Add git-number script to calculate generation numbers for commits.

Compatible with any git topology (multiple roots, weird branching/merging, etc.)
I can't get it to be any faster (in python). Suggestions welcome :).

On z600/linux, this takes 5.1s to calculate the initial count for 2e3de954ef0a
(HEAD on src.git at the time of writing). Subsequent lookups take ~0.06s. For
reference, this machine takes 3s to just list the revisions in sorted order
without any additional processing (using rev-list).

All calculations are stored in a git-notes-style ref with the exception that the
leaf 'tree' object which would normally be stored in a git-notes world is
replaced with a packed binary file which consists of records [hash int]. Each
run of this script will create only 1 commit object on this internal ref which
will have as its parents:
  * The previous git number commit
  * All of the target commits we calculated numbers for.
This ref is then excluded on subsequent invocations of rev-list, which means that
git-number will only ever process commit objects which it hasn't already
calculated a value for. It also prevents you from attempting to number this
special ref :).

This implementation only has a 1-byte fanout which seems to be the best
performance for the repos we're dealing with (i.e. on the order of 500k commit
objects).  Bumping this up to a 2-byte fanout became extremely slow (I suspect
the internal caching structures I'm using are not efficient in this mode and
could be improved). Using no fanout is slower than the 1 byte fanout for lookups
by about 30%.

R=agable@chromium.org, stip@chromium.org, szager@chromium.org
BUG=280154,309692,skia:1639

Review URL: https://codereview.chromium.org/26109002

git-svn-id: svn://svn.chromium.org/chrome/trunk/tools/depot_tools@236035 0039d316-1c4b-4281-b951-d872f2087c98

git-number

+#!/bin/sh
+# Copyright 2013 The Chromium Authors. All rights reserved.
+# Use of this source code is governed by a BSD-style license that can be
+# found in the LICENSE file.
+
+# git-number -- a git-command for calculating and displaying the generation
+# number of a commit.
+
+# Test if this script is running under a MSys install.  If it is, we will
+# hardcode the path to Python where possible.
+OUTPUT="$(uname | grep 'MINGW')"
+MINGW=$?
+
+if [ $MINGW = 0 ]; then
+  base_dir="${0%\\*}"
+else
+  base_dir=$(dirname "$0")
+fi
+
+# Uncomment this line if you never use gclient.
+# "$base_dir"/update_depot_tools
+
+if [ -e "$base_dir/python.bat" -a $MINGW = 0 ]; then
+  PYTHONDONTWRITEBYTECODE=1 cmd.exe //c "$base_dir\\python.bat" "$base_dir\\git_number.py" "$@"
+else
+  PYTHONDONTWRITEBYTECODE=1 exec "$base_dir/git_number.py" "$@"
+fi

git_common.py

+# Copyright 2013 The Chromium Authors. All rights reserved.
+# Use of this source code is governed by a BSD-style license that can be
+# found in the LICENSE file.
+
+# Monkeypatch IMapIterator so that Ctrl-C can kill everything properly.
+# Derived from https://gist.github.com/aljungberg/626518
+import multiprocessing.pool
+from multiprocessing.pool import IMapIterator
+def wrapper(func):
+  def wrap(self, timeout=None):
+    return func(self, timeout=timeout or 1e100)
+  return wrap
+IMapIterator.next = wrapper(IMapIterator.next)
+IMapIterator.__next__ = IMapIterator.next
+# TODO(iannucci): Monkeypatch all other 'wait' methods too.
+
+
+import binascii
+import contextlib
+import functools
+import logging
+import signal
+import sys
+import tempfile
+import threading
+
+import subprocess2
+
+
+GIT_EXE = 'git.bat' if sys.platform.startswith('win') else 'git'
+
+
+class BadCommitRefException(Exception):
+  def __init__(self, refs):
+    msg = ('one of %s does not seem to be a valid commitref.' %
+           str(refs))
+    super(BadCommitRefException, self).__init__(msg)
+
+
+def memoize_one(**kwargs):
+  """Memoizes a single-argument pure function.
+
+  Values of None are not cached.
+
+  Kwargs:
+    threadsafe (bool) - REQUIRED. Specifies whether to use locking around
+      cache manipulation functions. This is a kwarg so that users of memoize_one
+      are forced to explicitly and verbosely pick True or False.
+
+  Adds three methods to the decorated function:
+    * get(key, default=None) - Gets the value for this key from the cache.
+    * set(key, value) - Sets the value for this key from the cache.
+    * clear() - Drops the entire contents of the cache.  Useful for unittests.
+    * update(other) - Updates the contents of the cache from another dict.
+  """
+  assert 'threadsafe' in kwargs, 'Must specify threadsafe={True,False}'
+  threadsafe = kwargs['threadsafe']
+
+  if threadsafe:
+    def withlock(lock, f):
+      def inner(*args, **kwargs):
+        with lock:
+          return f(*args, **kwargs)
+      return inner
+  else:
+    def withlock(_lock, f):
+      return f
+
+  def decorator(f):
+    # Instantiate the lock in decorator, in case users of memoize_one do:
+    #
+    # memoizer = memoize_one(threadsafe=True)
+    #
+    # @memoizer
+    # def fn1(val): ...
+    #
+    # @memoizer
+    # def fn2(val): ...
+
+    lock = threading.Lock() if threadsafe else None
+    cache = {}
+    _get = withlock(lock, cache.get)
+    _set = withlock(lock, cache.__setitem__)
+
+    @functools.wraps(f)
+    def inner(arg):
+      ret = _get(arg)
+      if ret is None:
+        ret = f(arg)
+        if ret is not None:
+          _set(arg, ret)
+      return ret
+    inner.get = _get
+    inner.set = _set
+    inner.clear = withlock(lock, cache.clear)
+    inner.update = withlock(lock, cache.update)
+    return inner
+  return decorator
+
+
+def _ScopedPool_initer(orig, orig_args):  # pragma: no cover
+  """Initializer method for ScopedPool's subprocesses.
+
+  This helps ScopedPool handle Ctrl-C's correctly.
+  """
+  signal.signal(signal.SIGINT, signal.SIG_IGN)
+  if orig:
+    orig(*orig_args)
+
+
+@contextlib.contextmanager
+def ScopedPool(*args, **kwargs):
+  """Context Manager which returns a multiprocessing.pool instance which
+  correctly deals with thrown exceptions.
+
+  *args - Arguments to multiprocessing.pool
+
+  Kwargs:
+    kind ('threads', 'procs') - The type of underlying coprocess to use.
+    **etc - Arguments to multiprocessing.pool
+  """
+  if kwargs.pop('kind', None) == 'threads':
+    pool = multiprocessing.pool.ThreadPool(*args, **kwargs)
+  else:
+    orig, orig_args = kwargs.get('initializer'), kwargs.get('initargs', ())
+    kwargs['initializer'] = _ScopedPool_initer
+    kwargs['initargs'] = orig, orig_args
+    pool = multiprocessing.pool.Pool(*args, **kwargs)
+
+  try:
+    yield pool
+    pool.close()
+  except:
+    pool.terminate()
+    raise
+  finally:
+    pool.join()
+
+
+class ProgressPrinter(object):
+  """Threaded single-stat status message printer."""
+  def __init__(self, fmt, enabled=None, stream=sys.stderr, period=0.5):
+    """Create a ProgressPrinter.
+
+    Use it as a context manager which produces a simple 'increment' method:
+
+      with ProgressPrinter('(%%(count)d/%d)' % 1000) as inc:
+        for i in xrange(1000):
+          # do stuff
+          if i % 10 == 0:
+            inc(10)
+
+    Args:
+      fmt - String format with a single '%(count)d' where the counter value
+        should go.
+      enabled (bool) - If this is None, will default to True if
+        logging.getLogger() is set to INFO or more verbose.
+      stream (file-like) - The stream to print status messages to.
+      period (float) - The time in seconds for the printer thread to wait
+        between printing.
+    """
+    self.fmt = fmt
+    if enabled is None:  # pragma: no cover
+      self.enabled = logging.getLogger().isEnabledFor(logging.INFO)
+    else:
+      self.enabled = enabled
+
+    self._count = 0
+    self._dead = False
+    self._dead_cond = threading.Condition()
+    self._stream = stream
+    self._thread = threading.Thread(target=self._run)
+    self._period = period
+
+  def _emit(self, s):
+    if self.enabled:
+      self._stream.write('\r' + s)
+      self._stream.flush()
+
+  def _run(self):
+    with self._dead_cond:
+      while not self._dead:
+        self._emit(self.fmt % {'count': self._count})
+        self._dead_cond.wait(self._period)
+        self._emit((self.fmt + '\n') % {'count': self._count})
+
+  def inc(self, amount=1):
+    self._count += amount
+
+  def __enter__(self):
+    self._thread.start()
+    return self.inc
+
+  def __exit__(self, _exc_type, _exc_value, _traceback):
+    self._dead = True
+    with self._dead_cond:
+      self._dead_cond.notifyAll()
+    self._thread.join()
+    del self._thread
+
+
+def parse_commitrefs(*commitrefs):
+  """Returns binary encoded commit hashes for one or more commitrefs.
+
+  A commitref is anything which can resolve to a commit. Popular examples:
+    * 'HEAD'
+    * 'origin/master'
+    * 'cool_branch~2'
+  """
+  try:
+    return map(binascii.unhexlify, hashes(*commitrefs))
+  except subprocess2.CalledProcessError:
+    raise BadCommitRefException(commitrefs)
+
+
+def run(*cmd, **kwargs):
+  """Runs a git command. Returns stdout as a string.
+
+  If logging is DEBUG, we'll print the command before we run it.
+
+  kwargs
+    autostrip (bool) - Strip the output. Defaults to True.
+  Output string is always strip()'d.
+  """
+  autostrip = kwargs.pop('autostrip', True)
+  cmd = (GIT_EXE,) + cmd
+  logging.debug('Running %s', ' '.join(repr(tok) for tok in cmd))
+  ret = subprocess2.check_output(cmd, stderr=subprocess2.PIPE, **kwargs)
+  if autostrip:
+    ret = (ret or '').strip()
+  return ret
+
+
+def hashes(*reflike):
+  return run('rev-parse', *reflike).splitlines()
+
+
+def intern_f(f, kind='blob'):
+  """Interns a file object into the git object store.
+
+  Args:
+    f (file-like object) - The file-like object to intern
+    kind (git object type) - One of 'blob', 'commit', 'tree', 'tag'.
+
+  Returns the git hash of the interned object (hex encoded).
+  """
+  ret = run('hash-object', '-t', kind, '-w', '--stdin', stdin=f)
+  f.close()
+  return ret
+
+
+def tree(treeref, recurse=False):
+  """Returns a dict representation of a git tree object.
+
+  Args:
+    treeref (str) - a git ref which resolves to a tree (commits count as trees).
+    recurse (bool) - include all of the tree's decendants too. File names will
+      take the form of 'some/path/to/file'.
+
+  Return format:
+    { 'file_name': (mode, type, ref) }
+
+    mode is an integer where:
+      * 0040000 - Directory
+      * 0100644 - Regular non-executable file
+      * 0100664 - Regular non-executable group-writeable file
+      * 0100755 - Regular executable file
+      * 0120000 - Symbolic link
+      * 0160000 - Gitlink
+
+    type is a string where it's one of 'blob', 'commit', 'tree', 'tag'.
+
+    ref is the hex encoded hash of the entry.
+  """
+  ret = {}
+  opts = ['ls-tree', '--full-tree']
+  if recurse:
+    opts.append('-r')
+  opts.append(treeref)
+  try:
+    for line in run(*opts).splitlines():
+      mode, typ, ref, name = line.split(None, 3)
+      ret[name] = (mode, typ, ref)
+  except subprocess2.CalledProcessError:
+    return None
+  return ret
+
+
+def mktree(treedict):
+  """Makes a git tree object and returns its hash.
+
+  See |tree()| for the values of mode, type, and ref.
+
+  Args:
+    treedict - { name: (mode, type, ref) }
+  """
+  with tempfile.TemporaryFile() as f:
+    for name, (mode, typ, ref) in treedict.iteritems():
+      f.write('%s %s %s\t%s\0' % (mode, typ, ref, name))
+    f.seek(0)
+    return run('mktree', '-z', stdin=f)

git_number.py

+#!/usr/bin/env python
+# Copyright 2013 The Chromium Authors. All rights reserved.
+# Use of this source code is governed by a BSD-style license that can be
+# found in the LICENSE file.
+
+"""Usage: %prog [options] [<commitref>]*
+
+If no <commitref>'s are supplied, it defaults to HEAD.
+
+Calculates the generation number for one or more commits in a git repo.
+
+Generation number of a commit C with parents P is defined as:
+  generation_number(C, []) = 0
+  generation_number(C, P)  = max(map(generation_number, P)) + 1
+
+This number can be used to order commits relative to each other, as long as for
+any pair of the commits, one is an ancestor of the other.
+
+Since calculating the generation number of a commit requires walking that
+commit's entire history, this script caches all calculated data inside the git
+repo that it operates on in the ref 'refs/number/commits'.
+"""
+
+import binascii
+import collections
+import logging
+import optparse
+import os
+import struct
+import sys
+import tempfile
+
+import git_common as git
+import subprocess2
+
+CHUNK_FMT = '!20sL'
+CHUNK_SIZE = struct.calcsize(CHUNK_FMT)
+DIRTY_TREES = collections.defaultdict(int)
+REF = 'refs/number/commits'
+
+# Number of bytes to use for the prefix on our internal number structure.
+# 0 is slow to deserialize. 2 creates way too much bookeeping overhead (would
+# need to reimplement cache data structures to be a bit more sophisticated than
+# dicts. 1 seems to be just right.
+PREFIX_LEN = 1
+
+# Set this to 'threads' to gather coverage data while testing.
+POOL_KIND = 'procs'
+
+
+def pathlify(hash_prefix):
+  """Converts a binary object hash prefix into a posix path, one folder per
+  byte.
+
+  >>> pathlify('\xDE\xAD')
+  'de/ad'
+  """
+  return '/'.join('%02x' % ord(b) for b in hash_prefix)
+
+
+@git.memoize_one(threadsafe=False)
+def get_number_tree(prefix_bytes):
+  """Returns a dictionary of the git-number registry specified by
+  |prefix_bytes|.
+
+  This is in the form of {<full binary ref>: <gen num> ...}
+
+  >>> get_number_tree('\x83\xb4')
+  {'\x83\xb4\xe3\xe4W\xf9J*\x8f/c\x16\xecD\xd1\x04\x8b\xa9qz': 169, ...}
+  """
+  ref = '%s:%s' % (REF, pathlify(prefix_bytes))
+
+  try:
+    raw = buffer(git.run('cat-file', 'blob', ref, autostrip=False))
+    return dict(struct.unpack_from(CHUNK_FMT, raw, i * CHUNK_SIZE)
+                for i in xrange(len(raw) / CHUNK_SIZE))
+  except subprocess2.CalledProcessError:
+    return {}
+
+
+@git.memoize_one(threadsafe=False)
+def get_num(commit_hash):
+  """Returns the generation number for a commit.
+
+  Returns None if the generation number for this commit hasn't been calculated
+  yet (see load_generation_numbers()).
+  """
+  return get_number_tree(commit_hash[:PREFIX_LEN]).get(commit_hash)
+
+
+def clear_caches(on_disk=False):
+  """Clears in-process caches for e.g. unit testing."""
+  get_number_tree.clear()
+  get_num.clear()
+  if on_disk:
+    git.run('update-ref', '-d', REF)
+
+
+def intern_number_tree(tree):
+  """Transforms a number tree (in the form returned by |get_number_tree|) into
+  a git blob.
+
+  Returns the git blob id as hex-encoded string.
+
+  >>> d = {'\x83\xb4\xe3\xe4W\xf9J*\x8f/c\x16\xecD\xd1\x04\x8b\xa9qz': 169}
+  >>> intern_number_tree(d)
+  'c552317aa95ca8c3f6aae3357a4be299fbcb25ce'
+  """
+  with tempfile.TemporaryFile() as f:
+    for k, v in sorted(tree.iteritems()):
+      f.write(struct.pack(CHUNK_FMT, k, v))
+    f.seek(0)
+    return git.intern_f(f)
+
+
+def leaf_map_fn((pre, tree)):
+  """Converts a prefix and number tree into a git index line."""
+  return '100644 blob %s\t%s\0' % (intern_number_tree(tree), pathlify(pre))
+
+
+def finalize(targets):
+  """Saves all cache data to the git repository.
+
+  After calculating the generation number for |targets|, call finalize() to
+  save all the work to the git repository.
+
+  This in particular saves the trees referred to by DIRTY_TREES.
+  """
+  if not DIRTY_TREES:
+    return
+
+  msg = 'git-number Added %s numbers' % sum(DIRTY_TREES.itervalues())
+
+  idx = os.path.join(git.run('rev-parse', '--git-dir'), 'number.idx')
+  env = os.environ.copy()
+  env['GIT_INDEX_FILE'] = idx
+
+  progress_message = 'Finalizing: (%%(count)d/%d)' % len(DIRTY_TREES)
+  with git.ProgressPrinter(progress_message) as inc:
+    git.run('read-tree', REF, env=env)
+
+    prefixes_trees = ((p, get_number_tree(p)) for p in sorted(DIRTY_TREES))
+    updater = subprocess2.Popen(['git', 'update-index', '-z', '--index-info'],
+                                stdin=subprocess2.PIPE, env=env)
+
+    with git.ScopedPool(kind=POOL_KIND) as leaf_pool:
+      for item in leaf_pool.imap(leaf_map_fn, prefixes_trees):
+        updater.stdin.write(item)
+        inc()
+
+    updater.stdin.close()
+    updater.wait()
+    assert updater.returncode == 0
+
+    tree_id = git.run('write-tree', env=env)
+    commit_cmd = ['commit-tree', '-m', msg, '-p'] + git.hashes(REF)
+    for t in targets:
+      commit_cmd.extend(['-p', binascii.hexlify(t)])
+    commit_cmd.append(tree_id)
+    commit_hash = git.run(*commit_cmd)
+    git.run('update-ref', REF, commit_hash)
+  DIRTY_TREES.clear()
+
+
+def preload_tree(prefix):
+  """Returns the prefix and parsed tree object for the specified prefix."""
+  return prefix, get_number_tree(prefix)
+
+
+def all_prefixes(depth=PREFIX_LEN):
+  for x in (chr(i) for i in xrange(255)):
+    # This isn't covered because PREFIX_LEN currently == 1
+    if depth > 1:  # pragma: no cover
+      for r in all_prefixes(depth - 1):
+        yield x + r
+    else:
+      yield x
+
+
+def load_generation_numbers(targets):
+  """Populates the caches of get_num and get_number_tree so they contain
+  the results for |targets|.
+
+  Loads cached numbers from disk, and calculates missing numbers if one or
+  more of |targets| is newer than the cached calculations.
+
+  Args:
+    targets - An iterable of binary-encoded full git commit hashes.
+  """
+  # In case they pass us a generator, listify targets.
+  targets = list(targets)
+
+  if all(get_num(t) is not None for t in targets):
+    return
+
+  if git.tree(REF) is None:
+    empty = git.mktree({})
+    commit_hash = git.run('commit-tree', '-m', 'Initial commit from git-number',
+                          empty)
+    git.run('update-ref', REF, commit_hash)
+
+  with git.ScopedPool(kind=POOL_KIND) as pool:
+    preload_iter = pool.imap_unordered(preload_tree, all_prefixes())
+
+    rev_list = []
+
+    with git.ProgressPrinter('Loading commits: %(count)d') as inc:
+      # Curiously, buffering the list into memory seems to be the fastest
+      # approach in python (as opposed to iterating over the lines in the
+      # stdout as they're produced). GIL strikes again :/
+      cmd = [
+        'rev-list', '--topo-order', '--parents', '--reverse', '^' + REF,
+      ] + map(binascii.hexlify, targets)
+      for line in git.run(*cmd).splitlines():
+        tokens = map(binascii.unhexlify, line.split())
+        rev_list.append((tokens[0], tokens[1:]))
+        inc()
+
+    get_number_tree.update(preload_iter)
+
+  with git.ProgressPrinter('Counting: %%(count)d/%d' % len(rev_list)) as inc:
+    for commit_hash, pars in rev_list:
+      num = max(map(get_num, pars)) + 1 if pars else 0
+
+      prefix = commit_hash[:PREFIX_LEN]
+      get_number_tree(prefix)[commit_hash] = num
+      DIRTY_TREES[prefix] += 1
+      get_num.set(commit_hash, num)
+
+      inc()
+
+
+def main():  # pragma: no cover
+  parser = optparse.OptionParser(usage=sys.modules[__name__].__doc__)
+  parser.add_option('--no-cache', action='store_true',
+                    help='Do not actually cache anything we calculate.')
+  parser.add_option('--reset', action='store_true',
+                    help='Reset the generation number cache and quit.')
+  parser.add_option('-v', '--verbose', action='count', default=0,
+                    help='Be verbose. Use more times for more verbosity.')
+  opts, args = parser.parse_args()
+
+  levels = [logging.ERROR, logging.INFO, logging.DEBUG]
+  logging.basicConfig(level=levels[min(opts.verbose, len(levels) - 1)])
+
+  try:
+    if opts.reset:
+      clear_caches(on_disk=True)
+      return
+
+    try:
+      targets = git.parse_commitrefs(*(args or ['HEAD']))
+    except git.BadCommitRefException as e:
+      parser.error(e)
+
+    load_generation_numbers(targets)
+    if not opts.no_cache:
+      finalize(targets)
+
+    print '\n'.join(map(str, map(get_num, targets)))
+    return 0
+  except KeyboardInterrupt:
+    return 1
+
+
+if __name__ == '__main__':  # pragma: no cover
+  sys.exit(main())

testing_support/coverage_utils.py

+# Copyright 2013 The Chromium Authors. All rights reserved.
+# Use of this source code is governed by a BSD-style license that can be
+# found in the LICENSE file.
+
+import distutils.version
+import os
+import sys
+import textwrap
+import unittest
+
+ROOT_PATH = os.path.abspath(os.path.join(
+    os.path.dirname(os.path.dirname(__file__))))
+
+
+def native_error(msg, version):
+  print textwrap.dedent("""\
+  ERROR: Native python-coverage (version: %s) is required to be
+  installed on your PYTHONPATH to run this test. Recommendation:
+     sudo pip install python-coverage
+  %s""") % (version, msg)
+  sys.exit(1)
+
+def covered_main(includes, require_native=None):
+  """Equivalent of unittest.main(), except that it gathers coverage data, and
+  asserts if the test is not at 100% coverage.
+
+  Args:
+    includes (list(str) or str) - List of paths to include in coverage report.
+      May also be a single path instead of a list.
+    require_native (str) - If non-None, will require that
+      at least |require_native| version of coverage is installed on the
+      system with CTracer.
+  """
+  try:
+    import coverage
+    if require_native is not None:
+      got_ver = coverage.__version__
+      if not coverage.collector.CTracer:
+        native_error((
+            "Native python-coverage module required.\n"
+            "Pure-python implementation (version: %s) found: %s"
+          ) % (got_ver, coverage), require_native)
+      if got_ver < distutils.version.LooseVersion(require_native):
+        native_error("Wrong version (%s) found: %s" % (got_ver, coverage),
+                     require_native)
+  except ImportError:
+    if require_native is None:
+      sys.path.insert(0, os.path.join(ROOT_PATH, 'third_party'))
+      import coverage
+    else:
+      print ("ERROR: python-coverage (%s) is required to be installed on your "
+             "PYTHONPATH to run this test." % require_native)
+      sys.exit(1)
+
+  COVERAGE = coverage.coverage(include=includes)
+  COVERAGE.start()
+
+  retcode = 0
+  try:
+    unittest.main()
+  except SystemExit as e:
+    retcode = e.code or retcode
+
+  COVERAGE.stop()
+  if COVERAGE.report() != 100.0:
+    print 'FATAL: not at 100% coverage.'
+    retcode = 2
+
+  return retcode

testing_support/git_test_utils.py

+# Copyright 2013 The Chromium Authors. All rights reserved.
+# Use of this source code is governed by a BSD-style license that can be
+# found in the LICENSE file.
+
+import atexit
+import collections
+import copy
+import datetime
+import hashlib
+import os
+import shutil
+import subprocess
+import tempfile
+import unittest
+
+
+def git_hash_data(data, typ='blob'):
+  """Calculate the git-style SHA1 for some data.
+
+  Only supports 'blob' type data at the moment.
+  """
+  assert typ == 'blob', 'Only support blobs for now'
+  return hashlib.sha1('blob %s\0%s' % (len(data), data)).hexdigest()
+
+
+class OrderedSet(collections.MutableSet):
+  # from http://code.activestate.com/recipes/576694/
+  def __init__(self, iterable=None):
+    self.end = end = []
+    end += [None, end, end]         # sentinel node for doubly linked list
+    self.data = {}                  # key --> [key, prev, next]
+    if iterable is not None:
+      self |= iterable
+
+  def __contains__(self, key):
+    return key in self.data
+
+  def __eq__(self, other):
+    if isinstance(other, OrderedSet):
+      return len(self) == len(other) and list(self) == list(other)
+    return set(self) == set(other)
+
+  def __ne__(self, other):
+    if isinstance(other, OrderedSet):
+      return len(self) != len(other) or list(self) != list(other)
+    return set(self) != set(other)
+
+  def __len__(self):
+    return len(self.data)
+
+  def __iter__(self):
+    end = self.end
+    curr = end[2]
+    while curr is not end:
+      yield curr[0]
+      curr = curr[2]
+
+  def __repr__(self):
+    if not self:
+      return '%s()' % (self.__class__.__name__,)
+    return '%s(%r)' % (self.__class__.__name__, list(self))
+
+  def __reversed__(self):
+    end = self.end
+    curr = end[1]
+    while curr is not end:
+      yield curr[0]
+      curr = curr[1]
+
+  def add(self, key):
+    if key not in self.data:
+      end = self.end
+      curr = end[1]
+      curr[2] = end[1] = self.data[key] = [key, curr, end]
+
+  def difference_update(self, *others):
+    for other in others:
+      for i in other:
+        self.discard(i)
+
+  def discard(self, key):
+    if key in self.data:
+      key, prev, nxt = self.data.pop(key)
+      prev[2] = nxt
+      nxt[1] = prev
+
+  def pop(self, last=True):  # pylint: disable=W0221
+    if not self:
+      raise KeyError('set is empty')
+    key = self.end[1][0] if last else self.end[2][0]
+    self.discard(key)
+    return key
+
+
+class GitRepoSchema(object):
+  """A declarative git testing repo.
+
+  Pass a schema to __init__ in the form of:
+     A B C D
+       B E D
+
+  This is the repo
+
+     A - B - C - D
+           \ E /
+
+  Whitespace doesn't matter. Each line is a declaration of which commits come
+  before which other commits.
+
+  Every commit gets a tag 'tag_%(commit)s'
+  Every unique terminal commit gets a branch 'branch_%(commit)s'
+  Last commit in First line is the branch 'master'
+  Root commits get a ref 'root_%(commit)s'
+
+  Timestamps are in topo order, earlier commits (as indicated by their presence
+  in the schema) get earlier timestamps. Stamps start at the Unix Epoch, and
+  increment by 1 day each.
+  """
+  COMMIT = collections.namedtuple('COMMIT', 'name parents is_branch is_root')
+
+  def __init__(self, repo_schema='',
+               content_fn=lambda v: {v: {'data': v}}):
+    """Builds a new GitRepoSchema.
+
+    Args:
+      repo_schema (str) - Initial schema for this repo. See class docstring for
+        info on the schema format.
+      content_fn ((commit_name) -> commit_data) - A function which will be
+        lazily called to obtain data for each commit. The results of this
+        function are cached (i.e. it will never be called twice for the same
+        commit_name). See the docstring on the GitRepo class for the format of
+        the data returned by this function.
+    """
+    self.master = None
+    self.par_map = {}
+    self.data_cache = {}
+    self.content_fn = content_fn
+    self.add_commits(repo_schema)
+
+  def walk(self):
+    """(Generator) Walks the repo schema from roots to tips.
+
+    Generates GitRepoSchema.COMMIT objects for each commit.
+
+    Throws an AssertionError if it detects a cycle.
+    """
+    is_root = True
+    par_map = copy.deepcopy(self.par_map)
+    while par_map:
+      empty_keys = set(k for k, v in par_map.iteritems() if not v)
+      assert empty_keys, 'Cycle detected! %s' % par_map
+
+      for k in sorted(empty_keys):
+        yield self.COMMIT(k, self.par_map[k],
+                          not any(k in v for v in self.par_map.itervalues()),
+                          is_root)
+        del par_map[k]
+      for v in par_map.itervalues():
+        v.difference_update(empty_keys)
+      is_root = False
+
+  def add_commits(self, schema):
+    """Adds more commits from a schema into the existing Schema.
+
+    Args:
+      schema (str) - See class docstring for info on schema format.
+
+    Throws an AssertionError if it detects a cycle.
+    """
+    for commits in (l.split() for l in schema.splitlines() if l.strip()):
+      parent = None
+      for commit in commits:
+        if commit not in self.par_map:
+          self.par_map[commit] = OrderedSet()
+        if parent is not None:
+          self.par_map[commit].add(parent)
+        parent = commit
+      if parent and not self.master:
+        self.master = parent
+    for _ in self.walk():  # This will throw if there are any cycles.
+      pass
+
+  def reify(self):
+    """Returns a real GitRepo for this GitRepoSchema"""
+    return GitRepo(self)
+
+  def data_for(self, commit):
+    """Obtains the data for |commit|.
+
+    See the docstring on the GitRepo class for the format of the returned data.
+
+    Caches the result on this GitRepoSchema instance.
+    """
+    if commit not in self.data_cache:
+      self.data_cache[commit] = self.content_fn(commit)
+    return self.data_cache[commit]
+
+
+class GitRepo(object):
+  """Creates a real git repo for a GitRepoSchema.
+
+  Obtains schema and content information from the GitRepoSchema.
+
+  The format for the commit data supplied by GitRepoSchema.data_for is:
+    {
+      SPECIAL_KEY: special_value,
+      ...
+      "path/to/some/file": { 'data': "some data content for this file",
+                              'mode': 0755 },
+      ...
+    }
+
+  The SPECIAL_KEYs are the following attribues of the GitRepo class:
+    * AUTHOR_NAME
+    * AUTHOR_EMAIL
+    * AUTHOR_DATE - must be a datetime.datetime instance
+    * COMMITTER_NAME
+    * COMMITTER_EMAIL
+    * COMMITTER_DATE - must be a datetime.datetime instance
+
+  For file content, if 'data' is None, then this commit will `git rm` that file.
+  """
+  BASE_TEMP_DIR = tempfile.mkdtemp(suffix='base', prefix='git_repo')
+  atexit.register(shutil.rmtree, BASE_TEMP_DIR)
+
+  # Singleton objects to specify specific data in a commit dictionary.
+  AUTHOR_NAME = object()
+  AUTHOR_EMAIL = object()
+  AUTHOR_DATE = object()
+  COMMITTER_NAME = object()
+  COMMITTER_EMAIL = object()
+  COMMITTER_DATE = object()
+
+  DEFAULT_AUTHOR_NAME = 'Author McAuthorly'
+  DEFAULT_AUTHOR_EMAIL = 'author@example.com'
+  DEFAULT_COMMITTER_NAME = 'Charles Committish'
+  DEFAULT_COMMITTER_EMAIL = 'commitish@example.com'
+
+  COMMAND_OUTPUT = collections.namedtuple('COMMAND_OUTPUT', 'retcode stdout')
+
+  def __init__(self, schema):
+    """Makes new GitRepo.
+
+    Automatically creates a temp folder under GitRepo.BASE_TEMP_DIR. It's
+    recommended that you clean this repo up by calling nuke() on it, but if not,
+    GitRepo will automatically clean up all allocated repos at the exit of the
+    program (assuming a normal exit like with sys.exit)
+
+    Args:
+      schema - An instance of GitRepoSchema
+    """
+    self.repo_path = tempfile.mkdtemp(dir=self.BASE_TEMP_DIR)
+    self.commit_map = {}
+    self._date = datetime.datetime(1970, 1, 1)
+
+    self.git('init')
+    for commit in schema.walk():
+      self._add_schema_commit(commit, schema.data_for(commit.name))
+    if schema.master:
+      self.git('update-ref', 'master', self[schema.master])
+
+  def __getitem__(self, commit_name):
+    """Gets the hash of a commit by its schema name.
+
+    >>> r = GitRepo(GitRepoSchema('A B C'))
+    >>> r['B']
+    '7381febe1da03b09da47f009963ab7998a974935'
+    """
+    return self.commit_map[commit_name]
+
+  def _add_schema_commit(self, commit, data):
+    data = data or {}
+
+    if commit.parents:
+      parents = list(commit.parents)
+      self.git('checkout', '--detach', '-q', self[parents[0]])
+      if len(parents) > 1:
+        self.git('merge', '--no-commit', '-q', *[self[x] for x in parents[1:]])
+    else:
+      self.git('checkout', '--orphan', 'root_%s' % commit.name)
+      self.git('rm', '-rf', '.')
+
+    env = {}
+    for prefix in ('AUTHOR', 'COMMITTER'):
+      for suffix in ('NAME', 'EMAIL', 'DATE'):
+        singleton = '%s_%s' % (prefix, suffix)
+        key = getattr(self, singleton)
+        if key in data:
+          val = data[key]
+        else:
+          if suffix == 'DATE':
+            val = self._date
+            self._date += datetime.timedelta(days=1)
+          else:
+            val = getattr(self, 'DEFAULT_%s' % singleton)
+        env['GIT_%s' % singleton] = str(val)
+
+    for fname, file_data in data.iteritems():
+      deleted = False
+      if 'data' in file_data:
+        data = file_data.get('data')
+        if data is None:
+          deleted = True
+          self.git('rm', fname)
+        else:
+          path = os.path.join(self.repo_path, fname)
+          pardir = os.path.dirname(path)
+          if not os.path.exists(pardir):
+            os.makedirs(pardir)
+          with open(path, 'wb') as f:
+            f.write(data)
+
+      mode = file_data.get('mode')
+      if mode and not deleted:
+        os.chmod(path, mode)
+
+      self.git('add', fname)
+
+    rslt = self.git('commit', '--allow-empty', '-m', commit.name, env=env)
+    assert rslt.retcode == 0, 'Failed to commit %s' % str(commit)
+    self.commit_map[commit.name] = self.git('rev-parse', 'HEAD').stdout.strip()
+    self.git('tag', 'tag_%s' % commit.name, self[commit.name])
+    if commit.is_branch:
+      self.git('update-ref', 'branch_%s' % commit.name, self[commit.name])
+
+  def git(self, *args, **kwargs):
+    """Runs a git command specified by |args| in this repo."""
+    assert self.repo_path is not None
+    try:
+      with open(os.devnull, 'wb') as devnull:
+        output = subprocess.check_output(
+          ('git',) + args, cwd=self.repo_path, stderr=devnull, **kwargs)
+      return self.COMMAND_OUTPUT(0, output)
+    except subprocess.CalledProcessError as e:
+      return self.COMMAND_OUTPUT(e.returncode, e.output)
+
+  def nuke(self):
+    """Obliterates the git repo on disk.
+
+    Causes this GitRepo to be unusable.
+    """
+    shutil.rmtree(self.repo_path)
+    self.repo_path = None
+
+  def run(self, fn, *args, **kwargs):
+    """Run a python function with the given args and kwargs with the cwd set to
+    the git repo."""
+    assert self.repo_path is not None
+    curdir = os.getcwd()
+    try:
+      os.chdir(self.repo_path)
+      return fn(*args, **kwargs)
+    finally:
+      os.chdir(curdir)
+
+
+class GitRepoSchemaTestBase(unittest.TestCase):
+  """A TestCase with a built-in GitRepoSchema.
+
+  Expects a class variable REPO to be a GitRepoSchema string in the form
+  described by that class.
+
+  You may also set class variables in the form COMMIT_%(commit_name)s, which
+  provide the content for the given commit_name commits.
+
+  You probably will end up using either GitRepoReadOnlyTestBase or
+  GitRepoReadWriteTestBase for real tests.
+  """
+  REPO = None
+
+  @classmethod
+  def getRepoContent(cls, commit):
+    return getattr(cls, 'COMMIT_%s' % commit, None)
+
+  @classmethod
+  def setUpClass(cls):
+    super(GitRepoSchemaTestBase, cls).setUpClass()
+    assert cls.REPO is not None
+    cls.r_schema = GitRepoSchema(cls.REPO, cls.getRepoContent)
+
+
+class GitRepoReadOnlyTestBase(GitRepoSchemaTestBase):
+  """Injects a GitRepo object given the schema and content from
+  GitRepoSchemaTestBase into TestCase classes which subclass this.
+
+  This GitRepo will appear as self.repo, and will be deleted and recreated once
+  for the duration of all the tests in the subclass.
+  """
+  REPO = None
+
+  @classmethod
+  def setUpClass(cls):
+    super(GitRepoReadOnlyTestBase, cls).setUpClass()
+    assert cls.REPO is not None
+    cls.repo = cls.r_schema.reify()
+
+  @classmethod
+  def tearDownClass(cls):
+    cls.repo.nuke()
+    super(GitRepoReadOnlyTestBase, cls).tearDownClass()
+
+
+class GitRepoReadWriteTestBase(GitRepoSchemaTestBase):
+  """Injects a GitRepo object given the schema and content from
+  GitRepoSchemaTestBase into TestCase classes which subclass this.
+
+  This GitRepo will appear as self.repo, and will be deleted and recreated for
+  each test function in the subclass.
+  """
+  REPO = None
+
+  def setUp(self):
+    super(GitRepoReadWriteTestBase, self).setUp()
+    self.repo = self.r_schema.reify()
+
+  def tearDown(self):
+    self.repo.nuke()
+    super(GitRepoReadWriteTestBase, self).tearDown()

tests/git_common_test.py

+#!/usr/bin/env python
+# Copyright 2013 The Chromium Authors. All rights reserved.
+# Use of this source code is governed by a BSD-style license that can be
+# found in the LICENSE file.
+
+"""Unit tests for git_common.py"""
+
+import binascii
+import collections
+import os
+import signal
+import sys
+import tempfile
+import time
+import unittest
+
+DEPOT_TOOLS_ROOT = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
+sys.path.insert(0, DEPOT_TOOLS_ROOT)
+
+from testing_support import coverage_utils
+from testing_support import git_test_utils
+
+
+class GitCommonTestBase(unittest.TestCase):
+  @classmethod
+  def setUpClass(cls):
+    super(GitCommonTestBase, cls).setUpClass()
+    import git_common
+    cls.gc = git_common
+
+
+class Support(GitCommonTestBase):
+  def _testMemoizeOneBody(self, threadsafe):
+    calls = collections.defaultdict(int)
+    def double_if_even(val):
+      calls[val] += 1
+      return val * 2 if val % 2 == 0 else None
+    # Use this explicitly as a wrapper fn instead of a decorator. Otherwise
+    # pylint crashes (!!)
+    double_if_even = self.gc.memoize_one(threadsafe=threadsafe)(double_if_even)
+
+    self.assertEqual(4, double_if_even(2))
+    self.assertEqual(4, double_if_even(2))
+    self.assertEqual(None, double_if_even(1))
+    self.assertEqual(None, double_if_even(1))
+    self.assertDictEqual({1: 2, 2: 1}, calls)
+
+    double_if_even.set(10, 20)
+    self.assertEqual(20, double_if_even(10))
+    self.assertDictEqual({1: 2, 2: 1}, calls)
+
+    double_if_even.clear()
+    self.assertEqual(4, double_if_even(2))
+    self.assertEqual(4, double_if_even(2))
+    self.assertEqual(None, double_if_even(1))
+    self.assertEqual(None, double_if_even(1))
+    self.assertEqual(20, double_if_even(10))
+    self.assertDictEqual({1: 4, 2: 2, 10: 1}, calls)
+
+  def testMemoizeOne(self):
+    self._testMemoizeOneBody(threadsafe=False)
+
+  def testMemoizeOneThreadsafe(self):
+    self._testMemoizeOneBody(threadsafe=True)
+
+
+def slow_square(i):
+  """Helper for ScopedPoolTest.
+
+  Must be global because non top-level functions aren't pickleable.
+  """
+  return i ** 2
+
+
+class ScopedPoolTest(GitCommonTestBase):
+  CTRL_C = signal.CTRL_C_EVENT if sys.platform == 'win32' else signal.SIGINT
+
+  def testThreads(self):
+    result = []
+    with self.gc.ScopedPool(kind='threads') as pool:
+      result = list(pool.imap(slow_square, xrange(10)))
+    self.assertEqual([0, 1, 4, 9, 16, 25, 36, 49, 64, 81], result)
+
+  def testThreadsCtrlC(self):
+    result = []
+    with self.assertRaises(KeyboardInterrupt):
+      with self.gc.ScopedPool(kind='threads') as pool:
+        # Make sure this pool is interrupted in mid-swing
+        for i in pool.imap(slow_square, xrange(1000000)):
+          if i > 32:
+            os.kill(os.getpid(), self.CTRL_C)
+          result.append(i)
+    self.assertEqual([0, 1, 4, 9, 16, 25], result)
+
+  def testProcs(self):
+    result = []
+    with self.gc.ScopedPool() as pool:
+      result = list(pool.imap(slow_square, xrange(10)))
+    self.assertEqual([0, 1, 4, 9, 16, 25, 36, 49, 64, 81], result)
+
+  def testProcsCtrlC(self):
+    result = []
+    with self.assertRaises(KeyboardInterrupt):
+      with self.gc.ScopedPool() as pool:
+        # Make sure this pool is interrupted in mid-swing
+        for i in pool.imap(slow_square, xrange(1000000)):
+          if i > 32:
+            os.kill(os.getpid(), self.CTRL_C)
+          result.append(i)
+    self.assertEqual([0, 1, 4, 9, 16, 25], result)
+
+
+class ProgressPrinterTest(GitCommonTestBase):
+  class FakeStream(object):
+    def __init__(self):
+      self.data = set()
+      self.count = 0
+
+    def write(self, line):
+      self.data.add(line)
+
+    def flush(self):
+      self.count += 1
+
+  @unittest.expectedFailure
+  def testBasic(self):
+    """This test is probably racy, but I don't have a better alternative."""
+    fmt = '%(count)d/10'
+    stream = self.FakeStream()
+
+    pp = self.gc.ProgressPrinter(fmt, enabled=True, stream=stream, period=0.01)
+    with pp as inc:
+      for _ in xrange(10):
+        time.sleep(0.02)
+        inc()
+
+    filtered = set(x.strip() for x in stream.data)
+    rslt = set(fmt % {'count': i} for i in xrange(11))
+    self.assertSetEqual(filtered, rslt)
+    self.assertGreaterEqual(stream.count, 10)
+
+
+class GitReadOnlyFunctionsTest(git_test_utils.GitRepoReadOnlyTestBase,
+                               GitCommonTestBase):
+  REPO = """
+  A B C D
+    B E D
+  """
+
+  COMMIT_A = {
+    'some/files/file1': {'data': 'file1'},
+    'some/files/file2': {'data': 'file2'},
+    'some/files/file3': {'data': 'file3'},
+    'some/other/file':  {'data': 'otherfile'},
+  }
+
+  COMMIT_C = {
+    'some/files/file2': {
+      'mode': 0755,
+      'data': 'file2 - vanilla'},
+  }
+
+  COMMIT_E = {
+    'some/files/file2': {'data': 'file2 - merged'},
+  }
+
+  COMMIT_D = {
+    'some/files/file2': {'data': 'file2 - vanilla\nfile2 - merged'},
+  }
+
+  def testHashes(self):
+    ret = self.repo.run(
+      self.gc.hashes, *[
+        'master',
+        'master~3',
+        self.repo['E']+'~',
+        self.repo['D']+'^2',
+        'tag_C^{}',
+      ]
+    )
+    self.assertEqual([
+      self.repo['D'],
+      self.repo['A'],
+      self.repo['B'],
+      self.repo['E'],
+      self.repo['C'],
+    ], ret)
+
+  def testParseCommitrefs(self):
+    ret = self.repo.run(
+      self.gc.parse_commitrefs, *[
+        'master',
+        'master~3',
+        self.repo['E']+'~',
+        self.repo['D']+'^2',
+        'tag_C^{}',
+      ]
+    )
+    self.assertEqual(ret, map(binascii.unhexlify, [
+      self.repo['D'],
+      self.repo['A'],
+      self.repo['B'],
+      self.repo['E'],
+      self.repo['C'],
+    ]))
+
+    with self.assertRaisesRegexp(Exception, r"one of \('master', 'bananas'\)"):
+      self.repo.run(self.gc.parse_commitrefs, 'master', 'bananas')
+
+  def testTree(self):
+    tree = self.repo.run(self.gc.tree, 'master:some/files')
+    file1 = self.COMMIT_A['some/files/file1']['data']
+    file2 = self.COMMIT_D['some/files/file2']['data']
+    file3 = self.COMMIT_A['some/files/file3']['data']
+    self.assertEquals(
+        tree['file1'],
+        ('100644', 'blob', git_test_utils.git_hash_data(file1)))
+    self.assertEquals(
+        tree['file2'],
+        ('100755', 'blob', git_test_utils.git_hash_data(file2)))
+    self.assertEquals(
+        tree['file3'],
+        ('100644', 'blob', git_test_utils.git_hash_data(file3)))
+
+    tree = self.repo.run(self.gc.tree, 'master:some')
+    self.assertEquals(len(tree), 2)
+    # Don't check the tree hash because we're lazy :)
+    self.assertEquals(tree['files'][:2], ('040000', 'tree'))
+
+    tree = self.repo.run(self.gc.tree, 'master:wat')
+    self.assertEqual(tree, None)
+
+  def testTreeRecursive(self):
+    tree = self.repo.run(self.gc.tree, 'master:some', recurse=True)
+    file1 = self.COMMIT_A['some/files/file1']['data']
+    file2 = self.COMMIT_D['some/files/file2']['data']
+    file3 = self.COMMIT_A['some/files/file3']['data']
+    other = self.COMMIT_A['some/other/file']['data']
+    self.assertEquals(
+        tree['files/file1'],
+        ('100644', 'blob', git_test_utils.git_hash_data(file1)))
+    self.assertEquals(
+        tree['files/file2'],
+        ('100755', 'blob', git_test_utils.git_hash_data(file2)))
+    self.assertEquals(
+        tree['files/file3'],
+        ('100644', 'blob', git_test_utils.git_hash_data(file3)))
+    self.assertEquals(
+        tree['other/file'],
+        ('100644', 'blob', git_test_utils.git_hash_data(other)))
+
+
+class GitMutableFunctionsTest(git_test_utils.GitRepoReadWriteTestBase,
+                              GitCommonTestBase):
+  REPO = ''
+
+  def _intern_data(self, data):
+    with tempfile.TemporaryFile() as f:
+      f.write(data)
+      f.seek(0)
+      return self.repo.run(self.gc.intern_f, f)
+
+  def testInternF(self):
+    data = 'CoolBobcatsBro'
+    data_hash = self._intern_data(data)
+    self.assertEquals(git_test_utils.git_hash_data(data), data_hash)
+    self.assertEquals(data, self.repo.git('cat-file', 'blob', data_hash).stdout)
+
+  def testMkTree(self):
+    tree = {}
+    for i in 1, 2, 3:
+      name = 'file%d' % i
+      tree[name] = ('100644', 'blob', self._intern_data(name))
+    tree_hash = self.repo.run(self.gc.mktree, tree)
+    self.assertEquals('37b61866d6e061c4ba478e7eb525be7b5752737d', tree_hash)
+
+
+if __name__ == '__main__':
+  sys.exit(coverage_utils.covered_main(
+    os.path.join(DEPOT_TOOLS_ROOT, 'git_common.py')
+  ))

tests/git_number_test.py

+#!/usr/bin/env python
+# Copyright 2013 The Chromium Authors. All rights reserved.
+# Use of this source code is governed by a BSD-style license that can be
+# found in the LICENSE file.
+
+"""Unit tests for git_number.py"""
+
+import binascii
+import os
+import sys
+
+DEPOT_TOOLS_ROOT = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
+sys.path.insert(0, DEPOT_TOOLS_ROOT)
+
+from testing_support import git_test_utils
+from testing_support import coverage_utils
+
+
+class Basic(git_test_utils.GitRepoReadWriteTestBase):
+  REPO = """
+  A B C D E
+    B   F E
+  X Y     E
+  """
+
+  @classmethod
+  def setUpClass(cls):
+    super(Basic, cls).setUpClass()
+    import git_number
+    cls.gn = git_number
+    cls.old_POOL_KIND = cls.gn.POOL_KIND
+    cls.gn.POOL_KIND = 'threads'
+
+  @classmethod
+  def tearDownClass(cls):
+    cls.gn.POOL_KIND = cls.old_POOL_KIND
+    super(Basic, cls).tearDownClass()
+
+  def tearDown(self):
+    self.gn.clear_caches()
+    super(Basic, self).tearDown()
+
+  def _git_number(self, refs, cache=False):
+    refs = map(binascii.unhexlify, refs)
+    self.repo.run(self.gn.load_generation_numbers, refs)
+    if cache:
+      self.repo.run(self.gn.finalize, refs)
+    return map(self.gn.get_num, refs)
+
+  def testBasic(self):
+    self.assertEqual([0], self._git_number([self.repo['A']]))
+    self.assertEqual([2], self._git_number([self.repo['F']]))
+    self.assertEqual([0], self._git_number([self.repo['X']]))
+    self.assertEqual([4], self._git_number([self.repo['E']]))
+
+  def testInProcessCache(self):
+    self.assertEqual(
+        None,
+        self.repo.run(self.gn.get_num, binascii.unhexlify(self.repo['A'])))
+    self.assertEqual([4], self._git_number([self.repo['E']]))
+    self.assertEqual(
+        0,
+        self.repo.run(self.gn.get_num, binascii.unhexlify(self.repo['A'])))
+
+  def testOnDiskCache(self):
+    self.assertEqual(
+        None,
+        self.repo.run(self.gn.get_num, binascii.unhexlify(self.repo['A'])))
+    self.assertEqual([4], self._git_number([self.repo['E']], cache=True))
+    self.assertEqual([4], self._git_number([self.repo['E']], cache=True))
+    self.gn.clear_caches()
+    self.assertEqual(
+        0,
+        self.repo.run(self.gn.get_num, binascii.unhexlify(self.repo['A'])))
+    self.gn.clear_caches()
+    self.repo.run(self.gn.clear_caches, True)
+    self.assertEqual(
+        None,
+        self.repo.run(self.gn.get_num, binascii.unhexlify(self.repo['A'])))
+
+
+if __name__ == '__main__':
+  sys.exit(coverage_utils.covered_main(
+    os.path.join(DEPOT_TOOLS_ROOT, 'git_number.py'),
+    '3.7'
+  ))
\ No newline at end of file