Eric Snow | 32439d6 | 2015-05-02 19:15:18 -0600 | [diff] [blame] | 1 | """Core implementation of path-based import. |
| 2 | |
| 3 | This module is NOT meant to be directly imported! It has been designed such |
| 4 | that it can be bootstrapped into Python as the implementation of import. As |
| 5 | such it requires the injection of specific modules and attributes in order to |
| 6 | work. One should use importlib as the public-facing version of this module. |
| 7 | |
| 8 | """ |
| 9 | # |
| 10 | # IMPORTANT: Whenever making changes to this module, be sure to run |
| 11 | # a top-level make in order to get the frozen version of the module |
| 12 | # updated. Not doing so will result in the Makefile to fail for |
| 13 | # all others who don't have a ./python around to freeze the module |
| 14 | # in the early stages of compilation. |
| 15 | # |
| 16 | |
| 17 | # See importlib._setup() for what is injected into the global namespace. |
| 18 | |
| 19 | # When editing this code be aware that code executed at import time CANNOT |
| 20 | # reference any injected objects! This includes not only global code but also |
| 21 | # anything specified at the class level. |
| 22 | |
| 23 | # Bootstrap-related code ###################################################### |
| 24 | |
| 25 | _CASE_INSENSITIVE_PLATFORMS = 'win', 'cygwin', 'darwin' |
| 26 | |
| 27 | |
| 28 | def _make_relax_case(): |
| 29 | if sys.platform.startswith(_CASE_INSENSITIVE_PLATFORMS): |
| 30 | def _relax_case(): |
| 31 | """True if filenames must be checked case-insensitively.""" |
| 32 | return b'PYTHONCASEOK' in _os.environ |
| 33 | else: |
| 34 | def _relax_case(): |
| 35 | """True if filenames must be checked case-insensitively.""" |
| 36 | return False |
| 37 | return _relax_case |
| 38 | |
| 39 | |
| 40 | def _w_long(x): |
| 41 | """Convert a 32-bit integer to little-endian.""" |
| 42 | return (int(x) & 0xFFFFFFFF).to_bytes(4, 'little') |
| 43 | |
| 44 | |
| 45 | def _r_long(int_bytes): |
| 46 | """Convert 4 bytes in little-endian to an integer.""" |
| 47 | return int.from_bytes(int_bytes, 'little') |
| 48 | |
| 49 | |
| 50 | def _path_join(*path_parts): |
| 51 | """Replacement for os.path.join().""" |
| 52 | return path_sep.join([part.rstrip(path_separators) |
| 53 | for part in path_parts if part]) |
| 54 | |
| 55 | |
| 56 | def _path_split(path): |
| 57 | """Replacement for os.path.split().""" |
| 58 | if len(path_separators) == 1: |
| 59 | front, _, tail = path.rpartition(path_sep) |
| 60 | return front, tail |
| 61 | for x in reversed(path): |
| 62 | if x in path_separators: |
| 63 | front, tail = path.rsplit(x, maxsplit=1) |
| 64 | return front, tail |
| 65 | return '', path |
| 66 | |
| 67 | |
| 68 | def _path_stat(path): |
| 69 | """Stat the path. |
| 70 | |
| 71 | Made a separate function to make it easier to override in experiments |
| 72 | (e.g. cache stat results). |
| 73 | |
| 74 | """ |
| 75 | return _os.stat(path) |
| 76 | |
| 77 | |
| 78 | def _path_is_mode_type(path, mode): |
| 79 | """Test whether the path is the specified mode type.""" |
| 80 | try: |
| 81 | stat_info = _path_stat(path) |
| 82 | except OSError: |
| 83 | return False |
| 84 | return (stat_info.st_mode & 0o170000) == mode |
| 85 | |
| 86 | |
| 87 | def _path_isfile(path): |
| 88 | """Replacement for os.path.isfile.""" |
| 89 | return _path_is_mode_type(path, 0o100000) |
| 90 | |
| 91 | |
| 92 | def _path_isdir(path): |
| 93 | """Replacement for os.path.isdir.""" |
| 94 | if not path: |
| 95 | path = _os.getcwd() |
| 96 | return _path_is_mode_type(path, 0o040000) |
| 97 | |
| 98 | |
| 99 | def _write_atomic(path, data, mode=0o666): |
| 100 | """Best-effort function to write data to a path atomically. |
| 101 | Be prepared to handle a FileExistsError if concurrent writing of the |
| 102 | temporary file is attempted.""" |
| 103 | # id() is used to generate a pseudo-random filename. |
| 104 | path_tmp = '{}.{}'.format(path, id(path)) |
| 105 | fd = _os.open(path_tmp, |
| 106 | _os.O_EXCL | _os.O_CREAT | _os.O_WRONLY, mode & 0o666) |
| 107 | try: |
| 108 | # We first write data to a temporary file, and then use os.replace() to |
| 109 | # perform an atomic rename. |
| 110 | with _io.FileIO(fd, 'wb') as file: |
| 111 | file.write(data) |
| 112 | _os.replace(path_tmp, path) |
| 113 | except OSError: |
| 114 | try: |
| 115 | _os.unlink(path_tmp) |
| 116 | except OSError: |
| 117 | pass |
| 118 | raise |
| 119 | |
| 120 | |
| 121 | _code_type = type(_write_atomic.__code__) |
| 122 | |
| 123 | |
| 124 | # Finder/loader utility code ############################################### |
| 125 | |
| 126 | # Magic word to reject .pyc files generated by other Python versions. |
| 127 | # It should change for each incompatible change to the bytecode. |
| 128 | # |
| 129 | # The value of CR and LF is incorporated so if you ever read or write |
| 130 | # a .pyc file in text mode the magic number will be wrong; also, the |
| 131 | # Apple MPW compiler swaps their values, botching string constants. |
| 132 | # |
| 133 | # The magic numbers must be spaced apart at least 2 values, as the |
| 134 | # -U interpeter flag will cause MAGIC+1 being used. They have been |
| 135 | # odd numbers for some time now. |
| 136 | # |
| 137 | # There were a variety of old schemes for setting the magic number. |
| 138 | # The current working scheme is to increment the previous value by |
| 139 | # 10. |
| 140 | # |
| 141 | # Starting with the adoption of PEP 3147 in Python 3.2, every bump in magic |
| 142 | # number also includes a new "magic tag", i.e. a human readable string used |
| 143 | # to represent the magic number in __pycache__ directories. When you change |
| 144 | # the magic number, you must also set a new unique magic tag. Generally this |
| 145 | # can be named after the Python major version of the magic number bump, but |
| 146 | # it can really be anything, as long as it's different than anything else |
| 147 | # that's come before. The tags are included in the following table, starting |
| 148 | # with Python 3.2a0. |
| 149 | # |
| 150 | # Known values: |
| 151 | # Python 1.5: 20121 |
| 152 | # Python 1.5.1: 20121 |
| 153 | # Python 1.5.2: 20121 |
| 154 | # Python 1.6: 50428 |
| 155 | # Python 2.0: 50823 |
| 156 | # Python 2.0.1: 50823 |
| 157 | # Python 2.1: 60202 |
| 158 | # Python 2.1.1: 60202 |
| 159 | # Python 2.1.2: 60202 |
| 160 | # Python 2.2: 60717 |
| 161 | # Python 2.3a0: 62011 |
| 162 | # Python 2.3a0: 62021 |
| 163 | # Python 2.3a0: 62011 (!) |
| 164 | # Python 2.4a0: 62041 |
| 165 | # Python 2.4a3: 62051 |
| 166 | # Python 2.4b1: 62061 |
| 167 | # Python 2.5a0: 62071 |
| 168 | # Python 2.5a0: 62081 (ast-branch) |
| 169 | # Python 2.5a0: 62091 (with) |
| 170 | # Python 2.5a0: 62092 (changed WITH_CLEANUP opcode) |
| 171 | # Python 2.5b3: 62101 (fix wrong code: for x, in ...) |
| 172 | # Python 2.5b3: 62111 (fix wrong code: x += yield) |
| 173 | # Python 2.5c1: 62121 (fix wrong lnotab with for loops and |
| 174 | # storing constants that should have been removed) |
| 175 | # Python 2.5c2: 62131 (fix wrong code: for x, in ... in listcomp/genexp) |
| 176 | # Python 2.6a0: 62151 (peephole optimizations and STORE_MAP opcode) |
| 177 | # Python 2.6a1: 62161 (WITH_CLEANUP optimization) |
| 178 | # Python 2.7a0: 62171 (optimize list comprehensions/change LIST_APPEND) |
| 179 | # Python 2.7a0: 62181 (optimize conditional branches: |
| 180 | # introduce POP_JUMP_IF_FALSE and POP_JUMP_IF_TRUE) |
| 181 | # Python 2.7a0 62191 (introduce SETUP_WITH) |
| 182 | # Python 2.7a0 62201 (introduce BUILD_SET) |
| 183 | # Python 2.7a0 62211 (introduce MAP_ADD and SET_ADD) |
| 184 | # Python 3000: 3000 |
| 185 | # 3010 (removed UNARY_CONVERT) |
| 186 | # 3020 (added BUILD_SET) |
| 187 | # 3030 (added keyword-only parameters) |
| 188 | # 3040 (added signature annotations) |
| 189 | # 3050 (print becomes a function) |
| 190 | # 3060 (PEP 3115 metaclass syntax) |
| 191 | # 3061 (string literals become unicode) |
| 192 | # 3071 (PEP 3109 raise changes) |
| 193 | # 3081 (PEP 3137 make __file__ and __name__ unicode) |
| 194 | # 3091 (kill str8 interning) |
| 195 | # 3101 (merge from 2.6a0, see 62151) |
| 196 | # 3103 (__file__ points to source file) |
| 197 | # Python 3.0a4: 3111 (WITH_CLEANUP optimization). |
| 198 | # Python 3.0a5: 3131 (lexical exception stacking, including POP_EXCEPT) |
| 199 | # Python 3.1a0: 3141 (optimize list, set and dict comprehensions: |
| 200 | # change LIST_APPEND and SET_ADD, add MAP_ADD) |
| 201 | # Python 3.1a0: 3151 (optimize conditional branches: |
| 202 | # introduce POP_JUMP_IF_FALSE and POP_JUMP_IF_TRUE) |
| 203 | # Python 3.2a0: 3160 (add SETUP_WITH) |
| 204 | # tag: cpython-32 |
| 205 | # Python 3.2a1: 3170 (add DUP_TOP_TWO, remove DUP_TOPX and ROT_FOUR) |
| 206 | # tag: cpython-32 |
| 207 | # Python 3.2a2 3180 (add DELETE_DEREF) |
| 208 | # Python 3.3a0 3190 __class__ super closure changed |
| 209 | # Python 3.3a0 3200 (__qualname__ added) |
| 210 | # 3210 (added size modulo 2**32 to the pyc header) |
| 211 | # Python 3.3a1 3220 (changed PEP 380 implementation) |
| 212 | # Python 3.3a4 3230 (revert changes to implicit __class__ closure) |
| 213 | # Python 3.4a1 3250 (evaluate positional default arguments before |
| 214 | # keyword-only defaults) |
| 215 | # Python 3.4a1 3260 (add LOAD_CLASSDEREF; allow locals of class to override |
| 216 | # free vars) |
| 217 | # Python 3.4a1 3270 (various tweaks to the __class__ closure) |
| 218 | # Python 3.4a1 3280 (remove implicit class argument) |
| 219 | # Python 3.4a4 3290 (changes to __qualname__ computation) |
| 220 | # Python 3.4a4 3300 (more changes to __qualname__ computation) |
| 221 | # Python 3.4rc2 3310 (alter __qualname__ computation) |
| 222 | # Python 3.5a0 3320 (matrix multiplication operator) |
Benjamin Peterson | 025e9eb | 2015-05-05 20:16:41 -0400 | [diff] [blame] | 223 | # Python 3.5b1 3330 (PEP 448: Additional Unpacking Generalizations) |
Benjamin Peterson | ee85339 | 2015-05-28 14:30:26 -0500 | [diff] [blame] | 224 | # Python 3.5b2 3340 (fix dictionary display evaluation order #11205) |
Yury Selivanov | 5376ba9 | 2015-06-22 12:19:30 -0400 | [diff] [blame] | 225 | # Python 3.5b2 3350 (add GET_YIELD_FROM_ITER opcode #24400) |
Eric Snow | 32439d6 | 2015-05-02 19:15:18 -0600 | [diff] [blame] | 226 | # |
| 227 | # MAGIC must change whenever the bytecode emitted by the compiler may no |
| 228 | # longer be understood by older implementations of the eval loop (usually |
| 229 | # due to the addition of new opcodes). |
| 230 | |
Yury Selivanov | 5376ba9 | 2015-06-22 12:19:30 -0400 | [diff] [blame] | 231 | MAGIC_NUMBER = (3350).to_bytes(2, 'little') + b'\r\n' |
Eric Snow | 32439d6 | 2015-05-02 19:15:18 -0600 | [diff] [blame] | 232 | _RAW_MAGIC_NUMBER = int.from_bytes(MAGIC_NUMBER, 'little') # For import.c |
| 233 | |
| 234 | _PYCACHE = '__pycache__' |
| 235 | _OPT = 'opt-' |
| 236 | |
| 237 | SOURCE_SUFFIXES = ['.py'] # _setup() adds .pyw as needed. |
| 238 | |
| 239 | BYTECODE_SUFFIXES = ['.pyc'] |
| 240 | # Deprecated. |
| 241 | DEBUG_BYTECODE_SUFFIXES = OPTIMIZED_BYTECODE_SUFFIXES = BYTECODE_SUFFIXES |
| 242 | |
| 243 | def cache_from_source(path, debug_override=None, *, optimization=None): |
| 244 | """Given the path to a .py file, return the path to its .pyc file. |
| 245 | |
| 246 | The .py file does not need to exist; this simply returns the path to the |
| 247 | .pyc file calculated as if the .py file were imported. |
| 248 | |
| 249 | The 'optimization' parameter controls the presumed optimization level of |
| 250 | the bytecode file. If 'optimization' is not None, the string representation |
| 251 | of the argument is taken and verified to be alphanumeric (else ValueError |
| 252 | is raised). |
| 253 | |
| 254 | The debug_override parameter is deprecated. If debug_override is not None, |
| 255 | a True value is the same as setting 'optimization' to the empty string |
| 256 | while a False value is equivalent to setting 'optimization' to '1'. |
| 257 | |
| 258 | If sys.implementation.cache_tag is None then NotImplementedError is raised. |
| 259 | |
| 260 | """ |
| 261 | if debug_override is not None: |
| 262 | _warnings.warn('the debug_override parameter is deprecated; use ' |
| 263 | "'optimization' instead", DeprecationWarning) |
| 264 | if optimization is not None: |
| 265 | message = 'debug_override or optimization must be set to None' |
| 266 | raise TypeError(message) |
| 267 | optimization = '' if debug_override else 1 |
| 268 | head, tail = _path_split(path) |
| 269 | base, sep, rest = tail.rpartition('.') |
| 270 | tag = sys.implementation.cache_tag |
| 271 | if tag is None: |
| 272 | raise NotImplementedError('sys.implementation.cache_tag is None') |
| 273 | almost_filename = ''.join([(base if base else rest), sep, tag]) |
| 274 | if optimization is None: |
| 275 | if sys.flags.optimize == 0: |
| 276 | optimization = '' |
| 277 | else: |
| 278 | optimization = sys.flags.optimize |
| 279 | optimization = str(optimization) |
| 280 | if optimization != '': |
| 281 | if not optimization.isalnum(): |
| 282 | raise ValueError('{!r} is not alphanumeric'.format(optimization)) |
| 283 | almost_filename = '{}.{}{}'.format(almost_filename, _OPT, optimization) |
| 284 | return _path_join(head, _PYCACHE, almost_filename + BYTECODE_SUFFIXES[0]) |
| 285 | |
| 286 | |
| 287 | def source_from_cache(path): |
| 288 | """Given the path to a .pyc. file, return the path to its .py file. |
| 289 | |
| 290 | The .pyc file does not need to exist; this simply returns the path to |
| 291 | the .py file calculated to correspond to the .pyc file. If path does |
| 292 | not conform to PEP 3147/488 format, ValueError will be raised. If |
| 293 | sys.implementation.cache_tag is None then NotImplementedError is raised. |
| 294 | |
| 295 | """ |
| 296 | if sys.implementation.cache_tag is None: |
| 297 | raise NotImplementedError('sys.implementation.cache_tag is None') |
| 298 | head, pycache_filename = _path_split(path) |
| 299 | head, pycache = _path_split(head) |
| 300 | if pycache != _PYCACHE: |
| 301 | raise ValueError('{} not bottom-level directory in ' |
| 302 | '{!r}'.format(_PYCACHE, path)) |
| 303 | dot_count = pycache_filename.count('.') |
| 304 | if dot_count not in {2, 3}: |
| 305 | raise ValueError('expected only 2 or 3 dots in ' |
| 306 | '{!r}'.format(pycache_filename)) |
| 307 | elif dot_count == 3: |
| 308 | optimization = pycache_filename.rsplit('.', 2)[-2] |
| 309 | if not optimization.startswith(_OPT): |
| 310 | raise ValueError("optimization portion of filename does not start " |
| 311 | "with {!r}".format(_OPT)) |
| 312 | opt_level = optimization[len(_OPT):] |
| 313 | if not opt_level.isalnum(): |
| 314 | raise ValueError("optimization level {!r} is not an alphanumeric " |
| 315 | "value".format(optimization)) |
| 316 | base_filename = pycache_filename.partition('.')[0] |
| 317 | return _path_join(head, base_filename + SOURCE_SUFFIXES[0]) |
| 318 | |
| 319 | |
| 320 | def _get_sourcefile(bytecode_path): |
| 321 | """Convert a bytecode file path to a source path (if possible). |
| 322 | |
| 323 | This function exists purely for backwards-compatibility for |
| 324 | PyImport_ExecCodeModuleWithFilenames() in the C API. |
| 325 | |
| 326 | """ |
| 327 | if len(bytecode_path) == 0: |
| 328 | return None |
| 329 | rest, _, extension = bytecode_path.rpartition('.') |
| 330 | if not rest or extension.lower()[-3:-1] != 'py': |
| 331 | return bytecode_path |
| 332 | try: |
| 333 | source_path = source_from_cache(bytecode_path) |
| 334 | except (NotImplementedError, ValueError): |
| 335 | source_path = bytecode_path[:-1] |
| 336 | return source_path if _path_isfile(source_path) else bytecode_path |
| 337 | |
| 338 | |
| 339 | def _get_cached(filename): |
| 340 | if filename.endswith(tuple(SOURCE_SUFFIXES)): |
| 341 | try: |
| 342 | return cache_from_source(filename) |
| 343 | except NotImplementedError: |
| 344 | pass |
| 345 | elif filename.endswith(tuple(BYTECODE_SUFFIXES)): |
| 346 | return filename |
| 347 | else: |
| 348 | return None |
| 349 | |
| 350 | |
| 351 | def _calc_mode(path): |
| 352 | """Calculate the mode permissions for a bytecode file.""" |
| 353 | try: |
| 354 | mode = _path_stat(path).st_mode |
| 355 | except OSError: |
| 356 | mode = 0o666 |
| 357 | # We always ensure write access so we can update cached files |
| 358 | # later even when the source files are read-only on Windows (#6074) |
| 359 | mode |= 0o200 |
| 360 | return mode |
| 361 | |
| 362 | |
| 363 | def _verbose_message(message, *args, verbosity=1): |
| 364 | """Print the message to stderr if -v/PYTHONVERBOSE is turned on.""" |
| 365 | if sys.flags.verbose >= verbosity: |
| 366 | if not message.startswith(('#', 'import ')): |
| 367 | message = '# ' + message |
| 368 | print(message.format(*args), file=sys.stderr) |
| 369 | |
| 370 | |
| 371 | def _check_name(method): |
| 372 | """Decorator to verify that the module being requested matches the one the |
| 373 | loader can handle. |
| 374 | |
| 375 | The first argument (self) must define _name which the second argument is |
| 376 | compared against. If the comparison fails then ImportError is raised. |
| 377 | |
| 378 | """ |
| 379 | def _check_name_wrapper(self, name=None, *args, **kwargs): |
| 380 | if name is None: |
| 381 | name = self.name |
| 382 | elif self.name != name: |
Nick Coghlan | d5cacbb | 2015-05-23 22:24:10 +1000 | [diff] [blame] | 383 | raise ImportError('loader for %s cannot handle %s' % |
| 384 | (self.name, name), name=name) |
Eric Snow | 32439d6 | 2015-05-02 19:15:18 -0600 | [diff] [blame] | 385 | return method(self, name, *args, **kwargs) |
| 386 | try: |
| 387 | _wrap = _bootstrap._wrap |
| 388 | except NameError: |
| 389 | # XXX yuck |
| 390 | def _wrap(new, old): |
| 391 | for replace in ['__module__', '__name__', '__qualname__', '__doc__']: |
| 392 | if hasattr(old, replace): |
| 393 | setattr(new, replace, getattr(old, replace)) |
| 394 | new.__dict__.update(old.__dict__) |
| 395 | _wrap(_check_name_wrapper, method) |
| 396 | return _check_name_wrapper |
| 397 | |
| 398 | |
| 399 | def _find_module_shim(self, fullname): |
| 400 | """Try to find a loader for the specified module by delegating to |
| 401 | self.find_loader(). |
| 402 | |
| 403 | This method is deprecated in favor of finder.find_spec(). |
| 404 | |
| 405 | """ |
| 406 | # Call find_loader(). If it returns a string (indicating this |
| 407 | # is a namespace package portion), generate a warning and |
| 408 | # return None. |
| 409 | loader, portions = self.find_loader(fullname) |
| 410 | if loader is None and len(portions): |
| 411 | msg = 'Not importing directory {}: missing __init__' |
| 412 | _warnings.warn(msg.format(portions[0]), ImportWarning) |
| 413 | return loader |
| 414 | |
| 415 | |
Eric Snow | 32439d6 | 2015-05-02 19:15:18 -0600 | [diff] [blame] | 416 | def _validate_bytecode_header(data, source_stats=None, name=None, path=None): |
| 417 | """Validate the header of the passed-in bytecode against source_stats (if |
| 418 | given) and returning the bytecode that can be compiled by compile(). |
| 419 | |
| 420 | All other arguments are used to enhance error reporting. |
| 421 | |
| 422 | ImportError is raised when the magic number is incorrect or the bytecode is |
| 423 | found to be stale. EOFError is raised when the data is found to be |
| 424 | truncated. |
| 425 | |
| 426 | """ |
| 427 | exc_details = {} |
| 428 | if name is not None: |
| 429 | exc_details['name'] = name |
| 430 | else: |
| 431 | # To prevent having to make all messages have a conditional name. |
| 432 | name = '<bytecode>' |
| 433 | if path is not None: |
| 434 | exc_details['path'] = path |
| 435 | magic = data[:4] |
| 436 | raw_timestamp = data[4:8] |
| 437 | raw_size = data[8:12] |
| 438 | if magic != MAGIC_NUMBER: |
| 439 | message = 'bad magic number in {!r}: {!r}'.format(name, magic) |
| 440 | _verbose_message(message) |
| 441 | raise ImportError(message, **exc_details) |
| 442 | elif len(raw_timestamp) != 4: |
| 443 | message = 'reached EOF while reading timestamp in {!r}'.format(name) |
| 444 | _verbose_message(message) |
| 445 | raise EOFError(message) |
| 446 | elif len(raw_size) != 4: |
| 447 | message = 'reached EOF while reading size of source in {!r}'.format(name) |
| 448 | _verbose_message(message) |
| 449 | raise EOFError(message) |
| 450 | if source_stats is not None: |
| 451 | try: |
| 452 | source_mtime = int(source_stats['mtime']) |
| 453 | except KeyError: |
| 454 | pass |
| 455 | else: |
| 456 | if _r_long(raw_timestamp) != source_mtime: |
| 457 | message = 'bytecode is stale for {!r}'.format(name) |
| 458 | _verbose_message(message) |
| 459 | raise ImportError(message, **exc_details) |
| 460 | try: |
| 461 | source_size = source_stats['size'] & 0xFFFFFFFF |
| 462 | except KeyError: |
| 463 | pass |
| 464 | else: |
| 465 | if _r_long(raw_size) != source_size: |
| 466 | raise ImportError('bytecode is stale for {!r}'.format(name), |
| 467 | **exc_details) |
| 468 | return data[12:] |
| 469 | |
| 470 | |
| 471 | def _compile_bytecode(data, name=None, bytecode_path=None, source_path=None): |
| 472 | """Compile bytecode as returned by _validate_bytecode_header().""" |
| 473 | code = marshal.loads(data) |
| 474 | if isinstance(code, _code_type): |
| 475 | _verbose_message('code object from {!r}', bytecode_path) |
| 476 | if source_path is not None: |
| 477 | _imp._fix_co_filename(code, source_path) |
| 478 | return code |
| 479 | else: |
| 480 | raise ImportError('Non-code object in {!r}'.format(bytecode_path), |
| 481 | name=name, path=bytecode_path) |
| 482 | |
| 483 | def _code_to_bytecode(code, mtime=0, source_size=0): |
| 484 | """Compile a code object into bytecode for writing out to a byte-compiled |
| 485 | file.""" |
| 486 | data = bytearray(MAGIC_NUMBER) |
| 487 | data.extend(_w_long(mtime)) |
| 488 | data.extend(_w_long(source_size)) |
| 489 | data.extend(marshal.dumps(code)) |
| 490 | return data |
| 491 | |
| 492 | |
| 493 | def decode_source(source_bytes): |
| 494 | """Decode bytes representing source code and return the string. |
| 495 | |
| 496 | Universal newline support is used in the decoding. |
| 497 | """ |
| 498 | import tokenize # To avoid bootstrap issues. |
| 499 | source_bytes_readline = _io.BytesIO(source_bytes).readline |
| 500 | encoding = tokenize.detect_encoding(source_bytes_readline) |
| 501 | newline_decoder = _io.IncrementalNewlineDecoder(None, True) |
| 502 | return newline_decoder.decode(source_bytes.decode(encoding[0])) |
| 503 | |
| 504 | |
| 505 | # Module specifications ####################################################### |
| 506 | |
Eric Snow | 32439d6 | 2015-05-02 19:15:18 -0600 | [diff] [blame] | 507 | _POPULATE = object() |
| 508 | |
| 509 | |
| 510 | def spec_from_file_location(name, location=None, *, loader=None, |
| 511 | submodule_search_locations=_POPULATE): |
| 512 | """Return a module spec based on a file location. |
| 513 | |
| 514 | To indicate that the module is a package, set |
| 515 | submodule_search_locations to a list of directory paths. An |
| 516 | empty list is sufficient, though its not otherwise useful to the |
| 517 | import system. |
| 518 | |
| 519 | The loader must take a spec as its only __init__() arg. |
| 520 | |
| 521 | """ |
| 522 | if location is None: |
| 523 | # The caller may simply want a partially populated location- |
| 524 | # oriented spec. So we set the location to a bogus value and |
| 525 | # fill in as much as we can. |
| 526 | location = '<unknown>' |
| 527 | if hasattr(loader, 'get_filename'): |
| 528 | # ExecutionLoader |
| 529 | try: |
| 530 | location = loader.get_filename(name) |
| 531 | except ImportError: |
| 532 | pass |
| 533 | |
| 534 | # If the location is on the filesystem, but doesn't actually exist, |
| 535 | # we could return None here, indicating that the location is not |
| 536 | # valid. However, we don't have a good way of testing since an |
| 537 | # indirect location (e.g. a zip file or URL) will look like a |
| 538 | # non-existent file relative to the filesystem. |
| 539 | |
| 540 | spec = _bootstrap.ModuleSpec(name, loader, origin=location) |
| 541 | spec._set_fileattr = True |
| 542 | |
| 543 | # Pick a loader if one wasn't provided. |
| 544 | if loader is None: |
| 545 | for loader_class, suffixes in _get_supported_file_loaders(): |
| 546 | if location.endswith(tuple(suffixes)): |
| 547 | loader = loader_class(name, location) |
| 548 | spec.loader = loader |
| 549 | break |
| 550 | else: |
| 551 | return None |
| 552 | |
| 553 | # Set submodule_search_paths appropriately. |
| 554 | if submodule_search_locations is _POPULATE: |
| 555 | # Check the loader. |
| 556 | if hasattr(loader, 'is_package'): |
| 557 | try: |
| 558 | is_package = loader.is_package(name) |
| 559 | except ImportError: |
| 560 | pass |
| 561 | else: |
| 562 | if is_package: |
| 563 | spec.submodule_search_locations = [] |
| 564 | else: |
| 565 | spec.submodule_search_locations = submodule_search_locations |
| 566 | if spec.submodule_search_locations == []: |
| 567 | if location: |
| 568 | dirname = _path_split(location)[0] |
| 569 | spec.submodule_search_locations.append(dirname) |
| 570 | |
| 571 | return spec |
| 572 | |
| 573 | |
| 574 | # Loaders ##################################################################### |
| 575 | |
| 576 | class WindowsRegistryFinder: |
| 577 | |
| 578 | """Meta path finder for modules declared in the Windows registry.""" |
| 579 | |
| 580 | REGISTRY_KEY = ( |
| 581 | 'Software\\Python\\PythonCore\\{sys_version}' |
| 582 | '\\Modules\\{fullname}') |
| 583 | REGISTRY_KEY_DEBUG = ( |
| 584 | 'Software\\Python\\PythonCore\\{sys_version}' |
| 585 | '\\Modules\\{fullname}\\Debug') |
| 586 | DEBUG_BUILD = False # Changed in _setup() |
| 587 | |
| 588 | @classmethod |
| 589 | def _open_registry(cls, key): |
| 590 | try: |
| 591 | return _winreg.OpenKey(_winreg.HKEY_CURRENT_USER, key) |
| 592 | except OSError: |
| 593 | return _winreg.OpenKey(_winreg.HKEY_LOCAL_MACHINE, key) |
| 594 | |
| 595 | @classmethod |
| 596 | def _search_registry(cls, fullname): |
| 597 | if cls.DEBUG_BUILD: |
| 598 | registry_key = cls.REGISTRY_KEY_DEBUG |
| 599 | else: |
| 600 | registry_key = cls.REGISTRY_KEY |
| 601 | key = registry_key.format(fullname=fullname, |
| 602 | sys_version=sys.version[:3]) |
| 603 | try: |
| 604 | with cls._open_registry(key) as hkey: |
| 605 | filepath = _winreg.QueryValue(hkey, '') |
| 606 | except OSError: |
| 607 | return None |
| 608 | return filepath |
| 609 | |
| 610 | @classmethod |
| 611 | def find_spec(cls, fullname, path=None, target=None): |
| 612 | filepath = cls._search_registry(fullname) |
| 613 | if filepath is None: |
| 614 | return None |
| 615 | try: |
| 616 | _path_stat(filepath) |
| 617 | except OSError: |
| 618 | return None |
| 619 | for loader, suffixes in _get_supported_file_loaders(): |
| 620 | if filepath.endswith(tuple(suffixes)): |
Eric Snow | 183a941 | 2015-05-15 21:54:59 -0600 | [diff] [blame] | 621 | spec = _bootstrap.spec_from_loader(fullname, |
| 622 | loader(fullname, filepath), |
| 623 | origin=filepath) |
Eric Snow | 32439d6 | 2015-05-02 19:15:18 -0600 | [diff] [blame] | 624 | return spec |
| 625 | |
| 626 | @classmethod |
| 627 | def find_module(cls, fullname, path=None): |
| 628 | """Find module named in the registry. |
| 629 | |
| 630 | This method is deprecated. Use exec_module() instead. |
| 631 | |
| 632 | """ |
| 633 | spec = cls.find_spec(fullname, path) |
| 634 | if spec is not None: |
| 635 | return spec.loader |
| 636 | else: |
| 637 | return None |
| 638 | |
| 639 | |
| 640 | class _LoaderBasics: |
| 641 | |
| 642 | """Base class of common code needed by both SourceLoader and |
| 643 | SourcelessFileLoader.""" |
| 644 | |
| 645 | def is_package(self, fullname): |
| 646 | """Concrete implementation of InspectLoader.is_package by checking if |
| 647 | the path returned by get_filename has a filename of '__init__.py'.""" |
| 648 | filename = _path_split(self.get_filename(fullname))[1] |
| 649 | filename_base = filename.rsplit('.', 1)[0] |
| 650 | tail_name = fullname.rpartition('.')[2] |
| 651 | return filename_base == '__init__' and tail_name != '__init__' |
| 652 | |
| 653 | def create_module(self, spec): |
| 654 | """Use default semantics for module creation.""" |
| 655 | |
| 656 | def exec_module(self, module): |
| 657 | """Execute the module.""" |
| 658 | code = self.get_code(module.__name__) |
| 659 | if code is None: |
| 660 | raise ImportError('cannot load module {!r} when get_code() ' |
| 661 | 'returns None'.format(module.__name__)) |
| 662 | _bootstrap._call_with_frames_removed(exec, code, module.__dict__) |
| 663 | |
Eric Snow | 183a941 | 2015-05-15 21:54:59 -0600 | [diff] [blame] | 664 | def load_module(self, fullname): |
| 665 | return _bootstrap._load_module_shim(self, fullname) |
Eric Snow | 32439d6 | 2015-05-02 19:15:18 -0600 | [diff] [blame] | 666 | |
| 667 | |
| 668 | class SourceLoader(_LoaderBasics): |
| 669 | |
| 670 | def path_mtime(self, path): |
| 671 | """Optional method that returns the modification time (an int) for the |
| 672 | specified path, where path is a str. |
| 673 | |
| 674 | Raises IOError when the path cannot be handled. |
| 675 | """ |
| 676 | raise IOError |
| 677 | |
| 678 | def path_stats(self, path): |
| 679 | """Optional method returning a metadata dict for the specified path |
| 680 | to by the path (str). |
| 681 | Possible keys: |
| 682 | - 'mtime' (mandatory) is the numeric timestamp of last source |
| 683 | code modification; |
| 684 | - 'size' (optional) is the size in bytes of the source code. |
| 685 | |
| 686 | Implementing this method allows the loader to read bytecode files. |
| 687 | Raises IOError when the path cannot be handled. |
| 688 | """ |
| 689 | return {'mtime': self.path_mtime(path)} |
| 690 | |
| 691 | def _cache_bytecode(self, source_path, cache_path, data): |
| 692 | """Optional method which writes data (bytes) to a file path (a str). |
| 693 | |
| 694 | Implementing this method allows for the writing of bytecode files. |
| 695 | |
| 696 | The source path is needed in order to correctly transfer permissions |
| 697 | """ |
| 698 | # For backwards compatibility, we delegate to set_data() |
| 699 | return self.set_data(cache_path, data) |
| 700 | |
| 701 | def set_data(self, path, data): |
| 702 | """Optional method which writes data (bytes) to a file path (a str). |
| 703 | |
| 704 | Implementing this method allows for the writing of bytecode files. |
| 705 | """ |
| 706 | |
| 707 | |
| 708 | def get_source(self, fullname): |
| 709 | """Concrete implementation of InspectLoader.get_source.""" |
| 710 | path = self.get_filename(fullname) |
| 711 | try: |
| 712 | source_bytes = self.get_data(path) |
| 713 | except OSError as exc: |
| 714 | raise ImportError('source not available through get_data()', |
| 715 | name=fullname) from exc |
| 716 | return decode_source(source_bytes) |
| 717 | |
| 718 | def source_to_code(self, data, path, *, _optimize=-1): |
| 719 | """Return the code object compiled from source. |
| 720 | |
| 721 | The 'data' argument can be any object type that compile() supports. |
| 722 | """ |
| 723 | return _bootstrap._call_with_frames_removed(compile, data, path, 'exec', |
| 724 | dont_inherit=True, optimize=_optimize) |
| 725 | |
| 726 | def get_code(self, fullname): |
| 727 | """Concrete implementation of InspectLoader.get_code. |
| 728 | |
| 729 | Reading of bytecode requires path_stats to be implemented. To write |
| 730 | bytecode, set_data must also be implemented. |
| 731 | |
| 732 | """ |
| 733 | source_path = self.get_filename(fullname) |
| 734 | source_mtime = None |
| 735 | try: |
| 736 | bytecode_path = cache_from_source(source_path) |
| 737 | except NotImplementedError: |
| 738 | bytecode_path = None |
| 739 | else: |
| 740 | try: |
| 741 | st = self.path_stats(source_path) |
| 742 | except IOError: |
| 743 | pass |
| 744 | else: |
| 745 | source_mtime = int(st['mtime']) |
| 746 | try: |
| 747 | data = self.get_data(bytecode_path) |
| 748 | except OSError: |
| 749 | pass |
| 750 | else: |
| 751 | try: |
| 752 | bytes_data = _validate_bytecode_header(data, |
| 753 | source_stats=st, name=fullname, |
| 754 | path=bytecode_path) |
| 755 | except (ImportError, EOFError): |
| 756 | pass |
| 757 | else: |
| 758 | _verbose_message('{} matches {}', bytecode_path, |
| 759 | source_path) |
| 760 | return _compile_bytecode(bytes_data, name=fullname, |
| 761 | bytecode_path=bytecode_path, |
| 762 | source_path=source_path) |
| 763 | source_bytes = self.get_data(source_path) |
| 764 | code_object = self.source_to_code(source_bytes, source_path) |
| 765 | _verbose_message('code object from {}', source_path) |
| 766 | if (not sys.dont_write_bytecode and bytecode_path is not None and |
| 767 | source_mtime is not None): |
| 768 | data = _code_to_bytecode(code_object, source_mtime, |
| 769 | len(source_bytes)) |
| 770 | try: |
| 771 | self._cache_bytecode(source_path, bytecode_path, data) |
| 772 | _verbose_message('wrote {!r}', bytecode_path) |
| 773 | except NotImplementedError: |
| 774 | pass |
| 775 | return code_object |
| 776 | |
| 777 | |
| 778 | class FileLoader: |
| 779 | |
| 780 | """Base file loader class which implements the loader protocol methods that |
| 781 | require file system usage.""" |
| 782 | |
| 783 | def __init__(self, fullname, path): |
| 784 | """Cache the module name and the path to the file found by the |
| 785 | finder.""" |
| 786 | self.name = fullname |
| 787 | self.path = path |
| 788 | |
| 789 | def __eq__(self, other): |
| 790 | return (self.__class__ == other.__class__ and |
| 791 | self.__dict__ == other.__dict__) |
| 792 | |
| 793 | def __hash__(self): |
| 794 | return hash(self.name) ^ hash(self.path) |
| 795 | |
| 796 | @_check_name |
| 797 | def load_module(self, fullname): |
| 798 | """Load a module from a file. |
| 799 | |
| 800 | This method is deprecated. Use exec_module() instead. |
| 801 | |
| 802 | """ |
| 803 | # The only reason for this method is for the name check. |
| 804 | # Issue #14857: Avoid the zero-argument form of super so the implementation |
| 805 | # of that form can be updated without breaking the frozen module |
| 806 | return super(FileLoader, self).load_module(fullname) |
| 807 | |
| 808 | @_check_name |
| 809 | def get_filename(self, fullname): |
| 810 | """Return the path to the source file as found by the finder.""" |
| 811 | return self.path |
| 812 | |
| 813 | def get_data(self, path): |
| 814 | """Return the data from path as raw bytes.""" |
| 815 | with _io.FileIO(path, 'r') as file: |
| 816 | return file.read() |
| 817 | |
| 818 | |
| 819 | class SourceFileLoader(FileLoader, SourceLoader): |
| 820 | |
| 821 | """Concrete implementation of SourceLoader using the file system.""" |
| 822 | |
| 823 | def path_stats(self, path): |
| 824 | """Return the metadata for the path.""" |
| 825 | st = _path_stat(path) |
| 826 | return {'mtime': st.st_mtime, 'size': st.st_size} |
| 827 | |
| 828 | def _cache_bytecode(self, source_path, bytecode_path, data): |
| 829 | # Adapt between the two APIs |
| 830 | mode = _calc_mode(source_path) |
| 831 | return self.set_data(bytecode_path, data, _mode=mode) |
| 832 | |
| 833 | def set_data(self, path, data, *, _mode=0o666): |
| 834 | """Write bytes data to a file.""" |
| 835 | parent, filename = _path_split(path) |
| 836 | path_parts = [] |
| 837 | # Figure out what directories are missing. |
| 838 | while parent and not _path_isdir(parent): |
| 839 | parent, part = _path_split(parent) |
| 840 | path_parts.append(part) |
| 841 | # Create needed directories. |
| 842 | for part in reversed(path_parts): |
| 843 | parent = _path_join(parent, part) |
| 844 | try: |
| 845 | _os.mkdir(parent) |
| 846 | except FileExistsError: |
| 847 | # Probably another Python process already created the dir. |
| 848 | continue |
| 849 | except OSError as exc: |
| 850 | # Could be a permission error, read-only filesystem: just forget |
| 851 | # about writing the data. |
| 852 | _verbose_message('could not create {!r}: {!r}', parent, exc) |
| 853 | return |
| 854 | try: |
| 855 | _write_atomic(path, data, _mode) |
| 856 | _verbose_message('created {!r}', path) |
| 857 | except OSError as exc: |
| 858 | # Same as above: just don't write the bytecode. |
| 859 | _verbose_message('could not create {!r}: {!r}', path, exc) |
| 860 | |
| 861 | |
| 862 | class SourcelessFileLoader(FileLoader, _LoaderBasics): |
| 863 | |
| 864 | """Loader which handles sourceless file imports.""" |
| 865 | |
| 866 | def get_code(self, fullname): |
| 867 | path = self.get_filename(fullname) |
| 868 | data = self.get_data(path) |
| 869 | bytes_data = _validate_bytecode_header(data, name=fullname, path=path) |
| 870 | return _compile_bytecode(bytes_data, name=fullname, bytecode_path=path) |
| 871 | |
| 872 | def get_source(self, fullname): |
| 873 | """Return None as there is no source code.""" |
| 874 | return None |
| 875 | |
| 876 | |
| 877 | # Filled in by _setup(). |
| 878 | EXTENSION_SUFFIXES = [] |
| 879 | |
| 880 | |
Nick Coghlan | d5cacbb | 2015-05-23 22:24:10 +1000 | [diff] [blame] | 881 | class ExtensionFileLoader(FileLoader, _LoaderBasics): |
Eric Snow | 32439d6 | 2015-05-02 19:15:18 -0600 | [diff] [blame] | 882 | |
| 883 | """Loader for extension modules. |
| 884 | |
| 885 | The constructor is designed to work with FileFinder. |
| 886 | |
| 887 | """ |
| 888 | |
| 889 | def __init__(self, name, path): |
| 890 | self.name = name |
| 891 | self.path = path |
| 892 | |
| 893 | def __eq__(self, other): |
| 894 | return (self.__class__ == other.__class__ and |
| 895 | self.__dict__ == other.__dict__) |
| 896 | |
| 897 | def __hash__(self): |
| 898 | return hash(self.name) ^ hash(self.path) |
| 899 | |
Nick Coghlan | d5cacbb | 2015-05-23 22:24:10 +1000 | [diff] [blame] | 900 | def create_module(self, spec): |
| 901 | """Create an unitialized extension module""" |
| 902 | module = _bootstrap._call_with_frames_removed( |
| 903 | _imp.create_dynamic, spec) |
| 904 | _verbose_message('extension module {!r} loaded from {!r}', |
| 905 | spec.name, self.path) |
Eric Snow | 32439d6 | 2015-05-02 19:15:18 -0600 | [diff] [blame] | 906 | return module |
| 907 | |
Nick Coghlan | d5cacbb | 2015-05-23 22:24:10 +1000 | [diff] [blame] | 908 | def exec_module(self, module): |
| 909 | """Initialize an extension module""" |
| 910 | _bootstrap._call_with_frames_removed(_imp.exec_dynamic, module) |
| 911 | _verbose_message('extension module {!r} executed from {!r}', |
| 912 | self.name, self.path) |
| 913 | |
Eric Snow | 32439d6 | 2015-05-02 19:15:18 -0600 | [diff] [blame] | 914 | def is_package(self, fullname): |
| 915 | """Return True if the extension module is a package.""" |
| 916 | file_name = _path_split(self.path)[1] |
| 917 | return any(file_name == '__init__' + suffix |
| 918 | for suffix in EXTENSION_SUFFIXES) |
| 919 | |
| 920 | def get_code(self, fullname): |
| 921 | """Return None as an extension module cannot create a code object.""" |
| 922 | return None |
| 923 | |
| 924 | def get_source(self, fullname): |
| 925 | """Return None as extension modules have no source code.""" |
| 926 | return None |
| 927 | |
| 928 | @_check_name |
| 929 | def get_filename(self, fullname): |
| 930 | """Return the path to the source file as found by the finder.""" |
| 931 | return self.path |
| 932 | |
| 933 | |
| 934 | class _NamespacePath: |
| 935 | """Represents a namespace package's path. It uses the module name |
| 936 | to find its parent module, and from there it looks up the parent's |
| 937 | __path__. When this changes, the module's own path is recomputed, |
| 938 | using path_finder. For top-level modules, the parent module's path |
| 939 | is sys.path.""" |
| 940 | |
| 941 | def __init__(self, name, path, path_finder): |
| 942 | self._name = name |
| 943 | self._path = path |
| 944 | self._last_parent_path = tuple(self._get_parent_path()) |
| 945 | self._path_finder = path_finder |
| 946 | |
| 947 | def _find_parent_path_names(self): |
| 948 | """Returns a tuple of (parent-module-name, parent-path-attr-name)""" |
| 949 | parent, dot, me = self._name.rpartition('.') |
| 950 | if dot == '': |
| 951 | # This is a top-level module. sys.path contains the parent path. |
| 952 | return 'sys', 'path' |
| 953 | # Not a top-level module. parent-module.__path__ contains the |
| 954 | # parent path. |
| 955 | return parent, '__path__' |
| 956 | |
| 957 | def _get_parent_path(self): |
| 958 | parent_module_name, path_attr_name = self._find_parent_path_names() |
| 959 | return getattr(sys.modules[parent_module_name], path_attr_name) |
| 960 | |
| 961 | def _recalculate(self): |
| 962 | # If the parent's path has changed, recalculate _path |
| 963 | parent_path = tuple(self._get_parent_path()) # Make a copy |
| 964 | if parent_path != self._last_parent_path: |
| 965 | spec = self._path_finder(self._name, parent_path) |
| 966 | # Note that no changes are made if a loader is returned, but we |
| 967 | # do remember the new parent path |
| 968 | if spec is not None and spec.loader is None: |
| 969 | if spec.submodule_search_locations: |
| 970 | self._path = spec.submodule_search_locations |
| 971 | self._last_parent_path = parent_path # Save the copy |
| 972 | return self._path |
| 973 | |
| 974 | def __iter__(self): |
| 975 | return iter(self._recalculate()) |
| 976 | |
| 977 | def __len__(self): |
| 978 | return len(self._recalculate()) |
| 979 | |
| 980 | def __repr__(self): |
| 981 | return '_NamespacePath({!r})'.format(self._path) |
| 982 | |
| 983 | def __contains__(self, item): |
| 984 | return item in self._recalculate() |
| 985 | |
| 986 | def append(self, item): |
| 987 | self._path.append(item) |
| 988 | |
| 989 | |
| 990 | # We use this exclusively in module_from_spec() for backward-compatibility. |
| 991 | class _NamespaceLoader: |
| 992 | def __init__(self, name, path, path_finder): |
| 993 | self._path = _NamespacePath(name, path, path_finder) |
| 994 | |
| 995 | @classmethod |
| 996 | def module_repr(cls, module): |
| 997 | """Return repr for the module. |
| 998 | |
| 999 | The method is deprecated. The import machinery does the job itself. |
| 1000 | |
| 1001 | """ |
| 1002 | return '<module {!r} (namespace)>'.format(module.__name__) |
| 1003 | |
| 1004 | def is_package(self, fullname): |
| 1005 | return True |
| 1006 | |
| 1007 | def get_source(self, fullname): |
| 1008 | return '' |
| 1009 | |
| 1010 | def get_code(self, fullname): |
| 1011 | return compile('', '<string>', 'exec', dont_inherit=True) |
| 1012 | |
| 1013 | def create_module(self, spec): |
| 1014 | """Use default semantics for module creation.""" |
| 1015 | |
| 1016 | def exec_module(self, module): |
| 1017 | pass |
| 1018 | |
| 1019 | def load_module(self, fullname): |
| 1020 | """Load a namespace module. |
| 1021 | |
| 1022 | This method is deprecated. Use exec_module() instead. |
| 1023 | |
| 1024 | """ |
| 1025 | # The import system never calls this method. |
| 1026 | _verbose_message('namespace module loaded with path {!r}', self._path) |
Eric Snow | 183a941 | 2015-05-15 21:54:59 -0600 | [diff] [blame] | 1027 | return _bootstrap._load_module_shim(self, fullname) |
Eric Snow | 32439d6 | 2015-05-02 19:15:18 -0600 | [diff] [blame] | 1028 | |
| 1029 | |
| 1030 | # Finders ##################################################################### |
| 1031 | |
| 1032 | class PathFinder: |
| 1033 | |
| 1034 | """Meta path finder for sys.path and package __path__ attributes.""" |
| 1035 | |
| 1036 | @classmethod |
| 1037 | def invalidate_caches(cls): |
| 1038 | """Call the invalidate_caches() method on all path entry finders |
| 1039 | stored in sys.path_importer_caches (where implemented).""" |
| 1040 | for finder in sys.path_importer_cache.values(): |
| 1041 | if hasattr(finder, 'invalidate_caches'): |
| 1042 | finder.invalidate_caches() |
| 1043 | |
| 1044 | @classmethod |
| 1045 | def _path_hooks(cls, path): |
| 1046 | """Search sequence of hooks for a finder for 'path'. |
| 1047 | |
| 1048 | If 'hooks' is false then use sys.path_hooks. |
| 1049 | |
| 1050 | """ |
| 1051 | if sys.path_hooks is not None and not sys.path_hooks: |
| 1052 | _warnings.warn('sys.path_hooks is empty', ImportWarning) |
| 1053 | for hook in sys.path_hooks: |
| 1054 | try: |
| 1055 | return hook(path) |
| 1056 | except ImportError: |
| 1057 | continue |
| 1058 | else: |
| 1059 | return None |
| 1060 | |
| 1061 | @classmethod |
| 1062 | def _path_importer_cache(cls, path): |
| 1063 | """Get the finder for the path entry from sys.path_importer_cache. |
| 1064 | |
| 1065 | If the path entry is not in the cache, find the appropriate finder |
| 1066 | and cache it. If no finder is available, store None. |
| 1067 | |
| 1068 | """ |
| 1069 | if path == '': |
| 1070 | try: |
| 1071 | path = _os.getcwd() |
| 1072 | except FileNotFoundError: |
| 1073 | # Don't cache the failure as the cwd can easily change to |
| 1074 | # a valid directory later on. |
| 1075 | return None |
| 1076 | try: |
| 1077 | finder = sys.path_importer_cache[path] |
| 1078 | except KeyError: |
| 1079 | finder = cls._path_hooks(path) |
| 1080 | sys.path_importer_cache[path] = finder |
| 1081 | return finder |
| 1082 | |
| 1083 | @classmethod |
| 1084 | def _legacy_get_spec(cls, fullname, finder): |
| 1085 | # This would be a good place for a DeprecationWarning if |
| 1086 | # we ended up going that route. |
| 1087 | if hasattr(finder, 'find_loader'): |
| 1088 | loader, portions = finder.find_loader(fullname) |
| 1089 | else: |
| 1090 | loader = finder.find_module(fullname) |
| 1091 | portions = [] |
| 1092 | if loader is not None: |
Eric Snow | 183a941 | 2015-05-15 21:54:59 -0600 | [diff] [blame] | 1093 | return _bootstrap.spec_from_loader(fullname, loader) |
Eric Snow | 32439d6 | 2015-05-02 19:15:18 -0600 | [diff] [blame] | 1094 | spec = _bootstrap.ModuleSpec(fullname, None) |
| 1095 | spec.submodule_search_locations = portions |
| 1096 | return spec |
| 1097 | |
| 1098 | @classmethod |
| 1099 | def _get_spec(cls, fullname, path, target=None): |
| 1100 | """Find the loader or namespace_path for this module/package name.""" |
| 1101 | # If this ends up being a namespace package, namespace_path is |
| 1102 | # the list of paths that will become its __path__ |
| 1103 | namespace_path = [] |
| 1104 | for entry in path: |
| 1105 | if not isinstance(entry, (str, bytes)): |
| 1106 | continue |
| 1107 | finder = cls._path_importer_cache(entry) |
| 1108 | if finder is not None: |
| 1109 | if hasattr(finder, 'find_spec'): |
| 1110 | spec = finder.find_spec(fullname, target) |
| 1111 | else: |
| 1112 | spec = cls._legacy_get_spec(fullname, finder) |
| 1113 | if spec is None: |
| 1114 | continue |
| 1115 | if spec.loader is not None: |
| 1116 | return spec |
| 1117 | portions = spec.submodule_search_locations |
| 1118 | if portions is None: |
| 1119 | raise ImportError('spec missing loader') |
| 1120 | # This is possibly part of a namespace package. |
| 1121 | # Remember these path entries (if any) for when we |
| 1122 | # create a namespace package, and continue iterating |
| 1123 | # on path. |
| 1124 | namespace_path.extend(portions) |
| 1125 | else: |
| 1126 | spec = _bootstrap.ModuleSpec(fullname, None) |
| 1127 | spec.submodule_search_locations = namespace_path |
| 1128 | return spec |
| 1129 | |
| 1130 | @classmethod |
| 1131 | def find_spec(cls, fullname, path=None, target=None): |
| 1132 | """find the module on sys.path or 'path' based on sys.path_hooks and |
| 1133 | sys.path_importer_cache.""" |
| 1134 | if path is None: |
| 1135 | path = sys.path |
| 1136 | spec = cls._get_spec(fullname, path, target) |
| 1137 | if spec is None: |
| 1138 | return None |
| 1139 | elif spec.loader is None: |
| 1140 | namespace_path = spec.submodule_search_locations |
| 1141 | if namespace_path: |
| 1142 | # We found at least one namespace path. Return a |
| 1143 | # spec which can create the namespace package. |
| 1144 | spec.origin = 'namespace' |
| 1145 | spec.submodule_search_locations = _NamespacePath(fullname, namespace_path, cls._get_spec) |
| 1146 | return spec |
| 1147 | else: |
| 1148 | return None |
| 1149 | else: |
| 1150 | return spec |
| 1151 | |
| 1152 | @classmethod |
| 1153 | def find_module(cls, fullname, path=None): |
| 1154 | """find the module on sys.path or 'path' based on sys.path_hooks and |
| 1155 | sys.path_importer_cache. |
| 1156 | |
| 1157 | This method is deprecated. Use find_spec() instead. |
| 1158 | |
| 1159 | """ |
| 1160 | spec = cls.find_spec(fullname, path) |
| 1161 | if spec is None: |
| 1162 | return None |
| 1163 | return spec.loader |
| 1164 | |
| 1165 | |
| 1166 | class FileFinder: |
| 1167 | |
| 1168 | """File-based finder. |
| 1169 | |
| 1170 | Interactions with the file system are cached for performance, being |
| 1171 | refreshed when the directory the finder is handling has been modified. |
| 1172 | |
| 1173 | """ |
| 1174 | |
| 1175 | def __init__(self, path, *loader_details): |
| 1176 | """Initialize with the path to search on and a variable number of |
| 1177 | 2-tuples containing the loader and the file suffixes the loader |
| 1178 | recognizes.""" |
| 1179 | loaders = [] |
| 1180 | for loader, suffixes in loader_details: |
| 1181 | loaders.extend((suffix, loader) for suffix in suffixes) |
| 1182 | self._loaders = loaders |
| 1183 | # Base (directory) path |
| 1184 | self.path = path or '.' |
| 1185 | self._path_mtime = -1 |
| 1186 | self._path_cache = set() |
| 1187 | self._relaxed_path_cache = set() |
| 1188 | |
| 1189 | def invalidate_caches(self): |
| 1190 | """Invalidate the directory mtime.""" |
| 1191 | self._path_mtime = -1 |
| 1192 | |
| 1193 | find_module = _find_module_shim |
| 1194 | |
| 1195 | def find_loader(self, fullname): |
| 1196 | """Try to find a loader for the specified module, or the namespace |
| 1197 | package portions. Returns (loader, list-of-portions). |
| 1198 | |
| 1199 | This method is deprecated. Use find_spec() instead. |
| 1200 | |
| 1201 | """ |
| 1202 | spec = self.find_spec(fullname) |
| 1203 | if spec is None: |
| 1204 | return None, [] |
| 1205 | return spec.loader, spec.submodule_search_locations or [] |
| 1206 | |
| 1207 | def _get_spec(self, loader_class, fullname, path, smsl, target): |
| 1208 | loader = loader_class(fullname, path) |
| 1209 | return spec_from_file_location(fullname, path, loader=loader, |
| 1210 | submodule_search_locations=smsl) |
| 1211 | |
| 1212 | def find_spec(self, fullname, target=None): |
| 1213 | """Try to find a loader for the specified module, or the namespace |
| 1214 | package portions. Returns (loader, list-of-portions).""" |
| 1215 | is_namespace = False |
| 1216 | tail_module = fullname.rpartition('.')[2] |
| 1217 | try: |
| 1218 | mtime = _path_stat(self.path or _os.getcwd()).st_mtime |
| 1219 | except OSError: |
| 1220 | mtime = -1 |
| 1221 | if mtime != self._path_mtime: |
| 1222 | self._fill_cache() |
| 1223 | self._path_mtime = mtime |
| 1224 | # tail_module keeps the original casing, for __file__ and friends |
| 1225 | if _relax_case(): |
| 1226 | cache = self._relaxed_path_cache |
| 1227 | cache_module = tail_module.lower() |
| 1228 | else: |
| 1229 | cache = self._path_cache |
| 1230 | cache_module = tail_module |
| 1231 | # Check if the module is the name of a directory (and thus a package). |
| 1232 | if cache_module in cache: |
| 1233 | base_path = _path_join(self.path, tail_module) |
| 1234 | for suffix, loader_class in self._loaders: |
| 1235 | init_filename = '__init__' + suffix |
| 1236 | full_path = _path_join(base_path, init_filename) |
| 1237 | if _path_isfile(full_path): |
| 1238 | return self._get_spec(loader_class, fullname, full_path, [base_path], target) |
| 1239 | else: |
| 1240 | # If a namespace package, return the path if we don't |
| 1241 | # find a module in the next section. |
| 1242 | is_namespace = _path_isdir(base_path) |
| 1243 | # Check for a file w/ a proper suffix exists. |
| 1244 | for suffix, loader_class in self._loaders: |
| 1245 | full_path = _path_join(self.path, tail_module + suffix) |
| 1246 | _verbose_message('trying {}'.format(full_path), verbosity=2) |
| 1247 | if cache_module + suffix in cache: |
| 1248 | if _path_isfile(full_path): |
| 1249 | return self._get_spec(loader_class, fullname, full_path, None, target) |
| 1250 | if is_namespace: |
| 1251 | _verbose_message('possible namespace for {}'.format(base_path)) |
| 1252 | spec = _bootstrap.ModuleSpec(fullname, None) |
| 1253 | spec.submodule_search_locations = [base_path] |
| 1254 | return spec |
| 1255 | return None |
| 1256 | |
| 1257 | def _fill_cache(self): |
| 1258 | """Fill the cache of potential modules and packages for this directory.""" |
| 1259 | path = self.path |
| 1260 | try: |
| 1261 | contents = _os.listdir(path or _os.getcwd()) |
| 1262 | except (FileNotFoundError, PermissionError, NotADirectoryError): |
| 1263 | # Directory has either been removed, turned into a file, or made |
| 1264 | # unreadable. |
| 1265 | contents = [] |
| 1266 | # We store two cached versions, to handle runtime changes of the |
| 1267 | # PYTHONCASEOK environment variable. |
| 1268 | if not sys.platform.startswith('win'): |
| 1269 | self._path_cache = set(contents) |
| 1270 | else: |
| 1271 | # Windows users can import modules with case-insensitive file |
| 1272 | # suffixes (for legacy reasons). Make the suffix lowercase here |
| 1273 | # so it's done once instead of for every import. This is safe as |
| 1274 | # the specified suffixes to check against are always specified in a |
| 1275 | # case-sensitive manner. |
| 1276 | lower_suffix_contents = set() |
| 1277 | for item in contents: |
| 1278 | name, dot, suffix = item.partition('.') |
| 1279 | if dot: |
| 1280 | new_name = '{}.{}'.format(name, suffix.lower()) |
| 1281 | else: |
| 1282 | new_name = name |
| 1283 | lower_suffix_contents.add(new_name) |
| 1284 | self._path_cache = lower_suffix_contents |
| 1285 | if sys.platform.startswith(_CASE_INSENSITIVE_PLATFORMS): |
| 1286 | self._relaxed_path_cache = {fn.lower() for fn in contents} |
| 1287 | |
| 1288 | @classmethod |
| 1289 | def path_hook(cls, *loader_details): |
| 1290 | """A class method which returns a closure to use on sys.path_hook |
| 1291 | which will return an instance using the specified loaders and the path |
| 1292 | called on the closure. |
| 1293 | |
| 1294 | If the path called on the closure is not a directory, ImportError is |
| 1295 | raised. |
| 1296 | |
| 1297 | """ |
| 1298 | def path_hook_for_FileFinder(path): |
| 1299 | """Path hook for importlib.machinery.FileFinder.""" |
| 1300 | if not _path_isdir(path): |
| 1301 | raise ImportError('only directories are supported', path=path) |
| 1302 | return cls(path, *loader_details) |
| 1303 | |
| 1304 | return path_hook_for_FileFinder |
| 1305 | |
| 1306 | def __repr__(self): |
| 1307 | return 'FileFinder({!r})'.format(self.path) |
| 1308 | |
| 1309 | |
| 1310 | # Import setup ############################################################### |
| 1311 | |
| 1312 | def _fix_up_module(ns, name, pathname, cpathname=None): |
| 1313 | # This function is used by PyImport_ExecCodeModuleObject(). |
| 1314 | loader = ns.get('__loader__') |
| 1315 | spec = ns.get('__spec__') |
| 1316 | if not loader: |
| 1317 | if spec: |
| 1318 | loader = spec.loader |
| 1319 | elif pathname == cpathname: |
| 1320 | loader = SourcelessFileLoader(name, pathname) |
| 1321 | else: |
| 1322 | loader = SourceFileLoader(name, pathname) |
| 1323 | if not spec: |
| 1324 | spec = spec_from_file_location(name, pathname, loader=loader) |
| 1325 | try: |
| 1326 | ns['__spec__'] = spec |
| 1327 | ns['__loader__'] = loader |
| 1328 | ns['__file__'] = pathname |
| 1329 | ns['__cached__'] = cpathname |
| 1330 | except Exception: |
| 1331 | # Not important enough to report. |
| 1332 | pass |
| 1333 | |
| 1334 | |
| 1335 | def _get_supported_file_loaders(): |
| 1336 | """Returns a list of file-based module loaders. |
| 1337 | |
| 1338 | Each item is a tuple (loader, suffixes). |
| 1339 | """ |
| 1340 | extensions = ExtensionFileLoader, _imp.extension_suffixes() |
| 1341 | source = SourceFileLoader, SOURCE_SUFFIXES |
| 1342 | bytecode = SourcelessFileLoader, BYTECODE_SUFFIXES |
| 1343 | return [extensions, source, bytecode] |
| 1344 | |
| 1345 | |
| 1346 | def _setup(_bootstrap_module): |
| 1347 | """Setup the path-based importers for importlib by importing needed |
| 1348 | built-in modules and injecting them into the global namespace. |
| 1349 | |
| 1350 | Other components are extracted from the core bootstrap module. |
| 1351 | |
| 1352 | """ |
| 1353 | global sys, _imp, _bootstrap |
| 1354 | _bootstrap = _bootstrap_module |
| 1355 | sys = _bootstrap.sys |
| 1356 | _imp = _bootstrap._imp |
| 1357 | |
| 1358 | # Directly load built-in modules needed during bootstrap. |
| 1359 | self_module = sys.modules[__name__] |
| 1360 | for builtin_name in ('_io', '_warnings', 'builtins', 'marshal'): |
| 1361 | if builtin_name not in sys.modules: |
| 1362 | builtin_module = _bootstrap._builtin_from_name(builtin_name) |
| 1363 | else: |
| 1364 | builtin_module = sys.modules[builtin_name] |
| 1365 | setattr(self_module, builtin_name, builtin_module) |
| 1366 | |
| 1367 | # Directly load the os module (needed during bootstrap). |
| 1368 | os_details = ('posix', ['/']), ('nt', ['\\', '/']) |
| 1369 | for builtin_os, path_separators in os_details: |
| 1370 | # Assumption made in _path_join() |
| 1371 | assert all(len(sep) == 1 for sep in path_separators) |
| 1372 | path_sep = path_separators[0] |
| 1373 | if builtin_os in sys.modules: |
| 1374 | os_module = sys.modules[builtin_os] |
| 1375 | break |
| 1376 | else: |
| 1377 | try: |
| 1378 | os_module = _bootstrap._builtin_from_name(builtin_os) |
| 1379 | break |
| 1380 | except ImportError: |
| 1381 | continue |
| 1382 | else: |
| 1383 | raise ImportError('importlib requires posix or nt') |
| 1384 | setattr(self_module, '_os', os_module) |
| 1385 | setattr(self_module, 'path_sep', path_sep) |
| 1386 | setattr(self_module, 'path_separators', ''.join(path_separators)) |
| 1387 | |
| 1388 | # Directly load the _thread module (needed during bootstrap). |
| 1389 | try: |
| 1390 | thread_module = _bootstrap._builtin_from_name('_thread') |
| 1391 | except ImportError: |
| 1392 | # Python was built without threads |
| 1393 | thread_module = None |
| 1394 | setattr(self_module, '_thread', thread_module) |
| 1395 | |
| 1396 | # Directly load the _weakref module (needed during bootstrap). |
| 1397 | weakref_module = _bootstrap._builtin_from_name('_weakref') |
| 1398 | setattr(self_module, '_weakref', weakref_module) |
| 1399 | |
| 1400 | # Directly load the winreg module (needed during bootstrap). |
| 1401 | if builtin_os == 'nt': |
| 1402 | winreg_module = _bootstrap._builtin_from_name('winreg') |
| 1403 | setattr(self_module, '_winreg', winreg_module) |
| 1404 | |
| 1405 | # Constants |
| 1406 | setattr(self_module, '_relax_case', _make_relax_case()) |
| 1407 | EXTENSION_SUFFIXES.extend(_imp.extension_suffixes()) |
| 1408 | if builtin_os == 'nt': |
| 1409 | SOURCE_SUFFIXES.append('.pyw') |
| 1410 | if '_d.pyd' in EXTENSION_SUFFIXES: |
| 1411 | WindowsRegistryFinder.DEBUG_BUILD = True |
| 1412 | |
| 1413 | |
| 1414 | def _install(_bootstrap_module): |
| 1415 | """Install the path-based import components.""" |
| 1416 | _setup(_bootstrap_module) |
| 1417 | supported_loaders = _get_supported_file_loaders() |
| 1418 | sys.path_hooks.extend([FileFinder.path_hook(*supported_loaders)]) |
| 1419 | if _os.__name__ == 'nt': |
| 1420 | sys.meta_path.append(WindowsRegistryFinder) |
| 1421 | sys.meta_path.append(PathFinder) |
| 1422 | |
| 1423 | # XXX We expose a couple of classes in _bootstrap for the sake of |
| 1424 | # a setuptools bug (https://bitbucket.org/pypa/setuptools/issue/378). |
| 1425 | _bootstrap_module.FileFinder = FileFinder |
| 1426 | _bootstrap_module.SourceFileLoader = SourceFileLoader |