1 """Module/script to byte-compile all .py files to .pyc files.
2
3 When called as a script with arguments, this compiles the directories
4 given as arguments recursively; the -l option prevents it from
5 recursing into directories.
6
7 Without arguments, it compiles all modules on sys.path, without
8 recursing into subdirectories. (Even though it should do so for
9 packages -- for now, you'll have to deal with packages separately.)
10
11 See module py_compile for details of the actual byte-compilation.
12 """
13 import os
14 import sys
15 import importlib.util
16 import py_compile
17 import struct
18 import filecmp
19
20 from functools import partial
21 from pathlib import Path
22
23 __all__ = ["compile_dir","compile_file","compile_path"]
24
25 def _walk_dir(dir, maxlevels, quiet=0):
26 if quiet < 2 and isinstance(dir, os.PathLike):
27 dir = os.fspath(dir)
28 if not quiet:
29 print('Listing {!r}...'.format(dir))
30 try:
31 names = os.listdir(dir)
32 except OSError:
33 if quiet < 2:
34 print("Can't list {!r}".format(dir))
35 names = []
36 names.sort()
37 for name in names:
38 if name == '__pycache__':
39 continue
40 fullname = os.path.join(dir, name)
41 if not os.path.isdir(fullname):
42 yield fullname
43 elif (maxlevels > 0 and name != os.curdir and name != os.pardir and
44 os.path.isdir(fullname) and not os.path.islink(fullname)):
45 yield from _walk_dir(fullname, maxlevels=maxlevels - 1,
46 quiet=quiet)
47
48 def compile_dir(dir, maxlevels=None, ddir=None, force=False,
49 rx=None, quiet=0, legacy=False, optimize=-1, workers=1,
50 invalidation_mode=None, *, stripdir=None,
51 prependdir=None, limit_sl_dest=None, hardlink_dupes=False):
52 """Byte-compile all modules in the given directory tree.
53
54 Arguments (only dir is required):
55
56 dir: the directory to byte-compile
57 maxlevels: maximum recursion level (default `sys.getrecursionlimit()`)
58 ddir: the directory that will be prepended to the path to the
59 file as it is compiled into each byte-code file.
60 force: if True, force compilation, even if timestamps are up-to-date
61 quiet: full output with False or 0, errors only with 1,
62 no output with 2
63 legacy: if True, produce legacy pyc paths instead of PEP 3147 paths
64 optimize: int or list of optimization levels or -1 for level of
65 the interpreter. Multiple levels leads to multiple compiled
66 files each with one optimization level.
67 workers: maximum number of parallel workers
68 invalidation_mode: how the up-to-dateness of the pyc will be checked
69 stripdir: part of path to left-strip from source file path
70 prependdir: path to prepend to beginning of original file path, applied
71 after stripdir
72 limit_sl_dest: ignore symlinks if they are pointing outside of
73 the defined path
74 hardlink_dupes: hardlink duplicated pyc files
75 """
76 ProcessPoolExecutor = None
77 if ddir is not None and (stripdir is not None or prependdir is not None):
78 raise ValueError(("Destination dir (ddir) cannot be used "
79 "in combination with stripdir or prependdir"))
80 if ddir is not None:
81 stripdir = dir
82 prependdir = ddir
83 ddir = None
84 if workers < 0:
85 raise ValueError('workers must be greater or equal to 0')
86 if workers != 1:
87 # Check if this is a system where ProcessPoolExecutor can function.
88 from concurrent.futures.process import _check_system_limits
89 try:
90 _check_system_limits()
91 except NotImplementedError:
92 workers = 1
93 else:
94 from concurrent.futures import ProcessPoolExecutor
95 if maxlevels is None:
96 maxlevels = sys.getrecursionlimit()
97 files = _walk_dir(dir, quiet=quiet, maxlevels=maxlevels)
98 success = True
99 if workers != 1 and ProcessPoolExecutor is not None:
100 import multiprocessing
101 if multiprocessing.get_start_method() == 'fork':
102 mp_context = multiprocessing.get_context('forkserver')
103 else:
104 mp_context = None
105 # If workers == 0, let ProcessPoolExecutor choose
106 workers = workers or None
107 with ProcessPoolExecutor(max_workers=workers,
108 mp_context=mp_context) as executor:
109 results = executor.map(partial(compile_file,
110 ddir=ddir, force=force,
111 rx=rx, quiet=quiet,
112 legacy=legacy,
113 optimize=optimize,
114 invalidation_mode=invalidation_mode,
115 stripdir=stripdir,
116 prependdir=prependdir,
117 limit_sl_dest=limit_sl_dest,
118 hardlink_dupes=hardlink_dupes),
119 files)
120 success = min(results, default=True)
121 else:
122 for file in files:
123 if not compile_file(file, ddir, force, rx, quiet,
124 legacy, optimize, invalidation_mode,
125 stripdir=stripdir, prependdir=prependdir,
126 limit_sl_dest=limit_sl_dest,
127 hardlink_dupes=hardlink_dupes):
128 success = False
129 return success
130
131 def compile_file(fullname, ddir=None, force=False, rx=None, quiet=0,
132 legacy=False, optimize=-1,
133 invalidation_mode=None, *, stripdir=None, prependdir=None,
134 limit_sl_dest=None, hardlink_dupes=False):
135 """Byte-compile one file.
136
137 Arguments (only fullname is required):
138
139 fullname: the file to byte-compile
140 ddir: if given, the directory name compiled in to the
141 byte-code file.
142 force: if True, force compilation, even if timestamps are up-to-date
143 quiet: full output with False or 0, errors only with 1,
144 no output with 2
145 legacy: if True, produce legacy pyc paths instead of PEP 3147 paths
146 optimize: int or list of optimization levels or -1 for level of
147 the interpreter. Multiple levels leads to multiple compiled
148 files each with one optimization level.
149 invalidation_mode: how the up-to-dateness of the pyc will be checked
150 stripdir: part of path to left-strip from source file path
151 prependdir: path to prepend to beginning of original file path, applied
152 after stripdir
153 limit_sl_dest: ignore symlinks if they are pointing outside of
154 the defined path.
155 hardlink_dupes: hardlink duplicated pyc files
156 """
157
158 if ddir is not None and (stripdir is not None or prependdir is not None):
159 raise ValueError(("Destination dir (ddir) cannot be used "
160 "in combination with stripdir or prependdir"))
161
162 success = True
163 fullname = os.fspath(fullname)
164 stripdir = os.fspath(stripdir) if stripdir is not None else None
165 name = os.path.basename(fullname)
166
167 dfile = None
168
169 if ddir is not None:
170 dfile = os.path.join(ddir, name)
171
172 if stripdir is not None:
173 fullname_parts = fullname.split(os.path.sep)
174 stripdir_parts = stripdir.split(os.path.sep)
175 ddir_parts = list(fullname_parts)
176
177 for spart, opart in zip(stripdir_parts, fullname_parts):
178 if spart == opart:
179 ddir_parts.remove(spart)
180
181 dfile = os.path.join(*ddir_parts)
182
183 if prependdir is not None:
184 if dfile is None:
185 dfile = os.path.join(prependdir, fullname)
186 else:
187 dfile = os.path.join(prependdir, dfile)
188
189 if isinstance(optimize, int):
190 optimize = [optimize]
191
192 # Use set() to remove duplicates.
193 # Use sorted() to create pyc files in a deterministic order.
194 optimize = sorted(set(optimize))
195
196 if hardlink_dupes and len(optimize) < 2:
197 raise ValueError("Hardlinking of duplicated bytecode makes sense "
198 "only for more than one optimization level")
199
200 if rx is not None:
201 mo = rx.search(fullname)
202 if mo:
203 return success
204
205 if limit_sl_dest is not None and os.path.islink(fullname):
206 if Path(limit_sl_dest).resolve() not in Path(fullname).resolve().parents:
207 return success
208
209 opt_cfiles = {}
210
211 if os.path.isfile(fullname):
212 for opt_level in optimize:
213 if legacy:
214 opt_cfiles[opt_level] = fullname + 'c'
215 else:
216 if opt_level >= 0:
217 opt = opt_level if opt_level >= 1 else ''
218 cfile = (importlib.util.cache_from_source(
219 fullname, optimization=opt))
220 opt_cfiles[opt_level] = cfile
221 else:
222 cfile = importlib.util.cache_from_source(fullname)
223 opt_cfiles[opt_level] = cfile
224
225 head, tail = name[:-3], name[-3:]
226 if tail == '.py':
227 if not force:
228 try:
229 mtime = int(os.stat(fullname).st_mtime)
230 expect = struct.pack('<4sLL', importlib.util.MAGIC_NUMBER,
231 0, mtime & 0xFFFF_FFFF)
232 for cfile in opt_cfiles.values():
233 with open(cfile, 'rb') as chandle:
234 actual = chandle.read(12)
235 if expect != actual:
236 break
237 else:
238 return success
239 except OSError:
240 pass
241 if not quiet:
242 print('Compiling {!r}...'.format(fullname))
243 try:
244 for index, opt_level in enumerate(optimize):
245 cfile = opt_cfiles[opt_level]
246 ok = py_compile.compile(fullname, cfile, dfile, True,
247 optimize=opt_level,
248 invalidation_mode=invalidation_mode)
249 if index > 0 and hardlink_dupes:
250 previous_cfile = opt_cfiles[optimize[index - 1]]
251 if filecmp.cmp(cfile, previous_cfile, shallow=False):
252 os.unlink(cfile)
253 os.link(previous_cfile, cfile)
254 except py_compile.PyCompileError as err:
255 success = False
256 if quiet >= 2:
257 return success
258 elif quiet:
259 print('*** Error compiling {!r}...'.format(fullname))
260 else:
261 print('*** ', end='')
262 # escape non-printable characters in msg
263 encoding = sys.stdout.encoding or sys.getdefaultencoding()
264 msg = err.msg.encode(encoding, errors='backslashreplace').decode(encoding)
265 print(msg)
266 except (SyntaxError, UnicodeError, OSError) as e:
267 success = False
268 if quiet >= 2:
269 return success
270 elif quiet:
271 print('*** Error compiling {!r}...'.format(fullname))
272 else:
273 print('*** ', end='')
274 print(e.__class__.__name__ + ':', e)
275 else:
276 if ok == 0:
277 success = False
278 return success
279
280 def compile_path(skip_curdir=1, maxlevels=0, force=False, quiet=0,
281 legacy=False, optimize=-1,
282 invalidation_mode=None):
283 """Byte-compile all module on sys.path.
284
285 Arguments (all optional):
286
287 skip_curdir: if true, skip current directory (default True)
288 maxlevels: max recursion level (default 0)
289 force: as for compile_dir() (default False)
290 quiet: as for compile_dir() (default 0)
291 legacy: as for compile_dir() (default False)
292 optimize: as for compile_dir() (default -1)
293 invalidation_mode: as for compiler_dir()
294 """
295 success = True
296 for dir in sys.path:
297 if (not dir or dir == os.curdir) and skip_curdir:
298 if quiet < 2:
299 print('Skipping current directory')
300 else:
301 success = success and compile_dir(
302 dir,
303 maxlevels,
304 None,
305 force,
306 quiet=quiet,
307 legacy=legacy,
308 optimize=optimize,
309 invalidation_mode=invalidation_mode,
310 )
311 return success
312
313
314 def main():
315 """Script main program."""
316 import argparse
317
318 parser = argparse.ArgumentParser(
319 description='Utilities to support installing Python libraries.')
320 parser.add_argument('-l', action='store_const', const=0,
321 default=None, dest='maxlevels',
322 help="don't recurse into subdirectories")
323 parser.add_argument('-r', type=int, dest='recursion',
324 help=('control the maximum recursion level. '
325 'if `-l` and `-r` options are specified, '
326 'then `-r` takes precedence.'))
327 parser.add_argument('-f', action='store_true', dest='force',
328 help='force rebuild even if timestamps are up to date')
329 parser.add_argument('-q', action='count', dest='quiet', default=0,
330 help='output only error messages; -qq will suppress '
331 'the error messages as well.')
332 parser.add_argument('-b', action='store_true', dest='legacy',
333 help='use legacy (pre-PEP3147) compiled file locations')
334 parser.add_argument('-d', metavar='DESTDIR', dest='ddir', default=None,
335 help=('directory to prepend to file paths for use in '
336 'compile-time tracebacks and in runtime '
337 'tracebacks in cases where the source file is '
338 'unavailable'))
339 parser.add_argument('-s', metavar='STRIPDIR', dest='stripdir',
340 default=None,
341 help=('part of path to left-strip from path '
342 'to source file - for example buildroot. '
343 '`-d` and `-s` options cannot be '
344 'specified together.'))
345 parser.add_argument('-p', metavar='PREPENDDIR', dest='prependdir',
346 default=None,
347 help=('path to add as prefix to path '
348 'to source file - for example / to make '
349 'it absolute when some part is removed '
350 'by `-s` option. '
351 '`-d` and `-p` options cannot be '
352 'specified together.'))
353 parser.add_argument('-x', metavar='REGEXP', dest='rx', default=None,
354 help=('skip files matching the regular expression; '
355 'the regexp is searched for in the full path '
356 'of each file considered for compilation'))
357 parser.add_argument('-i', metavar='FILE', dest='flist',
358 help=('add all the files and directories listed in '
359 'FILE to the list considered for compilation; '
360 'if "-", names are read from stdin'))
361 parser.add_argument('compile_dest', metavar='FILE|DIR', nargs='*',
362 help=('zero or more file and directory names '
363 'to compile; if no arguments given, defaults '
364 'to the equivalent of -l sys.path'))
365 parser.add_argument('-j', '--workers', default=1,
366 type=int, help='Run compileall concurrently')
367 invalidation_modes = [mode.name.lower().replace('_', '-')
368 for mode in py_compile.PycInvalidationMode]
369 parser.add_argument('--invalidation-mode',
370 choices=sorted(invalidation_modes),
371 help=('set .pyc invalidation mode; defaults to '
372 '"checked-hash" if the SOURCE_DATE_EPOCH '
373 'environment variable is set, and '
374 '"timestamp" otherwise.'))
375 parser.add_argument('-o', action='append', type=int, dest='opt_levels',
376 help=('Optimization levels to run compilation with. '
377 'Default is -1 which uses the optimization level '
378 'of the Python interpreter itself (see -O).'))
379 parser.add_argument('-e', metavar='DIR', dest='limit_sl_dest',
380 help='Ignore symlinks pointing outsite of the DIR')
381 parser.add_argument('--hardlink-dupes', action='store_true',
382 dest='hardlink_dupes',
383 help='Hardlink duplicated pyc files')
384
385 args = parser.parse_args()
386 compile_dests = args.compile_dest
387
388 if args.rx:
389 import re
390 args.rx = re.compile(args.rx)
391
392 if args.limit_sl_dest == "":
393 args.limit_sl_dest = None
394
395 if args.recursion is not None:
396 maxlevels = args.recursion
397 else:
398 maxlevels = args.maxlevels
399
400 if args.opt_levels is None:
401 args.opt_levels = [-1]
402
403 if len(args.opt_levels) == 1 and args.hardlink_dupes:
404 parser.error(("Hardlinking of duplicated bytecode makes sense "
405 "only for more than one optimization level."))
406
407 if args.ddir is not None and (
408 args.stripdir is not None or args.prependdir is not None
409 ):
410 parser.error("-d cannot be used in combination with -s or -p")
411
412 # if flist is provided then load it
413 if args.flist:
414 try:
415 with (sys.stdin if args.flist=='-' else
416 open(args.flist, encoding="utf-8")) as f:
417 for line in f:
418 compile_dests.append(line.strip())
419 except OSError:
420 if args.quiet < 2:
421 print("Error reading file list {}".format(args.flist))
422 return False
423
424 if args.invalidation_mode:
425 ivl_mode = args.invalidation_mode.replace('-', '_').upper()
426 invalidation_mode = py_compile.PycInvalidationMode[ivl_mode]
427 else:
428 invalidation_mode = None
429
430 success = True
431 try:
432 if compile_dests:
433 for dest in compile_dests:
434 if os.path.isfile(dest):
435 if not compile_file(dest, args.ddir, args.force, args.rx,
436 args.quiet, args.legacy,
437 invalidation_mode=invalidation_mode,
438 stripdir=args.stripdir,
439 prependdir=args.prependdir,
440 optimize=args.opt_levels,
441 limit_sl_dest=args.limit_sl_dest,
442 hardlink_dupes=args.hardlink_dupes):
443 success = False
444 else:
445 if not compile_dir(dest, maxlevels, args.ddir,
446 args.force, args.rx, args.quiet,
447 args.legacy, workers=args.workers,
448 invalidation_mode=invalidation_mode,
449 stripdir=args.stripdir,
450 prependdir=args.prependdir,
451 optimize=args.opt_levels,
452 limit_sl_dest=args.limit_sl_dest,
453 hardlink_dupes=args.hardlink_dupes):
454 success = False
455 return success
456 else:
457 return compile_path(legacy=args.legacy, force=args.force,
458 quiet=args.quiet,
459 invalidation_mode=invalidation_mode)
460 except KeyboardInterrupt:
461 if args.quiet < 2:
462 print("\n[interrupted]")
463 return False
464 return True
465
466
467 if __name__ == '__main__':
468 exit_status = int(not main())
469 sys.exit(exit_status)