 """The abstract base class for all I/O classes, acting on streams of
Benjamin Peterson7bb4d2d2008-04-13 02:01:27 +0000297 bytes. There is no public constructor.
Christian Heimes1a6387e2008-03-26 12:49:49 +0000298
299 This class provides dummy implementations for many methods that
Benjamin Peterson7bb4d2d2008-04-13 02:01:27 +0000300 derived classes can override selectively; the default implementations
301 represent a file that cannot be read, written or seeked.
Christian Heimes1a6387e2008-03-26 12:49:49 +0000302
Benjamin Peterson7bb4d2d2008-04-13 02:01:27 +0000303 Even though IOBase does not declare read, readinto, or write because
304 their signatures will vary, implementations and clients should
305 consider those methods part of the interface. Also, implementations
306 may raise a IOError when operations they do not support are called.
Christian Heimes1a6387e2008-03-26 12:49:49 +0000307
Benjamin Peterson7bb4d2d2008-04-13 02:01:27 +0000308 The basic type used for binary data read from or written to a file is
309 bytes. bytearrays are accepted too, and in some cases (such as
310 readinto) needed. Text I/O classes work with str data.
311
312 Note that calling any method (even inquiries) on a closed stream is
313 undefined. Implementations may raise IOError in this case.
314
315 IOBase (and its subclasses) support the iterator protocol, meaning
316 that an IOBase object can be iterated over yielding the lines in a
317 stream.
318
319 IOBase also supports the :keyword:`with` statement. In this example,
320 fp is closed after the suite of the with statment is complete:
321
322 with open('spam.txt', 'r') as fp:
323 fp.write('Spam and eggs!')
Christian Heimes1a6387e2008-03-26 12:49:49 +0000324 """
325
326 __metaclass__ = abc.ABCMeta
327
328 ### Internal ###
329
330 def _unsupported(self, name):
331 """Internal: raise an exception for unsupported operations."""
332 raise UnsupportedOperation("%s.%s() not supported" %
333 (self.__class__.__name__, name))
334
335 ### Positioning ###
336
337 def seek(self, pos, whence = 0):
Benjamin Peterson9ae080e2008-05-04 22:39:33 +0000338 """Change stream position.
Christian Heimes1a6387e2008-03-26 12:49:49 +0000339
Benjamin Peterson7bb4d2d2008-04-13 02:01:27 +0000340 Change the stream position to byte offset offset. offset is
341 interpreted relative to the position indicated by whence. Values
342 for whence are:
343
344 * 0 -- start of stream (the default); offset should be zero or positive
345 * 1 -- current stream position; offset may be negative
346 * 2 -- end of stream; offset is usually negative
347
348 Return the new absolute position.
Christian Heimes1a6387e2008-03-26 12:49:49 +0000349 """
350 self._unsupported("seek")
351
352 def tell(self):
Benjamin Peterson9ae080e2008-05-04 22:39:33 +0000353 """Return current stream position."""
Christian Heimes1a6387e2008-03-26 12:49:49 +0000354 return self.seek(0, 1)
355
356 def truncate(self, pos = None):
Benjamin Peterson9ae080e2008-05-04 22:39:33 +0000357 """Truncate file to size bytes.
Christian Heimes1a6387e2008-03-26 12:49:49 +0000358
Benjamin Peterson9ae080e2008-05-04 22:39:33 +0000359 Size defaults to the current IO position as reported by tell(). Return
360 the new size.
Christian Heimes1a6387e2008-03-26 12:49:49 +0000361 """
362 self._unsupported("truncate")
363
364 ### Flush and close ###
365
366 def flush(self):
Benjamin Peterson9ae080e2008-05-04 22:39:33 +0000367 """Flush write buffers, if applicable.
Christian Heimes1a6387e2008-03-26 12:49:49 +0000368
Benjamin Peterson7bb4d2d2008-04-13 02:01:27 +0000369 This is not implemented for read-only and non-blocking streams.
Christian Heimes1a6387e2008-03-26 12:49:49 +0000370 """
371 # XXX Should this return the number of bytes written???
372
373 __closed = False
374
375 def close(self):
Benjamin Peterson9ae080e2008-05-04 22:39:33 +0000376 """Flush and close the IO object.
Christian Heimes1a6387e2008-03-26 12:49:49 +0000377
Benjamin Peterson7bb4d2d2008-04-13 02:01:27 +0000378 This method has no effect if the file is already closed.
Christian Heimes1a6387e2008-03-26 12:49:49 +0000379 """
380 if not self.__closed:
381 try:
382 self.flush()
383 except IOError:
384 pass # If flush() fails, just give up
385 self.__closed = True
386
387 def __del__(self):
388 """Destructor. Calls close()."""
389 # The try/except block is in case this is called at program
390 # exit time, when it's possible that globals have already been
391 # deleted, and then the close() call might fail. Since
392 # there's nothing we can do about such failures and they annoy
393 # the end users, we suppress the traceback.
394 try:
395 self.close()
396 except:
397 pass
398
399 ### Inquiries ###
400
401 def seekable(self):
Benjamin Peterson9ae080e2008-05-04 22:39:33 +0000402 """Return whether object supports random access.
Christian Heimes1a6387e2008-03-26 12:49:49 +0000403
404 If False, seek(), tell() and truncate() will raise IOError.
405 This method may need to do a test seek().
406 """
407 return False
408
409 def _checkSeekable(self, msg=None):
410 """Internal: raise an IOError if file is not seekable
411 """
412 if not self.seekable():
413 raise IOError("File or stream is not seekable."
414 if msg is None else msg)
415
416
417 def readable(self):
Benjamin Peterson9ae080e2008-05-04 22:39:33 +0000418 """Return whether object was opened for reading.
Christian Heimes1a6387e2008-03-26 12:49:49 +0000419
420 If False, read() will raise IOError.
421 """
422 return False
423
424 def _checkReadable(self, msg=None):
425 """Internal: raise an IOError if file is not readable
426 """
427 if not self.readable():
428 raise IOError("File or stream is not readable."
429 if msg is None else msg)
430
431 def writable(self):
Benjamin Peterson9ae080e2008-05-04 22:39:33 +0000432 """Return whether object was opened for writing.
Christian Heimes1a6387e2008-03-26 12:49:49 +0000433
434 If False, write() and truncate() will raise IOError.
435 """
436 return False
437
438 def _checkWritable(self, msg=None):
439 """Internal: raise an IOError if file is not writable
440 """
441 if not self.writable():
442 raise IOError("File or stream is not writable."
443 if msg is None else msg)
444
445 @property
446 def closed(self):
447 """closed: bool. True iff the file has been closed.
448
449 For backwards compatibility, this is a property, not a predicate.
450 """
451 return self.__closed
452
453 def _checkClosed(self, msg=None):
454 """Internal: raise an ValueError if file is closed
455 """
456 if self.closed:
457 raise ValueError("I/O operation on closed file."
458 if msg is None else msg)
459
460 ### Context manager ###
461
462 def __enter__(self):
463 """Context management protocol. Returns self."""
464 self._checkClosed()
465 return self
466
467 def __exit__(self, *args):
468 """Context management protocol. Calls close()"""
469 self.close()
470
471 ### Lower-level APIs ###
472
473 # XXX Should these be present even if unimplemented?
474
475 def fileno(self):
Benjamin Peterson9ae080e2008-05-04 22:39:33 +0000476 """Returns underlying file descriptor if one exists.
Christian Heimes1a6387e2008-03-26 12:49:49 +0000477
Benjamin Peterson9ae080e2008-05-04 22:39:33 +0000478 An IOError is raised if the IO object does not use a file descriptor.
Christian Heimes1a6387e2008-03-26 12:49:49 +0000479 """
480 self._unsupported("fileno")
481
482 def isatty(self):
Benjamin Peterson9ae080e2008-05-04 22:39:33 +0000483 """Return whether this is an 'interactive' stream.
Christian Heimes1a6387e2008-03-26 12:49:49 +0000484
Benjamin Peterson9ae080e2008-05-04 22:39:33 +0000485 Return False if it can't be determined.
Christian Heimes1a6387e2008-03-26 12:49:49 +0000486 """
487 self._checkClosed()
488 return False
489
490 ### Readline[s] and writelines ###
491
492 def readline(self, limit = -1):
Benjamin Peterson9ae080e2008-05-04 22:39:33 +0000493 r"""Read and return a line from the stream.
Benjamin Peterson7bb4d2d2008-04-13 02:01:27 +0000494
495 If limit is specified, at most limit bytes will be read.
496
497 The line terminator is always b'\n' for binary files; for text
498 files, the newlines argument to open can be used to select the line
499 terminator(s) recognized.
500 """
Alexandre Vassalotti1aed6242008-05-09 21:49:43 +0000501 self._checkClosed()
Christian Heimes1a6387e2008-03-26 12:49:49 +0000502 if hasattr(self, "peek"):
503 def nreadahead():
504 readahead = self.peek(1)
505 if not readahead:
506 return 1
507 n = (readahead.find(b"\n") + 1) or len(readahead)
508 if limit >= 0:
509 n = min(n, limit)
510 return n
511 else:
512 def nreadahead():
513 return 1
514 if limit is None:
515 limit = -1
Alexandre Vassalotti1aed6242008-05-09 21:49:43 +0000516 if not isinstance(limit, (int, long)):
517 raise TypeError("limit must be an integer")
Christian Heimes1a6387e2008-03-26 12:49:49 +0000518 res = bytearray()
519 while limit < 0 or len(res) < limit:
520 b = self.read(nreadahead())
521 if not b:
522 break
523 res += b
524 if res.endswith(b"\n"):
525 break
526 return bytes(res)
527
528 def __iter__(self):
529 self._checkClosed()
530 return self
531
532 def next(self):
533 line = self.readline()
534 if not line:
535 raise StopIteration
536 return line
537
538 def readlines(self, hint=None):
Benjamin Peterson9ae080e2008-05-04 22:39:33 +0000539 """Return a list of lines from the stream.
Benjamin Peterson7bb4d2d2008-04-13 02:01:27 +0000540
541 hint can be specified to control the number of lines read: no more
542 lines will be read if the total size (in bytes/characters) of all
543 lines so far exceeds hint.
544 """
Christian Heimes1a6387e2008-03-26 12:49:49 +0000545 if hint is None:
Alexandre Vassalotti1aed6242008-05-09 21:49:43 +0000546 hint = -1
547 if not isinstance(hint, (int, long)):
548 raise TypeError("hint must be an integer")
549 if hint <= 0:
Christian Heimes1a6387e2008-03-26 12:49:49 +0000550 return list(self)
551 n = 0
552 lines = []
553 for line in self:
554 lines.append(line)
555 n += len(line)
556 if n >= hint:
557 break
558 return lines
559
560 def writelines(self, lines):
561 self._checkClosed()
562 for line in lines:
563 self.write(line)
564
565
566class RawIOBase(IOBase):
567
Benjamin Peterson7bb4d2d2008-04-13 02:01:27 +0000568 """Base class for raw binary I/O."""
Christian Heimes1a6387e2008-03-26 12:49:49 +0000569
Benjamin Peterson7bb4d2d2008-04-13 02:01:27 +0000570 # The read() method is implemented by calling readinto(); derived
571 # classes that want to support read() only need to implement
572 # readinto() as a primitive operation. In general, readinto() can be
573 # more efficient than read().
Christian Heimes1a6387e2008-03-26 12:49:49 +0000574
Benjamin Peterson7bb4d2d2008-04-13 02:01:27 +0000575 # (It would be tempting to also provide an implementation of
576 # readinto() in terms of read(), in case the latter is a more suitable
577 # primitive operation, but that would lead to nasty recursion in case
578 # a subclass doesn't implement either.)
Christian Heimes1a6387e2008-03-26 12:49:49 +0000579
580 def read(self, n = -1):
Benjamin Peterson9ae080e2008-05-04 22:39:33 +0000581 """Read and return up to n bytes.
Christian Heimes1a6387e2008-03-26 12:49:49 +0000582
583 Returns an empty bytes array on EOF, or None if the object is
584 set not to block and has no data to read.
585 """
586 if n is None:
587 n = -1
588 if n < 0:
589 return self.readall()
590 b = bytearray(n.__index__())
591 n = self.readinto(b)
592 del b[n:]
593 return bytes(b)
594
595 def readall(self):
Benjamin Peterson9ae080e2008-05-04 22:39:33 +0000596 """Read until EOF, using multiple read() call."""
Christian Heimes1a6387e2008-03-26 12:49:49 +0000597 res = bytearray()
598 while True:
599 data = self.read(DEFAULT_BUFFER_SIZE)
600 if not data:
601 break
602 res += data
603 return bytes(res)
604
605 def readinto(self, b):
Benjamin Peterson9ae080e2008-05-04 22:39:33 +0000606 """Read up to len(b) bytes into b.
Christian Heimes1a6387e2008-03-26 12:49:49 +0000607
608 Returns number of bytes read (0 for EOF), or None if the object
609 is set not to block as has no data to read.
610 """
611 self._unsupported("readinto")
612
613 def write(self, b):
Benjamin Peterson9ae080e2008-05-04 22:39:33 +0000614 """Write the given buffer to the IO stream.
Christian Heimes1a6387e2008-03-26 12:49:49 +0000615
616 Returns the number of bytes written, which may be less than len(b).
617 """
618 self._unsupported("write")
619
620
621class FileIO(_fileio._FileIO, RawIOBase):
622
Benjamin Peterson7bb4d2d2008-04-13 02:01:27 +0000623 """Raw I/O implementation for OS files."""
Christian Heimes1a6387e2008-03-26 12:49:49 +0000624
Benjamin Peterson7bb4d2d2008-04-13 02:01:27 +0000625 # This multiply inherits from _FileIO and RawIOBase to make
626 # isinstance(io.FileIO(), io.RawIOBase) return True without requiring
627 # that _fileio._FileIO inherits from io.RawIOBase (which would be hard
628 # to do since _fileio.c is written in C).
Christian Heimes1a6387e2008-03-26 12:49:49 +0000629
Georg Brandld2094602008-12-05 08:51:30 +0000630 def __init__(self, name, mode="r", closefd=True):
631 _fileio._FileIO.__init__(self, name, mode, closefd)
632 self._name = name
633
Christian Heimes1a6387e2008-03-26 12:49:49 +0000634 def close(self):
635 _fileio._FileIO.close(self)
636 RawIOBase.close(self)
637
638 @property
639 def name(self):
640 return self._name
641
Christian Heimes1a6387e2008-03-26 12:49:49 +0000642
643class BufferedIOBase(IOBase):
644
645 """Base class for buffered IO objects.
646
647 The main difference with RawIOBase is that the read() method
648 supports omitting the size argument, and does not have a default
649 implementation that defers to readinto().
650
651 In addition, read(), readinto() and write() may raise
652 BlockingIOError if the underlying raw stream is in non-blocking
653 mode and not ready; unlike their raw counterparts, they will never
654 return None.
655
656 A typical implementation should not inherit from a RawIOBase
657 implementation, but wrap one.
658 """
659
660 def read(self, n = None):
Benjamin Peterson9ae080e2008-05-04 22:39:33 +0000661 """Read and return up to n bytes.
Christian Heimes1a6387e2008-03-26 12:49:49 +0000662
663 If the argument is omitted, None, or negative, reads and
664 returns all data until EOF.
665
666 If the argument is positive, and the underlying raw stream is
667 not 'interactive', multiple raw reads may be issued to satisfy
668 the byte count (unless EOF is reached first). But for
669 interactive raw streams (XXX and for pipes?), at most one raw
670 read will be issued, and a short result does not imply that
671 EOF is imminent.
672
673 Returns an empty bytes array on EOF.
674
675 Raises BlockingIOError if the underlying raw stream has no
676 data at the moment.
677 """
678 self._unsupported("read")
679
680 def readinto(self, b):
Benjamin Peterson9ae080e2008-05-04 22:39:33 +0000681 """Read up to len(b) bytes into b.
Christian Heimes1a6387e2008-03-26 12:49:49 +0000682
Benjamin Peterson7bb4d2d2008-04-13 02:01:27 +0000683 Like read(), this may issue multiple reads to the underlying raw
684 stream, unless the latter is 'interactive'.
Christian Heimes1a6387e2008-03-26 12:49:49 +0000685
686 Returns the number of bytes read (0 for EOF).
687
688 Raises BlockingIOError if the underlying raw stream has no
689 data at the moment.
690 """
691 # XXX This ought to work with anything that supports the buffer API
692 data = self.read(len(b))
693 n = len(data)
694 try:
695 b[:n] = data
696 except TypeError as err:
697 import array
698 if not isinstance(b, array.array):
699 raise err
Alexandre Vassalotti1aed6242008-05-09 21:49:43 +0000700 b[:n] = array.array(b'b', data)
Christian Heimes1a6387e2008-03-26 12:49:49 +0000701 return n
702
703 def write(self, b):
Benjamin Peterson9ae080e2008-05-04 22:39:33 +0000704 """Write the given buffer to the IO stream.
Christian Heimes1a6387e2008-03-26 12:49:49 +0000705
Benjamin Peterson9ae080e2008-05-04 22:39:33 +0000706 Return the number of bytes written, which is never less than
Christian Heimes1a6387e2008-03-26 12:49:49 +0000707 len(b).
708
709 Raises BlockingIOError if the buffer is full and the
710 underlying raw stream cannot accept more data at the moment.
711 """
712 self._unsupported("write")
713
714
715class _BufferedIOMixin(BufferedIOBase):
716
717 """A mixin implementation of BufferedIOBase with an underlying raw stream.
718
719 This passes most requests on to the underlying raw stream. It
720 does *not* provide implementations of read(), readinto() or
721 write().
722 """
723
724 def __init__(self, raw):
725 self.raw = raw
726
727 ### Positioning ###
728
729 def seek(self, pos, whence=0):
730 return self.raw.seek(pos, whence)
731
732 def tell(self):
733 return self.raw.tell()
734
735 def truncate(self, pos=None):
736 # Flush the stream. We're mixing buffered I/O with lower-level I/O,
737 # and a flush may be necessary to synch both views of the current
738 # file state.
739 self.flush()
740
741 if pos is None:
742 pos = self.tell()
Alexandre Vassalotti1aed6242008-05-09 21:49:43 +0000743 # XXX: Should seek() be used, instead of passing the position
744 # XXX directly to truncate?
Christian Heimes1a6387e2008-03-26 12:49:49 +0000745 return self.raw.truncate(pos)
746
747 ### Flush and close ###
748
749 def flush(self):
750 self.raw.flush()
751
752 def close(self):
753 if not self.closed:
754 try:
755 self.flush()
756 except IOError:
757 pass # If flush() fails, just give up
758 self.raw.close()
759
760 ### Inquiries ###
761
762 def seekable(self):
763 return self.raw.seekable()
764
765 def readable(self):
766 return self.raw.readable()
767
768 def writable(self):
769 return self.raw.writable()
770
771 @property
772 def closed(self):
773 return self.raw.closed
774
Georg Brandld2094602008-12-05 08:51:30 +0000775 @property
776 def name(self):
777 return self.raw.name
778
779 @property
780 def mode(self):
781 return self.raw.mode
782
Christian Heimes1a6387e2008-03-26 12:49:49 +0000783 ### Lower-level APIs ###
784
785 def fileno(self):
786 return self.raw.fileno()
787
788 def isatty(self):
789 return self.raw.isatty()
790
791
Alexandre Vassalotti1aed6242008-05-09 21:49:43 +0000792class _BytesIO(BufferedIOBase):
Christian Heimes1a6387e2008-03-26 12:49:49 +0000793
794 """Buffered I/O implementation using an in-memory bytes buffer."""
795
796 # XXX More docs
797
798 def __init__(self, initial_bytes=None):
799 buf = bytearray()
800 if initial_bytes is not None:
Alexandre Vassalotti1aed6242008-05-09 21:49:43 +0000801 buf += bytearray(initial_bytes)
Christian Heimes1a6387e2008-03-26 12:49:49 +0000802 self._buffer = buf
803 self._pos = 0
804
805 def getvalue(self):
Benjamin Peterson9ae080e2008-05-04 22:39:33 +0000806 """Return the bytes value (contents) of the buffer
Benjamin Peterson7bb4d2d2008-04-13 02:01:27 +0000807 """
Alexandre Vassalotti1aed6242008-05-09 21:49:43 +0000808 if self.closed:
809 raise ValueError("getvalue on closed file")
Christian Heimes1a6387e2008-03-26 12:49:49 +0000810 return bytes(self._buffer)
811
812 def read(self, n=None):
Alexandre Vassalotti1aed6242008-05-09 21:49:43 +0000813 if self.closed:
814 raise ValueError("read from closed file")
Christian Heimes1a6387e2008-03-26 12:49:49 +0000815 if n is None:
816 n = -1
Alexandre Vassalotti1aed6242008-05-09 21:49:43 +0000817 if not isinstance(n, (int, long)):
818 raise TypeError("argument must be an integer")
Christian Heimes1a6387e2008-03-26 12:49:49 +0000819 if n < 0:
820 n = len(self._buffer)
Alexandre Vassalotti1aed6242008-05-09 21:49:43 +0000821 if len(self._buffer) <= self._pos:
822 return b""
Christian Heimes1a6387e2008-03-26 12:49:49 +0000823 newpos = min(len(self._buffer), self._pos + n)
824 b = self._buffer[self._pos : newpos]
825 self._pos = newpos
826 return bytes(b)
827
828 def read1(self, n):
Benjamin Peterson9ae080e2008-05-04 22:39:33 +0000829 """this is the same as read.
Benjamin Peterson7bb4d2d2008-04-13 02:01:27 +0000830 """
Christian Heimes1a6387e2008-03-26 12:49:49 +0000831 return self.read(n)
832
833 def write(self, b):
834 if self.closed:
835 raise ValueError("write to closed file")
836 if isinstance(b, unicode):
837 raise TypeError("can't write unicode to binary stream")
838 n = len(b)
Alexandre Vassalotti1aed6242008-05-09 21:49:43 +0000839 if n == 0:
840 return 0
Alexandre Vassalotti844f7572008-05-10 19:59:16 +0000841 pos = self._pos
842 if pos > len(self._buffer):
Christian Heimes1a6387e2008-03-26 12:49:49 +0000843 # Inserts null bytes between the current end of the file
844 # and the new write position.
Alexandre Vassalotti844f7572008-05-10 19:59:16 +0000845 padding = b'\x00' * (pos - len(self._buffer))
846 self._buffer += padding
847 self._buffer[pos:pos + n] = b
848 self._pos += n
Christian Heimes1a6387e2008-03-26 12:49:49 +0000849 return n
850
851 def seek(self, pos, whence=0):
Alexandre Vassalotti1aed6242008-05-09 21:49:43 +0000852 if self.closed:
853 raise ValueError("seek on closed file")
Christian Heimes1a6387e2008-03-26 12:49:49 +0000854 try:
855 pos = pos.__index__()
856 except AttributeError as err:
857 raise TypeError("an integer is required") # from err
858 if whence == 0:
Alexandre Vassalotti1aed6242008-05-09 21:49:43 +0000859 if pos < 0:
860 raise ValueError("negative seek position %r" % (pos,))
861 self._pos = pos
Christian Heimes1a6387e2008-03-26 12:49:49 +0000862 elif whence == 1:
863 self._pos = max(0, self._pos + pos)
864 elif whence == 2:
865 self._pos = max(0, len(self._buffer) + pos)
866 else:
Alexandre Vassalotti1aed6242008-05-09 21:49:43 +0000867 raise ValueError("invalid whence value")
Christian Heimes1a6387e2008-03-26 12:49:49 +0000868 return self._pos
869
870 def tell(self):
Alexandre Vassalotti1aed6242008-05-09 21:49:43 +0000871 if self.closed:
872 raise ValueError("tell on closed file")
Christian Heimes1a6387e2008-03-26 12:49:49 +0000873 return self._pos
874
875 def truncate(self, pos=None):
Alexandre Vassalotti1aed6242008-05-09 21:49:43 +0000876 if self.closed:
877 raise ValueError("truncate on closed file")
Christian Heimes1a6387e2008-03-26 12:49:49 +0000878 if pos is None:
879 pos = self._pos
Alexandre Vassalotti1aed6242008-05-09 21:49:43 +0000880 elif pos < 0:
881 raise ValueError("negative truncate position %r" % (pos,))
Christian Heimes1a6387e2008-03-26 12:49:49 +0000882 del self._buffer[pos:]
Alexandre Vassalotti1aed6242008-05-09 21:49:43 +0000883 return self.seek(pos)
Christian Heimes1a6387e2008-03-26 12:49:49 +0000884
885 def readable(self):
886 return True
887
888 def writable(self):
889 return True
890
891 def seekable(self):
892 return True
893
Alexandre Vassalotti1aed6242008-05-09 21:49:43 +0000894# Use the faster implementation of BytesIO if available
895try:
896 import _bytesio
897
898 class BytesIO(_bytesio._BytesIO, BufferedIOBase):
899 __doc__ = _bytesio._BytesIO.__doc__
900
901except ImportError:
902 BytesIO = _BytesIO
903
Christian Heimes1a6387e2008-03-26 12:49:49 +0000904
905class BufferedReader(_BufferedIOMixin):
906
Benjamin Peterson7bb4d2d2008-04-13 02:01:27 +0000907 """BufferedReader(raw[, buffer_size])
908
909 A buffer for a readable, sequential BaseRawIO object.
910
911 The constructor creates a BufferedReader for the given readable raw
912 stream and buffer_size. If buffer_size is omitted, DEFAULT_BUFFER_SIZE
913 is used.
914 """
Christian Heimes1a6387e2008-03-26 12:49:49 +0000915
916 def __init__(self, raw, buffer_size=DEFAULT_BUFFER_SIZE):
917 """Create a new buffered reader using the given readable raw IO object.
918 """
919 raw._checkReadable()
920 _BufferedIOMixin.__init__(self, raw)
Christian Heimes1a6387e2008-03-26 12:49:49 +0000921 self.buffer_size = buffer_size
Benjamin Peterson01a24322008-07-28 23:35:27 +0000922 self._reset_read_buf()
Antoine Pitrou11ec65d2008-08-14 21:04:30 +0000923 self._read_lock = threading.Lock()
Benjamin Peterson01a24322008-07-28 23:35:27 +0000924
925 def _reset_read_buf(self):
926 self._read_buf = b""
927 self._read_pos = 0
Christian Heimes1a6387e2008-03-26 12:49:49 +0000928
929 def read(self, n=None):
930 """Read n bytes.
931
932 Returns exactly n bytes of data unless the underlying raw IO
933 stream reaches EOF or if the call would block in non-blocking
934 mode. If n is negative, read until EOF or until read() would
935 block.
936 """
Antoine Pitrou11ec65d2008-08-14 21:04:30 +0000937 with self._read_lock:
938 return self._read_unlocked(n)
939
940 def _read_unlocked(self, n=None):
Christian Heimes1a6387e2008-03-26 12:49:49 +0000941 nodata_val = b""
Benjamin Peterson01a24322008-07-28 23:35:27 +0000942 empty_values = (b"", None)
943 buf = self._read_buf
944 pos = self._read_pos
945
946 # Special case for when the number of bytes to read is unspecified.
947 if n is None or n == -1:
948 self._reset_read_buf()
949 chunks = [buf[pos:]] # Strip the consumed bytes.
950 current_size = 0
951 while True:
952 # Read until EOF or until read() would block.
953 chunk = self.raw.read()
954 if chunk in empty_values:
955 nodata_val = chunk
956 break
957 current_size += len(chunk)
958 chunks.append(chunk)
959 return b"".join(chunks) or nodata_val
960
961 # The number of bytes to read is specified, return at most n bytes.
962 avail = len(buf) - pos # Length of the available buffered data.
963 if n <= avail:
964 # Fast path: the data to read is fully buffered.
965 self._read_pos += n
966 return buf[pos:pos+n]
967 # Slow path: read from the stream until enough bytes are read,
968 # or until an EOF occurs or until read() would block.
969 chunks = [buf[pos:]]
970 wanted = max(self.buffer_size, n)
971 while avail < n:
972 chunk = self.raw.read(wanted)
973 if chunk in empty_values:
974 nodata_val = chunk
Christian Heimes1a6387e2008-03-26 12:49:49 +0000975 break
Benjamin Peterson01a24322008-07-28 23:35:27 +0000976 avail += len(chunk)
977 chunks.append(chunk)
978 # n is more then avail only when an EOF occurred or when
979 # read() would have blocked.
980 n = min(n, avail)
981 out = b"".join(chunks)
982 self._read_buf = out[n:] # Save the extra data in the buffer.
983 self._read_pos = 0
984 return out[:n] if out else nodata_val
Christian Heimes1a6387e2008-03-26 12:49:49 +0000985
986 def peek(self, n=0):
987 """Returns buffered bytes without advancing the position.
988
989 The argument indicates a desired minimal number of bytes; we
990 do at most one raw read to satisfy it. We never return more
991 than self.buffer_size.
992 """
Antoine Pitrou11ec65d2008-08-14 21:04:30 +0000993 with self._read_lock:
994 return self._peek_unlocked(n)
995
996 def _peek_unlocked(self, n=0):
Christian Heimes1a6387e2008-03-26 12:49:49 +0000997 want = min(n, self.buffer_size)
Benjamin Peterson01a24322008-07-28 23:35:27 +0000998 have = len(self._read_buf) - self._read_pos
Christian Heimes1a6387e2008-03-26 12:49:49 +0000999 if have < want:
1000 to_read = self.buffer_size - have
1001 current = self.raw.read(to_read)
1002 if current:
Benjamin Peterson01a24322008-07-28 23:35:27 +00001003 self._read_buf = self._read_buf[self._read_pos:] + current
1004 self._read_pos = 0
1005 return self._read_buf[self._read_pos:]
Christian Heimes1a6387e2008-03-26 12:49:49 +00001006
1007 def read1(self, n):
Benjamin Peterson7bb4d2d2008-04-13 02:01:27 +00001008 """Reads up to n bytes, with at most one read() system call."""
1009 # Returns up to n bytes. If at least one byte is buffered, we
1010 # only return buffered bytes. Otherwise, we do one raw read.
Christian Heimes1a6387e2008-03-26 12:49:49 +00001011 if n <= 0:
1012 return b""
Antoine Pitrou11ec65d2008-08-14 21:04:30 +00001013 with self._read_lock:
1014 self._peek_unlocked(1)
1015 return self._read_unlocked(
1016 min(n, len(self._read_buf) - self._read_pos))
Christian Heimes1a6387e2008-03-26 12:49:49 +00001017
1018 def tell(self):
Benjamin Peterson01a24322008-07-28 23:35:27 +00001019 return self.raw.tell() - len(self._read_buf) + self._read_pos
Christian Heimes1a6387e2008-03-26 12:49:49 +00001020
1021 def seek(self, pos, whence=0):
Antoine Pitrou11ec65d2008-08-14 21:04:30 +00001022 with self._read_lock:
1023 if whence == 1:
1024 pos -= len(self._read_buf) - self._read_pos
1025 pos = self.raw.seek(pos, whence)
1026 self._reset_read_buf()
1027 return pos
Christian Heimes1a6387e2008-03-26 12:49:49 +00001028
1029
1030class BufferedWriter(_BufferedIOMixin):
1031
Benjamin Peterson9ae080e2008-05-04 22:39:33 +00001032 """A buffer for a writeable sequential RawIO object.
Benjamin Peterson7bb4d2d2008-04-13 02:01:27 +00001033
1034 The constructor creates a BufferedWriter for the given writeable raw
1035 stream. If the buffer_size is not given, it defaults to
1036 DEAFULT_BUFFER_SIZE. If max_buffer_size is omitted, it defaults to
1037 twice the buffer size.
1038 """
Christian Heimes1a6387e2008-03-26 12:49:49 +00001039
1040 def __init__(self, raw,
1041 buffer_size=DEFAULT_BUFFER_SIZE, max_buffer_size=None):
1042 raw._checkWritable()
1043 _BufferedIOMixin.__init__(self, raw)
1044 self.buffer_size = buffer_size
1045 self.max_buffer_size = (2*buffer_size
1046 if max_buffer_size is None
1047 else max_buffer_size)
1048 self._write_buf = bytearray()
Antoine Pitrou11ec65d2008-08-14 21:04:30 +00001049 self._write_lock = threading.Lock()
Christian Heimes1a6387e2008-03-26 12:49:49 +00001050
1051 def write(self, b):
1052 if self.closed:
1053 raise ValueError("write to closed file")
1054 if isinstance(b, unicode):
1055 raise TypeError("can't write unicode to binary stream")
Antoine Pitrou11ec65d2008-08-14 21:04:30 +00001056 with self._write_lock:
1057 # XXX we can implement some more tricks to try and avoid
1058 # partial writes
1059 if len(self._write_buf) > self.buffer_size:
1060 # We're full, so let's pre-flush the buffer
1061 try:
1062 self._flush_unlocked()
1063 except BlockingIOError as e:
1064 # We can't accept anything else.
1065 # XXX Why not just let the exception pass through?
1066 raise BlockingIOError(e.errno, e.strerror, 0)
1067 before = len(self._write_buf)
1068 self._write_buf.extend(b)
1069 written = len(self._write_buf) - before
1070 if len(self._write_buf) > self.buffer_size:
1071 try:
1072 self._flush_unlocked()
1073 except BlockingIOError as e:
1074 if len(self._write_buf) > self.max_buffer_size:
1075 # We've hit max_buffer_size. We have to accept a
1076 # partial write and cut back our buffer.
1077 overage = len(self._write_buf) - self.max_buffer_size
1078 self._write_buf = self._write_buf[:self.max_buffer_size]
1079 raise BlockingIOError(e.errno, e.strerror, overage)
1080 return written
Christian Heimes1a6387e2008-03-26 12:49:49 +00001081
Alexandre Vassalotti1aed6242008-05-09 21:49:43 +00001082 def truncate(self, pos=None):
Antoine Pitrou11ec65d2008-08-14 21:04:30 +00001083 with self._write_lock:
1084 self._flush_unlocked()
1085 if pos is None:
1086 pos = self.raw.tell()
1087 return self.raw.truncate(pos)
Alexandre Vassalotti1aed6242008-05-09 21:49:43 +00001088
Christian Heimes1a6387e2008-03-26 12:49:49 +00001089 def flush(self):
Antoine Pitrou11ec65d2008-08-14 21:04:30 +00001090 with self._write_lock:
1091 self._flush_unlocked()
1092
1093 def _flush_unlocked(self):
Christian Heimes1a6387e2008-03-26 12:49:49 +00001094 if self.closed:
1095 raise ValueError("flush of closed file")
1096 written = 0
1097 try:
1098 while self._write_buf:
1099 n = self.raw.write(self._write_buf)
1100 del self._write_buf[:n]
1101 written += n
1102 except BlockingIOError as e:
1103 n = e.characters_written
1104 del self._write_buf[:n]
1105 written += n
1106 raise BlockingIOError(e.errno, e.strerror, written)
1107
1108 def tell(self):
1109 return self.raw.tell() + len(self._write_buf)
1110
1111 def seek(self, pos, whence=0):
Antoine Pitrou11ec65d2008-08-14 21:04:30 +00001112 with self._write_lock:
1113 self._flush_unlocked()
1114 return self.raw.seek(pos, whence)
Christian Heimes1a6387e2008-03-26 12:49:49 +00001115
1116
1117class BufferedRWPair(BufferedIOBase):
1118
1119 """A buffered reader and writer object together.
1120
Benjamin Peterson7bb4d2d2008-04-13 02:01:27 +00001121 A buffered reader object and buffered writer object put together to
1122 form a sequential IO object that can read and write. This is typically
1123 used with a socket or two-way pipe.
Christian Heimes1a6387e2008-03-26 12:49:49 +00001124
Benjamin Peterson7bb4d2d2008-04-13 02:01:27 +00001125 reader and writer are RawIOBase objects that are readable and
1126 writeable respectively. If the buffer_size is omitted it defaults to
1127 DEFAULT_BUFFER_SIZE. The max_buffer_size (for the buffered writer)
1128 defaults to twice the buffer size.
Christian Heimes1a6387e2008-03-26 12:49:49 +00001129 """
1130
Benjamin Peterson7bb4d2d2008-04-13 02:01:27 +00001131 # XXX The usefulness of this (compared to having two separate IO
1132 # objects) is questionable.
1133
Christian Heimes1a6387e2008-03-26 12:49:49 +00001134 def __init__(self, reader, writer,
1135 buffer_size=DEFAULT_BUFFER_SIZE, max_buffer_size=None):
1136 """Constructor.
1137
1138 The arguments are two RawIO instances.
1139 """
1140 reader._checkReadable()
1141 writer._checkWritable()
1142 self.reader = BufferedReader(reader, buffer_size)
1143 self.writer = BufferedWriter(writer, buffer_size, max_buffer_size)
1144
1145 def read(self, n=None):
1146 if n is None:
1147 n = -1
1148 return self.reader.read(n)
1149
1150 def readinto(self, b):
1151 return self.reader.readinto(b)
1152
1153 def write(self, b):
1154 return self.writer.write(b)
1155
1156 def peek(self, n=0):
1157 return self.reader.peek(n)
1158
1159 def read1(self, n):
1160 return self.reader.read1(n)
1161
1162 def readable(self):
1163 return self.reader.readable()
1164
1165 def writable(self):
1166 return self.writer.writable()
1167
1168 def flush(self):
1169 return self.writer.flush()
1170
1171 def close(self):
1172 self.writer.close()
1173 self.reader.close()
1174
1175 def isatty(self):
1176 return self.reader.isatty() or self.writer.isatty()
1177
1178 @property
1179 def closed(self):
Benjamin Peterson828a7062008-12-27 17:05:29 +00001180 return self.writer.closed
Christian Heimes1a6387e2008-03-26 12:49:49 +00001181
1182
1183class BufferedRandom(BufferedWriter, BufferedReader):
1184
Benjamin Peterson9ae080e2008-05-04 22:39:33 +00001185 """A buffered interface to random access streams.
Benjamin Peterson7bb4d2d2008-04-13 02:01:27 +00001186
1187 The constructor creates a reader and writer for a seekable stream,
1188 raw, given in the first argument. If the buffer_size is omitted it
1189 defaults to DEFAULT_BUFFER_SIZE. The max_buffer_size (for the buffered
1190 writer) defaults to twice the buffer size.
1191 """
Christian Heimes1a6387e2008-03-26 12:49:49 +00001192
1193 def __init__(self, raw,
1194 buffer_size=DEFAULT_BUFFER_SIZE, max_buffer_size=None):
1195 raw._checkSeekable()
1196 BufferedReader.__init__(self, raw, buffer_size)
1197 BufferedWriter.__init__(self, raw, buffer_size, max_buffer_size)
1198
1199 def seek(self, pos, whence=0):
1200 self.flush()
1201 # First do the raw seek, then empty the read buffer, so that
1202 # if the raw seek fails, we don't lose buffered data forever.
1203 pos = self.raw.seek(pos, whence)
Antoine Pitrou11ec65d2008-08-14 21:04:30 +00001204 with self._read_lock:
1205 self._reset_read_buf()
Christian Heimes1a6387e2008-03-26 12:49:49 +00001206 return pos
1207
1208 def tell(self):
Benjamin Peterson01a24322008-07-28 23:35:27 +00001209 if self._write_buf:
Christian Heimes1a6387e2008-03-26 12:49:49 +00001210 return self.raw.tell() + len(self._write_buf)
1211 else:
Benjamin Peterson01a24322008-07-28 23:35:27 +00001212 return BufferedReader.tell(self)
Christian Heimes1a6387e2008-03-26 12:49:49 +00001213
Alexandre Vassalotti1aed6242008-05-09 21:49:43 +00001214 def truncate(self, pos=None):
1215 if pos is None:
1216 pos = self.tell()
1217 # Use seek to flush the read buffer.
1218 self.seek(pos)
1219 return BufferedWriter.truncate(self)
1220
Christian Heimes1a6387e2008-03-26 12:49:49 +00001221 def read(self, n=None):
1222 if n is None:
1223 n = -1
1224 self.flush()
1225 return BufferedReader.read(self, n)
1226
1227 def readinto(self, b):
1228 self.flush()
1229 return BufferedReader.readinto(self, b)
1230
1231 def peek(self, n=0):
1232 self.flush()
1233 return BufferedReader.peek(self, n)
1234
1235 def read1(self, n):
1236 self.flush()
1237 return BufferedReader.read1(self, n)
1238
1239 def write(self, b):
1240 if self._read_buf:
Benjamin Peterson01a24322008-07-28 23:35:27 +00001241 # Undo readahead
Antoine Pitrou11ec65d2008-08-14 21:04:30 +00001242 with self._read_lock:
1243 self.raw.seek(self._read_pos - len(self._read_buf), 1)
1244 self._reset_read_buf()
Christian Heimes1a6387e2008-03-26 12:49:49 +00001245 return BufferedWriter.write(self, b)
1246
1247
1248class TextIOBase(IOBase):
1249
1250 """Base class for text I/O.
1251
Benjamin Peterson7bb4d2d2008-04-13 02:01:27 +00001252 This class provides a character and line based interface to stream
1253 I/O. There is no readinto method because Python's character strings
1254 are immutable. There is no public constructor.
Christian Heimes1a6387e2008-03-26 12:49:49 +00001255 """
1256
1257 def read(self, n = -1):
Benjamin Peterson9ae080e2008-05-04 22:39:33 +00001258 """Read at most n characters from stream.
Christian Heimes1a6387e2008-03-26 12:49:49 +00001259
1260 Read from underlying buffer until we have n characters or we hit EOF.
1261 If n is negative or omitted, read until EOF.
1262 """
1263 self._unsupported("read")
1264
1265 def write(self, s):
Benjamin Peterson9ae080e2008-05-04 22:39:33 +00001266 """Write string s to stream."""
Christian Heimes1a6387e2008-03-26 12:49:49 +00001267 self._unsupported("write")
1268
1269 def truncate(self, pos = None):
Benjamin Peterson9ae080e2008-05-04 22:39:33 +00001270 """Truncate size to pos."""
Alexandre Vassalotti1aed6242008-05-09 21:49:43 +00001271 self._unsupported("truncate")
Christian Heimes1a6387e2008-03-26 12:49:49 +00001272
1273 def readline(self):
Benjamin Peterson9ae080e2008-05-04 22:39:33 +00001274 """Read until newline or EOF.
Christian Heimes1a6387e2008-03-26 12:49:49 +00001275
1276 Returns an empty string if EOF is hit immediately.
1277 """
1278 self._unsupported("readline")
1279
1280 @property
1281 def encoding(self):
1282 """Subclasses should override."""
1283 return None
1284
1285 @property
1286 def newlines(self):
Benjamin Peterson9ae080e2008-05-04 22:39:33 +00001287 """Line endings translated so far.
Christian Heimes1a6387e2008-03-26 12:49:49 +00001288
1289 Only line endings translated during reading are considered.
1290
1291 Subclasses should override.
1292 """
1293 return None
1294
1295
1296class IncrementalNewlineDecoder(codecs.IncrementalDecoder):
1297 """Codec used when reading a file in universal newlines mode.
1298 It wraps another incremental decoder, translating \\r\\n and \\r into \\n.
1299 It also records the types of newlines encountered.
1300 When used with translate=False, it ensures that the newline sequence is
1301 returned in one piece.
1302 """
1303 def __init__(self, decoder, translate, errors='strict'):
1304 codecs.IncrementalDecoder.__init__(self, errors=errors)
Christian Heimes1a6387e2008-03-26 12:49:49 +00001305 self.translate = translate
1306 self.decoder = decoder
1307 self.seennl = 0
Antoine Pitrouf8638a82008-12-14 18:08:37 +00001308 self.pendingcr = False
Christian Heimes1a6387e2008-03-26 12:49:49 +00001309
1310 def decode(self, input, final=False):
1311 # decode input (with the eventual \r from a previous pass)
Christian Heimes1a6387e2008-03-26 12:49:49 +00001312 output = self.decoder.decode(input, final=final)
Antoine Pitrouf8638a82008-12-14 18:08:37 +00001313 if self.pendingcr and (output or final):
1314 output = "\r" + output
1315 self.pendingcr = False
Christian Heimes1a6387e2008-03-26 12:49:49 +00001316
1317 # retain last \r even when not translating data:
1318 # then readline() is sure to get \r\n in one pass
1319 if output.endswith("\r") and not final:
1320 output = output[:-1]
Antoine Pitrouf8638a82008-12-14 18:08:37 +00001321 self.pendingcr = True
Christian Heimes1a6387e2008-03-26 12:49:49 +00001322
1323 # Record which newlines are read
1324 crlf = output.count('\r\n')
1325 cr = output.count('\r') - crlf
1326 lf = output.count('\n') - crlf
1327 self.seennl |= (lf and self._LF) | (cr and self._CR) \
1328 | (crlf and self._CRLF)
1329
1330 if self.translate:
1331 if crlf:
1332 output = output.replace("\r\n", "\n")
1333 if cr:
1334 output = output.replace("\r", "\n")
1335
1336 return output
1337
1338 def getstate(self):
1339 buf, flag = self.decoder.getstate()
Antoine Pitrouf8638a82008-12-14 18:08:37 +00001340 flag <<= 1
1341 if self.pendingcr:
1342 flag |= 1
1343 return buf, flag
Christian Heimes1a6387e2008-03-26 12:49:49 +00001344
1345 def setstate(self, state):
1346 buf, flag = state
Antoine Pitrouf8638a82008-12-14 18:08:37 +00001347 self.pendingcr = bool(flag & 1)
1348 self.decoder.setstate((buf, flag >> 1))
Christian Heimes1a6387e2008-03-26 12:49:49 +00001349
1350 def reset(self):
1351 self.seennl = 0
Antoine Pitrouf8638a82008-12-14 18:08:37 +00001352 self.pendingcr = False
Christian Heimes1a6387e2008-03-26 12:49:49 +00001353 self.decoder.reset()
1354
1355 _LF = 1
1356 _CR = 2
1357 _CRLF = 4
1358
1359 @property
1360 def newlines(self):
1361 return (None,
1362 "\n",
1363 "\r",
1364 ("\r", "\n"),
1365 "\r\n",
1366 ("\n", "\r\n"),
1367 ("\r", "\r\n"),
1368 ("\r", "\n", "\r\n")
1369 )[self.seennl]
1370
1371
1372class TextIOWrapper(TextIOBase):
1373
Benjamin Peterson9ae080e2008-05-04 22:39:33 +00001374 r"""Character and line based layer over a BufferedIOBase object, buffer.
Benjamin Peterson7bb4d2d2008-04-13 02:01:27 +00001375
1376 encoding gives the name of the encoding that the stream will be
1377 decoded or encoded with. It defaults to locale.getpreferredencoding.
1378
1379 errors determines the strictness of encoding and decoding (see the
1380 codecs.register) and defaults to "strict".
1381
1382 newline can be None, '', '\n', '\r', or '\r\n'. It controls the
1383 handling of line endings. If it is None, universal newlines is
1384 enabled. With this enabled, on input, the lines endings '\n', '\r',
1385 or '\r\n' are translated to '\n' before being returned to the
1386 caller. Conversely, on output, '\n' is translated to the system
Jesus Cea585ad8a2009-07-02 15:37:21 +00001387 default line separator, os.linesep. If newline is any other of its
Benjamin Peterson7bb4d2d2008-04-13 02:01:27 +00001388 legal values, that newline becomes the newline when the file is read
1389 and it is returned untranslated. On output, '\n' is converted to the
1390 newline.
1391
1392 If line_buffering is True, a call to flush is implied when a call to
1393 write contains a newline character.
Christian Heimes1a6387e2008-03-26 12:49:49 +00001394 """
1395
1396 _CHUNK_SIZE = 128
1397
1398 def __init__(self, buffer, encoding=None, errors=None, newline=None,
1399 line_buffering=False):
1400 if newline not in (None, "", "\n", "\r", "\r\n"):
1401 raise ValueError("illegal newline value: %r" % (newline,))
1402 if encoding is None:
1403 try:
1404 encoding = os.device_encoding(buffer.fileno())
1405 except (AttributeError, UnsupportedOperation):
1406 pass
1407 if encoding is None:
1408 try:
1409 import locale
1410 except ImportError:
1411 # Importing locale may fail if Python is being built
1412 encoding = "ascii"
1413 else:
1414 encoding = locale.getpreferredencoding()
1415
Christian Heimes3784c6b2008-03-26 23:13:59 +00001416 if not isinstance(encoding, basestring):
Christian Heimes1a6387e2008-03-26 12:49:49 +00001417 raise ValueError("invalid encoding: %r" % encoding)
1418
1419 if errors is None:
1420 errors = "strict"
1421 else:
Christian Heimes3784c6b2008-03-26 23:13:59 +00001422 if not isinstance(errors, basestring):
Christian Heimes1a6387e2008-03-26 12:49:49 +00001423 raise ValueError("invalid errors: %r" % errors)
1424
1425 self.buffer = buffer
1426 self._line_buffering = line_buffering
1427 self._encoding = encoding
1428 self._errors = errors
1429 self._readuniversal = not newline
1430 self._readtranslate = newline is None
1431 self._readnl = newline
1432 self._writetranslate = newline != ''
1433 self._writenl = newline or os.linesep
1434 self._encoder = None
1435 self._decoder = None
1436 self._decoded_chars = '' # buffer for text returned from decoder
1437 self._decoded_chars_used = 0 # offset into _decoded_chars for read()
1438 self._snapshot = None # info for reconstructing decoder state
1439 self._seekable = self._telling = self.buffer.seekable()
1440
1441 # self._snapshot is either None, or a tuple (dec_flags, next_input)
1442 # where dec_flags is the second (integer) item of the decoder state
1443 # and next_input is the chunk of input bytes that comes next after the
1444 # snapshot point. We use this to reconstruct decoder states in tell().
1445
1446 # Naming convention:
1447 # - "bytes_..." for integer variables that count input bytes
1448 # - "chars_..." for integer variables that count decoded characters
1449
Christian Heimes1a6387e2008-03-26 12:49:49 +00001450 @property
1451 def encoding(self):
1452 return self._encoding
1453
1454 @property
1455 def errors(self):
1456 return self._errors
1457
1458 @property
1459 def line_buffering(self):
1460 return self._line_buffering
1461
1462 def seekable(self):
1463 return self._seekable
1464
Alexandre Vassalotti1aed6242008-05-09 21:49:43 +00001465 def readable(self):
1466 return self.buffer.readable()
1467
1468 def writable(self):
1469 return self.buffer.writable()
1470
Christian Heimes1a6387e2008-03-26 12:49:49 +00001471 def flush(self):
1472 self.buffer.flush()
1473 self._telling = self._seekable
1474
1475 def close(self):
1476 try:
1477 self.flush()
1478 except:
1479 pass # If flush() fails, just give up
1480 self.buffer.close()
1481
1482 @property
1483 def closed(self):
1484 return self.buffer.closed
1485
Georg Brandld2094602008-12-05 08:51:30 +00001486 @property
1487 def name(self):
1488 return self.buffer.name
1489
Christian Heimes1a6387e2008-03-26 12:49:49 +00001490 def fileno(self):
1491 return self.buffer.fileno()
1492
1493 def isatty(self):
1494 return self.buffer.isatty()
1495
1496 def write(self, s):
1497 if self.closed:
1498 raise ValueError("write to closed file")
1499 if not isinstance(s, unicode):
1500 raise TypeError("can't write %s to text stream" %
1501 s.__class__.__name__)
1502 length = len(s)
1503 haslf = (self._writetranslate or self._line_buffering) and "\n" in s
1504 if haslf and self._writetranslate and self._writenl != "\n":
1505 s = s.replace("\n", self._writenl)
1506 encoder = self._encoder or self._get_encoder()
1507 # XXX What if we were just reading?
1508 b = encoder.encode(s)
1509 self.buffer.write(b)
1510 if self._line_buffering and (haslf or "\r" in s):
1511 self.flush()
1512 self._snapshot = None
1513 if self._decoder:
1514 self._decoder.reset()
1515 return length
1516
1517 def _get_encoder(self):
1518 make_encoder = codecs.getincrementalencoder(self._encoding)
1519 self._encoder = make_encoder(self._errors)
1520 return self._encoder
1521
1522 def _get_decoder(self):
1523 make_decoder = codecs.getincrementaldecoder(self._encoding)
1524 decoder = make_decoder(self._errors)
1525 if self._readuniversal:
1526 decoder = IncrementalNewlineDecoder(decoder, self._readtranslate)
1527 self._decoder = decoder
1528 return decoder
1529
1530 # The following three methods implement an ADT for _decoded_chars.
1531 # Text returned from the decoder is buffered here until the client
1532 # requests it by calling our read() or readline() method.
1533 def _set_decoded_chars(self, chars):
1534 """Set the _decoded_chars buffer."""
1535 self._decoded_chars = chars
1536 self._decoded_chars_used = 0
1537
1538 def _get_decoded_chars(self, n=None):
1539 """Advance into the _decoded_chars buffer."""
1540 offset = self._decoded_chars_used
1541 if n is None:
1542 chars = self._decoded_chars[offset:]
1543 else:
1544 chars = self._decoded_chars[offset:offset + n]
1545 self._decoded_chars_used += len(chars)
1546 return chars
1547
1548 def _rewind_decoded_chars(self, n):
1549 """Rewind the _decoded_chars buffer."""
1550 if self._decoded_chars_used < n:
1551 raise AssertionError("rewind decoded_chars out of bounds")
1552 self._decoded_chars_used -= n
1553
1554 def _read_chunk(self):
1555 """
1556 Read and decode the next chunk of data from the BufferedReader.
1557
1558 The return value is True unless EOF was reached. The decoded string
1559 is placed in self._decoded_chars (replacing its previous value).
1560 The entire input chunk is sent to the decoder, though some of it
1561 may remain buffered in the decoder, yet to be converted.
1562 """
1563
1564 if self._decoder is None:
1565 raise ValueError("no decoder")
1566
1567 if self._telling:
1568 # To prepare for tell(), we need to snapshot a point in the
1569 # file where the decoder's input buffer is empty.
1570
1571 dec_buffer, dec_flags = self._decoder.getstate()
1572 # Given this, we know there was a valid snapshot point
1573 # len(dec_buffer) bytes ago with decoder state (b'', dec_flags).
1574
1575 # Read a chunk, decode it, and put the result in self._decoded_chars.
1576 input_chunk = self.buffer.read1(self._CHUNK_SIZE)
1577 eof = not input_chunk
1578 self._set_decoded_chars(self._decoder.decode(input_chunk, eof))
1579
1580 if self._telling:
1581 # At the snapshot point, len(dec_buffer) bytes before the read,
1582 # the next input to be decoded is dec_buffer + input_chunk.
1583 self._snapshot = (dec_flags, dec_buffer + input_chunk)
1584
1585 return not eof
1586
1587 def _pack_cookie(self, position, dec_flags=0,
1588 bytes_to_feed=0, need_eof=0, chars_to_skip=0):
1589 # The meaning of a tell() cookie is: seek to position, set the
1590 # decoder flags to dec_flags, read bytes_to_feed bytes, feed them
1591 # into the decoder with need_eof as the EOF flag, then skip
1592 # chars_to_skip characters of the decoded result. For most simple
1593 # decoders, tell() will often just give a byte offset in the file.
1594 return (position | (dec_flags<<64) | (bytes_to_feed<<128) |
1595 (chars_to_skip<<192) | bool(need_eof)<<256)
1596
1597 def _unpack_cookie(self, bigint):
1598 rest, position = divmod(bigint, 1<<64)
1599 rest, dec_flags = divmod(rest, 1<<64)
1600 rest, bytes_to_feed = divmod(rest, 1<<64)
1601 need_eof, chars_to_skip = divmod(rest, 1<<64)
1602 return position, dec_flags, bytes_to_feed, need_eof, chars_to_skip
1603
1604 def tell(self):
1605 if not self._seekable:
1606 raise IOError("underlying stream is not seekable")
1607 if not self._telling:
1608 raise IOError("telling position disabled by next() call")
1609 self.flush()
1610 position = self.buffer.tell()
1611 decoder = self._decoder
1612 if decoder is None or self._snapshot is None:
1613 if self._decoded_chars:
1614 # This should never happen.
1615 raise AssertionError("pending decoded text")
1616 return position
1617
1618 # Skip backward to the snapshot point (see _read_chunk).
1619 dec_flags, next_input = self._snapshot
1620 position -= len(next_input)
1621
1622 # How many decoded characters have been used up since the snapshot?
1623 chars_to_skip = self._decoded_chars_used
1624 if chars_to_skip == 0:
1625 # We haven't moved from the snapshot point.
1626 return self._pack_cookie(position, dec_flags)
1627
1628 # Starting from the snapshot position, we will walk the decoder
1629 # forward until it gives us enough decoded characters.
1630 saved_state = decoder.getstate()
1631 try:
1632 # Note our initial start point.
1633 decoder.setstate((b'', dec_flags))
1634 start_pos = position
1635 start_flags, bytes_fed, chars_decoded = dec_flags, 0, 0
1636 need_eof = 0
1637
1638 # Feed the decoder one byte at a time. As we go, note the
1639 # nearest "safe start point" before the current location
1640 # (a point where the decoder has nothing buffered, so seek()
1641 # can safely start from there and advance to this location).
Amaury Forgeot d'Arc7684f852008-05-03 12:21:13 +00001642 for next_byte in next_input:
Christian Heimes1a6387e2008-03-26 12:49:49 +00001643 bytes_fed += 1
1644 chars_decoded += len(decoder.decode(next_byte))
1645 dec_buffer, dec_flags = decoder.getstate()
1646 if not dec_buffer and chars_decoded <= chars_to_skip:
1647 # Decoder buffer is empty, so this is a safe start point.
1648 start_pos += bytes_fed
1649 chars_to_skip -= chars_decoded
1650 start_flags, bytes_fed, chars_decoded = dec_flags, 0, 0
1651 if chars_decoded >= chars_to_skip:
1652 break
1653 else:
1654 # We didn't get enough decoded data; signal EOF to get more.
1655 chars_decoded += len(decoder.decode(b'', final=True))
1656 need_eof = 1
1657 if chars_decoded < chars_to_skip:
1658 raise IOError("can't reconstruct logical file position")
1659
1660 # The returned cookie corresponds to the last safe start point.
1661 return self._pack_cookie(
1662 start_pos, start_flags, bytes_fed, need_eof, chars_to_skip)
1663 finally:
1664 decoder.setstate(saved_state)
1665
Alexandre Vassalotti1aed6242008-05-09 21:49:43 +00001666 def truncate(self, pos=None):
1667 self.flush()
1668 if pos is None:
1669 pos = self.tell()
1670 self.seek(pos)
1671 return self.buffer.truncate()
1672
Christian Heimes1a6387e2008-03-26 12:49:49 +00001673 def seek(self, cookie, whence=0):
Alexandre Vassalotti1aed6242008-05-09 21:49:43 +00001674 if self.closed:
1675 raise ValueError("tell on closed file")
Christian Heimes1a6387e2008-03-26 12:49:49 +00001676 if not self._seekable:
1677 raise IOError("underlying stream is not seekable")
1678 if whence == 1: # seek relative to current position
1679 if cookie != 0:
1680 raise IOError("can't do nonzero cur-relative seeks")
1681 # Seeking to the current position should attempt to
1682 # sync the underlying buffer with the current position.
1683 whence = 0
1684 cookie = self.tell()
1685 if whence == 2: # seek relative to end of file
1686 if cookie != 0:
1687 raise IOError("can't do nonzero end-relative seeks")
1688 self.flush()
1689 position = self.buffer.seek(0, 2)
1690 self._set_decoded_chars('')
1691 self._snapshot = None
1692 if self._decoder:
1693 self._decoder.reset()
1694 return position
1695 if whence != 0:
1696 raise ValueError("invalid whence (%r, should be 0, 1 or 2)" %
1697 (whence,))
1698 if cookie < 0:
1699 raise ValueError("negative seek position %r" % (cookie,))
1700 self.flush()
1701
1702 # The strategy of seek() is to go back to the safe start point
1703 # and replay the effect of read(chars_to_skip) from there.
1704 start_pos, dec_flags, bytes_to_feed, need_eof, chars_to_skip = \
1705 self._unpack_cookie(cookie)
1706
1707 # Seek back to the safe start point.
1708 self.buffer.seek(start_pos)
1709 self._set_decoded_chars('')
1710 self._snapshot = None
1711
1712 # Restore the decoder to its state from the safe start point.
1713 if self._decoder or dec_flags or chars_to_skip:
1714 self._decoder = self._decoder or self._get_decoder()
1715 self._decoder.setstate((b'', dec_flags))
1716 self._snapshot = (dec_flags, b'')
1717
1718 if chars_to_skip:
1719 # Just like _read_chunk, feed the decoder and save a snapshot.
1720 input_chunk = self.buffer.read(bytes_to_feed)
1721 self._set_decoded_chars(
1722 self._decoder.decode(input_chunk, need_eof))
1723 self._snapshot = (dec_flags, input_chunk)
1724
1725 # Skip chars_to_skip of the decoded characters.
1726 if len(self._decoded_chars) < chars_to_skip:
1727 raise IOError("can't restore logical file position")
1728 self._decoded_chars_used = chars_to_skip
1729
1730 return cookie
1731
1732 def read(self, n=None):
1733 if n is None:
1734 n = -1
1735 decoder = self._decoder or self._get_decoder()
1736 if n < 0:
1737 # Read everything.
1738 result = (self._get_decoded_chars() +
1739 decoder.decode(self.buffer.read(), final=True))
1740 self._set_decoded_chars('')
1741 self._snapshot = None
1742 return result
1743 else:
1744 # Keep reading chunks until we have n characters to return.
1745 eof = False
1746 result = self._get_decoded_chars(n)
1747 while len(result) < n and not eof:
1748 eof = not self._read_chunk()
1749 result += self._get_decoded_chars(n - len(result))
1750 return result
1751
1752 def next(self):
1753 self._telling = False
1754 line = self.readline()
1755 if not line:
1756 self._snapshot = None
1757 self._telling = self._seekable
1758 raise StopIteration
1759 return line
1760
1761 def readline(self, limit=None):
Alexandre Vassalotti1aed6242008-05-09 21:49:43 +00001762 if self.closed:
1763 raise ValueError("read from closed file")
Christian Heimes1a6387e2008-03-26 12:49:49 +00001764 if limit is None:
1765 limit = -1
Alexandre Vassalotti1aed6242008-05-09 21:49:43 +00001766 if not isinstance(limit, (int, long)):
1767 raise TypeError("limit must be an integer")
Christian Heimes1a6387e2008-03-26 12:49:49 +00001768
1769 # Grab all the decoded text (we will rewind any extra bits later).
1770 line = self._get_decoded_chars()
1771
1772 start = 0
1773 decoder = self._decoder or self._get_decoder()
1774
1775 pos = endpos = None
1776 while True:
1777 if self._readtranslate:
1778 # Newlines are already translated, only search for \n
1779 pos = line.find('\n', start)
1780 if pos >= 0:
1781 endpos = pos + 1
1782 break
1783 else:
1784 start = len(line)
1785
1786 elif self._readuniversal:
1787 # Universal newline search. Find any of \r, \r\n, \n
1788 # The decoder ensures that \r\n are not split in two pieces
1789
1790 # In C we'd look for these in parallel of course.
1791 nlpos = line.find("\n", start)
1792 crpos = line.find("\r", start)
1793 if crpos == -1:
1794 if nlpos == -1:
1795 # Nothing found
1796 start = len(line)
1797 else:
1798 # Found \n
1799 endpos = nlpos + 1
1800 break
1801 elif nlpos == -1:
1802 # Found lone \r
1803 endpos = crpos + 1
1804 break
1805 elif nlpos < crpos:
1806 # Found \n
1807 endpos = nlpos + 1
1808 break
1809 elif nlpos == crpos + 1:
1810 # Found \r\n
1811 endpos = crpos + 2
1812 break
1813 else:
1814 # Found \r
1815 endpos = crpos + 1
1816 break
1817 else:
1818 # non-universal
1819 pos = line.find(self._readnl)
1820 if pos >= 0:
1821 endpos = pos + len(self._readnl)
1822 break
1823
1824 if limit >= 0 and len(line) >= limit:
1825 endpos = limit # reached length limit
1826 break
1827
1828 # No line ending seen yet - get more data
1829 more_line = ''
1830 while self._read_chunk():
1831 if self._decoded_chars:
1832 break
1833 if self._decoded_chars:
1834 line += self._get_decoded_chars()
1835 else:
1836 # end of file
1837 self._set_decoded_chars('')
1838 self._snapshot = None
1839 return line
1840
1841 if limit >= 0 and endpos > limit:
1842 endpos = limit # don't exceed limit
1843
1844 # Rewind _decoded_chars to just after the line ending we found.
1845 self._rewind_decoded_chars(len(line) - endpos)
1846 return line[:endpos]
1847
1848 @property
1849 def newlines(self):
1850 return self._decoder.newlines if self._decoder else None
1851
1852class StringIO(TextIOWrapper):
1853
Benjamin Peterson9ae080e2008-05-04 22:39:33 +00001854 """An in-memory stream for text. The initial_value argument sets the
Benjamin Peterson7bb4d2d2008-04-13 02:01:27 +00001855 value of object. The other arguments are like those of TextIOWrapper's
1856 constructor.
1857 """
Christian Heimes1a6387e2008-03-26 12:49:49 +00001858
1859 def __init__(self, initial_value="", encoding="utf-8",
1860 errors="strict", newline="\n"):
1861 super(StringIO, self).__init__(BytesIO(),
1862 encoding=encoding,
1863 errors=errors,
1864 newline=newline)
Alexandre Vassalottidd0b90a2009-06-12 21:20:23 +00001865 # Issue #5645: make universal newlines semantics the same as in the
1866 # C version, even under Windows.
1867 if newline is None:
1868 self._writetranslate = False
Christian Heimes1a6387e2008-03-26 12:49:49 +00001869 if initial_value:
1870 if not isinstance(initial_value, unicode):
1871 initial_value = unicode(initial_value)
1872 self.write(initial_value)
1873 self.seek(0)
1874
1875 def getvalue(self):
1876 self.flush()
1877 return self.buffer.getvalue().decode(self._encoding, self._errors)