OLD | NEW |
---|---|
(Empty) | |
1 # Copyright (c) 2013 The Chromium Authors. All rights reserved. | |
2 # Use of this source code is governed by a BSD-style license that can be | |
3 # found in the LICENSE file. | |
4 | |
5 # Monkeypatch IMapIterator so that Ctrl-C can kill everything properly. | |
6 # Derived from https://gist.github.com/aljungberg/626518 | |
7 import multiprocessing.pool | |
8 from multiprocessing.pool import IMapIterator | |
9 def wrapper(func): | |
10 def wrap(self, timeout=None): | |
11 return func(self, timeout=timeout or 1e100) | |
12 return wrap | |
13 IMapIterator.next = wrapper(IMapIterator.next) | |
14 IMapIterator.__next__ = IMapIterator.next | |
15 | |
16 | |
17 import binascii | |
18 import contextlib | |
19 import functools | |
20 import logging | |
21 import signal | |
22 import subprocess | |
23 import sys | |
24 import tempfile | |
25 import threading | |
26 | |
27 | |
28 GIT_EXE = 'git.bat' if sys.platform.startswith('win') else 'git' | |
29 | |
30 | |
31 class CalledProcessError(Exception): | |
32 def __init__(self, returncode, cmd): | |
33 super(CalledProcessError, self).__init__() | |
34 self.returncode = returncode | |
35 self.cmd = cmd | |
36 | |
37 def __str__(self): | |
38 return ( | |
39 'Command "%s" returned non-zero exit status %d' % | |
40 (self.cmd, self.returncode)) | |
41 | |
42 | |
43 def memoize_one(**kwargs): | |
44 """Memoizes a single-argument pure function. | |
45 | |
46 Values of None are not cached. | |
47 | |
48 Kwargs: | |
49 threadsafe (bool) - REQUIRED. Specifies whether to use locking around | |
50 cache manipulation functions. This is a kwarg so that users of memoize_one | |
51 are forced to explicitly and verbosely pick True or False. | |
52 | |
53 Adds three methods to the decorated function: | |
54 * get(key, default=None) - Gets the value for this key from the cache. | |
55 * set(key, value) - Sets the value for this key from the cache. | |
56 * clear() - Drops the entire contents of the cache. Useful for unittests. | |
57 * update(other) - Updates the contents of the cache from another dict. | |
58 """ | |
59 assert 'threadsafe' in kwargs, "Must specify threadsafe={True,False}" | |
60 threadsafe = kwargs['threadsafe'] | |
61 | |
62 if threadsafe: | |
63 def withlock(lock, f): | |
64 def inner(*args, **kwargs): | |
65 with lock: | |
66 return f(*args, **kwargs) | |
67 return inner | |
68 else: | |
69 def withlock(_lock, f): | |
70 return f | |
71 | |
72 def decorator(f): | |
73 # Instantiate the lock in decorator, in case users of memoize_one do: | |
74 # | |
75 # memoizer = memoize_one(threadsafe=True) | |
76 # | |
77 # @memoizer | |
78 # def fn1(val): ... | |
79 # | |
80 # @memoizer | |
81 # def fn2(val): ... | |
82 | |
83 lock = threading.Lock() if threadsafe else None | |
84 cache = {} | |
85 _get = withlock(lock, cache.get) | |
86 _set = withlock(lock, cache.__setitem__) | |
87 | |
88 @functools.wraps(f) | |
89 def inner(arg): | |
90 ret = _get(arg) | |
91 if ret is None: | |
92 ret = f(arg) | |
93 if ret is not None: | |
94 _set(arg, ret) | |
95 return ret | |
96 inner.get = _get | |
97 inner.set = _set | |
98 inner.clear = withlock(lock, cache.clear) | |
99 inner.update = withlock(lock, cache.update) | |
100 return inner | |
101 return decorator | |
102 | |
103 | |
104 def _ScopedPool_initer(orig, orig_args): # pragma: no cover | |
105 """Initializer method for ScopedPool's subprocesses. | |
106 | |
107 This helps ScopedPool handle Ctrl-C's correctly. | |
108 """ | |
109 signal.signal(signal.SIGINT, signal.SIG_IGN) | |
110 if orig: | |
111 orig(*orig_args) | |
112 | |
113 | |
114 @contextlib.contextmanager | |
115 def ScopedPool(*args, **kwargs): | |
116 """Context Manager which returns a multiprocessing.pool instance which | |
117 correctly deals with thrown exceptions. | |
118 | |
119 *args - Arguments to multiprocessing.pool | |
120 | |
121 Kwargs: | |
122 kind ('threads', 'procs') - The type of underlying coprocess to use. | |
123 **etc - Arguments to multiprocessing.pool | |
124 """ | |
125 if kwargs.pop('kind', None) == 'threads': | |
126 pool = multiprocessing.pool.ThreadPool(*args, **kwargs) | |
127 else: | |
128 orig, orig_args = kwargs.get('initializer'), kwargs.get('initargs', ()) | |
129 kwargs['initializer'] = _ScopedPool_initer | |
130 kwargs['initargs'] = orig, orig_args | |
131 pool = multiprocessing.pool.Pool(*args, **kwargs) | |
132 | |
133 try: | |
134 yield pool | |
135 pool.close() | |
136 except: | |
137 pool.terminate() | |
138 raise | |
139 finally: | |
140 pool.join() | |
141 | |
142 | |
143 class ProgressPrinter(object): | |
M-A Ruel
2013/11/13 19:55:00
FTR, I prefer https://code.google.com/p/swarming/s
iannucci
2013/11/14 07:06:29
Nice :D.
Yeah we should have a standard library o
| |
144 """Threaded single-stat status message printer.""" | |
145 def __init__(self, fmt, enabled=None, stream=sys.stderr, period=0.5): | |
146 """Create a ProgressPrinter. | |
147 | |
148 Use it as a context manager which produces a simple 'increment' method: | |
149 | |
150 with ProgressPrinter('(%%(count)d/%d)' % 1000) as inc: | |
151 for i in xrange(1000): | |
152 # do stuff | |
153 if i % 10 == 0: | |
154 inc(10) | |
155 | |
156 Args: | |
157 fmt - String format with a single '%(count)d' where the counter value | |
158 should go. | |
159 enabled (bool) - If this is None, will default to True if | |
160 logging.getLogger() is set to INFO or more verbose. | |
161 stream (file-like) - The stream to print status messages to. | |
162 period (float) - The time in seconds for the printer thread to wait | |
163 between printing. | |
164 """ | |
165 self.fmt = fmt | |
166 if enabled is None: # pragma: no cover | |
167 self.enabled = logging.getLogger().isEnabledFor(logging.INFO) | |
168 else: | |
169 self.enabled = enabled | |
170 | |
171 self._count = 0 | |
172 self._dead = False | |
173 self._dead_cond = threading.Condition() | |
174 self._stream = stream | |
175 self._thread = threading.Thread(target=self._run) | |
176 self._period = period | |
177 | |
178 def _emit(self, s): | |
179 if self.enabled: | |
180 self._stream.write('\r'+s) | |
M-A Ruel
2013/11/13 19:55:00
self._stream.write('\r' + s)
iannucci
2013/11/14 07:06:29
d'oh. Why doesn't our presubmit barf on this?
M-A Ruel
2013/11/14 17:27:24
No idea.
| |
181 self._stream.flush() | |
182 | |
183 def _run(self): | |
184 with self._dead_cond: | |
185 while not self._dead: | |
186 self._emit(self.fmt % {'count': self._count}) | |
187 self._dead_cond.wait(self._period) | |
188 self._emit((self.fmt+'\n') % {'count': self._count}) | |
189 | |
190 def inc(self, amount=1): | |
191 self._count += amount | |
192 | |
193 def __enter__(self): | |
194 self._thread.start() | |
195 return self.inc | |
196 | |
197 def __exit__(self, _exc_type, _exc_value, _traceback): | |
198 self._dead = True | |
199 with self._dead_cond: | |
200 self._dead_cond.notifyAll() | |
201 self._thread.join() | |
202 del self._thread | |
203 | |
204 | |
205 def parse_commitrefs(*commitrefs): | |
206 """Returns binary encoded commit hashes for one or more commitrefs. | |
207 | |
208 A commitref is anything which can resolve to a commit. Popular examples: | |
209 * "HEAD" | |
210 * "origin/master" | |
211 * "cool_branch~2" | |
212 """ | |
213 try: | |
214 return map(binascii.unhexlify, hashes(*commitrefs)) | |
215 except CalledProcessError: | |
216 raise Exception('one of %s does not seem to be a valid commitref.' % | |
217 str(commitrefs)) | |
M-A Ruel
2013/11/13 19:55:00
str() is not needed by definition.
iannucci
2013/11/14 07:06:29
Well... it is because commitrefs is a tuple. I don
| |
218 | |
219 | |
220 def _check_output(*popenargs, **kwargs): | |
M-A Ruel
2013/11/13 19:55:00
Why not subprocess2.check_output() ? It works on 2
iannucci
2013/11/14 07:06:29
Good question. I'll use that.
| |
221 """Runs a Popen command, and return the stdout as a string. | |
222 | |
223 Throws CalledProcessError if the command returns non-zero. | |
224 | |
225 kwargs: | |
226 indata (str) - Data to provide to the command on stdin. Mutually exclusive | |
227 with the Popen kwarg 'stdin'. | |
228 | |
229 Other than that, popenargs is *args to Popen, and **kwargs is... **kwargs to | |
230 Popen. | |
231 """ | |
232 kwargs.setdefault('stdout', subprocess.PIPE) | |
233 kwargs.setdefault('stderr', subprocess.PIPE) | |
234 indata = kwargs.pop('indata', None) | |
235 if indata is not None: | |
236 kwargs['stdin'] = subprocess.PIPE | |
237 process = subprocess.Popen(*popenargs, **kwargs) | |
238 output, _ = process.communicate(indata) | |
239 if process.returncode: | |
240 cmd = kwargs.get('args') | |
241 if cmd is None: | |
242 cmd = popenargs[0] | |
243 raise CalledProcessError(process.returncode, cmd) | |
244 return output | |
245 | |
246 | |
247 def run(*cmd, **kwargs): | |
248 """Runs a git command. Returns stdout as a string. | |
249 | |
250 If logging is DEBUG, we'll print the command before we run it. | |
251 | |
252 kwargs | |
253 autostrip (bool) - Strip the output. Defaults to True. | |
254 Output string is always strip()'d. | |
255 """ | |
256 autostrip = kwargs.pop('autostrip', True) | |
257 cmd = (GIT_EXE,) + cmd | |
258 logging.debug('running: %s', " ".join(repr(tok) for tok in cmd)) | |
259 ret = _check_output(cmd, **kwargs) | |
260 if autostrip: | |
261 ret = (ret or '').strip() | |
262 return ret | |
263 | |
264 | |
265 def hashes(*reflike): | |
266 return run('rev-parse', *reflike).splitlines() | |
267 | |
268 | |
269 def intern_f(f, kind='blob'): | |
270 """Interns a file object into the git object store. | |
271 | |
272 Args: | |
273 f (file-like object) - The file-like object to intern | |
274 kind (git object type) - One of 'blob', 'commit', 'tree', 'tag'. | |
275 | |
276 Returns the git hash of the interned object (hex encoded). | |
277 """ | |
278 ret = run('hash-object', '-t', kind, '-w', '--stdin', stdin=f) | |
279 f.close() | |
280 return ret | |
281 | |
282 | |
283 def tree(treeref, recurse=False): | |
284 """Returns a dict representation of a git tree object. | |
285 | |
286 Args: | |
287 treeref (str) - a git ref which resolves to a tree (commits count as trees). | |
288 recurse (bool) - include all of the tree's decendants too. File names will | |
289 take the form of 'some/path/to/file'. | |
290 | |
291 Return format: | |
292 { 'file_name': (mode, type, ref) } | |
293 | |
294 mode is an integer where: | |
295 * 0040000 - Directory | |
296 * 0100644 - Regular non-executable file | |
297 * 0100664 - Regular non-executable group-writeable file | |
298 * 0100755 - Regular executable file | |
299 * 0120000 - Symbolic link | |
300 * 0160000 - Gitlink | |
301 | |
302 type is a string where it's one of 'blob', 'commit', 'tree', 'tag'. | |
303 | |
304 ref is the hex encoded hash of the entry. | |
305 """ | |
306 ret = {} | |
307 opts = ['ls-tree', '--full-tree'] | |
308 if recurse: | |
309 opts += ['-r'] | |
310 opts.append(treeref) | |
311 try: | |
312 for line in run(*opts).splitlines(): | |
313 mode, typ, ref, name = line.split(None, 3) | |
314 ret[name] = (mode, typ, ref) | |
315 except CalledProcessError: | |
316 return None | |
317 return ret | |
318 | |
319 | |
320 def mktree(treedict): | |
321 """Makes a git tree object and returns its hash. | |
322 | |
323 See |tree()| for the values of mode, type, and ref. | |
324 | |
325 Args: | |
326 treedict - { name: (mode, type, ref) } | |
327 """ | |
328 with tempfile.TemporaryFile() as f: | |
329 for name, (mode, typ, ref) in treedict.iteritems(): | |
330 f.write('%s %s %s\t%s\0' % (mode, typ, ref, name)) | |
331 f.seek(0) | |
332 return run('mktree', '-z', stdin=f) | |
OLD | NEW |