def IndexContentFiles(job: scrape_repos_pb2.ImportWorker) -> None:
"""Index content files."""
relpath = job.abspath[len(str(job.clone_dir)) + 1 :]
try:
texts = preprocessors.Preprocess(
pathlib.Path(job.clone_dir),
relpath,
job.all_files_relpaths,
job.preprocessors,
)
for i, text in enumerate(texts):
sha256 = hashlib.sha256(text.encode("utf-8"))
proto = scrape_repos_pb2.ContentFile(
clone_from_url=job.clone_from_url,
relpath=relpath,
artifact_index=i,
sha256=sha256.digest(),
charcount=len(text),
linecount=len(text.split("\n")),
text=text,
)
path = pathlib.Path(job.index_dir) / (
binascii.hexlify(proto.sha256).decode("utf-8") + ".pbtxt"
)
pbutil.ToFile(proto, path)
except UnicodeDecodeError:
app.Warning("Failed to decode %s", relpath)
def test_Preprocess_mock_preprocessor(tempdir):
"""Test unmodified output if no preprocessors."""
MakeFile(tempdir, 'a', 'hello')
assert preprocessors.Preprocess(tempdir, 'a', ['a'], [
'datasets.github.scrape_repos.preprocessors.preprocessors_test'
':MockPreprocessor'
]) == ['PREPROCESSED']
def test_Preprocess_mock_preprocessor_exception(tempdir):
"""Test that an exception is propagated."""
MakeFile(tempdir, 'a', 'hello')
with pytest.raises(ValueError):
preprocessors.Preprocess(tempdir, 'a', ['a'], [
'datasets.github.scrape_repos.preprocessors.preprocessors_test'
':MockPreprocessorInternalError'
])
def test_Preprocess_mock_preprocessor(tempdir):
"""Test unmodified output if no preprocessors."""
MakeFile(tempdir, "a", "hello")
assert preprocessors.Preprocess(
tempdir,
"a",
["a"],
[
"datasets.github.scrape_repos.preprocessors.preprocessors_test"
":MockPreprocessor"
],
) == ["PREPROCESSED"]
def test_Preprocess_mock_preprocessor_exception(tempdir):
"""Test that an exception is propagated."""
MakeFile(tempdir, "a", "hello")
with test.Raises(ValueError):
preprocessors.Preprocess(
tempdir,
"a",
["a"],
[
"datasets.github.scrape_repos.preprocessors.preprocessors_test"
":MockPreprocessorInternalError"
],
)
def ImportWorker(
job: scrape_repos_pb2.ImportWorker
) -> typing.List[contentfiles.ContentFile]:
"""Import a content file."""
relpath = job.abspath[len(str(job.clone_dir)) + 1:]
outputs: typing.List[contentfiles.ContentFile] = []
try:
texts = preprocessors.Preprocess(pathlib.Path(job.clone_dir), relpath,
job.all_files_relpaths, job.preprocessors)
for i, text in enumerate(texts):
sha256 = hashlib.sha256(text.encode('utf-8'))
outputs.append(contentfiles.ContentFile(
clone_from_url=job.clone_from_url,
relpath=relpath, artifact_index=i,
sha256=sha256.digest(), charcount=len(text),
linecount=len(text.split('\n')), text=text))
except UnicodeDecodeError:
logging.warning('Failed to decode %s', relpath)
return outputs
def test_Preprocess_no_preprocessors(tempdir):
"""Test unmodified output if no preprocessors."""
MakeFile(tempdir, 'a', 'hello')
assert preprocessors.Preprocess(tempdir, 'a', ['a'], []) == ['hello']
def test_Preprocess_no_preprocessors(tempdir): """Test unmodified output if no preprocessors.""" MakeFile(tempdir, "a", "hello") assert preprocessors.Preprocess(tempdir, "a", ["a"], []) == ["hello"]
