def test_superfluous_copies_in_ws_dir(self):
"""
https://github.com/OCR-D/core/issues/227
"""
def find_recursive(root):
ret = []
for _, _, f in walk(root):
for file in f:
ret.append(file)
return ret
with TemporaryDirectory() as wsdir:
with open(
assets.path_to(
'SBB0000F29300010000/data/mets_one_file.xml'),
'r') as f_in:
with open(join(wsdir, 'mets.xml'), 'w') as f_out:
f_out.write(f_in.read())
self.assertEqual(len(find_recursive(wsdir)), 1)
ws1 = Workspace(self.resolver, wsdir)
for file in ws1.mets.find_files():
ws1.download_file(file)
self.assertEqual(len(find_recursive(wsdir)), 2)
self.assertTrue(
exists(join(wsdir, 'OCR-D-IMG/FILE_0005_IMAGE.tif')))
def workspace_from_url(self, mets_url, dst_dir=None, clobber_mets=False, mets_basename=None, download=False, src_baseurl=None):
"""
Create a workspace from a METS by URL (i.e. clone it).
Sets the mets.xml file
Arguments:
mets_url (string): Source mets URL
dst_dir (string, None): Target directory for the workspace
clobber_mets (boolean, False): Whether to overwrite existing mets.xml. By default existing mets.xml will raise an exception.
download (boolean, False): Whether to download all the files
src_baseurl (string, None): Base URL for resolving relative file locations
Returns:
Workspace
"""
if mets_url is None:
raise ValueError("Must pass 'mets_url' workspace_from_url")
# if mets_url is a relative filename, make it absolute
if is_local_filename(mets_url) and not Path(mets_url).is_absolute():
mets_url = str(Path(Path.cwd() / mets_url))
# if mets_basename is not given, use the last URL segment of the mets_url
if mets_basename is None:
mets_basename = nth_url_segment(mets_url, -1)
# If src_baseurl wasn't given, determine from mets_url by removing last url segment
if not src_baseurl:
last_segment = nth_url_segment(mets_url)
src_baseurl = remove_non_path_from_url(remove_non_path_from_url(mets_url)[:-len(last_segment)])
# resolve dst_dir
if not dst_dir:
if is_local_filename(mets_url):
log.debug("Deriving dst_dir %s from %s", Path(mets_url).parent, mets_url)
dst_dir = Path(mets_url).parent
else:
log.debug("Creating ephemeral workspace '%s' for METS @ <%s>", dst_dir, mets_url)
dst_dir = tempfile.mkdtemp(prefix=TMP_PREFIX)
# XXX Path.resolve is always strict in Python <= 3.5, so create dst_dir unless it exists consistently
if not Path(dst_dir).exists():
Path(dst_dir).mkdir(parents=True, exist_ok=False)
dst_dir = str(Path(dst_dir).resolve())
log.debug("workspace_from_url\nmets_basename='%s'\nmets_url='%s'\nsrc_baseurl='%s'\ndst_dir='%s'",
mets_basename, mets_url, src_baseurl, dst_dir)
self.download_to_directory(dst_dir, mets_url, basename=mets_basename, if_exists='overwrite' if clobber_mets else 'raise')
workspace = Workspace(self, dst_dir, mets_basename=mets_basename, baseurl=src_baseurl)
if download:
for f in workspace.mets.find_files():
workspace.download_file(f)
return workspace
def workspace_from_url(self, mets_url, directory=None, clobber_mets=False, mets_basename=None, download=False, download_local=False):
"""
Create a workspace from a METS by URL.
Sets the mets.xml file
"""
if directory is not None and not directory.startswith('/'):
directory = os.path.abspath(directory)
if mets_url is None:
if directory is None:
raise Exception("Must pass mets_url and/or directory to workspace_from_url")
else:
mets_url = 'file://%s/%s' % (directory, mets_basename)
if mets_url.find('://') == -1:
# resolve to absolute
mets_url = os.path.abspath(mets_url)
mets_url = 'file://' + mets_url
if directory is None:
# if mets_url is a file-url assume working directory to be where
# the mets.xml resides
if mets_url.startswith('file://'):
# if directory was not given and mets_url is a file assume that
# directory should be the directory where the mets.xml resides
directory = os.path.dirname(mets_url[len('file://'):])
else:
directory = tempfile.mkdtemp(prefix=TMP_PREFIX)
log.debug("Creating workspace '%s' for METS @ <%s>", directory, mets_url)
# if mets_basename is not given, use the last URL segment of the mets_url
if mets_basename is None:
mets_basename = mets_url \
.rsplit('/', 1)[-1] \
.split('?')[0] \
.split('#')[0]
mets_fpath = os.path.join(directory, mets_basename)
log.debug("Copying mets url '%s' to '%s'", mets_url, mets_fpath)
if 'file://' + mets_fpath == mets_url:
log.debug("Target and source mets are identical")
else:
if os.path.exists(mets_fpath) and not clobber_mets:
raise Exception("File '%s' already exists but clobber_mets is false" % mets_fpath)
else:
self.download_to_directory(directory, mets_url, basename=mets_basename)
workspace = Workspace(self, directory, mets_basename=mets_basename)
if download_local or download:
for file_grp in workspace.mets.file_groups:
if download_local:
for f in workspace.mets.find_files(fileGrp=file_grp, local_only=True):
workspace.download_file(f, subdir=file_grp)
else:
workspace.download_files_in_group(file_grp)
return workspace
def test_superfluous_copies_in_ws_dir(tmp_path):
"""
https://github.com/OCR-D/core/issues/227
"""
# arrange
src_path = assets.path_to('SBB0000F29300010000/data/mets_one_file.xml')
dst_path = join(tmp_path, 'mets.xml')
copyfile(src_path, dst_path)
ws1 = Workspace(Resolver(), tmp_path)
# assert directory files
assert count_files(tmp_path) == 1
# act
for file in ws1.mets.find_all_files():
ws1.download_file(file)
# assert
assert count_files(tmp_path) == 2
assert exists(join(tmp_path, 'OCR-D-IMG/FILE_0005_IMAGE.tif'))
def test_227_1(self):
def find_recursive(root):
ret = []
for _, _, f in os.walk(root):
for file in f:
ret.append(file)
return ret
with TemporaryDirectory() as wsdir:
with open(
assets.path_to(
'SBB0000F29300010000/data/mets_one_file.xml'),
'r') as f_in:
with open(join(wsdir, 'mets.xml'), 'w') as f_out:
f_out.write(f_in.read())
self.assertEqual(len(find_recursive(wsdir)), 1)
ws1 = Workspace(self.resolver, wsdir)
for file in ws1.mets.find_files():
ws1.download_file(file)
self.assertEqual(len(find_recursive(wsdir)), 2)
self.assertTrue(exists(join(wsdir, 'OCR-D-IMG/FILE_0005_IMAGE')))
def workspace_from_url(self,
mets_url,
dst_dir=None,
clobber_mets=False,
mets_basename=None,
download=False,
src_baseurl=None):
"""
Create a workspace from a METS by URL (i.e. clone if :py:attr:`mets_url` is remote or :py:attr:`dst_dir` is given).
Arguments:
mets_url (string): Source METS URL or filesystem path
Keyword Arguments:
dst_dir (string, None): Target directory for the workspace. \
By default create a temporary directory under :py:data:`ocrd.constants.TMP_PREFIX`. \
(The resulting path can be retrieved via :py:attr:`ocrd.Workspace.directory`.)
clobber_mets (boolean, False): Whether to overwrite existing ``mets.xml``. \
By default existing ``mets.xml`` will raise an exception.
download (boolean, False): Whether to also download all the files referenced by the METS
src_baseurl (string, None): Base URL for resolving relative file locations
Download (clone) :py:attr:`mets_url` to ``mets.xml`` in :py:attr:`dst_dir`, unless
the former is already local and the latter is ``none`` or already identical to its directory name.
Returns:
a new :py:class:`~ocrd.workspace.Workspace`
"""
log = getLogger('ocrd.resolver.workspace_from_url')
if mets_url is None:
raise ValueError("Must pass 'mets_url' workspace_from_url")
# if mets_url is a relative filename, make it absolute
if is_local_filename(mets_url) and not Path(mets_url).is_absolute():
mets_url = str(Path(Path.cwd() / mets_url))
# if mets_basename is not given, use the last URL segment of the mets_url
if mets_basename is None:
mets_basename = nth_url_segment(mets_url, -1)
# If src_baseurl wasn't given, determine from mets_url by removing last url segment
if not src_baseurl:
last_segment = nth_url_segment(mets_url)
src_baseurl = remove_non_path_from_url(
remove_non_path_from_url(mets_url)[:-len(last_segment)])
# resolve dst_dir
if not dst_dir:
if is_local_filename(mets_url):
log.debug("Deriving dst_dir %s from %s",
Path(mets_url).parent, mets_url)
dst_dir = Path(mets_url).parent
else:
log.debug("Creating ephemeral workspace '%s' for METS @ <%s>",
dst_dir, mets_url)
dst_dir = mkdtemp(prefix=TMP_PREFIX)
# XXX Path.resolve is always strict in Python <= 3.5, so create dst_dir unless it exists consistently
if not Path(dst_dir).exists():
Path(dst_dir).mkdir(parents=True, exist_ok=False)
dst_dir = str(Path(dst_dir).resolve())
log.debug(
"workspace_from_url\nmets_basename='%s'\nmets_url='%s'\nsrc_baseurl='%s'\ndst_dir='%s'",
mets_basename, mets_url, src_baseurl, dst_dir)
self.download_to_directory(
dst_dir,
mets_url,
basename=mets_basename,
if_exists='overwrite' if clobber_mets else 'skip')
workspace = Workspace(self,
dst_dir,
mets_basename=mets_basename,
baseurl=src_baseurl)
if download:
for f in workspace.mets.find_files():
workspace.download_file(f)
return workspace
def workspace_from_url(self,
mets_url,
dst_dir=None,
clobber_mets=False,
mets_basename=None,
download=False,
baseurl=None):
"""
Create a workspace from a METS by URL.
Sets the mets.xml file
Arguments:
mets_url (string): Source mets URL
dst_dir (string, None): Target directory for the workspace
clobber_mets (boolean, False): Whether to overwrite existing mets.xml. By default existing mets.xml will raise an exception.
download (boolean, False): Whether to download all the files
baseurl (string, None): Base URL for resolving relative file locations
Returns:
Workspace
"""
if dst_dir and not dst_dir.startswith('/'):
dst_dir = abspath(dst_dir)
if mets_url is None:
if baseurl is None:
raise Exception(
"Must pass mets_url and/or baseurl to workspace_from_url")
else:
mets_url = 'file://%s/%s' % (baseurl, mets_basename
if mets_basename else 'mets.xml')
if baseurl is None:
baseurl = mets_url.rsplit('/', 1)[0]
log.debug(
"workspace_from_url\nmets_url='%s'\nbaseurl='%s'\ndst_dir='%s'",
mets_url, baseurl, dst_dir)
# resolve to absolute
if '://' not in mets_url:
mets_url = 'file://%s' % abspath(mets_url)
if dst_dir is None:
# if mets_url is a file-url assume working directory is source directory
if mets_url.startswith('file://'):
# if dst_dir was not given and mets_url is a file assume that
# dst_dir should be the directory where the mets.xml resides
dst_dir = dirname(mets_url[len('file://'):])
else:
dst_dir = tempfile.mkdtemp(prefix=TMP_PREFIX)
log.debug("Creating workspace '%s' for METS @ <%s>", dst_dir,
mets_url)
# if mets_basename is not given, use the last URL segment of the mets_url
if mets_basename is None:
mets_basename = mets_url \
.rsplit('/', 1)[-1] \
.split('?')[0] \
.split('#')[0]
dst_mets = join(dst_dir, mets_basename)
log.debug("Copying mets url '%s' to '%s'", mets_url, dst_mets)
if 'file://' + dst_mets == mets_url:
log.debug("Target and source mets are identical")
else:
if exists(dst_mets) and not clobber_mets:
raise Exception(
"File '%s' already exists but clobber_mets is false" %
dst_mets)
else:
self.download_to_directory(dst_dir,
mets_url,
basename=mets_basename)
workspace = Workspace(self,
dst_dir,
mets_basename=mets_basename,
baseurl=baseurl)
if download:
for f in workspace.mets.find_files():
workspace.download_file(f)
return workspace