def download_and_analyze_file(self, file_id: int,
address_pipeline: AddressPipeline,
fallback_city: str) -> bool:
"""
Downloads and analyses a single file, i.e. extracting text, locations and persons.
Returns False for http errors on downloading and True otherwise.
"""
file = File.objects.get(id=file_id)
url = file.get_oparl_url()
with NamedTemporaryFile() as tmpfile:
try:
content, content_type = self.loader.load_file(url)
if content_type and file.mime_type and content_type != file.mime_type:
logger.warning(
"Diverging mime types: Expected {}, got {}".format(
file.mime_type, content_type))
file.mime_type = content_type or file.mime_type
tmpfile.write(content)
tmpfile.file.seek(0)
file.filesize = len(content)
except HTTPError:
logger.exception("File {}: Failed to download {}".format(
file.id, url))
return False
logger.debug("File {}: Downloaded {} ({}, {})".format(
file.id, url, file.mime_type, filesizeformat(file.filesize)))
minio_client().put_object(
minio_file_bucket,
str(file.id),
tmpfile.file,
file.filesize,
content_type=file.mime_type,
)
# If the api has text, keep that
if self.download_files and not file.parsed_text:
file.parsed_text, file.page_count = extract_from_file(
tmpfile.file, tmpfile.name, file.mime_type, file.id)
if file.parsed_text:
locations = extract_locations(file.parsed_text,
pipeline=address_pipeline,
fallback_city=fallback_city)
file.locations.set(locations)
persons = extract_persons(file.name + "\n" +
(file.parsed_text or "") + "\n")
file.mentioned_persons.set(persons)
logger.debug("File {}: Found {} locations and {} persons".format(
file.id, len(locations), len(persons)))
else:
logger.warning("File {}: Couldn't get any text".format(file.id))
db.connections.close_all()
file.save()
return True
def test_person_extraction(self):
frank = Person.objects.get(pk=1)
doug = Person.objects.get(pk=4)
will = Person.objects.get(pk=7)
text = "A text \nabout Frank Underwood, Stamper, Doug, and a \nmisspelled WilliamConway."
persons = extract_persons(text)
self.assertTrue(doug in persons)
self.assertTrue(frank in persons)
self.assertFalse(will in persons)
text = "Also the more formal name, \"Underwood, Francis\" should be found."
persons = extract_persons(text)
self.assertFalse(doug in persons)
self.assertTrue(frank in persons)
self.assertFalse(will in persons)
def test_person_extraction(self):
frank = Person.objects.get(pk=1)
doug = Person.objects.get(pk=4)
will = Person.objects.get(pk=7)
text = "A text \nabout Frank Underwood, Stamper, Doug, and a \nmisspelled WilliamConway."
persons = extract_persons(text)
self.assertTrue(doug in persons)
self.assertTrue(frank in persons)
self.assertTrue(will not in persons)
text = 'Also the more formal name, "Underwood, Francis" should be found.'
persons = extract_persons(text)
self.assertTrue(doug not in persons)
self.assertTrue(frank in persons)
self.assertTrue(will not in persons)
text = "We should check word boundaries like Doug Stampering something."
persons = extract_persons(text)
self.assertTrue(doug not in persons)
def parse_file(self, file: File, fallback_city: str):
logging.info("- Parsing: " + str(file.id) + " (" + file.name + ")")
with minio_client().get_object(minio_file_bucket,
str(file.id)) as file_handle:
recognized_text = get_ocr_text_from_pdf(file_handle.read())
if len(recognized_text) > 0:
file.parsed_text = cleanup_extracted_text(recognized_text)
file.mentioned_persons = extract_persons(file.name + "\n" +
(recognized_text or "") +
"\n")
file.locations.set(
extract_locations(file.parsed_text, fallback_city))
file.save()
else:
logging.warning("Nothing recognized")
def parse_file(self, file: File):
logging.info("- Parsing: " + str(file.id) + " (" + file.name + ")")
file_path = os.path.abspath(os.path.dirname(__name__))
file_path = os.path.join(file_path, settings.MEDIA_ROOT,
file.storage_filename)
recognized_text = get_ocr_text_from_pdf(file_path)
if len(recognized_text) > 0:
file.parsed_text = cleanup_extracted_text(recognized_text)
file.mentioned_persons = extract_persons(file.name + "\n" +
(recognized_text or "") +
"\n")
file.locations = extract_locations(recognized_text)
file.save()
else:
logging.warning("Nothing recognized")
def file(self, libobject: OParl.File):
file, do_update = self.check_for_modification(libobject, File)
if not file or not do_update:
return file
self.logger.info("Processing File {}".format(libobject.get_id()))
if libobject.get_file_name():
displayed_filename = libobject.get_file_name()
elif libobject.get_name():
extension = mimetypes.guess_extension("application/pdf") or ""
length = self.filename_length_cutoff - len(extension)
displayed_filename = slugify(
libobject.get_name())[:length] + extension
else:
displayed_filename = slugify(
libobject.get_access_url())[-self.filename_length_cutoff:]
parsed_text_before = file.parsed_text
file_name_before = file.name
file.oparl_id = libobject.get_id()
file.name = libobject.get_name()
file.displayed_filename = displayed_filename
file.mime_type = libobject.get_mime_type(
) or "application/octet-stream"
file.legal_date = self.glib_datetime_to_python_date(
libobject.get_date())
file.sort_date = file.created
file.oparl_access_url = libobject.get_access_url()
file.oparl_download_url = libobject.get_download_url()
# If no text comes from the API, don't overwrite previously extracted PDF-content with an empty string
if libobject.get_text():
file.parsed_text = libobject.get_text()
if self.download_files:
self.download_file(file, libobject)
else:
file.storage_filename = ""
file.filesize = -1
parsed_text = file.parsed_text
if file.storage_filename and not file.parsed_text:
parsed_text = self.extract_text_from_file(file)
file = self.call_custom_hook("sanitize_file", file)
if len(file.name) > 200:
file.name = textwrap.wrap(file.name, 199)[0] + "\u2026"
file.save()
if file_name_before != file.name or parsed_text_before != file.parsed_text:
# These two operations are rather CPU-intensive, so we only perform them if something relevant has changed
file.locations = extract_locations(parsed_text)
file.mentioned_persons = extract_persons(file.name + "\n" +
(parsed_text or "") +
"\n")
file.save()
return file
def download_and_analyze_file(self, file_id: int,
address_pipeline: AddressPipeline,
fallback_city: str) -> bool:
"""
Downloads and analyses a single file, i.e. extracting text, locations and persons.
Returns False for http errors on downloading and True otherwise.
"""
file = File.objects.get(id=file_id)
url = file.get_oparl_url()
with NamedTemporaryFile() as tmp_file:
try:
content, content_type = self.loader.load_file(url)
if content_type and file.mime_type and content_type != file.mime_type:
logger.warning(
"Diverging mime types: Expected {}, got {}".format(
file.mime_type, content_type))
if content_type and content_type.split(";")[0] == "text/html":
logger.error(
f"File {file.id}: Content type was {content_type}, this seems to be a silent error"
)
return False
file.mime_type = content_type or file.mime_type
tmp_file.write(content)
tmp_file.file.seek(0)
file.filesize = len(content)
except RequestException as e:
# Normal server error
if e.response and 400 <= e.response.status_code < 600:
logger.error(
f"File {file.id}: Failed to download {url} with error {e.response.status_code}"
)
else:
logger.exception(
f"File {file.id}: Failed to download {url}")
return False
logger.debug("File {}: Downloaded {} ({}, {})".format(
file.id, url, file.mime_type, filesizeformat(file.filesize)))
if not settings.PROXY_ONLY_TEMPLATE:
minio_client().put_object(
minio_file_bucket,
str(file.id),
tmp_file.file,
file.filesize,
content_type=file.mime_type,
)
# If the api has text, keep that
if self.download_files and not file.parsed_text:
file.parsed_text, file.page_count = extract_from_file(
tmp_file.file, tmp_file.name, file.mime_type, file.id)
if file.parsed_text:
locations = extract_locations(file.parsed_text,
pipeline=address_pipeline,
fallback_city=fallback_city)
file.locations.set(locations)
persons = extract_persons(file.name + "\n" +
(file.parsed_text or "") + "\n")
file.mentioned_persons.set(persons)
logger.debug("File {}: Found {} locations and {} persons".format(
file.id, len(locations), len(persons)))
else:
logger.warning(f"File {file.id}: Couldn't get any text")
try:
db.connections.close_all()
file.save()
except (ElasticsearchException, DatabaseError) as e:
logger.exception(f"File {file.id}: Failed to save: {e}")
return False
return True
def parse_file(self, file: File):
logging.info("- Parsing: " + str(file.id) + " (" + file.name + ")")
file.mentioned_persons = extract_persons(file.name + "\n" + (file.parsed_text or "") + "\n")
file.save()
