Antoine Pitrou | 37dc5f8 | 2011-04-03 17:05:46 +0200 | [diff] [blame] | 1 | """Interface to the libbzip2 compression library. |
| 2 | |
| 3 | This module provides a file interface, classes for incremental |
| 4 | (de)compression, and functions for one-shot (de)compression. |
| 5 | """ |
| 6 | |
| 7 | __all__ = ["BZ2File", "BZ2Compressor", "BZ2Decompressor", "compress", |
| 8 | "decompress"] |
| 9 | |
| 10 | __author__ = "Nadeem Vawda <nadeem.vawda@gmail.com>" |
| 11 | |
| 12 | import io |
| 13 | import threading |
| 14 | import warnings |
| 15 | |
| 16 | from _bz2 import BZ2Compressor, BZ2Decompressor |
| 17 | |
| 18 | |
| 19 | _MODE_CLOSED = 0 |
| 20 | _MODE_READ = 1 |
| 21 | _MODE_READ_EOF = 2 |
| 22 | _MODE_WRITE = 3 |
| 23 | |
| 24 | _BUFFER_SIZE = 8192 |
| 25 | |
| 26 | |
| 27 | class BZ2File(io.BufferedIOBase): |
| 28 | |
| 29 | """A file object providing transparent bzip2 (de)compression. |
| 30 | |
| 31 | A BZ2File can act as a wrapper for an existing file object, or refer |
| 32 | directly to a named file on disk. |
| 33 | |
| 34 | Note that BZ2File provides a *binary* file interface - data read is |
| 35 | returned as bytes, and data to be written should be given as bytes. |
| 36 | """ |
| 37 | |
| 38 | def __init__(self, filename=None, mode="r", buffering=None, |
| 39 | compresslevel=9, fileobj=None): |
| 40 | """Open a bzip2-compressed file. |
| 41 | |
| 42 | If filename is given, open the named file. Otherwise, operate on |
| 43 | the file object given by fileobj. Exactly one of these two |
| 44 | parameters should be provided. |
| 45 | |
| 46 | mode can be 'r' for reading (default), or 'w' for writing. |
| 47 | |
| 48 | buffering is ignored. Its use is deprecated. |
| 49 | |
| 50 | If mode is 'w', compresslevel can be a number between 1 and 9 |
| 51 | specifying the level of compression: 1 produces the least |
| 52 | compression, and 9 (default) produces the most compression. |
| 53 | """ |
| 54 | # This lock must be recursive, so that BufferedIOBase's |
| 55 | # readline(), readlines() and writelines() don't deadlock. |
| 56 | self._lock = threading.RLock() |
| 57 | self._fp = None |
| 58 | self._closefp = False |
| 59 | self._mode = _MODE_CLOSED |
| 60 | self._pos = 0 |
| 61 | self._size = -1 |
| 62 | |
| 63 | if buffering is not None: |
| 64 | warnings.warn("Use of 'buffering' argument is deprecated", |
| 65 | DeprecationWarning) |
| 66 | |
| 67 | if not (1 <= compresslevel <= 9): |
| 68 | raise ValueError("compresslevel must be between 1 and 9") |
| 69 | |
| 70 | if mode in ("", "r", "rb"): |
| 71 | mode = "rb" |
| 72 | mode_code = _MODE_READ |
| 73 | self._decompressor = BZ2Decompressor() |
| 74 | self._buffer = None |
| 75 | elif mode in ("w", "wb"): |
| 76 | mode = "wb" |
| 77 | mode_code = _MODE_WRITE |
Nadeem Vawda | 249ab5e | 2011-09-11 22:38:11 +0200 | [diff] [blame] | 78 | self._compressor = BZ2Compressor(compresslevel) |
Nadeem Vawda | 55b4338 | 2011-05-27 01:52:15 +0200 | [diff] [blame] | 79 | elif mode in ("a", "ab"): |
| 80 | mode = "ab" |
| 81 | mode_code = _MODE_WRITE |
Nadeem Vawda | 249ab5e | 2011-09-11 22:38:11 +0200 | [diff] [blame] | 82 | self._compressor = BZ2Compressor(compresslevel) |
Antoine Pitrou | 37dc5f8 | 2011-04-03 17:05:46 +0200 | [diff] [blame] | 83 | else: |
| 84 | raise ValueError("Invalid mode: {!r}".format(mode)) |
| 85 | |
| 86 | if filename is not None and fileobj is None: |
| 87 | self._fp = open(filename, mode) |
| 88 | self._closefp = True |
| 89 | self._mode = mode_code |
| 90 | elif fileobj is not None and filename is None: |
| 91 | self._fp = fileobj |
| 92 | self._mode = mode_code |
| 93 | else: |
| 94 | raise ValueError("Must give exactly one of filename and fileobj") |
| 95 | |
| 96 | def close(self): |
| 97 | """Flush and close the file. |
| 98 | |
| 99 | May be called more than once without error. Once the file is |
| 100 | closed, any other operation on it will raise a ValueError. |
| 101 | """ |
| 102 | with self._lock: |
| 103 | if self._mode == _MODE_CLOSED: |
| 104 | return |
| 105 | try: |
| 106 | if self._mode in (_MODE_READ, _MODE_READ_EOF): |
| 107 | self._decompressor = None |
| 108 | elif self._mode == _MODE_WRITE: |
| 109 | self._fp.write(self._compressor.flush()) |
| 110 | self._compressor = None |
| 111 | finally: |
Antoine Pitrou | 24ce386 | 2011-04-03 17:08:49 +0200 | [diff] [blame] | 112 | try: |
Antoine Pitrou | 37dc5f8 | 2011-04-03 17:05:46 +0200 | [diff] [blame] | 113 | if self._closefp: |
| 114 | self._fp.close() |
| 115 | finally: |
| 116 | self._fp = None |
| 117 | self._closefp = False |
| 118 | self._mode = _MODE_CLOSED |
| 119 | self._buffer = None |
| 120 | |
| 121 | @property |
| 122 | def closed(self): |
| 123 | """True if this file is closed.""" |
| 124 | return self._mode == _MODE_CLOSED |
| 125 | |
| 126 | def fileno(self): |
| 127 | """Return the file descriptor for the underlying file.""" |
Nadeem Vawda | 44ae4a2 | 2011-11-30 17:39:30 +0200 | [diff] [blame] | 128 | self._check_not_closed() |
Antoine Pitrou | 37dc5f8 | 2011-04-03 17:05:46 +0200 | [diff] [blame] | 129 | return self._fp.fileno() |
| 130 | |
| 131 | def seekable(self): |
| 132 | """Return whether the file supports seeking.""" |
| 133 | return self.readable() |
| 134 | |
| 135 | def readable(self): |
| 136 | """Return whether the file was opened for reading.""" |
Nadeem Vawda | 44ae4a2 | 2011-11-30 17:39:30 +0200 | [diff] [blame] | 137 | self._check_not_closed() |
Antoine Pitrou | 37dc5f8 | 2011-04-03 17:05:46 +0200 | [diff] [blame] | 138 | return self._mode in (_MODE_READ, _MODE_READ_EOF) |
| 139 | |
| 140 | def writable(self): |
| 141 | """Return whether the file was opened for writing.""" |
Nadeem Vawda | 44ae4a2 | 2011-11-30 17:39:30 +0200 | [diff] [blame] | 142 | self._check_not_closed() |
Antoine Pitrou | 37dc5f8 | 2011-04-03 17:05:46 +0200 | [diff] [blame] | 143 | return self._mode == _MODE_WRITE |
| 144 | |
| 145 | # Mode-checking helper functions. |
| 146 | |
| 147 | def _check_not_closed(self): |
| 148 | if self.closed: |
| 149 | raise ValueError("I/O operation on closed file") |
| 150 | |
| 151 | def _check_can_read(self): |
| 152 | if not self.readable(): |
Antoine Pitrou | 37dc5f8 | 2011-04-03 17:05:46 +0200 | [diff] [blame] | 153 | raise io.UnsupportedOperation("File not open for reading") |
| 154 | |
| 155 | def _check_can_write(self): |
| 156 | if not self.writable(): |
Antoine Pitrou | 37dc5f8 | 2011-04-03 17:05:46 +0200 | [diff] [blame] | 157 | raise io.UnsupportedOperation("File not open for writing") |
| 158 | |
| 159 | def _check_can_seek(self): |
| 160 | if not self.seekable(): |
Antoine Pitrou | 37dc5f8 | 2011-04-03 17:05:46 +0200 | [diff] [blame] | 161 | raise io.UnsupportedOperation("Seeking is only supported " |
Nadeem Vawda | f1a1af2 | 2011-05-25 00:32:08 +0200 | [diff] [blame] | 162 | "on files open for reading") |
Antoine Pitrou | 37dc5f8 | 2011-04-03 17:05:46 +0200 | [diff] [blame] | 163 | |
| 164 | # Fill the readahead buffer if it is empty. Returns False on EOF. |
| 165 | def _fill_buffer(self): |
| 166 | if self._buffer: |
| 167 | return True |
Nadeem Vawda | 55b4338 | 2011-05-27 01:52:15 +0200 | [diff] [blame] | 168 | |
| 169 | if self._decompressor.unused_data: |
| 170 | rawblock = self._decompressor.unused_data |
| 171 | else: |
| 172 | rawblock = self._fp.read(_BUFFER_SIZE) |
| 173 | |
Antoine Pitrou | 37dc5f8 | 2011-04-03 17:05:46 +0200 | [diff] [blame] | 174 | if not rawblock: |
Nadeem Vawda | 55b4338 | 2011-05-27 01:52:15 +0200 | [diff] [blame] | 175 | if self._decompressor.eof: |
| 176 | self._mode = _MODE_READ_EOF |
| 177 | self._size = self._pos |
| 178 | return False |
| 179 | else: |
| 180 | raise EOFError("Compressed file ended before the " |
| 181 | "end-of-stream marker was reached") |
| 182 | |
| 183 | # Continue to next stream. |
| 184 | if self._decompressor.eof: |
| 185 | self._decompressor = BZ2Decompressor() |
| 186 | |
Antoine Pitrou | 37dc5f8 | 2011-04-03 17:05:46 +0200 | [diff] [blame] | 187 | self._buffer = self._decompressor.decompress(rawblock) |
| 188 | return True |
| 189 | |
| 190 | # Read data until EOF. |
| 191 | # If return_data is false, consume the data without returning it. |
| 192 | def _read_all(self, return_data=True): |
| 193 | blocks = [] |
| 194 | while self._fill_buffer(): |
| 195 | if return_data: |
| 196 | blocks.append(self._buffer) |
| 197 | self._pos += len(self._buffer) |
| 198 | self._buffer = None |
| 199 | if return_data: |
| 200 | return b"".join(blocks) |
| 201 | |
| 202 | # Read a block of up to n bytes. |
| 203 | # If return_data is false, consume the data without returning it. |
| 204 | def _read_block(self, n, return_data=True): |
| 205 | blocks = [] |
| 206 | while n > 0 and self._fill_buffer(): |
| 207 | if n < len(self._buffer): |
| 208 | data = self._buffer[:n] |
| 209 | self._buffer = self._buffer[n:] |
| 210 | else: |
| 211 | data = self._buffer |
| 212 | self._buffer = None |
| 213 | if return_data: |
| 214 | blocks.append(data) |
| 215 | self._pos += len(data) |
| 216 | n -= len(data) |
| 217 | if return_data: |
| 218 | return b"".join(blocks) |
| 219 | |
| 220 | def peek(self, n=0): |
| 221 | """Return buffered data without advancing the file position. |
| 222 | |
| 223 | Always returns at least one byte of data, unless at EOF. |
| 224 | The exact number of bytes returned is unspecified. |
| 225 | """ |
| 226 | with self._lock: |
| 227 | self._check_can_read() |
| 228 | if self._mode == _MODE_READ_EOF or not self._fill_buffer(): |
| 229 | return b"" |
| 230 | return self._buffer |
| 231 | |
| 232 | def read(self, size=-1): |
| 233 | """Read up to size uncompressed bytes from the file. |
| 234 | |
| 235 | If size is negative or omitted, read until EOF is reached. |
| 236 | Returns b'' if the file is already at EOF. |
| 237 | """ |
| 238 | with self._lock: |
| 239 | self._check_can_read() |
| 240 | if self._mode == _MODE_READ_EOF or size == 0: |
| 241 | return b"" |
| 242 | elif size < 0: |
| 243 | return self._read_all() |
| 244 | else: |
| 245 | return self._read_block(size) |
| 246 | |
| 247 | def read1(self, size=-1): |
| 248 | """Read up to size uncompressed bytes with at most one read |
| 249 | from the underlying stream. |
| 250 | |
| 251 | Returns b'' if the file is at EOF. |
| 252 | """ |
| 253 | with self._lock: |
| 254 | self._check_can_read() |
| 255 | if (size == 0 or self._mode == _MODE_READ_EOF or |
| 256 | not self._fill_buffer()): |
| 257 | return b"" |
| 258 | if 0 < size < len(self._buffer): |
| 259 | data = self._buffer[:size] |
| 260 | self._buffer = self._buffer[size:] |
| 261 | else: |
| 262 | data = self._buffer |
| 263 | self._buffer = None |
| 264 | self._pos += len(data) |
| 265 | return data |
| 266 | |
| 267 | def readinto(self, b): |
| 268 | """Read up to len(b) bytes into b. |
Antoine Pitrou | 24ce386 | 2011-04-03 17:08:49 +0200 | [diff] [blame] | 269 | |
Antoine Pitrou | 37dc5f8 | 2011-04-03 17:05:46 +0200 | [diff] [blame] | 270 | Returns the number of bytes read (0 for EOF). |
| 271 | """ |
| 272 | with self._lock: |
| 273 | return io.BufferedIOBase.readinto(self, b) |
| 274 | |
| 275 | def readline(self, size=-1): |
| 276 | """Read a line of uncompressed bytes from the file. |
| 277 | |
| 278 | The terminating newline (if present) is retained. If size is |
| 279 | non-negative, no more than size bytes will be read (in which |
| 280 | case the line may be incomplete). Returns b'' if already at EOF. |
| 281 | """ |
| 282 | if not hasattr(size, "__index__"): |
| 283 | raise TypeError("Integer argument expected") |
| 284 | size = size.__index__() |
| 285 | with self._lock: |
| 286 | return io.BufferedIOBase.readline(self, size) |
| 287 | |
| 288 | def readlines(self, size=-1): |
| 289 | """Read a list of lines of uncompressed bytes from the file. |
| 290 | |
| 291 | size can be specified to control the number of lines read: no |
| 292 | further lines will be read once the total size of the lines read |
| 293 | so far equals or exceeds size. |
| 294 | """ |
| 295 | if not hasattr(size, "__index__"): |
| 296 | raise TypeError("Integer argument expected") |
| 297 | size = size.__index__() |
| 298 | with self._lock: |
| 299 | return io.BufferedIOBase.readlines(self, size) |
| 300 | |
| 301 | def write(self, data): |
| 302 | """Write a byte string to the file. |
| 303 | |
| 304 | Returns the number of uncompressed bytes written, which is |
| 305 | always len(data). Note that due to buffering, the file on disk |
| 306 | may not reflect the data written until close() is called. |
| 307 | """ |
| 308 | with self._lock: |
| 309 | self._check_can_write() |
| 310 | compressed = self._compressor.compress(data) |
| 311 | self._fp.write(compressed) |
| 312 | self._pos += len(data) |
| 313 | return len(data) |
| 314 | |
| 315 | def writelines(self, seq): |
| 316 | """Write a sequence of byte strings to the file. |
| 317 | |
| 318 | Returns the number of uncompressed bytes written. |
| 319 | seq can be any iterable yielding byte strings. |
| 320 | |
| 321 | Line separators are not added between the written byte strings. |
| 322 | """ |
| 323 | with self._lock: |
| 324 | return io.BufferedIOBase.writelines(self, seq) |
| 325 | |
| 326 | # Rewind the file to the beginning of the data stream. |
| 327 | def _rewind(self): |
| 328 | self._fp.seek(0, 0) |
| 329 | self._mode = _MODE_READ |
| 330 | self._pos = 0 |
| 331 | self._decompressor = BZ2Decompressor() |
| 332 | self._buffer = None |
| 333 | |
| 334 | def seek(self, offset, whence=0): |
| 335 | """Change the file position. |
| 336 | |
| 337 | The new position is specified by offset, relative to the |
| 338 | position indicated by whence. Values for whence are: |
| 339 | |
| 340 | 0: start of stream (default); offset must not be negative |
| 341 | 1: current stream position |
| 342 | 2: end of stream; offset must not be positive |
| 343 | |
| 344 | Returns the new file position. |
| 345 | |
| 346 | Note that seeking is emulated, so depending on the parameters, |
| 347 | this operation may be extremely slow. |
| 348 | """ |
| 349 | with self._lock: |
| 350 | self._check_can_seek() |
| 351 | |
| 352 | # Recalculate offset as an absolute file position. |
| 353 | if whence == 0: |
| 354 | pass |
| 355 | elif whence == 1: |
| 356 | offset = self._pos + offset |
| 357 | elif whence == 2: |
| 358 | # Seeking relative to EOF - we need to know the file's size. |
| 359 | if self._size < 0: |
| 360 | self._read_all(return_data=False) |
| 361 | offset = self._size + offset |
| 362 | else: |
| 363 | raise ValueError("Invalid value for whence: {}".format(whence)) |
| 364 | |
| 365 | # Make it so that offset is the number of bytes to skip forward. |
| 366 | if offset < self._pos: |
| 367 | self._rewind() |
| 368 | else: |
| 369 | offset -= self._pos |
| 370 | |
| 371 | # Read and discard data until we reach the desired position. |
| 372 | if self._mode != _MODE_READ_EOF: |
| 373 | self._read_block(offset, return_data=False) |
| 374 | |
| 375 | return self._pos |
| 376 | |
| 377 | def tell(self): |
| 378 | """Return the current file position.""" |
| 379 | with self._lock: |
| 380 | self._check_not_closed() |
| 381 | return self._pos |
| 382 | |
| 383 | |
| 384 | def compress(data, compresslevel=9): |
| 385 | """Compress a block of data. |
| 386 | |
| 387 | compresslevel, if given, must be a number between 1 and 9. |
| 388 | |
| 389 | For incremental compression, use a BZ2Compressor object instead. |
| 390 | """ |
| 391 | comp = BZ2Compressor(compresslevel) |
| 392 | return comp.compress(data) + comp.flush() |
| 393 | |
| 394 | |
| 395 | def decompress(data): |
| 396 | """Decompress a block of data. |
| 397 | |
| 398 | For incremental decompression, use a BZ2Decompressor object instead. |
| 399 | """ |
| 400 | if len(data) == 0: |
| 401 | return b"" |
Nadeem Vawda | 55b4338 | 2011-05-27 01:52:15 +0200 | [diff] [blame] | 402 | |
Nadeem Vawda | 98838ba | 2011-05-30 01:12:24 +0200 | [diff] [blame] | 403 | results = [] |
Nadeem Vawda | 55b4338 | 2011-05-27 01:52:15 +0200 | [diff] [blame] | 404 | while True: |
| 405 | decomp = BZ2Decompressor() |
Nadeem Vawda | 98838ba | 2011-05-30 01:12:24 +0200 | [diff] [blame] | 406 | results.append(decomp.decompress(data)) |
Nadeem Vawda | 55b4338 | 2011-05-27 01:52:15 +0200 | [diff] [blame] | 407 | if not decomp.eof: |
| 408 | raise ValueError("Compressed data ended before the " |
| 409 | "end-of-stream marker was reached") |
| 410 | if not decomp.unused_data: |
Nadeem Vawda | 98838ba | 2011-05-30 01:12:24 +0200 | [diff] [blame] | 411 | return b"".join(results) |
Nadeem Vawda | 55b4338 | 2011-05-27 01:52:15 +0200 | [diff] [blame] | 412 | # There is unused data left over. Proceed to next stream. |
| 413 | data = decomp.unused_data |