Show More
@@ -1,402 +1,403 b'' | |||
|
1 | 1 | # __init__.py - Startup and module loading logic for Mercurial. |
|
2 | 2 | # |
|
3 | 3 | # Copyright 2015 Gregory Szorc <gregory.szorc@gmail.com> |
|
4 | 4 | # |
|
5 | 5 | # This software may be used and distributed according to the terms of the |
|
6 | 6 | # GNU General Public License version 2 or any later version. |
|
7 | 7 | |
|
8 | 8 | from __future__ import absolute_import |
|
9 | 9 | |
|
10 | 10 | import imp |
|
11 | 11 | import os |
|
12 | 12 | import sys |
|
13 | 13 | import zipimport |
|
14 | 14 | |
|
15 | 15 | from . import ( |
|
16 | 16 | policy |
|
17 | 17 | ) |
|
18 | 18 | |
|
19 | 19 | __all__ = [] |
|
20 | 20 | |
|
21 | 21 | modulepolicy = policy.policy |
|
22 | 22 | |
|
23 | 23 | # Modules that have both Python and C implementations. See also the |
|
24 | 24 | # set of .py files under mercurial/pure/. |
|
25 | 25 | _dualmodules = set([ |
|
26 | 26 | 'mercurial.base85', |
|
27 | 27 | 'mercurial.bdiff', |
|
28 | 28 | 'mercurial.diffhelpers', |
|
29 | 29 | 'mercurial.mpatch', |
|
30 | 30 | 'mercurial.osutil', |
|
31 | 31 | 'mercurial.parsers', |
|
32 | 32 | ]) |
|
33 | 33 | |
|
34 | 34 | class hgimporter(object): |
|
35 | 35 | """Object that conforms to import hook interface defined in PEP-302.""" |
|
36 | 36 | def find_module(self, name, path=None): |
|
37 | 37 | # We only care about modules that have both C and pure implementations. |
|
38 | 38 | if name in _dualmodules: |
|
39 | 39 | return self |
|
40 | 40 | return None |
|
41 | 41 | |
|
42 | 42 | def load_module(self, name): |
|
43 | 43 | mod = sys.modules.get(name, None) |
|
44 | 44 | if mod: |
|
45 | 45 | return mod |
|
46 | 46 | |
|
47 | 47 | mercurial = sys.modules['mercurial'] |
|
48 | 48 | |
|
49 | 49 | # The zip importer behaves sufficiently differently from the default |
|
50 | 50 | # importer to warrant its own code path. |
|
51 | 51 | loader = getattr(mercurial, '__loader__', None) |
|
52 | 52 | if isinstance(loader, zipimport.zipimporter): |
|
53 | 53 | def ziploader(*paths): |
|
54 | 54 | """Obtain a zipimporter for a directory under the main zip.""" |
|
55 | 55 | path = os.path.join(loader.archive, *paths) |
|
56 | 56 | zl = sys.path_importer_cache.get(path) |
|
57 | 57 | if not zl: |
|
58 | 58 | zl = zipimport.zipimporter(path) |
|
59 | 59 | return zl |
|
60 | 60 | |
|
61 | 61 | try: |
|
62 | 62 | if modulepolicy in policy.policynoc: |
|
63 | 63 | raise ImportError() |
|
64 | 64 | |
|
65 | 65 | zl = ziploader('mercurial') |
|
66 | 66 | mod = zl.load_module(name) |
|
67 | 67 | # Unlike imp, ziploader doesn't expose module metadata that |
|
68 | 68 | # indicates the type of module. So just assume what we found |
|
69 | 69 | # is OK (even though it could be a pure Python module). |
|
70 | 70 | except ImportError: |
|
71 | 71 | if modulepolicy == b'c': |
|
72 | 72 | raise |
|
73 | 73 | zl = ziploader('mercurial', 'pure') |
|
74 | 74 | mod = zl.load_module(name) |
|
75 | 75 | |
|
76 | 76 | sys.modules[name] = mod |
|
77 | 77 | return mod |
|
78 | 78 | |
|
79 | 79 | # Unlike the default importer which searches special locations and |
|
80 | 80 | # sys.path, we only look in the directory where "mercurial" was |
|
81 | 81 | # imported from. |
|
82 | 82 | |
|
83 | 83 | # imp.find_module doesn't support submodules (modules with "."). |
|
84 | 84 | # Instead you have to pass the parent package's __path__ attribute |
|
85 | 85 | # as the path argument. |
|
86 | 86 | stem = name.split('.')[-1] |
|
87 | 87 | |
|
88 | 88 | try: |
|
89 | 89 | if modulepolicy in policy.policynoc: |
|
90 | 90 | raise ImportError() |
|
91 | 91 | |
|
92 | 92 | modinfo = imp.find_module(stem, mercurial.__path__) |
|
93 | 93 | |
|
94 | 94 | # The Mercurial installer used to copy files from |
|
95 | 95 | # mercurial/pure/*.py to mercurial/*.py. Therefore, it's possible |
|
96 | 96 | # for some installations to have .py files under mercurial/*. |
|
97 | 97 | # Loading Python modules when we expected C versions could result |
|
98 | 98 | # in a) poor performance b) loading a version from a previous |
|
99 | 99 | # Mercurial version, potentially leading to incompatibility. Either |
|
100 | 100 | # scenario is bad. So we verify that modules loaded from |
|
101 | 101 | # mercurial/* are C extensions. If the current policy allows the |
|
102 | 102 | # loading of .py modules, the module will be re-imported from |
|
103 | 103 | # mercurial/pure/* below. |
|
104 | 104 | if modinfo[2][2] != imp.C_EXTENSION: |
|
105 | 105 | raise ImportError('.py version of %s found where C ' |
|
106 | 106 | 'version should exist' % name) |
|
107 | 107 | |
|
108 | 108 | except ImportError: |
|
109 | 109 | if modulepolicy == b'c': |
|
110 | 110 | raise |
|
111 | 111 | |
|
112 | 112 | # Could not load the C extension and pure Python is allowed. So |
|
113 | 113 | # try to load them. |
|
114 | 114 | from . import pure |
|
115 | 115 | modinfo = imp.find_module(stem, pure.__path__) |
|
116 | 116 | if not modinfo: |
|
117 | 117 | raise ImportError('could not find mercurial module %s' % |
|
118 | 118 | name) |
|
119 | 119 | |
|
120 | 120 | mod = imp.load_module(name, *modinfo) |
|
121 | 121 | sys.modules[name] = mod |
|
122 | 122 | return mod |
|
123 | 123 | |
|
124 | 124 | # Python 3 uses a custom module loader that transforms source code between |
|
125 | 125 | # source file reading and compilation. This is done by registering a custom |
|
126 | 126 | # finder that changes the spec for Mercurial modules to use a custom loader. |
|
127 | 127 | if sys.version_info[0] >= 3: |
|
128 | 128 | from . import pure |
|
129 | 129 | import importlib |
|
130 | 130 | import io |
|
131 | 131 | import token |
|
132 | 132 | import tokenize |
|
133 | 133 | |
|
134 | 134 | class hgpathentryfinder(importlib.abc.MetaPathFinder): |
|
135 | 135 | """A sys.meta_path finder that uses a custom module loader.""" |
|
136 | 136 | def find_spec(self, fullname, path, target=None): |
|
137 | 137 | # Only handle Mercurial-related modules. |
|
138 | 138 | if not fullname.startswith(('mercurial.', 'hgext.', 'hgext3rd.')): |
|
139 | 139 | return None |
|
140 | 140 | # zstd is already dual-version clean, don't try and mangle it |
|
141 | 141 | if fullname.startswith('mercurial.zstd'): |
|
142 | 142 | return None |
|
143 | 143 | |
|
144 | 144 | # This assumes Python 3 doesn't support loading C modules. |
|
145 | 145 | if fullname in _dualmodules: |
|
146 | 146 | stem = fullname.split('.')[-1] |
|
147 | 147 | fullname = 'mercurial.pure.%s' % stem |
|
148 | 148 | target = pure |
|
149 | 149 | assert len(path) == 1 |
|
150 | 150 | path = [os.path.join(path[0], 'pure')] |
|
151 | 151 | |
|
152 | 152 | # Try to find the module using other registered finders. |
|
153 | 153 | spec = None |
|
154 | 154 | for finder in sys.meta_path: |
|
155 | 155 | if finder == self: |
|
156 | 156 | continue |
|
157 | 157 | |
|
158 | 158 | spec = finder.find_spec(fullname, path, target=target) |
|
159 | 159 | if spec: |
|
160 | 160 | break |
|
161 | 161 | |
|
162 | 162 | # This is a Mercurial-related module but we couldn't find it |
|
163 | 163 | # using the previously-registered finders. This likely means |
|
164 | 164 | # the module doesn't exist. |
|
165 | 165 | if not spec: |
|
166 | 166 | return None |
|
167 | 167 | |
|
168 | 168 | if fullname.startswith('mercurial.pure.'): |
|
169 | 169 | spec.name = spec.name.replace('.pure.', '.') |
|
170 | 170 | |
|
171 | 171 | # TODO need to support loaders from alternate specs, like zip |
|
172 | 172 | # loaders. |
|
173 | 173 | spec.loader = hgloader(spec.name, spec.origin) |
|
174 | 174 | return spec |
|
175 | 175 | |
|
176 | 176 | def replacetokens(tokens, fullname): |
|
177 | 177 | """Transform a stream of tokens from raw to Python 3. |
|
178 | 178 | |
|
179 | 179 | It is called by the custom module loading machinery to rewrite |
|
180 | 180 | source/tokens between source decoding and compilation. |
|
181 | 181 | |
|
182 | 182 | Returns a generator of possibly rewritten tokens. |
|
183 | 183 | |
|
184 | 184 | The input token list may be mutated as part of processing. However, |
|
185 | 185 | its changes do not necessarily match the output token stream. |
|
186 | 186 | |
|
187 | 187 | REMEMBER TO CHANGE ``BYTECODEHEADER`` WHEN CHANGING THIS FUNCTION |
|
188 | 188 | OR CACHED FILES WON'T GET INVALIDATED PROPERLY. |
|
189 | 189 | """ |
|
190 | 190 | futureimpline = False |
|
191 | 191 | |
|
192 | 192 | # The following utility functions access the tokens list and i index of |
|
193 | 193 | # the for i, t enumerate(tokens) loop below |
|
194 | 194 | def _isop(j, *o): |
|
195 | 195 | """Assert that tokens[j] is an OP with one of the given values""" |
|
196 | 196 | try: |
|
197 | 197 | return tokens[j].type == token.OP and tokens[j].string in o |
|
198 | 198 | except IndexError: |
|
199 | 199 | return False |
|
200 | 200 | |
|
201 | 201 | def _findargnofcall(n): |
|
202 | 202 | """Find arg n of a call expression (start at 0) |
|
203 | 203 | |
|
204 | 204 | Returns index of the first token of that argument, or None if |
|
205 | 205 | there is not that many arguments. |
|
206 | 206 | |
|
207 | 207 | Assumes that token[i + 1] is '('. |
|
208 | 208 | |
|
209 | 209 | """ |
|
210 | 210 | nested = 0 |
|
211 | 211 | for j in range(i + 2, len(tokens)): |
|
212 | 212 | if _isop(j, ')', ']', '}'): |
|
213 | 213 | # end of call, tuple, subscription or dict / set |
|
214 | 214 | nested -= 1 |
|
215 | 215 | if nested < 0: |
|
216 | 216 | return None |
|
217 | 217 | elif n == 0: |
|
218 | 218 | # this is the starting position of arg |
|
219 | 219 | return j |
|
220 | 220 | elif _isop(j, '(', '[', '{'): |
|
221 | 221 | nested += 1 |
|
222 | 222 | elif _isop(j, ',') and nested == 0: |
|
223 | 223 | n -= 1 |
|
224 | 224 | |
|
225 | 225 | return None |
|
226 | 226 | |
|
227 | 227 | def _ensureunicode(j): |
|
228 | 228 | """Make sure the token at j is a unicode string |
|
229 | 229 | |
|
230 | 230 | This rewrites a string token to include the unicode literal prefix |
|
231 | 231 | so the string transformer won't add the byte prefix. |
|
232 | 232 | |
|
233 | 233 | Ignores tokens that are not strings. Assumes bounds checking has |
|
234 | 234 | already been done. |
|
235 | 235 | |
|
236 | 236 | """ |
|
237 | 237 | st = tokens[j] |
|
238 | 238 | if st.type == token.STRING and st.string.startswith(("'", '"')): |
|
239 | 239 | tokens[j] = st._replace(string='u%s' % st.string) |
|
240 | 240 | |
|
241 | 241 | for i, t in enumerate(tokens): |
|
242 | 242 | # Convert most string literals to byte literals. String literals |
|
243 | 243 | # in Python 2 are bytes. String literals in Python 3 are unicode. |
|
244 | 244 | # Most strings in Mercurial are bytes and unicode strings are rare. |
|
245 | 245 | # Rather than rewrite all string literals to use ``b''`` to indicate |
|
246 | 246 | # byte strings, we apply this token transformer to insert the ``b`` |
|
247 | 247 | # prefix nearly everywhere. |
|
248 | 248 | if t.type == token.STRING: |
|
249 | 249 | s = t.string |
|
250 | 250 | |
|
251 | 251 | # Preserve docstrings as string literals. This is inconsistent |
|
252 | 252 | # with regular unprefixed strings. However, the |
|
253 | 253 | # "from __future__" parsing (which allows a module docstring to |
|
254 | 254 | # exist before it) doesn't properly handle the docstring if it |
|
255 | 255 | # is b''' prefixed, leading to a SyntaxError. We leave all |
|
256 | 256 | # docstrings as unprefixed to avoid this. This means Mercurial |
|
257 | 257 | # components touching docstrings need to handle unicode, |
|
258 | 258 | # unfortunately. |
|
259 | 259 | if s[0:3] in ("'''", '"""'): |
|
260 | 260 | yield t |
|
261 | 261 | continue |
|
262 | 262 | |
|
263 | 263 | # If the first character isn't a quote, it is likely a string |
|
264 | 264 | # prefixing character (such as 'b', 'u', or 'r'. Ignore. |
|
265 | 265 | if s[0] not in ("'", '"'): |
|
266 | 266 | yield t |
|
267 | 267 | continue |
|
268 | 268 | |
|
269 | 269 | # String literal. Prefix to make a b'' string. |
|
270 | 270 | yield t._replace(string='b%s' % t.string) |
|
271 | 271 | continue |
|
272 | 272 | |
|
273 | 273 | # Insert compatibility imports at "from __future__ import" line. |
|
274 | 274 | # No '\n' should be added to preserve line numbers. |
|
275 | 275 | if (t.type == token.NAME and t.string == 'import' and |
|
276 | 276 | all(u.type == token.NAME for u in tokens[i - 2:i]) and |
|
277 | 277 | [u.string for u in tokens[i - 2:i]] == ['from', '__future__']): |
|
278 | 278 | futureimpline = True |
|
279 | 279 | if t.type == token.NEWLINE and futureimpline: |
|
280 | 280 | futureimpline = False |
|
281 | 281 | if fullname == 'mercurial.pycompat': |
|
282 | 282 | yield t |
|
283 | 283 | continue |
|
284 | 284 | r, c = t.start |
|
285 | 285 | l = (b'; from mercurial.pycompat import ' |
|
286 |
b'delattr, getattr, hasattr, setattr, xrange, |
|
|
286 | b'delattr, getattr, hasattr, setattr, xrange, ' | |
|
287 | b'open, unicode\n') | |
|
287 | 288 | for u in tokenize.tokenize(io.BytesIO(l).readline): |
|
288 | 289 | if u.type in (tokenize.ENCODING, token.ENDMARKER): |
|
289 | 290 | continue |
|
290 | 291 | yield u._replace( |
|
291 | 292 | start=(r, c + u.start[1]), end=(r, c + u.end[1])) |
|
292 | 293 | continue |
|
293 | 294 | |
|
294 | 295 | # This looks like a function call. |
|
295 | 296 | if t.type == token.NAME and _isop(i + 1, '('): |
|
296 | 297 | fn = t.string |
|
297 | 298 | |
|
298 | 299 | # *attr() builtins don't accept byte strings to 2nd argument. |
|
299 | 300 | if (fn in ('getattr', 'setattr', 'hasattr', 'safehasattr') and |
|
300 | 301 | not _isop(i - 1, '.')): |
|
301 | 302 | arg1idx = _findargnofcall(1) |
|
302 | 303 | if arg1idx is not None: |
|
303 | 304 | _ensureunicode(arg1idx) |
|
304 | 305 | |
|
305 | 306 | # .encode() and .decode() on str/bytes/unicode don't accept |
|
306 | 307 | # byte strings on Python 3. |
|
307 | 308 | elif fn in ('encode', 'decode') and _isop(i - 1, '.'): |
|
308 | 309 | for argn in range(2): |
|
309 | 310 | argidx = _findargnofcall(argn) |
|
310 | 311 | if argidx is not None: |
|
311 | 312 | _ensureunicode(argidx) |
|
312 | 313 | |
|
313 | 314 | # It changes iteritems/values to items/values as they are not |
|
314 | 315 | # present in Python 3 world. |
|
315 | 316 | elif fn in ('iteritems', 'itervalues'): |
|
316 | 317 | yield t._replace(string=fn[4:]) |
|
317 | 318 | continue |
|
318 | 319 | |
|
319 | 320 | # Emit unmodified token. |
|
320 | 321 | yield t |
|
321 | 322 | |
|
322 | 323 | # Header to add to bytecode files. This MUST be changed when |
|
323 | 324 | # ``replacetoken`` or any mechanism that changes semantics of module |
|
324 | 325 | # loading is changed. Otherwise cached bytecode may get loaded without |
|
325 | 326 | # the new transformation mechanisms applied. |
|
326 |
BYTECODEHEADER = b'HG\x00\x0 |
|
|
327 | BYTECODEHEADER = b'HG\x00\x0a' | |
|
327 | 328 | |
|
328 | 329 | class hgloader(importlib.machinery.SourceFileLoader): |
|
329 | 330 | """Custom module loader that transforms source code. |
|
330 | 331 | |
|
331 | 332 | When the source code is converted to a code object, we transform |
|
332 | 333 | certain patterns to be Python 3 compatible. This allows us to write code |
|
333 | 334 | that is natively Python 2 and compatible with Python 3 without |
|
334 | 335 | making the code excessively ugly. |
|
335 | 336 | |
|
336 | 337 | We do this by transforming the token stream between parse and compile. |
|
337 | 338 | |
|
338 | 339 | Implementing transformations invalidates caching assumptions made |
|
339 | 340 | by the built-in importer. The built-in importer stores a header on |
|
340 | 341 | saved bytecode files indicating the Python/bytecode version. If the |
|
341 | 342 | version changes, the cached bytecode is ignored. The Mercurial |
|
342 | 343 | transformations could change at any time. This means we need to check |
|
343 | 344 | that cached bytecode was generated with the current transformation |
|
344 | 345 | code or there could be a mismatch between cached bytecode and what |
|
345 | 346 | would be generated from this class. |
|
346 | 347 | |
|
347 | 348 | We supplement the bytecode caching layer by wrapping ``get_data`` |
|
348 | 349 | and ``set_data``. These functions are called when the |
|
349 | 350 | ``SourceFileLoader`` retrieves and saves bytecode cache files, |
|
350 | 351 | respectively. We simply add an additional header on the file. As |
|
351 | 352 | long as the version in this file is changed when semantics change, |
|
352 | 353 | cached bytecode should be invalidated when transformations change. |
|
353 | 354 | |
|
354 | 355 | The added header has the form ``HG<VERSION>``. That is a literal |
|
355 | 356 | ``HG`` with 2 binary bytes indicating the transformation version. |
|
356 | 357 | """ |
|
357 | 358 | def get_data(self, path): |
|
358 | 359 | data = super(hgloader, self).get_data(path) |
|
359 | 360 | |
|
360 | 361 | if not path.endswith(tuple(importlib.machinery.BYTECODE_SUFFIXES)): |
|
361 | 362 | return data |
|
362 | 363 | |
|
363 | 364 | # There should be a header indicating the Mercurial transformation |
|
364 | 365 | # version. If it doesn't exist or doesn't match the current version, |
|
365 | 366 | # we raise an OSError because that is what |
|
366 | 367 | # ``SourceFileLoader.get_code()`` expects when loading bytecode |
|
367 | 368 | # paths to indicate the cached file is "bad." |
|
368 | 369 | if data[0:2] != b'HG': |
|
369 | 370 | raise OSError('no hg header') |
|
370 | 371 | if data[0:4] != BYTECODEHEADER: |
|
371 | 372 | raise OSError('hg header version mismatch') |
|
372 | 373 | |
|
373 | 374 | return data[4:] |
|
374 | 375 | |
|
375 | 376 | def set_data(self, path, data, *args, **kwargs): |
|
376 | 377 | if path.endswith(tuple(importlib.machinery.BYTECODE_SUFFIXES)): |
|
377 | 378 | data = BYTECODEHEADER + data |
|
378 | 379 | |
|
379 | 380 | return super(hgloader, self).set_data(path, data, *args, **kwargs) |
|
380 | 381 | |
|
381 | 382 | def source_to_code(self, data, path): |
|
382 | 383 | """Perform token transformation before compilation.""" |
|
383 | 384 | buf = io.BytesIO(data) |
|
384 | 385 | tokens = tokenize.tokenize(buf.readline) |
|
385 | 386 | data = tokenize.untokenize(replacetokens(list(tokens), self.name)) |
|
386 | 387 | # Python's built-in importer strips frames from exceptions raised |
|
387 | 388 | # for this code. Unfortunately, that mechanism isn't extensible |
|
388 | 389 | # and our frame will be blamed for the import failure. There |
|
389 | 390 | # are extremely hacky ways to do frame stripping. We haven't |
|
390 | 391 | # implemented them because they are very ugly. |
|
391 | 392 | return super(hgloader, self).source_to_code(data, path) |
|
392 | 393 | |
|
393 | 394 | # We automagically register our custom importer as a side-effect of loading. |
|
394 | 395 | # This is necessary to ensure that any entry points are able to import |
|
395 | 396 | # mercurial.* modules without having to perform this registration themselves. |
|
396 | 397 | if sys.version_info[0] >= 3: |
|
397 | 398 | _importercls = hgpathentryfinder |
|
398 | 399 | else: |
|
399 | 400 | _importercls = hgimporter |
|
400 | 401 | if not any(isinstance(x, _importercls) for x in sys.meta_path): |
|
401 | 402 | # meta_path is used before any implicit finders and before sys.path. |
|
402 | 403 | sys.meta_path.insert(0, _importercls()) |
@@ -1,409 +1,410 b'' | |||
|
1 | 1 | # pycompat.py - portability shim for python 3 |
|
2 | 2 | # |
|
3 | 3 | # This software may be used and distributed according to the terms of the |
|
4 | 4 | # GNU General Public License version 2 or any later version. |
|
5 | 5 | |
|
6 | 6 | """Mercurial portability shim for python 3. |
|
7 | 7 | |
|
8 | 8 | This contains aliases to hide python version-specific details from the core. |
|
9 | 9 | """ |
|
10 | 10 | |
|
11 | 11 | from __future__ import absolute_import |
|
12 | 12 | |
|
13 | 13 | import getopt |
|
14 | 14 | import os |
|
15 | 15 | import shlex |
|
16 | 16 | import sys |
|
17 | 17 | |
|
18 | 18 | ispy3 = (sys.version_info[0] >= 3) |
|
19 | 19 | |
|
20 | 20 | if not ispy3: |
|
21 | 21 | import cPickle as pickle |
|
22 | 22 | import httplib |
|
23 | 23 | import Queue as _queue |
|
24 | 24 | import SocketServer as socketserver |
|
25 | 25 | import xmlrpclib |
|
26 | 26 | else: |
|
27 | 27 | import http.client as httplib |
|
28 | 28 | import pickle |
|
29 | 29 | import queue as _queue |
|
30 | 30 | import socketserver |
|
31 | 31 | import xmlrpc.client as xmlrpclib |
|
32 | 32 | |
|
33 | 33 | def identity(a): |
|
34 | 34 | return a |
|
35 | 35 | |
|
36 | 36 | if ispy3: |
|
37 | 37 | import builtins |
|
38 | 38 | import functools |
|
39 | 39 | import io |
|
40 | 40 | import struct |
|
41 | 41 | |
|
42 | 42 | fsencode = os.fsencode |
|
43 | 43 | fsdecode = os.fsdecode |
|
44 | 44 | # A bytes version of os.name. |
|
45 | 45 | oslinesep = os.linesep.encode('ascii') |
|
46 | 46 | osname = os.name.encode('ascii') |
|
47 | 47 | ospathsep = os.pathsep.encode('ascii') |
|
48 | 48 | ossep = os.sep.encode('ascii') |
|
49 | 49 | osaltsep = os.altsep |
|
50 | 50 | if osaltsep: |
|
51 | 51 | osaltsep = osaltsep.encode('ascii') |
|
52 | 52 | # os.getcwd() on Python 3 returns string, but it has os.getcwdb() which |
|
53 | 53 | # returns bytes. |
|
54 | 54 | getcwd = os.getcwdb |
|
55 | 55 | sysplatform = sys.platform.encode('ascii') |
|
56 | 56 | sysexecutable = sys.executable |
|
57 | 57 | if sysexecutable: |
|
58 | 58 | sysexecutable = os.fsencode(sysexecutable) |
|
59 | 59 | stringio = io.BytesIO |
|
60 | 60 | maplist = lambda *args: list(map(*args)) |
|
61 | 61 | |
|
62 | 62 | # TODO: .buffer might not exist if std streams were replaced; we'll need |
|
63 | 63 | # a silly wrapper to make a bytes stream backed by a unicode one. |
|
64 | 64 | stdin = sys.stdin.buffer |
|
65 | 65 | stdout = sys.stdout.buffer |
|
66 | 66 | stderr = sys.stderr.buffer |
|
67 | 67 | |
|
68 | 68 | # Since Python 3 converts argv to wchar_t type by Py_DecodeLocale() on Unix, |
|
69 | 69 | # we can use os.fsencode() to get back bytes argv. |
|
70 | 70 | # |
|
71 | 71 | # https://hg.python.org/cpython/file/v3.5.1/Programs/python.c#l55 |
|
72 | 72 | # |
|
73 | 73 | # TODO: On Windows, the native argv is wchar_t, so we'll need a different |
|
74 | 74 | # workaround to simulate the Python 2 (i.e. ANSI Win32 API) behavior. |
|
75 | 75 | if getattr(sys, 'argv', None) is not None: |
|
76 | 76 | sysargv = list(map(os.fsencode, sys.argv)) |
|
77 | 77 | |
|
78 | 78 | bytechr = struct.Struct('>B').pack |
|
79 | 79 | |
|
80 | 80 | class bytestr(bytes): |
|
81 | 81 | """A bytes which mostly acts as a Python 2 str |
|
82 | 82 | |
|
83 | 83 | >>> bytestr(), bytestr(bytearray(b'foo')), bytestr(u'ascii'), bytestr(1) |
|
84 | 84 | (b'', b'foo', b'ascii', b'1') |
|
85 | 85 | >>> s = bytestr(b'foo') |
|
86 | 86 | >>> assert s is bytestr(s) |
|
87 | 87 | |
|
88 | 88 | There's no implicit conversion from non-ascii str as its encoding is |
|
89 | 89 | unknown: |
|
90 | 90 | |
|
91 | 91 | >>> bytestr(chr(0x80)) # doctest: +ELLIPSIS |
|
92 | 92 | Traceback (most recent call last): |
|
93 | 93 | ... |
|
94 | 94 | UnicodeEncodeError: ... |
|
95 | 95 | |
|
96 | 96 | Comparison between bytestr and bytes should work: |
|
97 | 97 | |
|
98 | 98 | >>> assert bytestr(b'foo') == b'foo' |
|
99 | 99 | >>> assert b'foo' == bytestr(b'foo') |
|
100 | 100 | >>> assert b'f' in bytestr(b'foo') |
|
101 | 101 | >>> assert bytestr(b'f') in b'foo' |
|
102 | 102 | |
|
103 | 103 | Sliced elements should be bytes, not integer: |
|
104 | 104 | |
|
105 | 105 | >>> s[1], s[:2] |
|
106 | 106 | (b'o', b'fo') |
|
107 | 107 | >>> list(s), list(reversed(s)) |
|
108 | 108 | ([b'f', b'o', b'o'], [b'o', b'o', b'f']) |
|
109 | 109 | |
|
110 | 110 | As bytestr type isn't propagated across operations, you need to cast |
|
111 | 111 | bytes to bytestr explicitly: |
|
112 | 112 | |
|
113 | 113 | >>> s = bytestr(b'foo').upper() |
|
114 | 114 | >>> t = bytestr(s) |
|
115 | 115 | >>> s[0], t[0] |
|
116 | 116 | (70, b'F') |
|
117 | 117 | |
|
118 | 118 | Be careful to not pass a bytestr object to a function which expects |
|
119 | 119 | bytearray-like behavior. |
|
120 | 120 | |
|
121 | 121 | >>> t = bytes(t) # cast to bytes |
|
122 | 122 | >>> assert type(t) is bytes |
|
123 | 123 | """ |
|
124 | 124 | |
|
125 | 125 | def __new__(cls, s=b''): |
|
126 | 126 | if isinstance(s, bytestr): |
|
127 | 127 | return s |
|
128 | 128 | if not isinstance(s, (bytes, bytearray)): |
|
129 | 129 | s = str(s).encode(u'ascii') |
|
130 | 130 | return bytes.__new__(cls, s) |
|
131 | 131 | |
|
132 | 132 | def __getitem__(self, key): |
|
133 | 133 | s = bytes.__getitem__(self, key) |
|
134 | 134 | if not isinstance(s, bytes): |
|
135 | 135 | s = bytechr(s) |
|
136 | 136 | return s |
|
137 | 137 | |
|
138 | 138 | def __iter__(self): |
|
139 | 139 | return iterbytestr(bytes.__iter__(self)) |
|
140 | 140 | |
|
141 | 141 | def iterbytestr(s): |
|
142 | 142 | """Iterate bytes as if it were a str object of Python 2""" |
|
143 | 143 | return map(bytechr, s) |
|
144 | 144 | |
|
145 | 145 | def sysbytes(s): |
|
146 | 146 | """Convert an internal str (e.g. keyword, __doc__) back to bytes |
|
147 | 147 | |
|
148 | 148 | This never raises UnicodeEncodeError, but only ASCII characters |
|
149 | 149 | can be round-trip by sysstr(sysbytes(s)). |
|
150 | 150 | """ |
|
151 | 151 | return s.encode(u'utf-8') |
|
152 | 152 | |
|
153 | 153 | def sysstr(s): |
|
154 | 154 | """Return a keyword str to be passed to Python functions such as |
|
155 | 155 | getattr() and str.encode() |
|
156 | 156 | |
|
157 | 157 | This never raises UnicodeDecodeError. Non-ascii characters are |
|
158 | 158 | considered invalid and mapped to arbitrary but unique code points |
|
159 | 159 | such that 'sysstr(a) != sysstr(b)' for all 'a != b'. |
|
160 | 160 | """ |
|
161 | 161 | if isinstance(s, builtins.str): |
|
162 | 162 | return s |
|
163 | 163 | return s.decode(u'latin-1') |
|
164 | 164 | |
|
165 | 165 | def _wrapattrfunc(f): |
|
166 | 166 | @functools.wraps(f) |
|
167 | 167 | def w(object, name, *args): |
|
168 | 168 | return f(object, sysstr(name), *args) |
|
169 | 169 | return w |
|
170 | 170 | |
|
171 | 171 | # these wrappers are automagically imported by hgloader |
|
172 | 172 | delattr = _wrapattrfunc(builtins.delattr) |
|
173 | 173 | getattr = _wrapattrfunc(builtins.getattr) |
|
174 | 174 | hasattr = _wrapattrfunc(builtins.hasattr) |
|
175 | 175 | setattr = _wrapattrfunc(builtins.setattr) |
|
176 | 176 | xrange = builtins.range |
|
177 | unicode = str | |
|
177 | 178 | |
|
178 | 179 | def open(name, mode='r', buffering=-1): |
|
179 | 180 | return builtins.open(name, sysstr(mode), buffering) |
|
180 | 181 | |
|
181 | 182 | # getopt.getopt() on Python 3 deals with unicodes internally so we cannot |
|
182 | 183 | # pass bytes there. Passing unicodes will result in unicodes as return |
|
183 | 184 | # values which we need to convert again to bytes. |
|
184 | 185 | def getoptb(args, shortlist, namelist): |
|
185 | 186 | args = [a.decode('latin-1') for a in args] |
|
186 | 187 | shortlist = shortlist.decode('latin-1') |
|
187 | 188 | namelist = [a.decode('latin-1') for a in namelist] |
|
188 | 189 | opts, args = getopt.getopt(args, shortlist, namelist) |
|
189 | 190 | opts = [(a[0].encode('latin-1'), a[1].encode('latin-1')) |
|
190 | 191 | for a in opts] |
|
191 | 192 | args = [a.encode('latin-1') for a in args] |
|
192 | 193 | return opts, args |
|
193 | 194 | |
|
194 | 195 | # keys of keyword arguments in Python need to be strings which are unicodes |
|
195 | 196 | # Python 3. This function takes keyword arguments, convert the keys to str. |
|
196 | 197 | def strkwargs(dic): |
|
197 | 198 | dic = dict((k.decode('latin-1'), v) for k, v in dic.iteritems()) |
|
198 | 199 | return dic |
|
199 | 200 | |
|
200 | 201 | # keys of keyword arguments need to be unicode while passing into |
|
201 | 202 | # a function. This function helps us to convert those keys back to bytes |
|
202 | 203 | # again as we need to deal with bytes. |
|
203 | 204 | def byteskwargs(dic): |
|
204 | 205 | dic = dict((k.encode('latin-1'), v) for k, v in dic.iteritems()) |
|
205 | 206 | return dic |
|
206 | 207 | |
|
207 | 208 | # shlex.split() accepts unicodes on Python 3. This function takes bytes |
|
208 | 209 | # argument, convert it into unicodes, pass into shlex.split(), convert the |
|
209 | 210 | # returned value to bytes and return that. |
|
210 | 211 | # TODO: handle shlex.shlex(). |
|
211 | 212 | def shlexsplit(s): |
|
212 | 213 | ret = shlex.split(s.decode('latin-1')) |
|
213 | 214 | return [a.encode('latin-1') for a in ret] |
|
214 | 215 | |
|
215 | 216 | else: |
|
216 | 217 | import cStringIO |
|
217 | 218 | |
|
218 | 219 | bytechr = chr |
|
219 | 220 | bytestr = str |
|
220 | 221 | iterbytestr = iter |
|
221 | 222 | sysbytes = identity |
|
222 | 223 | sysstr = identity |
|
223 | 224 | |
|
224 | 225 | # Partial backport from os.py in Python 3, which only accepts bytes. |
|
225 | 226 | # In Python 2, our paths should only ever be bytes, a unicode path |
|
226 | 227 | # indicates a bug. |
|
227 | 228 | def fsencode(filename): |
|
228 | 229 | if isinstance(filename, str): |
|
229 | 230 | return filename |
|
230 | 231 | else: |
|
231 | 232 | raise TypeError( |
|
232 | 233 | "expect str, not %s" % type(filename).__name__) |
|
233 | 234 | |
|
234 | 235 | # In Python 2, fsdecode() has a very chance to receive bytes. So it's |
|
235 | 236 | # better not to touch Python 2 part as it's already working fine. |
|
236 | 237 | fsdecode = identity |
|
237 | 238 | |
|
238 | 239 | def getoptb(args, shortlist, namelist): |
|
239 | 240 | return getopt.getopt(args, shortlist, namelist) |
|
240 | 241 | |
|
241 | 242 | strkwargs = identity |
|
242 | 243 | byteskwargs = identity |
|
243 | 244 | |
|
244 | 245 | oslinesep = os.linesep |
|
245 | 246 | osname = os.name |
|
246 | 247 | ospathsep = os.pathsep |
|
247 | 248 | ossep = os.sep |
|
248 | 249 | osaltsep = os.altsep |
|
249 | 250 | stdin = sys.stdin |
|
250 | 251 | stdout = sys.stdout |
|
251 | 252 | stderr = sys.stderr |
|
252 | 253 | if getattr(sys, 'argv', None) is not None: |
|
253 | 254 | sysargv = sys.argv |
|
254 | 255 | sysplatform = sys.platform |
|
255 | 256 | getcwd = os.getcwd |
|
256 | 257 | sysexecutable = sys.executable |
|
257 | 258 | shlexsplit = shlex.split |
|
258 | 259 | stringio = cStringIO.StringIO |
|
259 | 260 | maplist = map |
|
260 | 261 | |
|
261 | 262 | empty = _queue.Empty |
|
262 | 263 | queue = _queue.Queue |
|
263 | 264 | |
|
264 | 265 | class _pycompatstub(object): |
|
265 | 266 | def __init__(self): |
|
266 | 267 | self._aliases = {} |
|
267 | 268 | |
|
268 | 269 | def _registeraliases(self, origin, items): |
|
269 | 270 | """Add items that will be populated at the first access""" |
|
270 | 271 | items = map(sysstr, items) |
|
271 | 272 | self._aliases.update( |
|
272 | 273 | (item.replace(sysstr('_'), sysstr('')).lower(), (origin, item)) |
|
273 | 274 | for item in items) |
|
274 | 275 | |
|
275 | 276 | def _registeralias(self, origin, attr, name): |
|
276 | 277 | """Alias ``origin``.``attr`` as ``name``""" |
|
277 | 278 | self._aliases[sysstr(name)] = (origin, sysstr(attr)) |
|
278 | 279 | |
|
279 | 280 | def __getattr__(self, name): |
|
280 | 281 | try: |
|
281 | 282 | origin, item = self._aliases[name] |
|
282 | 283 | except KeyError: |
|
283 | 284 | raise AttributeError(name) |
|
284 | 285 | self.__dict__[name] = obj = getattr(origin, item) |
|
285 | 286 | return obj |
|
286 | 287 | |
|
287 | 288 | httpserver = _pycompatstub() |
|
288 | 289 | urlreq = _pycompatstub() |
|
289 | 290 | urlerr = _pycompatstub() |
|
290 | 291 | if not ispy3: |
|
291 | 292 | import BaseHTTPServer |
|
292 | 293 | import CGIHTTPServer |
|
293 | 294 | import SimpleHTTPServer |
|
294 | 295 | import urllib2 |
|
295 | 296 | import urllib |
|
296 | 297 | import urlparse |
|
297 | 298 | urlreq._registeraliases(urllib, ( |
|
298 | 299 | "addclosehook", |
|
299 | 300 | "addinfourl", |
|
300 | 301 | "ftpwrapper", |
|
301 | 302 | "pathname2url", |
|
302 | 303 | "quote", |
|
303 | 304 | "splitattr", |
|
304 | 305 | "splitpasswd", |
|
305 | 306 | "splitport", |
|
306 | 307 | "splituser", |
|
307 | 308 | "unquote", |
|
308 | 309 | "url2pathname", |
|
309 | 310 | "urlencode", |
|
310 | 311 | )) |
|
311 | 312 | urlreq._registeraliases(urllib2, ( |
|
312 | 313 | "AbstractHTTPHandler", |
|
313 | 314 | "BaseHandler", |
|
314 | 315 | "build_opener", |
|
315 | 316 | "FileHandler", |
|
316 | 317 | "FTPHandler", |
|
317 | 318 | "HTTPBasicAuthHandler", |
|
318 | 319 | "HTTPDigestAuthHandler", |
|
319 | 320 | "HTTPHandler", |
|
320 | 321 | "HTTPPasswordMgrWithDefaultRealm", |
|
321 | 322 | "HTTPSHandler", |
|
322 | 323 | "install_opener", |
|
323 | 324 | "ProxyHandler", |
|
324 | 325 | "Request", |
|
325 | 326 | "urlopen", |
|
326 | 327 | )) |
|
327 | 328 | urlreq._registeraliases(urlparse, ( |
|
328 | 329 | "urlparse", |
|
329 | 330 | "urlunparse", |
|
330 | 331 | )) |
|
331 | 332 | urlerr._registeraliases(urllib2, ( |
|
332 | 333 | "HTTPError", |
|
333 | 334 | "URLError", |
|
334 | 335 | )) |
|
335 | 336 | httpserver._registeraliases(BaseHTTPServer, ( |
|
336 | 337 | "HTTPServer", |
|
337 | 338 | "BaseHTTPRequestHandler", |
|
338 | 339 | )) |
|
339 | 340 | httpserver._registeraliases(SimpleHTTPServer, ( |
|
340 | 341 | "SimpleHTTPRequestHandler", |
|
341 | 342 | )) |
|
342 | 343 | httpserver._registeraliases(CGIHTTPServer, ( |
|
343 | 344 | "CGIHTTPRequestHandler", |
|
344 | 345 | )) |
|
345 | 346 | |
|
346 | 347 | else: |
|
347 | 348 | import urllib.parse |
|
348 | 349 | urlreq._registeraliases(urllib.parse, ( |
|
349 | 350 | "splitattr", |
|
350 | 351 | "splitpasswd", |
|
351 | 352 | "splitport", |
|
352 | 353 | "splituser", |
|
353 | 354 | "urlparse", |
|
354 | 355 | "urlunparse", |
|
355 | 356 | )) |
|
356 | 357 | urlreq._registeralias(urllib.parse, "unquote_to_bytes", "unquote") |
|
357 | 358 | import urllib.request |
|
358 | 359 | urlreq._registeraliases(urllib.request, ( |
|
359 | 360 | "AbstractHTTPHandler", |
|
360 | 361 | "BaseHandler", |
|
361 | 362 | "build_opener", |
|
362 | 363 | "FileHandler", |
|
363 | 364 | "FTPHandler", |
|
364 | 365 | "ftpwrapper", |
|
365 | 366 | "HTTPHandler", |
|
366 | 367 | "HTTPSHandler", |
|
367 | 368 | "install_opener", |
|
368 | 369 | "pathname2url", |
|
369 | 370 | "HTTPBasicAuthHandler", |
|
370 | 371 | "HTTPDigestAuthHandler", |
|
371 | 372 | "HTTPPasswordMgrWithDefaultRealm", |
|
372 | 373 | "ProxyHandler", |
|
373 | 374 | "Request", |
|
374 | 375 | "url2pathname", |
|
375 | 376 | "urlopen", |
|
376 | 377 | )) |
|
377 | 378 | import urllib.response |
|
378 | 379 | urlreq._registeraliases(urllib.response, ( |
|
379 | 380 | "addclosehook", |
|
380 | 381 | "addinfourl", |
|
381 | 382 | )) |
|
382 | 383 | import urllib.error |
|
383 | 384 | urlerr._registeraliases(urllib.error, ( |
|
384 | 385 | "HTTPError", |
|
385 | 386 | "URLError", |
|
386 | 387 | )) |
|
387 | 388 | import http.server |
|
388 | 389 | httpserver._registeraliases(http.server, ( |
|
389 | 390 | "HTTPServer", |
|
390 | 391 | "BaseHTTPRequestHandler", |
|
391 | 392 | "SimpleHTTPRequestHandler", |
|
392 | 393 | "CGIHTTPRequestHandler", |
|
393 | 394 | )) |
|
394 | 395 | |
|
395 | 396 | # urllib.parse.quote() accepts both str and bytes, decodes bytes |
|
396 | 397 | # (if necessary), and returns str. This is wonky. We provide a custom |
|
397 | 398 | # implementation that only accepts bytes and emits bytes. |
|
398 | 399 | def quote(s, safe=r'/'): |
|
399 | 400 | s = urllib.parse.quote_from_bytes(s, safe=safe) |
|
400 | 401 | return s.encode('ascii', 'strict') |
|
401 | 402 | |
|
402 | 403 | # urllib.parse.urlencode() returns str. We use this function to make |
|
403 | 404 | # sure we return bytes. |
|
404 | 405 | def urlencode(query, doseq=False): |
|
405 | 406 | s = urllib.parse.urlencode(query, doseq=doseq) |
|
406 | 407 | return s.encode('ascii') |
|
407 | 408 | |
|
408 | 409 | urlreq.quote = quote |
|
409 | 410 | urlreq.urlencode = urlencode |
General Comments 0
You need to be logged in to leave comments.
Login now