def download_file(self, f, _recursion_count=0):
"""
Download a :py:mod:`ocrd.model.ocrd_file.OcrdFile` to the workspace.
"""
log = getLogger('ocrd.workspace.download_file')
log.debug('download_file %s [_recursion_count=%s]' % (f, _recursion_count))
with pushd_popd(self.directory):
try:
# If the f.url is already a file path, and is within self.directory, do nothing
url_path = Path(f.url).resolve()
if not (url_path.exists() and url_path.relative_to(str(Path(self.directory).resolve()))):
raise Exception("Not already downloaded, moving on")
except Exception as e:
basename = '%s%s' % (f.ID, MIME_TO_EXT.get(f.mimetype, '')) if f.ID else f.basename
try:
f.url = self.resolver.download_to_directory(self.directory, f.url, subdir=f.fileGrp, basename=basename)
except FileNotFoundError as e:
if not self.baseurl:
raise Exception("No baseurl defined by workspace. Cannot retrieve '%s'" % f.url)
if _recursion_count >= 1:
raise Exception("Already tried prepending baseurl '%s'. Cannot retrieve '%s'" % (self.baseurl, f.url))
log.debug("First run of resolver.download_to_directory(%s) failed, try prepending baseurl '%s': %s", f.url, self.baseurl, e)
f.url = '%s/%s' % (self.baseurl, f.url)
f.url = self.download_file(f, _recursion_count + 1).local_filename
f.local_filename = f.url
return f
def download_file(self, f, _recursion_count=0):
"""
Download a :py:mod:`ocrd.model.ocrd_file.OcrdFile` to the workspace.
"""
log.debug('download_file %s [_recursion_count=%s]' %
(f, _recursion_count))
with pushd_popd(self.directory):
# XXX FIXME hacky
basename = '%s%s' % (f.ID, MIME_TO_EXT.get(
f.mimetype, '')) if f.ID else f.basename
try:
f.url = self.resolver.download_to_directory(self.directory,
f.url,
subdir=f.fileGrp,
basename=basename)
except FileNotFoundError as e:
if not self.baseurl:
raise Exception(
"No baseurl defined by workspace. Cannot retrieve '%s'"
% f.url)
if _recursion_count >= 1:
raise Exception(
"Already tried prepending baseurl '%s'. Cannot retrieve '%s'"
% (self.baseurl, f.url))
log.debug(
"First run of resolver.download_to_directory(%s) failed, try prepending baseurl '%s': %s",
f.url, self.baseurl, e)
f.url = '%s/%s' % (self.baseurl, f.url)
f.url = self.download_file(f,
_recursion_count + 1).local_filename
# XXX FIXME HACK
f.local_filename = f.url
return f
def process(self):
LOG = getLogger('ocrd.dummy')
assert_file_grp_cardinality(self.input_file_grp, 1)
assert_file_grp_cardinality(self.output_file_grp, 1)
for input_file in self.input_files:
input_file = self.workspace.download_file(input_file)
file_id = make_file_id(input_file, self.output_file_grp)
ext = MIME_TO_EXT.get(input_file.mimetype, '')
local_filename = join(self.output_file_grp, file_id + ext)
pcgts = page_from_file(self.workspace.download_file(input_file))
pcgts.set_pcGtsId(file_id)
self.add_metadata(pcgts)
LOG.info("cp %s %s # %s -> %s", input_file.url, local_filename, input_file.ID, file_id)
if input_file.mimetype == MIMETYPE_PAGE:
# Source file is PAGE-XML: Write out in-memory PcGtsType
self.workspace.add_file(
ID=file_id,
file_grp=self.output_file_grp,
pageId=input_file.pageId,
mimetype=input_file.mimetype,
local_filename=local_filename,
content=to_xml(pcgts).encode('utf-8'))
else:
# Source file is not PAGE-XML: Copy byte-by-byte
with open(input_file.local_filename, 'rb') as f:
content = f.read()
self.workspace.add_file(
ID=file_id,
file_grp=self.output_file_grp,
pageId=input_file.pageId,
mimetype=input_file.mimetype,
local_filename=local_filename,
content=content)
if input_file.mimetype.startswith('image/'):
# write out the PAGE-XML representation for this image
page_file_id = file_id + '_PAGE'
pcgts.set_pcGtsId(page_file_id)
pcgts.get_Page().set_imageFilename(local_filename)
page_filename = join(self.output_file_grp, file_id + '.xml')
LOG.info("Add PAGE-XML %s generated for %s at %s",
page_file_id, file_id, page_filename)
self.workspace.add_file(
ID=page_file_id,
file_grp=self.output_file_grp,
pageId=input_file.pageId,
mimetype=MIMETYPE_PAGE,
local_filename=page_filename,
content=to_xml(pcgts).encode('utf-8'))