def test_local_file_non_archive(self):
"""This local file should be categorized as not an archive."""
path = "/tmp/file.txt"
self.assertEqual(path_is_archive(path), False)
self.assertEqual(get_archive_ext(path), "")
with self.assertRaises(UsageError):
strip_archive_ext(path)
def test_url_non_archive(self):
"""This URL should be categorized as not an archive."""
path = "https://codalab.org/file.txt"
self.assertEqual(path_is_archive(path), False)
self.assertEqual(get_archive_ext(path), "")
with self.assertRaises(UsageError):
strip_archive_ext(path)
def upload_bundle(self, source_file, bundle_type, worksheet_uuid):
"""
Upload |source_file| (a stream) to |worksheet_uuid|.
"""
# Construct info for creating the bundle.
bundle_subclass = get_bundle_subclass(bundle_type) # program or data
metadata = metadata_util.fill_missing_metadata(bundle_subclass, {}, initial_metadata={'name': source_file.filename, 'description': 'Upload ' + source_file.filename})
info = {'bundle_type': bundle_type, 'metadata': metadata}
# Upload it by creating a file handle and copying source_file to it (see RemoteBundleClient.upload_bundle in the CLI).
remote_file_uuid = self.client.open_temp_file(metadata['name'])
try:
with closing(RPCFileHandle(remote_file_uuid, self.client.proxy)) as dest:
file_util.copy(source_file.file, dest, autoflush=False, print_status='Uploading %s' % metadata['name'])
pack = False # For now, always unpack (note: do this after set remote_file_uuid, which needs the extension)
if not pack and zip_util.path_is_archive(metadata['name']):
metadata['name'] = zip_util.strip_archive_ext(metadata['name'])
# Then tell the client that the uploaded file handle is there.
new_bundle_uuid = self.client.finish_upload_bundle(
[remote_file_uuid],
not pack, # unpack
info,
worksheet_uuid,
True) # add_to_worksheet
except:
self.client.finalize_file(remote_file_uuid)
raise
return new_bundle_uuid
def get_blob(uuid, path=''):
"""
API to download the contents of a bundle or a subpath within a bundle.
For directories this method always returns a tarred and gzipped archive of
the directory.
For files, if the request has an Accept-Encoding header containing gzip,
then the returned file is gzipped.
"""
check_bundles_have_read_permission(local.model, request.user, [uuid])
bundle = local.model.get_bundle(uuid)
target_info = local.download_manager.get_target_info(uuid, path, 0)
if target_info is None:
abort(httplib.NOT_FOUND, 'Not found.')
# Figure out the file name.
if not path and bundle.metadata.name:
filename = bundle.metadata.name
else:
filename = target_info['name']
if target_info['type'] == 'directory':
# Always tar and gzip directories.
filename = filename + '.tar.gz'
fileobj = local.download_manager.stream_tarred_gzipped_directory(
uuid, path)
elif target_info['type'] == 'file':
if not zip_util.path_is_archive(
filename) and request_accepts_gzip_encoding():
# Let's gzip to save bandwidth. The browser will transparently decode
# the file.
filename = filename + '.gz'
fileobj = local.download_manager.stream_file(uuid,
path,
gzipped=True)
else:
fileobj = local.download_manager.stream_file(uuid,
path,
gzipped=False)
else:
# Symlinks.
abort(httplib.FORBIDDEN, 'Cannot download files of this type.')
# Set headers.
mimetype, _ = mimetypes.guess_type(filename, strict=False)
response.set_header('Content-Type', mimetype or 'text/plain')
if zip_util.get_archive_ext(
filename) == '.gz' and request_accepts_gzip_encoding():
filename = zip_util.strip_archive_ext(filename)
response.set_header('Content-Encoding', 'gzip')
else:
response.set_header('Content-Encoding', 'identity')
response.set_header('Content-Disposition', 'filename="%s"' % filename)
return fileobj
def get_blob(uuid, path=''):
"""
API to download the contents of a bundle or a subpath within a bundle.
For directories this method always returns a tarred and gzipped archive of
the directory.
For files, if the request has an Accept-Encoding header containing gzip,
then the returned file is gzipped.
"""
check_bundles_have_read_permission(local.model, request.user, [uuid])
bundle = local.model.get_bundle(uuid)
target_info = local.download_manager.get_target_info(uuid, path, 0)
if target_info is None:
abort(httplib.NOT_FOUND, 'Not found.')
# Figure out the file name.
if not path and bundle.metadata.name:
filename = bundle.metadata.name
else:
filename = target_info['name']
if target_info['type'] == 'directory':
# Always tar and gzip directories.
filename = filename + '.tar.gz'
fileobj = local.download_manager.stream_tarred_gzipped_directory(uuid, path)
elif target_info['type'] == 'file':
if not zip_util.path_is_archive(filename) and request_accepts_gzip_encoding():
# Let's gzip to save bandwidth. The browser will transparently decode
# the file.
filename = filename + '.gz'
fileobj = local.download_manager.stream_file(uuid, path, gzipped=True)
else:
fileobj = local.download_manager.stream_file(uuid, path, gzipped=False)
else:
# Symlinks.
abort(httplib.FORBIDDEN, 'Cannot download files of this type.')
# Set headers.
mimetype, _ = mimetypes.guess_type(filename, strict=False)
response.set_header('Content-Type', mimetype or 'text/plain')
if zip_util.get_archive_ext(filename) == '.gz' and request_accepts_gzip_encoding():
filename = zip_util.strip_archive_ext(filename)
response.set_header('Content-Encoding', 'gzip')
else:
response.set_header('Content-Encoding', 'identity')
response.set_header('Content-Disposition', 'filename="%s"' % filename)
return fileobj
def upload_bundle(self, source_file, bundle_type, worksheet_uuid):
"""
Upload |source_file| (a stream) to |worksheet_uuid|.
"""
# Construct info for creating the bundle.
bundle_subclass = get_bundle_subclass(bundle_type) # program or data
metadata = metadata_util.fill_missing_metadata(
bundle_subclass, {},
initial_metadata={
'name': source_file.filename,
'description': 'Upload ' + source_file.filename
})
info = {'bundle_type': bundle_type, 'metadata': metadata}
# Upload it by creating a file handle and copying source_file to it (see RemoteBundleClient.upload_bundle in the CLI).
remote_file_uuid = self.client.open_temp_file(metadata['name'])
try:
with closing(RPCFileHandle(remote_file_uuid,
self.client.proxy)) as dest:
file_util.copy(source_file.file,
dest,
autoflush=False,
print_status='Uploading %s' % metadata['name'])
pack = False # For now, always unpack (note: do this after set remote_file_uuid, which needs the extension)
if not pack and zip_util.path_is_archive(metadata['name']):
metadata['name'] = zip_util.strip_archive_ext(metadata['name'])
# Then tell the client that the uploaded file handle is there.
new_bundle_uuid = self.client.finish_upload_bundle(
[remote_file_uuid],
not pack, # unpack
info,
worksheet_uuid,
True) # add_to_worksheet
except:
self.client.finalize_file(remote_file_uuid)
raise
return new_bundle_uuid
def upload(self, sources, follow_symlinks, exclude_patterns, git, unpack, remove_sources):
'''
|sources|: specifies the locations of the contents to upload. Each element is either a URL or a local path.
|follow_symlinks|: for local path(s), whether to follow (resolve) symlinks
|exclude_patterns|: for local path(s), don't upload these patterns (e.g., *.o)
|git|: for URL, whether |source| is a git repo to clone.
|unpack|: for each source in |sources|, whether to unpack it if it's an archive.
|remove_sources|: remove |sources|.
If |sources| contains one source, then the bundle contents will be that source.
Otherwise, the bundle contents will be a directory with each of the sources.
Exceptions:
- If |git|, then each source is replaced with the result of running 'git clone |source|'
- If |unpack| is True or a source is an archive (zip, tar.gz, etc.), then unpack the source.
Install the contents of the directory at |source| into
DATA_SUBDIRECTORY in a subdirectory named by a hash of the contents.
Return a (data_hash, metadata) pair, where the metadata is a dict mapping
keys to precomputed statistics about the new data directory.
'''
to_delete = []
# Create temporary directory as a staging area and put everything there.
temp_path = tempfile.mkdtemp('-bundle_store_upload')
temp_subpaths = []
for source in sources:
# Where to save |source| to (might change this value if we unpack).
temp_subpath = os.path.join(temp_path, os.path.basename(source))
if remove_sources:
to_delete.append(source)
source_unpack = unpack and zip_util.path_is_archive(source)
if path_util.path_is_url(source):
# Download the URL.
print_util.open_line('BundleStore.upload: downloading %s to %s' % (source, temp_path))
if git:
file_util.git_clone(source, temp_subpath)
else:
file_util.download_url(source, temp_subpath, print_status=True)
if source_unpack:
zip_util.unpack(temp_subpath, zip_util.strip_archive_ext(temp_subpath))
path_util.remove(temp_subpath)
temp_subpath = zip_util.strip_archive_ext(temp_subpath)
print_util.clear_line()
else:
# Copy the local path.
source_path = path_util.normalize(source)
path_util.check_isvalid(source_path, 'upload')
# Recursively copy the directory into a new BundleStore temp directory.
print_util.open_line('BundleStore.upload: %s => %s' % (source_path, temp_subpath))
if source_unpack:
zip_util.unpack(source_path, zip_util.strip_archive_ext(temp_subpath))
temp_subpath = zip_util.strip_archive_ext(temp_subpath)
else:
if remove_sources:
path_util.rename(source_path, temp_subpath)
else:
path_util.copy(source_path, temp_subpath, follow_symlinks=follow_symlinks, exclude_patterns=exclude_patterns)
print_util.clear_line()
temp_subpaths.append(temp_subpath)
# If exactly one source, then upload that directly.
if len(temp_subpaths) == 1:
to_delete.append(temp_path)
temp_path = temp_subpaths[0]
# Multiplex between uploading a directory and uploading a file here.
# All other path_util calls will use these lists of directories and files.
if os.path.isdir(temp_path):
dirs_and_files = path_util.recursive_ls(temp_path)
else:
dirs_and_files = ([], [temp_path])
# Hash the contents of the temporary directory, and then if there is no
# data with this hash value, move this directory into the data directory.
print_util.open_line('BundleStore.upload: hashing %s' % temp_path)
data_hash = '0x%s' % (path_util.hash_directory(temp_path, dirs_and_files),)
print_util.clear_line()
print_util.open_line('BundleStore.upload: computing size of %s' % temp_path)
data_size = path_util.get_size(temp_path, dirs_and_files)
print_util.clear_line()
final_path = os.path.join(self.data, data_hash)
if os.path.exists(final_path):
# Already exists, just delete it
path_util.remove(temp_path)
else:
print >>sys.stderr, 'BundleStore.upload: moving %s to %s' % (temp_path, final_path)
path_util.rename(temp_path, final_path)
# Delete paths.
for path in to_delete:
if os.path.exists(path):
path_util.remove(path)
# After this operation there should always be a directory at the final path.
assert(os.path.lexists(final_path)), 'Uploaded to %s failed!' % (final_path,)
return (data_hash, {'data_size': data_size})
def upload_to_bundle_store(self, bundle, sources, follow_symlinks, exclude_patterns, remove_sources, git, unpack, simplify_archives):
"""
Uploads contents for the given bundle to the bundle store.
|sources|: specifies the locations of the contents to upload. Each element is
either a URL, a local path or a tuple (filename, file-like object).
|follow_symlinks|: for local path(s), whether to follow (resolve) symlinks,
but only if remove_sources is False.
|exclude_patterns|: for local path(s), don't upload these patterns (e.g., *.o),
but only if remove_sources is False.
|remove_sources|: for local path(s), whether |sources| should be removed
|git|: for URLs, whether |source| is a git repo to clone.
|unpack|: for each source in |sources|, whether to unpack it if it's an archive.
|simplify_archives|: whether to simplify unpacked archives so that if they
contain a single file, the final path is just that file,
not a directory containing that file.
If |sources| contains one source, then the bundle contents will be that source.
Otherwise, the bundle contents will be a directory with each of the sources.
Exceptions:
- If |git|, then each source is replaced with the result of running 'git clone |source|'
- If |unpack| is True or a source is an archive (zip, tar.gz, etc.), then unpack the source.
"""
bundle_path = self._bundle_store.get_bundle_location(bundle.uuid)
try:
path_util.make_directory(bundle_path)
# Note that for uploads with a single source, the directory
# structure is simplified at the end.
for source in sources:
is_url, is_local_path, is_fileobj, filename = self._interpret_source(source)
source_output_path = os.path.join(bundle_path, filename)
if is_url:
if git:
source_output_path = file_util.strip_git_ext(source_output_path)
file_util.git_clone(source, source_output_path)
else:
file_util.download_url(source, source_output_path)
if unpack and self._can_unpack_file(source_output_path):
self._unpack_file(
source_output_path, zip_util.strip_archive_ext(source_output_path),
remove_source=True, simplify_archive=simplify_archives)
elif is_local_path:
source_path = path_util.normalize(source)
path_util.check_isvalid(source_path, 'upload')
if unpack and self._can_unpack_file(source_path):
self._unpack_file(
source_path, zip_util.strip_archive_ext(source_output_path),
remove_source=remove_sources, simplify_archive=simplify_archives)
elif remove_sources:
path_util.rename(source_path, source_output_path)
else:
path_util.copy(source_path, source_output_path, follow_symlinks=follow_symlinks, exclude_patterns=exclude_patterns)
elif is_fileobj:
if unpack and zip_util.path_is_archive(filename):
self._unpack_fileobj(
source[0], source[1],
zip_util.strip_archive_ext(source_output_path),
simplify_archive=simplify_archives)
else:
with open(source_output_path, 'wb') as out:
shutil.copyfileobj(source[1], out)
if len(sources) == 1:
self._simplify_directory(bundle_path)
except:
if os.path.exists(bundle_path):
path_util.remove(bundle_path)
raise
def upload_to_bundle_store(
self,
bundle,
sources,
follow_symlinks,
exclude_patterns,
remove_sources,
git,
unpack,
simplify_archives,
):
"""
Uploads contents for the given bundle to the bundle store.
|sources|: specifies the locations of the contents to upload. Each element is
either a URL, a local path or a tuple (filename, binary file-like object).
|follow_symlinks|: for local path(s), whether to follow (resolve) symlinks,
but only if remove_sources is False.
|exclude_patterns|: for local path(s), don't upload these patterns (e.g., *.o),
but only if remove_sources is False.
|remove_sources|: for local path(s), whether |sources| should be removed
|git|: for URLs, whether |source| is a git repo to clone.
|unpack|: for each source in |sources|, whether to unpack it if it's an archive.
|simplify_archives|: whether to simplify unpacked archives so that if they
contain a single file, the final path is just that file,
not a directory containing that file.
If |sources| contains one source, then the bundle contents will be that source.
Otherwise, the bundle contents will be a directory with each of the sources.
Exceptions:
- If |git|, then each source is replaced with the result of running 'git clone |source|'
- If |unpack| is True or a source is an archive (zip, tar.gz, etc.), then unpack the source.
"""
exclude_patterns = (self._default_exclude_patterns +
exclude_patterns if exclude_patterns else
self._default_exclude_patterns)
bundle_link_url = getattr(bundle.metadata, "link_url", None)
if bundle_link_url:
# Don't do anything for linked bundles.
return
bundle_path = self._bundle_store.get_bundle_location(bundle.uuid)
try:
path_util.make_directory(bundle_path)
# Note that for uploads with a single source, the directory
# structure is simplified at the end.
for source in sources:
is_url, is_local_path, is_fileobj, filename = self._interpret_source(
source)
source_output_path = os.path.join(bundle_path, filename)
if is_url:
if git:
source_output_path = file_util.strip_git_ext(
source_output_path)
file_util.git_clone(source, source_output_path)
else:
file_util.download_url(source, source_output_path)
if unpack and self._can_unpack_file(
source_output_path):
self._unpack_file(
source_output_path,
zip_util.strip_archive_ext(source_output_path),
remove_source=True,
simplify_archive=simplify_archives,
)
elif is_local_path:
source_path = path_util.normalize(source)
path_util.check_isvalid(source_path, 'upload')
if unpack and self._can_unpack_file(source_path):
self._unpack_file(
source_path,
zip_util.strip_archive_ext(source_output_path),
remove_source=remove_sources,
simplify_archive=simplify_archives,
)
elif remove_sources:
path_util.rename(source_path, source_output_path)
else:
path_util.copy(
source_path,
source_output_path,
follow_symlinks=follow_symlinks,
exclude_patterns=exclude_patterns,
)
elif is_fileobj:
if unpack and zip_util.path_is_archive(filename):
self._unpack_fileobj(
source[0],
source[1],
zip_util.strip_archive_ext(source_output_path),
simplify_archive=simplify_archives,
)
else:
with open(source_output_path, 'wb') as out:
shutil.copyfileobj(source[1], out)
if len(sources) == 1:
self._simplify_directory(bundle_path)
except:
if os.path.exists(bundle_path):
path_util.remove(bundle_path)
raise
def _fetch_bundle_contents_blob(uuid, path=''):
"""
API to download the contents of a bundle or a subpath within a bundle.
For directories this method always returns a tarred and gzipped archive of
the directory.
For files, if the request has an Accept-Encoding header containing gzip,
then the returned file is gzipped.
"""
byte_range = get_request_range()
head_lines = query_get_type(int, 'head', default=0)
tail_lines = query_get_type(int, 'tail', default=0)
max_line_length = query_get_type(int, 'max_line_length', default=128)
check_bundles_have_read_permission(local.model, request.user, [uuid])
bundle = local.model.get_bundle(uuid)
target_info = local.download_manager.get_target_info(uuid, path, 0)
if target_info is None:
abort(httplib.NOT_FOUND, 'Not found.')
# Figure out the file name.
if not path and bundle.metadata.name:
filename = bundle.metadata.name
else:
filename = target_info['name']
if target_info['type'] == 'directory':
if byte_range:
abort(httplib.BAD_REQUEST,
'Range not supported for directory blobs.')
if head_lines:
abort(httplib.BAD_REQUEST,
'Head not supported for directory blobs.')
# Always tar and gzip directories.
filename = filename + '.tar.gz'
fileobj = local.download_manager.stream_tarred_gzipped_directory(
uuid, path)
elif target_info['type'] == 'file':
gzipped = False
if not zip_util.path_is_archive(
filename) and request_accepts_gzip_encoding():
# Let's gzip to save bandwidth. The browser will transparently decode
# the file.
filename = filename + '.gz'
gzipped = True
if byte_range and (head_lines or tail_lines):
abort(httplib.BAD_REQUEST,
'Head and range not supported on the same request.')
elif byte_range:
start, end = byte_range
fileobj = local.download_manager.read_file_section(
uuid, path, start, end - start + 1, gzipped)
elif head_lines or tail_lines:
fileobj = local.download_manager.summarize_file(
uuid, path, head_lines, tail_lines, max_line_length, None,
gzipped)
else:
fileobj = local.download_manager.stream_file(uuid, path, gzipped)
else:
# Symlinks.
abort(httplib.FORBIDDEN, 'Cannot download files of this type.')
# Set headers.
mimetype, _ = mimetypes.guess_type(filename, strict=False)
response.set_header('Content-Type', mimetype or 'text/plain')
if zip_util.get_archive_ext(
filename) == '.gz' and request_accepts_gzip_encoding():
filename = zip_util.strip_archive_ext(filename)
response.set_header('Content-Encoding', 'gzip')
else:
response.set_header('Content-Encoding', 'identity')
response.set_header('Content-Disposition', 'filename="%s"' % filename)
return fileobj
def test_url_archive(self):
"""This URL should be categorized as an archive."""
path = "https://codalab.org/file.tar.gz"
self.assertEqual(path_is_archive(path), True)
self.assertEqual(get_archive_ext(path), ".tar.gz")
self.assertEqual(strip_archive_ext(path), "https://codalab.org/file")
def test_local_file_archive(self):
"""This local should be categorized as an archive."""
path = "/tmp/file.tar.gz"
self.assertEqual(path_is_archive(path), True)
self.assertEqual(get_archive_ext(path), ".tar.gz")
self.assertEqual(strip_archive_ext(path), "/tmp/file")
def upload(self, sources, follow_symlinks, exclude_patterns, git, unpack, remove_sources, uuid):
"""
|sources|: specifies the locations of the contents to upload. Each element is either a URL or a local path.
|follow_symlinks|: for local path(s), whether to follow (resolve) symlinks
|exclude_patterns|: for local path(s), don't upload these patterns (e.g., *.o)
|git|: for URL, whether |source| is a git repo to clone.
|unpack|: for each source in |sources|, whether to unpack it if it's an archive.
|remove_sources|: remove |sources|.
If |sources| contains one source, then the bundle contents will be that source.
Otherwise, the bundle contents will be a directory with each of the sources.
Exceptions:
- If |git|, then each source is replaced with the result of running 'git clone |source|'
- If |unpack| is True or a source is an archive (zip, tar.gz, etc.), then unpack the source.
Install the contents of the directory at |source| into
DATA_SUBDIRECTORY in a subdirectory named by a hash of the contents.
Return a (data_hash, metadata) pair, where the metadata is a dict mapping
keys to precomputed statistics about the new data directory.
"""
to_delete = []
# If just a single file, set the final path to be equal to that file
single_path = len(sources) == 1
# Determine which disk this will go on
disk_choice = self.ring.get_node(uuid)
final_path = os.path.join(self.partitions, disk_choice, self.DATA_SUBDIRECTORY, uuid)
if os.path.exists(final_path):
raise UsageError('Path %s already present in bundle store' % final_path)
# Only make if not there
elif not single_path:
path_util.make_directory(final_path)
# Paths to resources
subpaths = []
for source in sources:
# Where to save |source| to (might change this value if we unpack).
if not single_path:
subpath = os.path.join(final_path, os.path.basename(source))
else:
subpath = final_path
if remove_sources:
to_delete.append(source)
source_unpack = unpack and zip_util.path_is_archive(source)
if source_unpack and single_path:
# Load the file into the bundle store under the given path
subpath += zip_util.get_archive_ext(source)
if path_util.path_is_url(source):
# Download the URL.
print_util.open_line('BundleStore.upload: downloading %s to %s' % (source, subpath))
if git:
file_util.git_clone(source, subpath)
else:
file_util.download_url(source, subpath, print_status=True)
if source_unpack:
zip_util.unpack(subpath, zip_util.strip_archive_ext(subpath))
path_util.remove(subpath)
subpath = zip_util.strip_archive_ext(subpath)
print_util.clear_line()
else:
# Copy the local path.
source_path = path_util.normalize(source)
path_util.check_isvalid(source_path, 'upload')
# Recursively copy the directory into the BundleStore
print_util.open_line('BundleStore.upload: %s => %s' % (source_path, subpath))
if source_unpack:
zip_util.unpack(source_path, zip_util.strip_archive_ext(subpath))
subpath = zip_util.strip_archive_ext(subpath)
else:
if remove_sources:
path_util.rename(source_path, subpath)
else:
path_util.copy(source_path, subpath, follow_symlinks=follow_symlinks, exclude_patterns=exclude_patterns)
print_util.clear_line()
subpaths.append(subpath)
dirs_and_files = None
if os.path.isdir(final_path):
dirs_and_files = path_util.recursive_ls(final_path)
else:
dirs_and_files = [], [final_path]
# Hash the contents of the bundle directory. Update the data_hash attribute
# for the bundle
print_util.open_line('BundleStore.upload: hashing %s' % final_path)
data_hash = '0x%s' % (path_util.hash_directory(final_path, dirs_and_files))
print_util.clear_line()
print_util.open_line('BundleStore.upload: computing size of %s' % final_path)
data_size = path_util.get_size(final_path, dirs_and_files)
print_util.clear_line()
# Delete paths.
for path in to_delete:
if os.path.exists(path):
path_util.remove(path)
# After this operation there should always be a directory at the final path.
assert (os.path.lexists(final_path)), 'Uploaded to %s failed!' % (final_path,)
return (data_hash, {'data_size': data_size})
