Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 1 | """ |
Benjamin Peterson | 7bb4d2d | 2008-04-13 02:01:27 +0000 | [diff] [blame] | 2 | The io module provides the Python interfaces to stream handling. The |
| 3 | builtin open function is defined in this module. |
| 4 | |
| 5 | At the top of the I/O hierarchy is the abstract base class IOBase. It |
| 6 | defines the basic interface to a stream. Note, however, that there is no |
| 7 | seperation between reading and writing to streams; implementations are |
| 8 | allowed to throw an IOError if they do not support a given operation. |
| 9 | |
| 10 | Extending IOBase is RawIOBase which deals simply with the reading and |
| 11 | writing of raw bytes to a stream. FileIO subclasses RawIOBase to provide |
| 12 | an interface to OS files. |
| 13 | |
| 14 | BufferedIOBase deals with buffering on a raw byte stream (RawIOBase). Its |
| 15 | subclasses, BufferedWriter, BufferedReader, and BufferedRWPair buffer |
| 16 | streams that are readable, writable, and both respectively. |
| 17 | BufferedRandom provides a buffered interface to random access |
| 18 | streams. BytesIO is a simple stream of in-memory bytes. |
| 19 | |
| 20 | Another IOBase subclass, TextIOBase, deals with the encoding and decoding |
| 21 | of streams into text. TextIOWrapper, which extends it, is a buffered text |
| 22 | interface to a buffered raw stream (`BufferedIOBase`). Finally, StringIO |
| 23 | is a in-memory stream for text. |
| 24 | |
| 25 | Argument names are not part of the specification, and only the arguments |
| 26 | of open() are intended to be used as keyword arguments. |
| 27 | |
| 28 | data: |
| 29 | |
| 30 | DEFAULT_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 Heimes | 3784c6b | 2008-03-26 23:13:59 +0000 | [diff] [blame] | 48 | from __future__ import print_function |
| 49 | from __future__ import unicode_literals |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 50 | |
| 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 | |
| 60 | import os |
| 61 | import abc |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 62 | import codecs |
| 63 | import _fileio |
Antoine Pitrou | 11ec65d | 2008-08-14 21:04:30 +0000 | [diff] [blame] | 64 | import threading |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 65 | |
| 66 | # open() uses st_blksize whenever we can |
| 67 | DEFAULT_BUFFER_SIZE = 8 * 1024 # bytes |
| 68 | |
| 69 | # py3k has only new style classes |
| 70 | __metaclass__ = type |
| 71 | |
| 72 | class 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 | |
| 81 | def open(file, mode="r", buffering=None, encoding=None, errors=None, |
| 82 | newline=None, closefd=True): |
Benjamin Peterson | 9ae080e | 2008-05-04 22:39:33 +0000 | [diff] [blame] | 83 | r"""Open file and return a stream. If the file cannot be opened, an IOError is |
| 84 | raised. |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 85 | |
Benjamin Peterson | 7bb4d2d | 2008-04-13 02:01:27 +0000 | [diff] [blame] | 86 | 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 Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 91 | |
Benjamin Peterson | 7bb4d2d | 2008-04-13 02:01:27 +0000 | [diff] [blame] | 92 | 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 Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 101 | |
Benjamin Peterson | 7bb4d2d | 2008-04-13 02:01:27 +0000 | [diff] [blame] | 102 | ========= =============================================================== |
| 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 Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 114 | |
Benjamin Peterson | 7bb4d2d | 2008-04-13 02:01:27 +0000 | [diff] [blame] | 115 | 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 Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 118 | |
Benjamin Peterson | 7bb4d2d | 2008-04-13 02:01:27 +0000 | [diff] [blame] | 119 | 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 Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 126 | |
Benjamin Peterson | 7bb4d2d | 2008-04-13 02:01:27 +0000 | [diff] [blame] | 127 | buffering is an optional integer used to set the buffering policy. By |
| 128 | default full buffering is on. Pass 0 to switch buffering off (only |
| 129 | allowed in binary mode), 1 to set line buffering, and an integer > 1 |
| 130 | for full buffering. |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 131 | |
Benjamin Peterson | 7bb4d2d | 2008-04-13 02:01:27 +0000 | [diff] [blame] | 132 | encoding is the name of the encoding used to decode or encode the |
| 133 | file. This should only be used in text mode. The default encoding is |
| 134 | platform dependent, but any encoding supported by Python can be |
| 135 | passed. See the codecs module for the list of supported encodings. |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 136 | |
Benjamin Peterson | 7bb4d2d | 2008-04-13 02:01:27 +0000 | [diff] [blame] | 137 | errors is an optional string that specifies how encoding errors are to |
| 138 | be handled---this argument should not be used in binary mode. Pass |
| 139 | 'strict' to raise a ValueError exception if there is an encoding error |
| 140 | (the default of None has the same effect), or pass 'ignore' to ignore |
| 141 | errors. (Note that ignoring encoding errors can lead to data loss.) |
| 142 | See the documentation for codecs.register for a list of the permitted |
| 143 | encoding error strings. |
| 144 | |
| 145 | newline controls how universal newlines works (it only applies to text |
| 146 | mode). It can be None, '', '\n', '\r', and '\r\n'. It works as |
| 147 | follows: |
| 148 | |
| 149 | * On input, if newline is None, universal newlines mode is |
| 150 | enabled. Lines in the input can end in '\n', '\r', or '\r\n', and |
| 151 | these are translated into '\n' before being returned to the |
| 152 | caller. If it is '', universal newline mode is enabled, but line |
| 153 | endings are returned to the caller untranslated. If it has any of |
| 154 | the other legal values, input lines are only terminated by the given |
| 155 | string, and the line ending is returned to the caller untranslated. |
| 156 | |
| 157 | * On output, if newline is None, any '\n' characters written are |
| 158 | translated to the system default line separator, os.linesep. If |
| 159 | newline is '', no translation takes place. If newline is any of the |
| 160 | other legal values, any '\n' characters written are translated to |
| 161 | the given string. |
| 162 | |
| 163 | If closefd is False, the underlying file descriptor will be kept open |
| 164 | when the file is closed. This does not work when a file name is given |
| 165 | and must be True in that case. |
| 166 | |
| 167 | open() returns a file object whose type depends on the mode, and |
| 168 | through which the standard file operations such as reading and writing |
| 169 | are performed. When open() is used to open a file in a text mode ('w', |
| 170 | 'r', 'wt', 'rt', etc.), it returns a TextIOWrapper. When used to open |
| 171 | a file in a binary mode, the returned class varies: in read binary |
| 172 | mode, it returns a BufferedReader; in write binary and append binary |
| 173 | modes, it returns a BufferedWriter, and in read/write mode, it returns |
| 174 | a BufferedRandom. |
| 175 | |
| 176 | It is also possible to use a string or bytearray as a file for both |
| 177 | reading and writing. For strings StringIO can be used like a file |
| 178 | opened in a text mode, and for bytes a BytesIO can be used like a file |
| 179 | opened in a binary mode. |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 180 | """ |
Christian Heimes | 3784c6b | 2008-03-26 23:13:59 +0000 | [diff] [blame] | 181 | if not isinstance(file, (basestring, int)): |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 182 | raise TypeError("invalid file: %r" % file) |
Christian Heimes | 3784c6b | 2008-03-26 23:13:59 +0000 | [diff] [blame] | 183 | if not isinstance(mode, basestring): |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 184 | raise TypeError("invalid mode: %r" % mode) |
| 185 | if buffering is not None and not isinstance(buffering, int): |
| 186 | raise TypeError("invalid buffering: %r" % buffering) |
Christian Heimes | 3784c6b | 2008-03-26 23:13:59 +0000 | [diff] [blame] | 187 | if encoding is not None and not isinstance(encoding, basestring): |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 188 | raise TypeError("invalid encoding: %r" % encoding) |
Christian Heimes | 3784c6b | 2008-03-26 23:13:59 +0000 | [diff] [blame] | 189 | if errors is not None and not isinstance(errors, basestring): |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 190 | raise TypeError("invalid errors: %r" % errors) |
| 191 | modes = set(mode) |
| 192 | if modes - set("arwb+tU") or len(mode) > len(modes): |
| 193 | raise ValueError("invalid mode: %r" % mode) |
| 194 | reading = "r" in modes |
| 195 | writing = "w" in modes |
| 196 | appending = "a" in modes |
| 197 | updating = "+" in modes |
| 198 | text = "t" in modes |
| 199 | binary = "b" in modes |
| 200 | if "U" in modes: |
| 201 | if writing or appending: |
| 202 | raise ValueError("can't use U and writing mode at once") |
| 203 | reading = True |
| 204 | if text and binary: |
| 205 | raise ValueError("can't have text and binary mode at once") |
| 206 | if reading + writing + appending > 1: |
| 207 | raise ValueError("can't have read/write/append mode at once") |
| 208 | if not (reading or writing or appending): |
| 209 | raise ValueError("must have exactly one of read/write/append mode") |
| 210 | if binary and encoding is not None: |
| 211 | raise ValueError("binary mode doesn't take an encoding argument") |
| 212 | if binary and errors is not None: |
| 213 | raise ValueError("binary mode doesn't take an errors argument") |
| 214 | if binary and newline is not None: |
| 215 | raise ValueError("binary mode doesn't take a newline argument") |
| 216 | raw = FileIO(file, |
| 217 | (reading and "r" or "") + |
| 218 | (writing and "w" or "") + |
| 219 | (appending and "a" or "") + |
| 220 | (updating and "+" or ""), |
| 221 | closefd) |
| 222 | if buffering is None: |
| 223 | buffering = -1 |
| 224 | line_buffering = False |
| 225 | if buffering == 1 or buffering < 0 and raw.isatty(): |
| 226 | buffering = -1 |
| 227 | line_buffering = True |
| 228 | if buffering < 0: |
| 229 | buffering = DEFAULT_BUFFER_SIZE |
| 230 | try: |
| 231 | bs = os.fstat(raw.fileno()).st_blksize |
| 232 | except (os.error, AttributeError): |
| 233 | pass |
| 234 | else: |
| 235 | if bs > 1: |
| 236 | buffering = bs |
| 237 | if buffering < 0: |
| 238 | raise ValueError("invalid buffering size") |
| 239 | if buffering == 0: |
| 240 | if binary: |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 241 | return raw |
| 242 | raise ValueError("can't have unbuffered text I/O") |
| 243 | if updating: |
| 244 | buffer = BufferedRandom(raw, buffering) |
| 245 | elif writing or appending: |
| 246 | buffer = BufferedWriter(raw, buffering) |
| 247 | elif reading: |
| 248 | buffer = BufferedReader(raw, buffering) |
| 249 | else: |
| 250 | raise ValueError("unknown mode: %r" % mode) |
| 251 | if binary: |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 252 | return buffer |
| 253 | text = TextIOWrapper(buffer, encoding, errors, newline, line_buffering) |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 254 | text.mode = mode |
| 255 | return text |
| 256 | |
| 257 | class _DocDescriptor: |
| 258 | """Helper for builtins.open.__doc__ |
| 259 | """ |
| 260 | def __get__(self, obj, typ): |
| 261 | return ( |
| 262 | "open(file, mode='r', buffering=None, encoding=None, " |
| 263 | "errors=None, newline=None, closefd=True)\n\n" + |
| 264 | open.__doc__) |
| 265 | |
| 266 | class OpenWrapper: |
| 267 | """Wrapper for builtins.open |
| 268 | |
| 269 | Trick so that open won't become a bound method when stored |
| 270 | as a class variable (as dumbdbm does). |
| 271 | |
| 272 | See initstdio() in Python/pythonrun.c. |
| 273 | """ |
| 274 | __doc__ = _DocDescriptor() |
| 275 | |
| 276 | def __new__(cls, *args, **kwargs): |
| 277 | return open(*args, **kwargs) |
| 278 | |
| 279 | |
| 280 | class UnsupportedOperation(ValueError, IOError): |
| 281 | pass |
| 282 | |
| 283 | |
| 284 | class IOBase(object): |
| 285 | |
Benjamin Peterson | 9ae080e | 2008-05-04 22:39:33 +0000 | [diff] [blame] | 286 | """The abstract base class for all I/O classes, acting on streams of |
Benjamin Peterson | 7bb4d2d | 2008-04-13 02:01:27 +0000 | [diff] [blame] | 287 | bytes. There is no public constructor. |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 288 | |
| 289 | This class provides dummy implementations for many methods that |
Benjamin Peterson | 7bb4d2d | 2008-04-13 02:01:27 +0000 | [diff] [blame] | 290 | derived classes can override selectively; the default implementations |
| 291 | represent a file that cannot be read, written or seeked. |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 292 | |
Benjamin Peterson | 7bb4d2d | 2008-04-13 02:01:27 +0000 | [diff] [blame] | 293 | Even though IOBase does not declare read, readinto, or write because |
| 294 | their signatures will vary, implementations and clients should |
| 295 | consider those methods part of the interface. Also, implementations |
| 296 | may raise a IOError when operations they do not support are called. |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 297 | |
Benjamin Peterson | 7bb4d2d | 2008-04-13 02:01:27 +0000 | [diff] [blame] | 298 | The basic type used for binary data read from or written to a file is |
| 299 | bytes. bytearrays are accepted too, and in some cases (such as |
| 300 | readinto) needed. Text I/O classes work with str data. |
| 301 | |
| 302 | Note that calling any method (even inquiries) on a closed stream is |
| 303 | undefined. Implementations may raise IOError in this case. |
| 304 | |
| 305 | IOBase (and its subclasses) support the iterator protocol, meaning |
| 306 | that an IOBase object can be iterated over yielding the lines in a |
| 307 | stream. |
| 308 | |
| 309 | IOBase also supports the :keyword:`with` statement. In this example, |
| 310 | fp is closed after the suite of the with statment is complete: |
| 311 | |
| 312 | with open('spam.txt', 'r') as fp: |
| 313 | fp.write('Spam and eggs!') |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 314 | """ |
| 315 | |
| 316 | __metaclass__ = abc.ABCMeta |
| 317 | |
| 318 | ### Internal ### |
| 319 | |
| 320 | def _unsupported(self, name): |
| 321 | """Internal: raise an exception for unsupported operations.""" |
| 322 | raise UnsupportedOperation("%s.%s() not supported" % |
| 323 | (self.__class__.__name__, name)) |
| 324 | |
| 325 | ### Positioning ### |
| 326 | |
| 327 | def seek(self, pos, whence = 0): |
Benjamin Peterson | 9ae080e | 2008-05-04 22:39:33 +0000 | [diff] [blame] | 328 | """Change stream position. |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 329 | |
Benjamin Peterson | 7bb4d2d | 2008-04-13 02:01:27 +0000 | [diff] [blame] | 330 | Change the stream position to byte offset offset. offset is |
| 331 | interpreted relative to the position indicated by whence. Values |
| 332 | for whence are: |
| 333 | |
| 334 | * 0 -- start of stream (the default); offset should be zero or positive |
| 335 | * 1 -- current stream position; offset may be negative |
| 336 | * 2 -- end of stream; offset is usually negative |
| 337 | |
| 338 | Return the new absolute position. |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 339 | """ |
| 340 | self._unsupported("seek") |
| 341 | |
| 342 | def tell(self): |
Benjamin Peterson | 9ae080e | 2008-05-04 22:39:33 +0000 | [diff] [blame] | 343 | """Return current stream position.""" |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 344 | return self.seek(0, 1) |
| 345 | |
| 346 | def truncate(self, pos = None): |
Benjamin Peterson | 9ae080e | 2008-05-04 22:39:33 +0000 | [diff] [blame] | 347 | """Truncate file to size bytes. |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 348 | |
Benjamin Peterson | 9ae080e | 2008-05-04 22:39:33 +0000 | [diff] [blame] | 349 | Size defaults to the current IO position as reported by tell(). Return |
| 350 | the new size. |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 351 | """ |
| 352 | self._unsupported("truncate") |
| 353 | |
| 354 | ### Flush and close ### |
| 355 | |
| 356 | def flush(self): |
Benjamin Peterson | 9ae080e | 2008-05-04 22:39:33 +0000 | [diff] [blame] | 357 | """Flush write buffers, if applicable. |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 358 | |
Benjamin Peterson | 7bb4d2d | 2008-04-13 02:01:27 +0000 | [diff] [blame] | 359 | This is not implemented for read-only and non-blocking streams. |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 360 | """ |
| 361 | # XXX Should this return the number of bytes written??? |
| 362 | |
| 363 | __closed = False |
| 364 | |
| 365 | def close(self): |
Benjamin Peterson | 9ae080e | 2008-05-04 22:39:33 +0000 | [diff] [blame] | 366 | """Flush and close the IO object. |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 367 | |
Benjamin Peterson | 7bb4d2d | 2008-04-13 02:01:27 +0000 | [diff] [blame] | 368 | This method has no effect if the file is already closed. |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 369 | """ |
| 370 | if not self.__closed: |
| 371 | try: |
| 372 | self.flush() |
| 373 | except IOError: |
| 374 | pass # If flush() fails, just give up |
| 375 | self.__closed = True |
| 376 | |
| 377 | def __del__(self): |
| 378 | """Destructor. Calls close().""" |
| 379 | # The try/except block is in case this is called at program |
| 380 | # exit time, when it's possible that globals have already been |
| 381 | # deleted, and then the close() call might fail. Since |
| 382 | # there's nothing we can do about such failures and they annoy |
| 383 | # the end users, we suppress the traceback. |
| 384 | try: |
| 385 | self.close() |
| 386 | except: |
| 387 | pass |
| 388 | |
| 389 | ### Inquiries ### |
| 390 | |
| 391 | def seekable(self): |
Benjamin Peterson | 9ae080e | 2008-05-04 22:39:33 +0000 | [diff] [blame] | 392 | """Return whether object supports random access. |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 393 | |
| 394 | If False, seek(), tell() and truncate() will raise IOError. |
| 395 | This method may need to do a test seek(). |
| 396 | """ |
| 397 | return False |
| 398 | |
| 399 | def _checkSeekable(self, msg=None): |
| 400 | """Internal: raise an IOError if file is not seekable |
| 401 | """ |
| 402 | if not self.seekable(): |
| 403 | raise IOError("File or stream is not seekable." |
| 404 | if msg is None else msg) |
| 405 | |
| 406 | |
| 407 | def readable(self): |
Benjamin Peterson | 9ae080e | 2008-05-04 22:39:33 +0000 | [diff] [blame] | 408 | """Return whether object was opened for reading. |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 409 | |
| 410 | If False, read() will raise IOError. |
| 411 | """ |
| 412 | return False |
| 413 | |
| 414 | def _checkReadable(self, msg=None): |
| 415 | """Internal: raise an IOError if file is not readable |
| 416 | """ |
| 417 | if not self.readable(): |
| 418 | raise IOError("File or stream is not readable." |
| 419 | if msg is None else msg) |
| 420 | |
| 421 | def writable(self): |
Benjamin Peterson | 9ae080e | 2008-05-04 22:39:33 +0000 | [diff] [blame] | 422 | """Return whether object was opened for writing. |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 423 | |
| 424 | If False, write() and truncate() will raise IOError. |
| 425 | """ |
| 426 | return False |
| 427 | |
| 428 | def _checkWritable(self, msg=None): |
| 429 | """Internal: raise an IOError if file is not writable |
| 430 | """ |
| 431 | if not self.writable(): |
| 432 | raise IOError("File or stream is not writable." |
| 433 | if msg is None else msg) |
| 434 | |
| 435 | @property |
| 436 | def closed(self): |
| 437 | """closed: bool. True iff the file has been closed. |
| 438 | |
| 439 | For backwards compatibility, this is a property, not a predicate. |
| 440 | """ |
| 441 | return self.__closed |
| 442 | |
| 443 | def _checkClosed(self, msg=None): |
| 444 | """Internal: raise an ValueError if file is closed |
| 445 | """ |
| 446 | if self.closed: |
| 447 | raise ValueError("I/O operation on closed file." |
| 448 | if msg is None else msg) |
| 449 | |
| 450 | ### Context manager ### |
| 451 | |
| 452 | def __enter__(self): |
| 453 | """Context management protocol. Returns self.""" |
| 454 | self._checkClosed() |
| 455 | return self |
| 456 | |
| 457 | def __exit__(self, *args): |
| 458 | """Context management protocol. Calls close()""" |
| 459 | self.close() |
| 460 | |
| 461 | ### Lower-level APIs ### |
| 462 | |
| 463 | # XXX Should these be present even if unimplemented? |
| 464 | |
| 465 | def fileno(self): |
Benjamin Peterson | 9ae080e | 2008-05-04 22:39:33 +0000 | [diff] [blame] | 466 | """Returns underlying file descriptor if one exists. |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 467 | |
Benjamin Peterson | 9ae080e | 2008-05-04 22:39:33 +0000 | [diff] [blame] | 468 | An IOError is raised if the IO object does not use a file descriptor. |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 469 | """ |
| 470 | self._unsupported("fileno") |
| 471 | |
| 472 | def isatty(self): |
Benjamin Peterson | 9ae080e | 2008-05-04 22:39:33 +0000 | [diff] [blame] | 473 | """Return whether this is an 'interactive' stream. |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 474 | |
Benjamin Peterson | 9ae080e | 2008-05-04 22:39:33 +0000 | [diff] [blame] | 475 | Return False if it can't be determined. |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 476 | """ |
| 477 | self._checkClosed() |
| 478 | return False |
| 479 | |
| 480 | ### Readline[s] and writelines ### |
| 481 | |
| 482 | def readline(self, limit = -1): |
Benjamin Peterson | 9ae080e | 2008-05-04 22:39:33 +0000 | [diff] [blame] | 483 | r"""Read and return a line from the stream. |
Benjamin Peterson | 7bb4d2d | 2008-04-13 02:01:27 +0000 | [diff] [blame] | 484 | |
| 485 | If limit is specified, at most limit bytes will be read. |
| 486 | |
| 487 | The line terminator is always b'\n' for binary files; for text |
| 488 | files, the newlines argument to open can be used to select the line |
| 489 | terminator(s) recognized. |
| 490 | """ |
Alexandre Vassalotti | 1aed624 | 2008-05-09 21:49:43 +0000 | [diff] [blame] | 491 | self._checkClosed() |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 492 | if hasattr(self, "peek"): |
| 493 | def nreadahead(): |
| 494 | readahead = self.peek(1) |
| 495 | if not readahead: |
| 496 | return 1 |
| 497 | n = (readahead.find(b"\n") + 1) or len(readahead) |
| 498 | if limit >= 0: |
| 499 | n = min(n, limit) |
| 500 | return n |
| 501 | else: |
| 502 | def nreadahead(): |
| 503 | return 1 |
| 504 | if limit is None: |
| 505 | limit = -1 |
Alexandre Vassalotti | 1aed624 | 2008-05-09 21:49:43 +0000 | [diff] [blame] | 506 | if not isinstance(limit, (int, long)): |
| 507 | raise TypeError("limit must be an integer") |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 508 | res = bytearray() |
| 509 | while limit < 0 or len(res) < limit: |
| 510 | b = self.read(nreadahead()) |
| 511 | if not b: |
| 512 | break |
| 513 | res += b |
| 514 | if res.endswith(b"\n"): |
| 515 | break |
| 516 | return bytes(res) |
| 517 | |
| 518 | def __iter__(self): |
| 519 | self._checkClosed() |
| 520 | return self |
| 521 | |
| 522 | def next(self): |
| 523 | line = self.readline() |
| 524 | if not line: |
| 525 | raise StopIteration |
| 526 | return line |
| 527 | |
| 528 | def readlines(self, hint=None): |
Benjamin Peterson | 9ae080e | 2008-05-04 22:39:33 +0000 | [diff] [blame] | 529 | """Return a list of lines from the stream. |
Benjamin Peterson | 7bb4d2d | 2008-04-13 02:01:27 +0000 | [diff] [blame] | 530 | |
| 531 | hint can be specified to control the number of lines read: no more |
| 532 | lines will be read if the total size (in bytes/characters) of all |
| 533 | lines so far exceeds hint. |
| 534 | """ |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 535 | if hint is None: |
Alexandre Vassalotti | 1aed624 | 2008-05-09 21:49:43 +0000 | [diff] [blame] | 536 | hint = -1 |
| 537 | if not isinstance(hint, (int, long)): |
| 538 | raise TypeError("hint must be an integer") |
| 539 | if hint <= 0: |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 540 | return list(self) |
| 541 | n = 0 |
| 542 | lines = [] |
| 543 | for line in self: |
| 544 | lines.append(line) |
| 545 | n += len(line) |
| 546 | if n >= hint: |
| 547 | break |
| 548 | return lines |
| 549 | |
| 550 | def writelines(self, lines): |
| 551 | self._checkClosed() |
| 552 | for line in lines: |
| 553 | self.write(line) |
| 554 | |
| 555 | |
| 556 | class RawIOBase(IOBase): |
| 557 | |
Benjamin Peterson | 7bb4d2d | 2008-04-13 02:01:27 +0000 | [diff] [blame] | 558 | """Base class for raw binary I/O.""" |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 559 | |
Benjamin Peterson | 7bb4d2d | 2008-04-13 02:01:27 +0000 | [diff] [blame] | 560 | # The read() method is implemented by calling readinto(); derived |
| 561 | # classes that want to support read() only need to implement |
| 562 | # readinto() as a primitive operation. In general, readinto() can be |
| 563 | # more efficient than read(). |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 564 | |
Benjamin Peterson | 7bb4d2d | 2008-04-13 02:01:27 +0000 | [diff] [blame] | 565 | # (It would be tempting to also provide an implementation of |
| 566 | # readinto() in terms of read(), in case the latter is a more suitable |
| 567 | # primitive operation, but that would lead to nasty recursion in case |
| 568 | # a subclass doesn't implement either.) |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 569 | |
| 570 | def read(self, n = -1): |
Benjamin Peterson | 9ae080e | 2008-05-04 22:39:33 +0000 | [diff] [blame] | 571 | """Read and return up to n bytes. |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 572 | |
| 573 | Returns an empty bytes array on EOF, or None if the object is |
| 574 | set not to block and has no data to read. |
| 575 | """ |
| 576 | if n is None: |
| 577 | n = -1 |
| 578 | if n < 0: |
| 579 | return self.readall() |
| 580 | b = bytearray(n.__index__()) |
| 581 | n = self.readinto(b) |
| 582 | del b[n:] |
| 583 | return bytes(b) |
| 584 | |
| 585 | def readall(self): |
Benjamin Peterson | 9ae080e | 2008-05-04 22:39:33 +0000 | [diff] [blame] | 586 | """Read until EOF, using multiple read() call.""" |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 587 | res = bytearray() |
| 588 | while True: |
| 589 | data = self.read(DEFAULT_BUFFER_SIZE) |
| 590 | if not data: |
| 591 | break |
| 592 | res += data |
| 593 | return bytes(res) |
| 594 | |
| 595 | def readinto(self, b): |
Benjamin Peterson | 9ae080e | 2008-05-04 22:39:33 +0000 | [diff] [blame] | 596 | """Read up to len(b) bytes into b. |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 597 | |
| 598 | Returns number of bytes read (0 for EOF), or None if the object |
| 599 | is set not to block as has no data to read. |
| 600 | """ |
| 601 | self._unsupported("readinto") |
| 602 | |
| 603 | def write(self, b): |
Benjamin Peterson | 9ae080e | 2008-05-04 22:39:33 +0000 | [diff] [blame] | 604 | """Write the given buffer to the IO stream. |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 605 | |
| 606 | Returns the number of bytes written, which may be less than len(b). |
| 607 | """ |
| 608 | self._unsupported("write") |
| 609 | |
| 610 | |
| 611 | class FileIO(_fileio._FileIO, RawIOBase): |
| 612 | |
Benjamin Peterson | 7bb4d2d | 2008-04-13 02:01:27 +0000 | [diff] [blame] | 613 | """Raw I/O implementation for OS files.""" |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 614 | |
Benjamin Peterson | 7bb4d2d | 2008-04-13 02:01:27 +0000 | [diff] [blame] | 615 | # This multiply inherits from _FileIO and RawIOBase to make |
| 616 | # isinstance(io.FileIO(), io.RawIOBase) return True without requiring |
| 617 | # that _fileio._FileIO inherits from io.RawIOBase (which would be hard |
| 618 | # to do since _fileio.c is written in C). |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 619 | |
Benjamin Peterson | ad100c3 | 2008-11-20 22:06:22 +0000 | [diff] [blame] | 620 | def __init__(self, name, mode="r", closefd=True): |
| 621 | _fileio._FileIO.__init__(self, name, mode, closefd) |
| 622 | self._name = name |
| 623 | |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 624 | def close(self): |
| 625 | _fileio._FileIO.close(self) |
| 626 | RawIOBase.close(self) |
| 627 | |
| 628 | @property |
| 629 | def name(self): |
| 630 | return self._name |
| 631 | |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 632 | |
| 633 | class BufferedIOBase(IOBase): |
| 634 | |
| 635 | """Base class for buffered IO objects. |
| 636 | |
| 637 | The main difference with RawIOBase is that the read() method |
| 638 | supports omitting the size argument, and does not have a default |
| 639 | implementation that defers to readinto(). |
| 640 | |
| 641 | In addition, read(), readinto() and write() may raise |
| 642 | BlockingIOError if the underlying raw stream is in non-blocking |
| 643 | mode and not ready; unlike their raw counterparts, they will never |
| 644 | return None. |
| 645 | |
| 646 | A typical implementation should not inherit from a RawIOBase |
| 647 | implementation, but wrap one. |
| 648 | """ |
| 649 | |
| 650 | def read(self, n = None): |
Benjamin Peterson | 9ae080e | 2008-05-04 22:39:33 +0000 | [diff] [blame] | 651 | """Read and return up to n bytes. |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 652 | |
| 653 | If the argument is omitted, None, or negative, reads and |
| 654 | returns all data until EOF. |
| 655 | |
| 656 | If the argument is positive, and the underlying raw stream is |
| 657 | not 'interactive', multiple raw reads may be issued to satisfy |
| 658 | the byte count (unless EOF is reached first). But for |
| 659 | interactive raw streams (XXX and for pipes?), at most one raw |
| 660 | read will be issued, and a short result does not imply that |
| 661 | EOF is imminent. |
| 662 | |
| 663 | Returns an empty bytes array on EOF. |
| 664 | |
| 665 | Raises BlockingIOError if the underlying raw stream has no |
| 666 | data at the moment. |
| 667 | """ |
| 668 | self._unsupported("read") |
| 669 | |
| 670 | def readinto(self, b): |
Benjamin Peterson | 9ae080e | 2008-05-04 22:39:33 +0000 | [diff] [blame] | 671 | """Read up to len(b) bytes into b. |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 672 | |
Benjamin Peterson | 7bb4d2d | 2008-04-13 02:01:27 +0000 | [diff] [blame] | 673 | Like read(), this may issue multiple reads to the underlying raw |
| 674 | stream, unless the latter is 'interactive'. |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 675 | |
| 676 | Returns the number of bytes read (0 for EOF). |
| 677 | |
| 678 | Raises BlockingIOError if the underlying raw stream has no |
| 679 | data at the moment. |
| 680 | """ |
| 681 | # XXX This ought to work with anything that supports the buffer API |
| 682 | data = self.read(len(b)) |
| 683 | n = len(data) |
| 684 | try: |
| 685 | b[:n] = data |
| 686 | except TypeError as err: |
| 687 | import array |
| 688 | if not isinstance(b, array.array): |
| 689 | raise err |
Alexandre Vassalotti | 1aed624 | 2008-05-09 21:49:43 +0000 | [diff] [blame] | 690 | b[:n] = array.array(b'b', data) |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 691 | return n |
| 692 | |
| 693 | def write(self, b): |
Benjamin Peterson | 9ae080e | 2008-05-04 22:39:33 +0000 | [diff] [blame] | 694 | """Write the given buffer to the IO stream. |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 695 | |
Benjamin Peterson | 9ae080e | 2008-05-04 22:39:33 +0000 | [diff] [blame] | 696 | Return the number of bytes written, which is never less than |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 697 | len(b). |
| 698 | |
| 699 | Raises BlockingIOError if the buffer is full and the |
| 700 | underlying raw stream cannot accept more data at the moment. |
| 701 | """ |
| 702 | self._unsupported("write") |
| 703 | |
| 704 | |
| 705 | class _BufferedIOMixin(BufferedIOBase): |
| 706 | |
| 707 | """A mixin implementation of BufferedIOBase with an underlying raw stream. |
| 708 | |
| 709 | This passes most requests on to the underlying raw stream. It |
| 710 | does *not* provide implementations of read(), readinto() or |
| 711 | write(). |
| 712 | """ |
| 713 | |
| 714 | def __init__(self, raw): |
| 715 | self.raw = raw |
| 716 | |
| 717 | ### Positioning ### |
| 718 | |
| 719 | def seek(self, pos, whence=0): |
| 720 | return self.raw.seek(pos, whence) |
| 721 | |
| 722 | def tell(self): |
| 723 | return self.raw.tell() |
| 724 | |
| 725 | def truncate(self, pos=None): |
| 726 | # Flush the stream. We're mixing buffered I/O with lower-level I/O, |
| 727 | # and a flush may be necessary to synch both views of the current |
| 728 | # file state. |
| 729 | self.flush() |
| 730 | |
| 731 | if pos is None: |
| 732 | pos = self.tell() |
Alexandre Vassalotti | 1aed624 | 2008-05-09 21:49:43 +0000 | [diff] [blame] | 733 | # XXX: Should seek() be used, instead of passing the position |
| 734 | # XXX directly to truncate? |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 735 | return self.raw.truncate(pos) |
| 736 | |
| 737 | ### Flush and close ### |
| 738 | |
| 739 | def flush(self): |
| 740 | self.raw.flush() |
| 741 | |
| 742 | def close(self): |
| 743 | if not self.closed: |
| 744 | try: |
| 745 | self.flush() |
| 746 | except IOError: |
| 747 | pass # If flush() fails, just give up |
| 748 | self.raw.close() |
| 749 | |
| 750 | ### Inquiries ### |
| 751 | |
| 752 | def seekable(self): |
| 753 | return self.raw.seekable() |
| 754 | |
| 755 | def readable(self): |
| 756 | return self.raw.readable() |
| 757 | |
| 758 | def writable(self): |
| 759 | return self.raw.writable() |
| 760 | |
| 761 | @property |
| 762 | def closed(self): |
| 763 | return self.raw.closed |
| 764 | |
Benjamin Peterson | ad100c3 | 2008-11-20 22:06:22 +0000 | [diff] [blame] | 765 | @property |
| 766 | def name(self): |
| 767 | return self.raw.name |
| 768 | |
| 769 | @property |
| 770 | def mode(self): |
| 771 | return self.raw.mode |
| 772 | |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 773 | ### Lower-level APIs ### |
| 774 | |
| 775 | def fileno(self): |
| 776 | return self.raw.fileno() |
| 777 | |
| 778 | def isatty(self): |
| 779 | return self.raw.isatty() |
| 780 | |
| 781 | |
Alexandre Vassalotti | 1aed624 | 2008-05-09 21:49:43 +0000 | [diff] [blame] | 782 | class _BytesIO(BufferedIOBase): |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 783 | |
| 784 | """Buffered I/O implementation using an in-memory bytes buffer.""" |
| 785 | |
| 786 | # XXX More docs |
| 787 | |
| 788 | def __init__(self, initial_bytes=None): |
| 789 | buf = bytearray() |
| 790 | if initial_bytes is not None: |
Alexandre Vassalotti | 1aed624 | 2008-05-09 21:49:43 +0000 | [diff] [blame] | 791 | buf += bytearray(initial_bytes) |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 792 | self._buffer = buf |
| 793 | self._pos = 0 |
| 794 | |
| 795 | def getvalue(self): |
Benjamin Peterson | 9ae080e | 2008-05-04 22:39:33 +0000 | [diff] [blame] | 796 | """Return the bytes value (contents) of the buffer |
Benjamin Peterson | 7bb4d2d | 2008-04-13 02:01:27 +0000 | [diff] [blame] | 797 | """ |
Alexandre Vassalotti | 1aed624 | 2008-05-09 21:49:43 +0000 | [diff] [blame] | 798 | if self.closed: |
| 799 | raise ValueError("getvalue on closed file") |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 800 | return bytes(self._buffer) |
| 801 | |
| 802 | def read(self, n=None): |
Alexandre Vassalotti | 1aed624 | 2008-05-09 21:49:43 +0000 | [diff] [blame] | 803 | if self.closed: |
| 804 | raise ValueError("read from closed file") |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 805 | if n is None: |
| 806 | n = -1 |
Alexandre Vassalotti | 1aed624 | 2008-05-09 21:49:43 +0000 | [diff] [blame] | 807 | if not isinstance(n, (int, long)): |
| 808 | raise TypeError("argument must be an integer") |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 809 | if n < 0: |
| 810 | n = len(self._buffer) |
Alexandre Vassalotti | 1aed624 | 2008-05-09 21:49:43 +0000 | [diff] [blame] | 811 | if len(self._buffer) <= self._pos: |
| 812 | return b"" |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 813 | newpos = min(len(self._buffer), self._pos + n) |
| 814 | b = self._buffer[self._pos : newpos] |
| 815 | self._pos = newpos |
| 816 | return bytes(b) |
| 817 | |
| 818 | def read1(self, n): |
Benjamin Peterson | 9ae080e | 2008-05-04 22:39:33 +0000 | [diff] [blame] | 819 | """this is the same as read. |
Benjamin Peterson | 7bb4d2d | 2008-04-13 02:01:27 +0000 | [diff] [blame] | 820 | """ |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 821 | return self.read(n) |
| 822 | |
| 823 | def write(self, b): |
| 824 | if self.closed: |
| 825 | raise ValueError("write to closed file") |
| 826 | if isinstance(b, unicode): |
| 827 | raise TypeError("can't write unicode to binary stream") |
| 828 | n = len(b) |
Alexandre Vassalotti | 1aed624 | 2008-05-09 21:49:43 +0000 | [diff] [blame] | 829 | if n == 0: |
| 830 | return 0 |
Alexandre Vassalotti | 844f757 | 2008-05-10 19:59:16 +0000 | [diff] [blame] | 831 | pos = self._pos |
| 832 | if pos > len(self._buffer): |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 833 | # Inserts null bytes between the current end of the file |
| 834 | # and the new write position. |
Alexandre Vassalotti | 844f757 | 2008-05-10 19:59:16 +0000 | [diff] [blame] | 835 | padding = b'\x00' * (pos - len(self._buffer)) |
| 836 | self._buffer += padding |
| 837 | self._buffer[pos:pos + n] = b |
| 838 | self._pos += n |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 839 | return n |
| 840 | |
| 841 | def seek(self, pos, whence=0): |
Alexandre Vassalotti | 1aed624 | 2008-05-09 21:49:43 +0000 | [diff] [blame] | 842 | if self.closed: |
| 843 | raise ValueError("seek on closed file") |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 844 | try: |
| 845 | pos = pos.__index__() |
| 846 | except AttributeError as err: |
| 847 | raise TypeError("an integer is required") # from err |
| 848 | if whence == 0: |
Alexandre Vassalotti | 1aed624 | 2008-05-09 21:49:43 +0000 | [diff] [blame] | 849 | if pos < 0: |
| 850 | raise ValueError("negative seek position %r" % (pos,)) |
| 851 | self._pos = pos |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 852 | elif whence == 1: |
| 853 | self._pos = max(0, self._pos + pos) |
| 854 | elif whence == 2: |
| 855 | self._pos = max(0, len(self._buffer) + pos) |
| 856 | else: |
Alexandre Vassalotti | 1aed624 | 2008-05-09 21:49:43 +0000 | [diff] [blame] | 857 | raise ValueError("invalid whence value") |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 858 | return self._pos |
| 859 | |
| 860 | def tell(self): |
Alexandre Vassalotti | 1aed624 | 2008-05-09 21:49:43 +0000 | [diff] [blame] | 861 | if self.closed: |
| 862 | raise ValueError("tell on closed file") |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 863 | return self._pos |
| 864 | |
| 865 | def truncate(self, pos=None): |
Alexandre Vassalotti | 1aed624 | 2008-05-09 21:49:43 +0000 | [diff] [blame] | 866 | if self.closed: |
| 867 | raise ValueError("truncate on closed file") |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 868 | if pos is None: |
| 869 | pos = self._pos |
Alexandre Vassalotti | 1aed624 | 2008-05-09 21:49:43 +0000 | [diff] [blame] | 870 | elif pos < 0: |
| 871 | raise ValueError("negative truncate position %r" % (pos,)) |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 872 | del self._buffer[pos:] |
Alexandre Vassalotti | 1aed624 | 2008-05-09 21:49:43 +0000 | [diff] [blame] | 873 | return self.seek(pos) |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 874 | |
| 875 | def readable(self): |
| 876 | return True |
| 877 | |
| 878 | def writable(self): |
| 879 | return True |
| 880 | |
| 881 | def seekable(self): |
| 882 | return True |
| 883 | |
Alexandre Vassalotti | 1aed624 | 2008-05-09 21:49:43 +0000 | [diff] [blame] | 884 | # Use the faster implementation of BytesIO if available |
| 885 | try: |
| 886 | import _bytesio |
| 887 | |
| 888 | class BytesIO(_bytesio._BytesIO, BufferedIOBase): |
| 889 | __doc__ = _bytesio._BytesIO.__doc__ |
| 890 | |
| 891 | except ImportError: |
| 892 | BytesIO = _BytesIO |
| 893 | |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 894 | |
| 895 | class BufferedReader(_BufferedIOMixin): |
| 896 | |
Benjamin Peterson | 7bb4d2d | 2008-04-13 02:01:27 +0000 | [diff] [blame] | 897 | """BufferedReader(raw[, buffer_size]) |
| 898 | |
| 899 | A buffer for a readable, sequential BaseRawIO object. |
| 900 | |
| 901 | The constructor creates a BufferedReader for the given readable raw |
| 902 | stream and buffer_size. If buffer_size is omitted, DEFAULT_BUFFER_SIZE |
| 903 | is used. |
| 904 | """ |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 905 | |
| 906 | def __init__(self, raw, buffer_size=DEFAULT_BUFFER_SIZE): |
| 907 | """Create a new buffered reader using the given readable raw IO object. |
| 908 | """ |
| 909 | raw._checkReadable() |
| 910 | _BufferedIOMixin.__init__(self, raw) |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 911 | self.buffer_size = buffer_size |
Benjamin Peterson | 01a2432 | 2008-07-28 23:35:27 +0000 | [diff] [blame] | 912 | self._reset_read_buf() |
Antoine Pitrou | 11ec65d | 2008-08-14 21:04:30 +0000 | [diff] [blame] | 913 | self._read_lock = threading.Lock() |
Benjamin Peterson | 01a2432 | 2008-07-28 23:35:27 +0000 | [diff] [blame] | 914 | |
| 915 | def _reset_read_buf(self): |
| 916 | self._read_buf = b"" |
| 917 | self._read_pos = 0 |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 918 | |
| 919 | def read(self, n=None): |
| 920 | """Read n bytes. |
| 921 | |
| 922 | Returns exactly n bytes of data unless the underlying raw IO |
| 923 | stream reaches EOF or if the call would block in non-blocking |
| 924 | mode. If n is negative, read until EOF or until read() would |
| 925 | block. |
| 926 | """ |
Antoine Pitrou | 11ec65d | 2008-08-14 21:04:30 +0000 | [diff] [blame] | 927 | with self._read_lock: |
| 928 | return self._read_unlocked(n) |
| 929 | |
| 930 | def _read_unlocked(self, n=None): |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 931 | nodata_val = b"" |
Benjamin Peterson | 01a2432 | 2008-07-28 23:35:27 +0000 | [diff] [blame] | 932 | empty_values = (b"", None) |
| 933 | buf = self._read_buf |
| 934 | pos = self._read_pos |
| 935 | |
| 936 | # Special case for when the number of bytes to read is unspecified. |
| 937 | if n is None or n == -1: |
| 938 | self._reset_read_buf() |
| 939 | chunks = [buf[pos:]] # Strip the consumed bytes. |
| 940 | current_size = 0 |
| 941 | while True: |
| 942 | # Read until EOF or until read() would block. |
| 943 | chunk = self.raw.read() |
| 944 | if chunk in empty_values: |
| 945 | nodata_val = chunk |
| 946 | break |
| 947 | current_size += len(chunk) |
| 948 | chunks.append(chunk) |
| 949 | return b"".join(chunks) or nodata_val |
| 950 | |
| 951 | # The number of bytes to read is specified, return at most n bytes. |
| 952 | avail = len(buf) - pos # Length of the available buffered data. |
| 953 | if n <= avail: |
| 954 | # Fast path: the data to read is fully buffered. |
| 955 | self._read_pos += n |
| 956 | return buf[pos:pos+n] |
| 957 | # Slow path: read from the stream until enough bytes are read, |
| 958 | # or until an EOF occurs or until read() would block. |
| 959 | chunks = [buf[pos:]] |
| 960 | wanted = max(self.buffer_size, n) |
| 961 | while avail < n: |
| 962 | chunk = self.raw.read(wanted) |
| 963 | if chunk in empty_values: |
| 964 | nodata_val = chunk |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 965 | break |
Benjamin Peterson | 01a2432 | 2008-07-28 23:35:27 +0000 | [diff] [blame] | 966 | avail += len(chunk) |
| 967 | chunks.append(chunk) |
| 968 | # n is more then avail only when an EOF occurred or when |
| 969 | # read() would have blocked. |
| 970 | n = min(n, avail) |
| 971 | out = b"".join(chunks) |
| 972 | self._read_buf = out[n:] # Save the extra data in the buffer. |
| 973 | self._read_pos = 0 |
| 974 | return out[:n] if out else nodata_val |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 975 | |
| 976 | def peek(self, n=0): |
| 977 | """Returns buffered bytes without advancing the position. |
| 978 | |
| 979 | The argument indicates a desired minimal number of bytes; we |
| 980 | do at most one raw read to satisfy it. We never return more |
| 981 | than self.buffer_size. |
| 982 | """ |
Antoine Pitrou | 11ec65d | 2008-08-14 21:04:30 +0000 | [diff] [blame] | 983 | with self._read_lock: |
| 984 | return self._peek_unlocked(n) |
| 985 | |
| 986 | def _peek_unlocked(self, n=0): |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 987 | want = min(n, self.buffer_size) |
Benjamin Peterson | 01a2432 | 2008-07-28 23:35:27 +0000 | [diff] [blame] | 988 | have = len(self._read_buf) - self._read_pos |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 989 | if have < want: |
| 990 | to_read = self.buffer_size - have |
| 991 | current = self.raw.read(to_read) |
| 992 | if current: |
Benjamin Peterson | 01a2432 | 2008-07-28 23:35:27 +0000 | [diff] [blame] | 993 | self._read_buf = self._read_buf[self._read_pos:] + current |
| 994 | self._read_pos = 0 |
| 995 | return self._read_buf[self._read_pos:] |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 996 | |
| 997 | def read1(self, n): |
Benjamin Peterson | 7bb4d2d | 2008-04-13 02:01:27 +0000 | [diff] [blame] | 998 | """Reads up to n bytes, with at most one read() system call.""" |
| 999 | # Returns up to n bytes. If at least one byte is buffered, we |
| 1000 | # only return buffered bytes. Otherwise, we do one raw read. |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 1001 | if n <= 0: |
| 1002 | return b"" |
Antoine Pitrou | 11ec65d | 2008-08-14 21:04:30 +0000 | [diff] [blame] | 1003 | with self._read_lock: |
| 1004 | self._peek_unlocked(1) |
| 1005 | return self._read_unlocked( |
| 1006 | min(n, len(self._read_buf) - self._read_pos)) |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 1007 | |
| 1008 | def tell(self): |
Benjamin Peterson | 01a2432 | 2008-07-28 23:35:27 +0000 | [diff] [blame] | 1009 | return self.raw.tell() - len(self._read_buf) + self._read_pos |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 1010 | |
| 1011 | def seek(self, pos, whence=0): |
Antoine Pitrou | 11ec65d | 2008-08-14 21:04:30 +0000 | [diff] [blame] | 1012 | with self._read_lock: |
| 1013 | if whence == 1: |
| 1014 | pos -= len(self._read_buf) - self._read_pos |
| 1015 | pos = self.raw.seek(pos, whence) |
| 1016 | self._reset_read_buf() |
| 1017 | return pos |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 1018 | |
| 1019 | |
| 1020 | class BufferedWriter(_BufferedIOMixin): |
| 1021 | |
Benjamin Peterson | 9ae080e | 2008-05-04 22:39:33 +0000 | [diff] [blame] | 1022 | """A buffer for a writeable sequential RawIO object. |
Benjamin Peterson | 7bb4d2d | 2008-04-13 02:01:27 +0000 | [diff] [blame] | 1023 | |
| 1024 | The constructor creates a BufferedWriter for the given writeable raw |
| 1025 | stream. If the buffer_size is not given, it defaults to |
| 1026 | DEAFULT_BUFFER_SIZE. If max_buffer_size is omitted, it defaults to |
| 1027 | twice the buffer size. |
| 1028 | """ |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 1029 | |
| 1030 | def __init__(self, raw, |
| 1031 | buffer_size=DEFAULT_BUFFER_SIZE, max_buffer_size=None): |
| 1032 | raw._checkWritable() |
| 1033 | _BufferedIOMixin.__init__(self, raw) |
| 1034 | self.buffer_size = buffer_size |
| 1035 | self.max_buffer_size = (2*buffer_size |
| 1036 | if max_buffer_size is None |
| 1037 | else max_buffer_size) |
| 1038 | self._write_buf = bytearray() |
Antoine Pitrou | 11ec65d | 2008-08-14 21:04:30 +0000 | [diff] [blame] | 1039 | self._write_lock = threading.Lock() |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 1040 | |
| 1041 | def write(self, b): |
| 1042 | if self.closed: |
| 1043 | raise ValueError("write to closed file") |
| 1044 | if isinstance(b, unicode): |
| 1045 | raise TypeError("can't write unicode to binary stream") |
Antoine Pitrou | 11ec65d | 2008-08-14 21:04:30 +0000 | [diff] [blame] | 1046 | with self._write_lock: |
| 1047 | # XXX we can implement some more tricks to try and avoid |
| 1048 | # partial writes |
| 1049 | if len(self._write_buf) > self.buffer_size: |
| 1050 | # We're full, so let's pre-flush the buffer |
| 1051 | try: |
| 1052 | self._flush_unlocked() |
| 1053 | except BlockingIOError as e: |
| 1054 | # We can't accept anything else. |
| 1055 | # XXX Why not just let the exception pass through? |
| 1056 | raise BlockingIOError(e.errno, e.strerror, 0) |
| 1057 | before = len(self._write_buf) |
| 1058 | self._write_buf.extend(b) |
| 1059 | written = len(self._write_buf) - before |
| 1060 | if len(self._write_buf) > self.buffer_size: |
| 1061 | try: |
| 1062 | self._flush_unlocked() |
| 1063 | except BlockingIOError as e: |
| 1064 | if len(self._write_buf) > self.max_buffer_size: |
| 1065 | # We've hit max_buffer_size. We have to accept a |
| 1066 | # partial write and cut back our buffer. |
| 1067 | overage = len(self._write_buf) - self.max_buffer_size |
| 1068 | self._write_buf = self._write_buf[:self.max_buffer_size] |
| 1069 | raise BlockingIOError(e.errno, e.strerror, overage) |
| 1070 | return written |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 1071 | |
Alexandre Vassalotti | 1aed624 | 2008-05-09 21:49:43 +0000 | [diff] [blame] | 1072 | def truncate(self, pos=None): |
Antoine Pitrou | 11ec65d | 2008-08-14 21:04:30 +0000 | [diff] [blame] | 1073 | with self._write_lock: |
| 1074 | self._flush_unlocked() |
| 1075 | if pos is None: |
| 1076 | pos = self.raw.tell() |
| 1077 | return self.raw.truncate(pos) |
Alexandre Vassalotti | 1aed624 | 2008-05-09 21:49:43 +0000 | [diff] [blame] | 1078 | |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 1079 | def flush(self): |
Antoine Pitrou | 11ec65d | 2008-08-14 21:04:30 +0000 | [diff] [blame] | 1080 | with self._write_lock: |
| 1081 | self._flush_unlocked() |
| 1082 | |
| 1083 | def _flush_unlocked(self): |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 1084 | if self.closed: |
| 1085 | raise ValueError("flush of closed file") |
| 1086 | written = 0 |
| 1087 | try: |
| 1088 | while self._write_buf: |
| 1089 | n = self.raw.write(self._write_buf) |
| 1090 | del self._write_buf[:n] |
| 1091 | written += n |
| 1092 | except BlockingIOError as e: |
| 1093 | n = e.characters_written |
| 1094 | del self._write_buf[:n] |
| 1095 | written += n |
| 1096 | raise BlockingIOError(e.errno, e.strerror, written) |
| 1097 | |
| 1098 | def tell(self): |
| 1099 | return self.raw.tell() + len(self._write_buf) |
| 1100 | |
| 1101 | def seek(self, pos, whence=0): |
Antoine Pitrou | 11ec65d | 2008-08-14 21:04:30 +0000 | [diff] [blame] | 1102 | with self._write_lock: |
| 1103 | self._flush_unlocked() |
| 1104 | return self.raw.seek(pos, whence) |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 1105 | |
| 1106 | |
| 1107 | class BufferedRWPair(BufferedIOBase): |
| 1108 | |
| 1109 | """A buffered reader and writer object together. |
| 1110 | |
Benjamin Peterson | 7bb4d2d | 2008-04-13 02:01:27 +0000 | [diff] [blame] | 1111 | A buffered reader object and buffered writer object put together to |
| 1112 | form a sequential IO object that can read and write. This is typically |
| 1113 | used with a socket or two-way pipe. |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 1114 | |
Benjamin Peterson | 7bb4d2d | 2008-04-13 02:01:27 +0000 | [diff] [blame] | 1115 | reader and writer are RawIOBase objects that are readable and |
| 1116 | writeable respectively. If the buffer_size is omitted it defaults to |
| 1117 | DEFAULT_BUFFER_SIZE. The max_buffer_size (for the buffered writer) |
| 1118 | defaults to twice the buffer size. |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 1119 | """ |
| 1120 | |
Benjamin Peterson | 7bb4d2d | 2008-04-13 02:01:27 +0000 | [diff] [blame] | 1121 | # XXX The usefulness of this (compared to having two separate IO |
| 1122 | # objects) is questionable. |
| 1123 | |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 1124 | def __init__(self, reader, writer, |
| 1125 | buffer_size=DEFAULT_BUFFER_SIZE, max_buffer_size=None): |
| 1126 | """Constructor. |
| 1127 | |
| 1128 | The arguments are two RawIO instances. |
| 1129 | """ |
| 1130 | reader._checkReadable() |
| 1131 | writer._checkWritable() |
| 1132 | self.reader = BufferedReader(reader, buffer_size) |
| 1133 | self.writer = BufferedWriter(writer, buffer_size, max_buffer_size) |
| 1134 | |
| 1135 | def read(self, n=None): |
| 1136 | if n is None: |
| 1137 | n = -1 |
| 1138 | return self.reader.read(n) |
| 1139 | |
| 1140 | def readinto(self, b): |
| 1141 | return self.reader.readinto(b) |
| 1142 | |
| 1143 | def write(self, b): |
| 1144 | return self.writer.write(b) |
| 1145 | |
| 1146 | def peek(self, n=0): |
| 1147 | return self.reader.peek(n) |
| 1148 | |
| 1149 | def read1(self, n): |
| 1150 | return self.reader.read1(n) |
| 1151 | |
| 1152 | def readable(self): |
| 1153 | return self.reader.readable() |
| 1154 | |
| 1155 | def writable(self): |
| 1156 | return self.writer.writable() |
| 1157 | |
| 1158 | def flush(self): |
| 1159 | return self.writer.flush() |
| 1160 | |
| 1161 | def close(self): |
| 1162 | self.writer.close() |
| 1163 | self.reader.close() |
| 1164 | |
| 1165 | def isatty(self): |
| 1166 | return self.reader.isatty() or self.writer.isatty() |
| 1167 | |
| 1168 | @property |
| 1169 | def closed(self): |
Benjamin Peterson | 54686e3 | 2008-12-24 15:10:27 +0000 | [diff] [blame] | 1170 | return self.writer.closed |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 1171 | |
| 1172 | |
| 1173 | class BufferedRandom(BufferedWriter, BufferedReader): |
| 1174 | |
Benjamin Peterson | 9ae080e | 2008-05-04 22:39:33 +0000 | [diff] [blame] | 1175 | """A buffered interface to random access streams. |
Benjamin Peterson | 7bb4d2d | 2008-04-13 02:01:27 +0000 | [diff] [blame] | 1176 | |
| 1177 | The constructor creates a reader and writer for a seekable stream, |
| 1178 | raw, given in the first argument. If the buffer_size is omitted it |
| 1179 | defaults to DEFAULT_BUFFER_SIZE. The max_buffer_size (for the buffered |
| 1180 | writer) defaults to twice the buffer size. |
| 1181 | """ |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 1182 | |
| 1183 | def __init__(self, raw, |
| 1184 | buffer_size=DEFAULT_BUFFER_SIZE, max_buffer_size=None): |
| 1185 | raw._checkSeekable() |
| 1186 | BufferedReader.__init__(self, raw, buffer_size) |
| 1187 | BufferedWriter.__init__(self, raw, buffer_size, max_buffer_size) |
| 1188 | |
| 1189 | def seek(self, pos, whence=0): |
| 1190 | self.flush() |
| 1191 | # First do the raw seek, then empty the read buffer, so that |
| 1192 | # if the raw seek fails, we don't lose buffered data forever. |
| 1193 | pos = self.raw.seek(pos, whence) |
Antoine Pitrou | 11ec65d | 2008-08-14 21:04:30 +0000 | [diff] [blame] | 1194 | with self._read_lock: |
| 1195 | self._reset_read_buf() |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 1196 | return pos |
| 1197 | |
| 1198 | def tell(self): |
Benjamin Peterson | 01a2432 | 2008-07-28 23:35:27 +0000 | [diff] [blame] | 1199 | if self._write_buf: |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 1200 | return self.raw.tell() + len(self._write_buf) |
| 1201 | else: |
Benjamin Peterson | 01a2432 | 2008-07-28 23:35:27 +0000 | [diff] [blame] | 1202 | return BufferedReader.tell(self) |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 1203 | |
Alexandre Vassalotti | 1aed624 | 2008-05-09 21:49:43 +0000 | [diff] [blame] | 1204 | def truncate(self, pos=None): |
| 1205 | if pos is None: |
| 1206 | pos = self.tell() |
| 1207 | # Use seek to flush the read buffer. |
| 1208 | self.seek(pos) |
| 1209 | return BufferedWriter.truncate(self) |
| 1210 | |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 1211 | def read(self, n=None): |
| 1212 | if n is None: |
| 1213 | n = -1 |
| 1214 | self.flush() |
| 1215 | return BufferedReader.read(self, n) |
| 1216 | |
| 1217 | def readinto(self, b): |
| 1218 | self.flush() |
| 1219 | return BufferedReader.readinto(self, b) |
| 1220 | |
| 1221 | def peek(self, n=0): |
| 1222 | self.flush() |
| 1223 | return BufferedReader.peek(self, n) |
| 1224 | |
| 1225 | def read1(self, n): |
| 1226 | self.flush() |
| 1227 | return BufferedReader.read1(self, n) |
| 1228 | |
| 1229 | def write(self, b): |
| 1230 | if self._read_buf: |
Benjamin Peterson | 01a2432 | 2008-07-28 23:35:27 +0000 | [diff] [blame] | 1231 | # Undo readahead |
Antoine Pitrou | 11ec65d | 2008-08-14 21:04:30 +0000 | [diff] [blame] | 1232 | with self._read_lock: |
| 1233 | self.raw.seek(self._read_pos - len(self._read_buf), 1) |
| 1234 | self._reset_read_buf() |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 1235 | return BufferedWriter.write(self, b) |
| 1236 | |
| 1237 | |
| 1238 | class TextIOBase(IOBase): |
| 1239 | |
| 1240 | """Base class for text I/O. |
| 1241 | |
Benjamin Peterson | 7bb4d2d | 2008-04-13 02:01:27 +0000 | [diff] [blame] | 1242 | This class provides a character and line based interface to stream |
| 1243 | I/O. There is no readinto method because Python's character strings |
| 1244 | are immutable. There is no public constructor. |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 1245 | """ |
| 1246 | |
| 1247 | def read(self, n = -1): |
Benjamin Peterson | 9ae080e | 2008-05-04 22:39:33 +0000 | [diff] [blame] | 1248 | """Read at most n characters from stream. |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 1249 | |
| 1250 | Read from underlying buffer until we have n characters or we hit EOF. |
| 1251 | If n is negative or omitted, read until EOF. |
| 1252 | """ |
| 1253 | self._unsupported("read") |
| 1254 | |
| 1255 | def write(self, s): |
Benjamin Peterson | 9ae080e | 2008-05-04 22:39:33 +0000 | [diff] [blame] | 1256 | """Write string s to stream.""" |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 1257 | self._unsupported("write") |
| 1258 | |
| 1259 | def truncate(self, pos = None): |
Benjamin Peterson | 9ae080e | 2008-05-04 22:39:33 +0000 | [diff] [blame] | 1260 | """Truncate size to pos.""" |
Alexandre Vassalotti | 1aed624 | 2008-05-09 21:49:43 +0000 | [diff] [blame] | 1261 | self._unsupported("truncate") |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 1262 | |
| 1263 | def readline(self): |
Benjamin Peterson | 9ae080e | 2008-05-04 22:39:33 +0000 | [diff] [blame] | 1264 | """Read until newline or EOF. |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 1265 | |
| 1266 | Returns an empty string if EOF is hit immediately. |
| 1267 | """ |
| 1268 | self._unsupported("readline") |
| 1269 | |
| 1270 | @property |
| 1271 | def encoding(self): |
| 1272 | """Subclasses should override.""" |
| 1273 | return None |
| 1274 | |
| 1275 | @property |
| 1276 | def newlines(self): |
Benjamin Peterson | 9ae080e | 2008-05-04 22:39:33 +0000 | [diff] [blame] | 1277 | """Line endings translated so far. |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 1278 | |
| 1279 | Only line endings translated during reading are considered. |
| 1280 | |
| 1281 | Subclasses should override. |
| 1282 | """ |
| 1283 | return None |
| 1284 | |
| 1285 | |
| 1286 | class IncrementalNewlineDecoder(codecs.IncrementalDecoder): |
| 1287 | """Codec used when reading a file in universal newlines mode. |
| 1288 | It wraps another incremental decoder, translating \\r\\n and \\r into \\n. |
| 1289 | It also records the types of newlines encountered. |
| 1290 | When used with translate=False, it ensures that the newline sequence is |
| 1291 | returned in one piece. |
| 1292 | """ |
| 1293 | def __init__(self, decoder, translate, errors='strict'): |
| 1294 | codecs.IncrementalDecoder.__init__(self, errors=errors) |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 1295 | self.translate = translate |
| 1296 | self.decoder = decoder |
| 1297 | self.seennl = 0 |
Antoine Pitrou | 655fbf1 | 2008-12-14 17:40:51 +0000 | [diff] [blame] | 1298 | self.pendingcr = False |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 1299 | |
| 1300 | def decode(self, input, final=False): |
| 1301 | # decode input (with the eventual \r from a previous pass) |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 1302 | output = self.decoder.decode(input, final=final) |
Antoine Pitrou | 655fbf1 | 2008-12-14 17:40:51 +0000 | [diff] [blame] | 1303 | if self.pendingcr and (output or final): |
| 1304 | output = "\r" + output |
| 1305 | self.pendingcr = False |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 1306 | |
| 1307 | # retain last \r even when not translating data: |
| 1308 | # then readline() is sure to get \r\n in one pass |
| 1309 | if output.endswith("\r") and not final: |
| 1310 | output = output[:-1] |
Antoine Pitrou | 655fbf1 | 2008-12-14 17:40:51 +0000 | [diff] [blame] | 1311 | self.pendingcr = True |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 1312 | |
| 1313 | # Record which newlines are read |
| 1314 | crlf = output.count('\r\n') |
| 1315 | cr = output.count('\r') - crlf |
| 1316 | lf = output.count('\n') - crlf |
| 1317 | self.seennl |= (lf and self._LF) | (cr and self._CR) \ |
| 1318 | | (crlf and self._CRLF) |
| 1319 | |
| 1320 | if self.translate: |
| 1321 | if crlf: |
| 1322 | output = output.replace("\r\n", "\n") |
| 1323 | if cr: |
| 1324 | output = output.replace("\r", "\n") |
| 1325 | |
| 1326 | return output |
| 1327 | |
| 1328 | def getstate(self): |
| 1329 | buf, flag = self.decoder.getstate() |
Antoine Pitrou | 655fbf1 | 2008-12-14 17:40:51 +0000 | [diff] [blame] | 1330 | flag <<= 1 |
| 1331 | if self.pendingcr: |
| 1332 | flag |= 1 |
| 1333 | return buf, flag |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 1334 | |
| 1335 | def setstate(self, state): |
| 1336 | buf, flag = state |
Antoine Pitrou | 655fbf1 | 2008-12-14 17:40:51 +0000 | [diff] [blame] | 1337 | self.pendingcr = bool(flag & 1) |
| 1338 | self.decoder.setstate((buf, flag >> 1)) |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 1339 | |
| 1340 | def reset(self): |
| 1341 | self.seennl = 0 |
Antoine Pitrou | 655fbf1 | 2008-12-14 17:40:51 +0000 | [diff] [blame] | 1342 | self.pendingcr = False |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 1343 | self.decoder.reset() |
| 1344 | |
| 1345 | _LF = 1 |
| 1346 | _CR = 2 |
| 1347 | _CRLF = 4 |
| 1348 | |
| 1349 | @property |
| 1350 | def newlines(self): |
| 1351 | return (None, |
| 1352 | "\n", |
| 1353 | "\r", |
| 1354 | ("\r", "\n"), |
| 1355 | "\r\n", |
| 1356 | ("\n", "\r\n"), |
| 1357 | ("\r", "\r\n"), |
| 1358 | ("\r", "\n", "\r\n") |
| 1359 | )[self.seennl] |
| 1360 | |
| 1361 | |
| 1362 | class TextIOWrapper(TextIOBase): |
| 1363 | |
Benjamin Peterson | 9ae080e | 2008-05-04 22:39:33 +0000 | [diff] [blame] | 1364 | r"""Character and line based layer over a BufferedIOBase object, buffer. |
Benjamin Peterson | 7bb4d2d | 2008-04-13 02:01:27 +0000 | [diff] [blame] | 1365 | |
| 1366 | encoding gives the name of the encoding that the stream will be |
| 1367 | decoded or encoded with. It defaults to locale.getpreferredencoding. |
| 1368 | |
| 1369 | errors determines the strictness of encoding and decoding (see the |
| 1370 | codecs.register) and defaults to "strict". |
| 1371 | |
| 1372 | newline can be None, '', '\n', '\r', or '\r\n'. It controls the |
| 1373 | handling of line endings. If it is None, universal newlines is |
| 1374 | enabled. With this enabled, on input, the lines endings '\n', '\r', |
| 1375 | or '\r\n' are translated to '\n' before being returned to the |
| 1376 | caller. Conversely, on output, '\n' is translated to the system |
| 1377 | default line seperator, os.linesep. If newline is any other of its |
| 1378 | legal values, that newline becomes the newline when the file is read |
| 1379 | and it is returned untranslated. On output, '\n' is converted to the |
| 1380 | newline. |
| 1381 | |
| 1382 | If line_buffering is True, a call to flush is implied when a call to |
| 1383 | write contains a newline character. |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 1384 | """ |
| 1385 | |
| 1386 | _CHUNK_SIZE = 128 |
| 1387 | |
| 1388 | def __init__(self, buffer, encoding=None, errors=None, newline=None, |
| 1389 | line_buffering=False): |
| 1390 | if newline not in (None, "", "\n", "\r", "\r\n"): |
| 1391 | raise ValueError("illegal newline value: %r" % (newline,)) |
| 1392 | if encoding is None: |
| 1393 | try: |
| 1394 | encoding = os.device_encoding(buffer.fileno()) |
| 1395 | except (AttributeError, UnsupportedOperation): |
| 1396 | pass |
| 1397 | if encoding is None: |
| 1398 | try: |
| 1399 | import locale |
| 1400 | except ImportError: |
| 1401 | # Importing locale may fail if Python is being built |
| 1402 | encoding = "ascii" |
| 1403 | else: |
| 1404 | encoding = locale.getpreferredencoding() |
| 1405 | |
Christian Heimes | 3784c6b | 2008-03-26 23:13:59 +0000 | [diff] [blame] | 1406 | if not isinstance(encoding, basestring): |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 1407 | raise ValueError("invalid encoding: %r" % encoding) |
| 1408 | |
| 1409 | if errors is None: |
| 1410 | errors = "strict" |
| 1411 | else: |
Christian Heimes | 3784c6b | 2008-03-26 23:13:59 +0000 | [diff] [blame] | 1412 | if not isinstance(errors, basestring): |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 1413 | raise ValueError("invalid errors: %r" % errors) |
| 1414 | |
| 1415 | self.buffer = buffer |
| 1416 | self._line_buffering = line_buffering |
| 1417 | self._encoding = encoding |
| 1418 | self._errors = errors |
| 1419 | self._readuniversal = not newline |
| 1420 | self._readtranslate = newline is None |
| 1421 | self._readnl = newline |
| 1422 | self._writetranslate = newline != '' |
| 1423 | self._writenl = newline or os.linesep |
| 1424 | self._encoder = None |
| 1425 | self._decoder = None |
| 1426 | self._decoded_chars = '' # buffer for text returned from decoder |
| 1427 | self._decoded_chars_used = 0 # offset into _decoded_chars for read() |
| 1428 | self._snapshot = None # info for reconstructing decoder state |
| 1429 | self._seekable = self._telling = self.buffer.seekable() |
| 1430 | |
| 1431 | # self._snapshot is either None, or a tuple (dec_flags, next_input) |
| 1432 | # where dec_flags is the second (integer) item of the decoder state |
| 1433 | # and next_input is the chunk of input bytes that comes next after the |
| 1434 | # snapshot point. We use this to reconstruct decoder states in tell(). |
| 1435 | |
| 1436 | # Naming convention: |
| 1437 | # - "bytes_..." for integer variables that count input bytes |
| 1438 | # - "chars_..." for integer variables that count decoded characters |
| 1439 | |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 1440 | @property |
| 1441 | def encoding(self): |
| 1442 | return self._encoding |
| 1443 | |
| 1444 | @property |
| 1445 | def errors(self): |
| 1446 | return self._errors |
| 1447 | |
| 1448 | @property |
| 1449 | def line_buffering(self): |
| 1450 | return self._line_buffering |
| 1451 | |
| 1452 | def seekable(self): |
| 1453 | return self._seekable |
| 1454 | |
Alexandre Vassalotti | 1aed624 | 2008-05-09 21:49:43 +0000 | [diff] [blame] | 1455 | def readable(self): |
| 1456 | return self.buffer.readable() |
| 1457 | |
| 1458 | def writable(self): |
| 1459 | return self.buffer.writable() |
| 1460 | |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 1461 | def flush(self): |
| 1462 | self.buffer.flush() |
| 1463 | self._telling = self._seekable |
| 1464 | |
| 1465 | def close(self): |
| 1466 | try: |
| 1467 | self.flush() |
| 1468 | except: |
| 1469 | pass # If flush() fails, just give up |
| 1470 | self.buffer.close() |
| 1471 | |
| 1472 | @property |
| 1473 | def closed(self): |
| 1474 | return self.buffer.closed |
| 1475 | |
Benjamin Peterson | ad100c3 | 2008-11-20 22:06:22 +0000 | [diff] [blame] | 1476 | @property |
| 1477 | def name(self): |
| 1478 | return self.buffer.name |
| 1479 | |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 1480 | def fileno(self): |
| 1481 | return self.buffer.fileno() |
| 1482 | |
| 1483 | def isatty(self): |
| 1484 | return self.buffer.isatty() |
| 1485 | |
| 1486 | def write(self, s): |
| 1487 | if self.closed: |
| 1488 | raise ValueError("write to closed file") |
| 1489 | if not isinstance(s, unicode): |
| 1490 | raise TypeError("can't write %s to text stream" % |
| 1491 | s.__class__.__name__) |
| 1492 | length = len(s) |
| 1493 | haslf = (self._writetranslate or self._line_buffering) and "\n" in s |
| 1494 | if haslf and self._writetranslate and self._writenl != "\n": |
| 1495 | s = s.replace("\n", self._writenl) |
| 1496 | encoder = self._encoder or self._get_encoder() |
| 1497 | # XXX What if we were just reading? |
| 1498 | b = encoder.encode(s) |
| 1499 | self.buffer.write(b) |
| 1500 | if self._line_buffering and (haslf or "\r" in s): |
| 1501 | self.flush() |
| 1502 | self._snapshot = None |
| 1503 | if self._decoder: |
| 1504 | self._decoder.reset() |
| 1505 | return length |
| 1506 | |
| 1507 | def _get_encoder(self): |
| 1508 | make_encoder = codecs.getincrementalencoder(self._encoding) |
| 1509 | self._encoder = make_encoder(self._errors) |
| 1510 | return self._encoder |
| 1511 | |
| 1512 | def _get_decoder(self): |
| 1513 | make_decoder = codecs.getincrementaldecoder(self._encoding) |
| 1514 | decoder = make_decoder(self._errors) |
| 1515 | if self._readuniversal: |
| 1516 | decoder = IncrementalNewlineDecoder(decoder, self._readtranslate) |
| 1517 | self._decoder = decoder |
| 1518 | return decoder |
| 1519 | |
| 1520 | # The following three methods implement an ADT for _decoded_chars. |
| 1521 | # Text returned from the decoder is buffered here until the client |
| 1522 | # requests it by calling our read() or readline() method. |
| 1523 | def _set_decoded_chars(self, chars): |
| 1524 | """Set the _decoded_chars buffer.""" |
| 1525 | self._decoded_chars = chars |
| 1526 | self._decoded_chars_used = 0 |
| 1527 | |
| 1528 | def _get_decoded_chars(self, n=None): |
| 1529 | """Advance into the _decoded_chars buffer.""" |
| 1530 | offset = self._decoded_chars_used |
| 1531 | if n is None: |
| 1532 | chars = self._decoded_chars[offset:] |
| 1533 | else: |
| 1534 | chars = self._decoded_chars[offset:offset + n] |
| 1535 | self._decoded_chars_used += len(chars) |
| 1536 | return chars |
| 1537 | |
| 1538 | def _rewind_decoded_chars(self, n): |
| 1539 | """Rewind the _decoded_chars buffer.""" |
| 1540 | if self._decoded_chars_used < n: |
| 1541 | raise AssertionError("rewind decoded_chars out of bounds") |
| 1542 | self._decoded_chars_used -= n |
| 1543 | |
| 1544 | def _read_chunk(self): |
| 1545 | """ |
| 1546 | Read and decode the next chunk of data from the BufferedReader. |
| 1547 | |
| 1548 | The return value is True unless EOF was reached. The decoded string |
| 1549 | is placed in self._decoded_chars (replacing its previous value). |
| 1550 | The entire input chunk is sent to the decoder, though some of it |
| 1551 | may remain buffered in the decoder, yet to be converted. |
| 1552 | """ |
| 1553 | |
| 1554 | if self._decoder is None: |
| 1555 | raise ValueError("no decoder") |
| 1556 | |
| 1557 | if self._telling: |
| 1558 | # To prepare for tell(), we need to snapshot a point in the |
| 1559 | # file where the decoder's input buffer is empty. |
| 1560 | |
| 1561 | dec_buffer, dec_flags = self._decoder.getstate() |
| 1562 | # Given this, we know there was a valid snapshot point |
| 1563 | # len(dec_buffer) bytes ago with decoder state (b'', dec_flags). |
| 1564 | |
| 1565 | # Read a chunk, decode it, and put the result in self._decoded_chars. |
| 1566 | input_chunk = self.buffer.read1(self._CHUNK_SIZE) |
| 1567 | eof = not input_chunk |
| 1568 | self._set_decoded_chars(self._decoder.decode(input_chunk, eof)) |
| 1569 | |
| 1570 | if self._telling: |
| 1571 | # At the snapshot point, len(dec_buffer) bytes before the read, |
| 1572 | # the next input to be decoded is dec_buffer + input_chunk. |
| 1573 | self._snapshot = (dec_flags, dec_buffer + input_chunk) |
| 1574 | |
| 1575 | return not eof |
| 1576 | |
| 1577 | def _pack_cookie(self, position, dec_flags=0, |
| 1578 | bytes_to_feed=0, need_eof=0, chars_to_skip=0): |
| 1579 | # The meaning of a tell() cookie is: seek to position, set the |
| 1580 | # decoder flags to dec_flags, read bytes_to_feed bytes, feed them |
| 1581 | # into the decoder with need_eof as the EOF flag, then skip |
| 1582 | # chars_to_skip characters of the decoded result. For most simple |
| 1583 | # decoders, tell() will often just give a byte offset in the file. |
| 1584 | return (position | (dec_flags<<64) | (bytes_to_feed<<128) | |
| 1585 | (chars_to_skip<<192) | bool(need_eof)<<256) |
| 1586 | |
| 1587 | def _unpack_cookie(self, bigint): |
| 1588 | rest, position = divmod(bigint, 1<<64) |
| 1589 | rest, dec_flags = divmod(rest, 1<<64) |
| 1590 | rest, bytes_to_feed = divmod(rest, 1<<64) |
| 1591 | need_eof, chars_to_skip = divmod(rest, 1<<64) |
| 1592 | return position, dec_flags, bytes_to_feed, need_eof, chars_to_skip |
| 1593 | |
| 1594 | def tell(self): |
| 1595 | if not self._seekable: |
| 1596 | raise IOError("underlying stream is not seekable") |
| 1597 | if not self._telling: |
| 1598 | raise IOError("telling position disabled by next() call") |
| 1599 | self.flush() |
| 1600 | position = self.buffer.tell() |
| 1601 | decoder = self._decoder |
| 1602 | if decoder is None or self._snapshot is None: |
| 1603 | if self._decoded_chars: |
| 1604 | # This should never happen. |
| 1605 | raise AssertionError("pending decoded text") |
| 1606 | return position |
| 1607 | |
| 1608 | # Skip backward to the snapshot point (see _read_chunk). |
| 1609 | dec_flags, next_input = self._snapshot |
| 1610 | position -= len(next_input) |
| 1611 | |
| 1612 | # How many decoded characters have been used up since the snapshot? |
| 1613 | chars_to_skip = self._decoded_chars_used |
| 1614 | if chars_to_skip == 0: |
| 1615 | # We haven't moved from the snapshot point. |
| 1616 | return self._pack_cookie(position, dec_flags) |
| 1617 | |
| 1618 | # Starting from the snapshot position, we will walk the decoder |
| 1619 | # forward until it gives us enough decoded characters. |
| 1620 | saved_state = decoder.getstate() |
| 1621 | try: |
| 1622 | # Note our initial start point. |
| 1623 | decoder.setstate((b'', dec_flags)) |
| 1624 | start_pos = position |
| 1625 | start_flags, bytes_fed, chars_decoded = dec_flags, 0, 0 |
| 1626 | need_eof = 0 |
| 1627 | |
| 1628 | # Feed the decoder one byte at a time. As we go, note the |
| 1629 | # nearest "safe start point" before the current location |
| 1630 | # (a point where the decoder has nothing buffered, so seek() |
| 1631 | # can safely start from there and advance to this location). |
Amaury Forgeot d'Arc | 7684f85 | 2008-05-03 12:21:13 +0000 | [diff] [blame] | 1632 | for next_byte in next_input: |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 1633 | bytes_fed += 1 |
| 1634 | chars_decoded += len(decoder.decode(next_byte)) |
| 1635 | dec_buffer, dec_flags = decoder.getstate() |
| 1636 | if not dec_buffer and chars_decoded <= chars_to_skip: |
| 1637 | # Decoder buffer is empty, so this is a safe start point. |
| 1638 | start_pos += bytes_fed |
| 1639 | chars_to_skip -= chars_decoded |
| 1640 | start_flags, bytes_fed, chars_decoded = dec_flags, 0, 0 |
| 1641 | if chars_decoded >= chars_to_skip: |
| 1642 | break |
| 1643 | else: |
| 1644 | # We didn't get enough decoded data; signal EOF to get more. |
| 1645 | chars_decoded += len(decoder.decode(b'', final=True)) |
| 1646 | need_eof = 1 |
| 1647 | if chars_decoded < chars_to_skip: |
| 1648 | raise IOError("can't reconstruct logical file position") |
| 1649 | |
| 1650 | # The returned cookie corresponds to the last safe start point. |
| 1651 | return self._pack_cookie( |
| 1652 | start_pos, start_flags, bytes_fed, need_eof, chars_to_skip) |
| 1653 | finally: |
| 1654 | decoder.setstate(saved_state) |
| 1655 | |
Alexandre Vassalotti | 1aed624 | 2008-05-09 21:49:43 +0000 | [diff] [blame] | 1656 | def truncate(self, pos=None): |
| 1657 | self.flush() |
| 1658 | if pos is None: |
| 1659 | pos = self.tell() |
| 1660 | self.seek(pos) |
| 1661 | return self.buffer.truncate() |
| 1662 | |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 1663 | def seek(self, cookie, whence=0): |
Alexandre Vassalotti | 1aed624 | 2008-05-09 21:49:43 +0000 | [diff] [blame] | 1664 | if self.closed: |
| 1665 | raise ValueError("tell on closed file") |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 1666 | if not self._seekable: |
| 1667 | raise IOError("underlying stream is not seekable") |
| 1668 | if whence == 1: # seek relative to current position |
| 1669 | if cookie != 0: |
| 1670 | raise IOError("can't do nonzero cur-relative seeks") |
| 1671 | # Seeking to the current position should attempt to |
| 1672 | # sync the underlying buffer with the current position. |
| 1673 | whence = 0 |
| 1674 | cookie = self.tell() |
| 1675 | if whence == 2: # seek relative to end of file |
| 1676 | if cookie != 0: |
| 1677 | raise IOError("can't do nonzero end-relative seeks") |
| 1678 | self.flush() |
| 1679 | position = self.buffer.seek(0, 2) |
| 1680 | self._set_decoded_chars('') |
| 1681 | self._snapshot = None |
| 1682 | if self._decoder: |
| 1683 | self._decoder.reset() |
| 1684 | return position |
| 1685 | if whence != 0: |
| 1686 | raise ValueError("invalid whence (%r, should be 0, 1 or 2)" % |
| 1687 | (whence,)) |
| 1688 | if cookie < 0: |
| 1689 | raise ValueError("negative seek position %r" % (cookie,)) |
| 1690 | self.flush() |
| 1691 | |
| 1692 | # The strategy of seek() is to go back to the safe start point |
| 1693 | # and replay the effect of read(chars_to_skip) from there. |
| 1694 | start_pos, dec_flags, bytes_to_feed, need_eof, chars_to_skip = \ |
| 1695 | self._unpack_cookie(cookie) |
| 1696 | |
| 1697 | # Seek back to the safe start point. |
| 1698 | self.buffer.seek(start_pos) |
| 1699 | self._set_decoded_chars('') |
| 1700 | self._snapshot = None |
| 1701 | |
| 1702 | # Restore the decoder to its state from the safe start point. |
| 1703 | if self._decoder or dec_flags or chars_to_skip: |
| 1704 | self._decoder = self._decoder or self._get_decoder() |
| 1705 | self._decoder.setstate((b'', dec_flags)) |
| 1706 | self._snapshot = (dec_flags, b'') |
| 1707 | |
| 1708 | if chars_to_skip: |
| 1709 | # Just like _read_chunk, feed the decoder and save a snapshot. |
| 1710 | input_chunk = self.buffer.read(bytes_to_feed) |
| 1711 | self._set_decoded_chars( |
| 1712 | self._decoder.decode(input_chunk, need_eof)) |
| 1713 | self._snapshot = (dec_flags, input_chunk) |
| 1714 | |
| 1715 | # Skip chars_to_skip of the decoded characters. |
| 1716 | if len(self._decoded_chars) < chars_to_skip: |
| 1717 | raise IOError("can't restore logical file position") |
| 1718 | self._decoded_chars_used = chars_to_skip |
| 1719 | |
| 1720 | return cookie |
| 1721 | |
| 1722 | def read(self, n=None): |
| 1723 | if n is None: |
| 1724 | n = -1 |
| 1725 | decoder = self._decoder or self._get_decoder() |
| 1726 | if n < 0: |
| 1727 | # Read everything. |
| 1728 | result = (self._get_decoded_chars() + |
| 1729 | decoder.decode(self.buffer.read(), final=True)) |
| 1730 | self._set_decoded_chars('') |
| 1731 | self._snapshot = None |
| 1732 | return result |
| 1733 | else: |
| 1734 | # Keep reading chunks until we have n characters to return. |
| 1735 | eof = False |
| 1736 | result = self._get_decoded_chars(n) |
| 1737 | while len(result) < n and not eof: |
| 1738 | eof = not self._read_chunk() |
| 1739 | result += self._get_decoded_chars(n - len(result)) |
| 1740 | return result |
| 1741 | |
| 1742 | def next(self): |
| 1743 | self._telling = False |
| 1744 | line = self.readline() |
| 1745 | if not line: |
| 1746 | self._snapshot = None |
| 1747 | self._telling = self._seekable |
| 1748 | raise StopIteration |
| 1749 | return line |
| 1750 | |
| 1751 | def readline(self, limit=None): |
Alexandre Vassalotti | 1aed624 | 2008-05-09 21:49:43 +0000 | [diff] [blame] | 1752 | if self.closed: |
| 1753 | raise ValueError("read from closed file") |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 1754 | if limit is None: |
| 1755 | limit = -1 |
Alexandre Vassalotti | 1aed624 | 2008-05-09 21:49:43 +0000 | [diff] [blame] | 1756 | if not isinstance(limit, (int, long)): |
| 1757 | raise TypeError("limit must be an integer") |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 1758 | |
| 1759 | # Grab all the decoded text (we will rewind any extra bits later). |
| 1760 | line = self._get_decoded_chars() |
| 1761 | |
| 1762 | start = 0 |
| 1763 | decoder = self._decoder or self._get_decoder() |
| 1764 | |
| 1765 | pos = endpos = None |
| 1766 | while True: |
| 1767 | if self._readtranslate: |
| 1768 | # Newlines are already translated, only search for \n |
| 1769 | pos = line.find('\n', start) |
| 1770 | if pos >= 0: |
| 1771 | endpos = pos + 1 |
| 1772 | break |
| 1773 | else: |
| 1774 | start = len(line) |
| 1775 | |
| 1776 | elif self._readuniversal: |
| 1777 | # Universal newline search. Find any of \r, \r\n, \n |
| 1778 | # The decoder ensures that \r\n are not split in two pieces |
| 1779 | |
| 1780 | # In C we'd look for these in parallel of course. |
| 1781 | nlpos = line.find("\n", start) |
| 1782 | crpos = line.find("\r", start) |
| 1783 | if crpos == -1: |
| 1784 | if nlpos == -1: |
| 1785 | # Nothing found |
| 1786 | start = len(line) |
| 1787 | else: |
| 1788 | # Found \n |
| 1789 | endpos = nlpos + 1 |
| 1790 | break |
| 1791 | elif nlpos == -1: |
| 1792 | # Found lone \r |
| 1793 | endpos = crpos + 1 |
| 1794 | break |
| 1795 | elif nlpos < crpos: |
| 1796 | # Found \n |
| 1797 | endpos = nlpos + 1 |
| 1798 | break |
| 1799 | elif nlpos == crpos + 1: |
| 1800 | # Found \r\n |
| 1801 | endpos = crpos + 2 |
| 1802 | break |
| 1803 | else: |
| 1804 | # Found \r |
| 1805 | endpos = crpos + 1 |
| 1806 | break |
| 1807 | else: |
| 1808 | # non-universal |
| 1809 | pos = line.find(self._readnl) |
| 1810 | if pos >= 0: |
| 1811 | endpos = pos + len(self._readnl) |
| 1812 | break |
| 1813 | |
| 1814 | if limit >= 0 and len(line) >= limit: |
| 1815 | endpos = limit # reached length limit |
| 1816 | break |
| 1817 | |
| 1818 | # No line ending seen yet - get more data |
| 1819 | more_line = '' |
| 1820 | while self._read_chunk(): |
| 1821 | if self._decoded_chars: |
| 1822 | break |
| 1823 | if self._decoded_chars: |
| 1824 | line += self._get_decoded_chars() |
| 1825 | else: |
| 1826 | # end of file |
| 1827 | self._set_decoded_chars('') |
| 1828 | self._snapshot = None |
| 1829 | return line |
| 1830 | |
| 1831 | if limit >= 0 and endpos > limit: |
| 1832 | endpos = limit # don't exceed limit |
| 1833 | |
| 1834 | # Rewind _decoded_chars to just after the line ending we found. |
| 1835 | self._rewind_decoded_chars(len(line) - endpos) |
| 1836 | return line[:endpos] |
| 1837 | |
| 1838 | @property |
| 1839 | def newlines(self): |
| 1840 | return self._decoder.newlines if self._decoder else None |
| 1841 | |
| 1842 | class StringIO(TextIOWrapper): |
| 1843 | |
Benjamin Peterson | 9ae080e | 2008-05-04 22:39:33 +0000 | [diff] [blame] | 1844 | """An in-memory stream for text. The initial_value argument sets the |
Benjamin Peterson | 7bb4d2d | 2008-04-13 02:01:27 +0000 | [diff] [blame] | 1845 | value of object. The other arguments are like those of TextIOWrapper's |
| 1846 | constructor. |
| 1847 | """ |
Christian Heimes | 1a6387e | 2008-03-26 12:49:49 +0000 | [diff] [blame] | 1848 | |
| 1849 | def __init__(self, initial_value="", encoding="utf-8", |
| 1850 | errors="strict", newline="\n"): |
| 1851 | super(StringIO, self).__init__(BytesIO(), |
| 1852 | encoding=encoding, |
| 1853 | errors=errors, |
| 1854 | newline=newline) |
| 1855 | if initial_value: |
| 1856 | if not isinstance(initial_value, unicode): |
| 1857 | initial_value = unicode(initial_value) |
| 1858 | self.write(initial_value) |
| 1859 | self.seek(0) |
| 1860 | |
| 1861 | def getvalue(self): |
| 1862 | self.flush() |
| 1863 | return self.buffer.getvalue().decode(self._encoding, self._errors) |