def test_corrupted_doc(self):
corrupted_doc_handle = FilesystemHandle.make_handle(
os.path.join(test_data_path, "msoffice/corrupted/test.trunc.doc"))
corrupted_doc = Source.from_handle(corrupted_doc_handle)
with SourceManager() as sm:
self.assertEqual(
list(corrupted_doc.handles(sm)), [],
"unrecognised CDFV2 document should be empty and wasn't")
def test_alternative_trimming(self):
alternative_source = MailSource(
FilesystemHandle.make_handle(
os.path.join(test_data_path, "alternative.eml")))
with SourceManager() as sm:
self.assertEqual(
len(list(alternative_source.handles(sm))),
1,
"text/plain trimming failed")
def test_libreoffice_size(self):
large_doc_handle = FilesystemHandle.make_handle(
os.path.join(test_data_path, "libreoffice/html-explosion.ods"))
large_doc = Source.from_handle(large_doc_handle)
with SourceManager() as sm:
for h in large_doc.handles(sm):
if h.name.endswith(".html"):
r = h.follow(sm)
self.assertLess(r.get_size().value, 1048576,
"LibreOffice HTML output was too big")
try:
cwd = Path(__file__).parent.absolute()
except:
cwd = Path().absolute()
fpath = cwd / '../data/files/document.docx'
#fpath = cwd / '../data/files/cpr-examples.odt'
reload_content = True
reload_content = False
try:
content
except:
reload_content = True
if reload_content:
h = FilesystemHandle.make_handle(fpath)
content = get_content_from_handle(h)
# newrule = CPRRule(modulus_11=True, ignore_irrelevant=False,
# examine_context=True)
# newrule.extract_surrounding_words = MethodType(extract_surrounding_words_fixed, newrule)
rules = [
(CPRSimple(modulus_11=True, ignore_irrelevant=False,
examine_context=True), "simple w. context"),
(CPRComplicated(modulus_11=True,
ignore_irrelevant=False,
examine_context=True), "'accepted' w. context"),
(CPRRule(modulus_11=True, ignore_irrelevant=False,
examine_context=True), "current w. context"),
# (CPRRule(modulus_11=True, ignore_irrelevant=False, examine_context=True),
def test_pdf_source(self):
self.run_rule(
PDFSource(
FilesystemHandle.make_handle(
os.path.join(test_data_path, "pdf/embedded-cpr.pdf"))))
def test_libreoffice_source(self):
self.run_rule(
LibreOfficeSource(
FilesystemHandle.make_handle(
os.path.join(test_data_path,
"libreoffice/embedded-cpr.odt"))))
import os.path
import unittest
from os2datascanner.engine2.model.core import SourceManager
from os2datascanner.engine2.model.file import FilesystemHandle
from os2datascanner.engine2.conversions.types import OutputType
from os2datascanner.engine2.conversions.registry import convert
here_path = os.path.dirname(__file__)
image_handle = FilesystemHandle.make_handle(
os.path.join(here_path, "data/ocr/cpr.png"))
html_handle = FilesystemHandle.make_handle(
os.path.join(here_path, "data/html/simple.html"))
class Engine2ConversionTest(unittest.TestCase):
def setUp(self):
self._sm = SourceManager()
self._ir = image_handle.follow(self._sm)
self._hr = html_handle.follow(self._sm)
def tearDown(self):
self._sm.clear()
def test_last_modified(self):
self.assertIsNotNone(
convert(self._ir, OutputType.LastModified).value)
import os.path
import unittest
from os2datascanner.engine2.model.core import SourceManager
from os2datascanner.engine2.model.file import FilesystemHandle
here_path = os.path.dirname(__file__)
doc_handle = FilesystemHandle.make_handle(
os.path.join(
here_path, "data", "msoffice", "test.doc"))
docx_handle = FilesystemHandle.make_handle(
os.path.join(
here_path, "data", "msoffice", "test.docx"))
class Engine2MIMETests(unittest.TestCase):
def test_doc_mime(self):
self.assertEqual(
doc_handle.guess_type(),
"application/msword",
".doc MIME guess is incorrect")
with SourceManager() as sm:
self.assertEqual(
doc_handle.follow(sm).compute_type(),
"application/msword",
".doc MIME computation is incorrect")
def test_docx_mime(self):
self.assertEqual(
docx_handle.guess_type(),
def test_xlsx(self):
self.run_rule_on_handle(
FilesystemHandle.make_handle(
os.path.join(test_data_path, "msoffice/test.xlsx")))
def test_ods(self):
self.run_rule_on_handle(
FilesystemHandle.make_handle(
os.path.join(test_data_path, "libreoffice/test.ods")))
def test_docx(self):
self.run_rule_on_handle(
FilesystemHandle.make_handle(
os.path.join(test_data_path, "msoffice/embedded-cpr.docx")))
def test_odt(self):
self.run_rule_on_handle(
FilesystemHandle.make_handle(
os.path.join(test_data_path, "libreoffice/embedded-cpr.odt")))
def handle(self, **kwargs):
with SourceManager() as sm:
for path in kwargs['FILE']:
guesses = guess_responsible_party(
FilesystemHandle.make_handle(path), sm)
print("{0}: {1}".format(path, guesses))