def GetFilesInDirectory(
directory: pathlib.Path,
exclude_patterns: typing.List[str]) -> typing.List[pathlib.Path]:
"""Recursively list all files in a directory.
Returns relative paths of all files in a directory which do not match the
exclude patterns. The list of exclude patterns supports UNIX style globbing.
Args:
directory: The path to the directory.
exclude_patterns: A list of patterns to exclude.
Returns:
A list of paths.
"""
exclude_patterns = set(exclude_patterns + ALWAYS_EXCLUDE_PATTERNS)
files = []
for path in sorted(fs.lsfiles(directory, recursive=True)):
for pattern in exclude_patterns:
if fnmatch.fnmatch(path, pattern):
logging.info('- %s', path)
break
else:
logging.info('+ %s', path)
files.append(pathlib.Path(path))
return files
def ReadClassificationsToTable(output_dir: pathlib.Path) -> pd.DataFrame:
rows = []
counters = {}
for f in fs.lsfiles(output_dir, recursive=True, abspaths=True):
path = pathlib.Path(f)
result_class, testbed_num, opt = path.parts[-4:-1]
t = testbed_num + opt
if t not in counters:
counters[t] = collections.defaultdict(int)
counters[t][result_class] += 1
for t, result_classes in counters.items():
rows.append([
t,
result_classes['bc'],
result_classes['bto'],
result_classes['abf'],
result_classes['arc'],
result_classes['awo'],
sum(result_classes.values()),
])
rows = sorted(rows, key=lambda x: (int(x[0][:-1]), x[0][-1]))
rows.append([
'Total',
len(fs.lsfiles(output_dir / 'bc', recursive=True)),
len(fs.lsfiles(output_dir / 'bto', recursive=True)),
len(fs.lsfiles(output_dir / 'abf', recursive=True)),
len(fs.lsfiles(output_dir / 'arc', recursive=True)),
len(fs.lsfiles(output_dir / 'awo', recursive=True)),
len(fs.lsfiles(output_dir / 'pass', recursive=True)),
])
df = pd.DataFrame(
rows, columns=['Testbed', 'bc', 'bto', 'abf', 'arc', 'awo', 'pass'])
df['Total'] = df.sum(axis=1)
return df
def test_lsfiles_recursive():
assert fs.lsfiles("labm8/data/test/testdir", recursive=True) == [
"a",
"b",
"c/e",
"c/f/f/i",
"c/f/h",
"c/g",
"d",
]
def main(argv):
"""Main entry point."""
if len(argv) > 1:
unknown_args = ', '.join(argv[1:])
raise app.UsageError(f'Unknown arguments "{unknown_args}"')
logging.info('Initializing datastore.')
config = pathlib.Path(FLAGS.datastore)
ds = datastore.DataStore.FromFile(config)
output_dir = pathlib.Path(FLAGS.output_directory)
# Make directories to write the classifications to. We use the same shorthand
# classification names as in Table 2 of the paper:
#
# http://chriscummins.cc/pub/2018-issta.pdf
(output_dir / 'bc').mkdir(parents=True, exist_ok=True)
(output_dir / 'bto').mkdir(exist_ok=True)
(output_dir / 'abf').mkdir(exist_ok=True)
(output_dir / 'arc').mkdir(exist_ok=True)
(output_dir / 'awo').mkdir(exist_ok=True)
(output_dir / 'pass').mkdir(exist_ok=True)
result_dirs = [
pathlib.Path(x) for x in FLAGS.input_directories
if pathlib.Path(x).is_dir()
]
results_paths = labtypes.flatten([
pathlib.Path(x) for x in fs.lsfiles(x, recursive=True, abspaths=True)
] for x in result_dirs)
logging.info('Importing %d results into datastore ...', len(results_paths))
with ds.Session(commit=True) as s:
for path in progressbar.ProgressBar()(results_paths):
# Instantiating a result from file has the side effect of adding the
# result object to the datastore's session.
result.Result.FromFile(s, path)
with ds.Session() as s:
testcases = s.query(testcase.Testcase)
logging.info('Difftesting the results of %d testcases ...',
testcases.count())
for t in progressbar.ProgressBar(
max_value=testcases.count())(testcases):
DifftestTestcase(s, t, output_dir)
df = ReadClassificationsToTable(output_dir)
print()
print(
'Table of results. For each testbed, this shows the number of results')
print('of each class, using the same shortand as in Table 2 of the paper:')
print('http://chriscummins.cc/pub/2018-issta.pdf')
print()
print(df.to_string(index=False))
print()
print('Individual classified programs are written to: '
f"'{output_dir}/<class>/<device>/'")
def main(argv):
"""Main entry point."""
if len(argv) > 1:
raise app.UsageError("Unknown arguments: '{}'.".format(' '.join(
argv[1:])))
if not FLAGS.export_path:
raise app.UsageError('--export_path must be a directory')
export_path = pathlib.Path(FLAGS.export_path)
if export_path.is_file():
raise app.UsageError('--export_path must be a directory')
# Make a directory for each outcome class.
for key in fish_pb2.CompilerCrashDiscriminatorTrainingExample.Outcome.keys(
):
(export_path / key.lower()).mkdir(parents=True, exist_ok=True)
logging.info('Connecting to MySQL database')
credentials = GetMySqlCredentials()
cnx = MySQLdb.connect(database='dsmith_04_opencl',
host='cc1',
user=credentials[0],
password=credentials[1])
cursor = cnx.cursor()
logging.info('Determining last export ID')
ids = sorted([
int(pathlib.Path(f).stem)
for f in fs.lsfiles(export_path, recursive=True, abspaths=True)
])
last_export_id = ids[-1] if ids else 0
logging.info('Exporting results from ID %s', last_export_id)
ExportOpenCLResults(cursor, last_export_id, export_path)
cursor.close()
cnx.close()
logging.info('Exported training set of %s files to %s',
humanize.intcomma(len(list(export_path.iterdir()))),
export_path)
def test_lsfiles_single_file(self):
self._test(["a"], fs.lsfiles("tests/data/testdir/a"))
def test_lsfiles_bad_path(self):
with self.assertRaises(OSError):
fs.lsfiles("/not/a/real/path/bro")
def test_lsfiles_recursive(self):
self._test(["a", "b", "c/e", "c/f/f/i", "c/f/h", "c/g", "d"],
fs.lsfiles("tests/data/testdir", recursive=True))
def test_lsfiles(self):
self._test(["a", "b", "d"], fs.lsfiles("tests/data/testdir"))
def test_lsfiles_single_file():
assert fs.lsfiles("labm8/data/test/testdir/a") == ["a"]
def test_lsfiles_bad_path():
with pytest.raises(OSError):
fs.lsfiles("/not/a/real/path/bro")
def test_lsfiles():
assert fs.lsfiles("labm8/data/test/testdir") == ["a", "b", "d"]