def test_convert_docx(self):
converter = TikaConverter()
filename = 'lorem.docx'
with open(os.path.join(ASSETS, filename)) as file_:
result = converter.convert_server(file_)
self.assertEquals(
'Lorem Ipsum',
strip_word_bookmarks(result.strip(), filename=filename))
def test_convert_docx(self):
converter = TikaConverter()
filename = 'lorem.docx'
with open(os.path.join(ASSETS, filename)) as file_:
result = converter.convert_server(file_)
self.assertEquals(
'Lorem Ipsum',
strip_word_bookmarks(result.strip(), filename=filename))
def convert(self, orig, data, filename='', mimetype=None, **kwargs):
converter = TikaConverter()
try:
plain_text = converter.convert(orig, filename=filename)
except (ConflictError, KeyboardInterrupt):
raise
except TikaConversionError, exc:
self._log_conversion_error(exc, mimetype=mimetype)
plain_text = ''
def convert(self, orig, data, filename='', mimetype=None, **kwargs):
converter = TikaConverter()
try:
plain_text = converter.convert(orig, filename=filename)
except (ConflictError, KeyboardInterrupt):
raise
except TikaConversionError, exc:
self._log_conversion_error(exc, mimetype=mimetype)
plain_text = ''
def test_converter_accepts_file_like_stream_object(self):
sample_text = 'TEXT'
# Patch run_process to just return sample output
mock_run_proc = self.mocker.replace('ftw.tika.converter.run_process')
self.expect(mock_run_proc(ARGS)).result((sample_text, 'stderr'))
self.replay()
with tempfile.NamedTemporaryFile() as tmp_file:
tika_converter = TikaConverter(path=tmp_file.name)
plain_text = tika_converter.convert(StringIO(sample_text))
self.assertEquals(plain_text, sample_text)
def test_process_error_causes_coverter_to_raise_conversion_error(self):
# Patch run_process to just raise a ProcessError
def raise_process_error(cmd):
raise ProcessError
mock_run_proc = self.mocker.replace('ftw.tika.converter.run_process')
self.expect(mock_run_proc(ARGS)).call(raise_process_error)
self.replay()
with tempfile.NamedTemporaryFile() as tmp_file:
tika_converter = TikaConverter(path=tmp_file.name)
with self.assertRaises(TikaConversionError):
tika_converter.convert('')
def test_converter_builds_correct_command_line(self):
# Patch run_process to just return stderr and the command line given
def return_cmd_line(cmd):
return (cmd, '')
mock_run_proc = self.mocker.replace('ftw.tika.converter.run_process')
self.expect(mock_run_proc(ARGS)).call(return_cmd_line)
self.replay()
jar_path = '/bin/ls'
tika_converter = TikaConverter(path=jar_path)
cmd = tika_converter.convert('')
cmd_without_doc_filename = cmd.split()[:-1]
self.assertEquals(cmd_without_doc_filename,
['java', '-jar', jar_path, '-t'])
def convert(self, orig, data, filename='', mimetype=None, **kwargs):
converter = TikaConverter()
try:
plain_text = converter.convert(orig, filename=filename)
except (ConflictError, KeyboardInterrupt):
raise
except TikaConversionError as exc:
self._log_conversion_error(exc, mimetype=mimetype)
plain_text = ''
except Exception as exc:
logger.warn(exc)
plain_text = ''
data.setData(plain_text)
return data
def test_missing_jar_path_causes_converter_to_raise(self):
# Since the path is configured in the IZCMLTikaConfig from the ZCML
# loaded in the testing layer, we need to unregister the config for
# this test to verify the exception.
config = getUtility(IZCMLTikaConfig)
getGlobalSiteManager().unregisterUtility(provided=IZCMLTikaConfig)
try:
with self.assertRaises(TikaJarNotConfigured):
TikaConverter().convert('')
finally:
getGlobalSiteManager().registerUtility(component=config)
def test_convert_docx(self):
converter = TikaConverter()
with open(os.path.join(ASSETS, "lorem.docx")) as file_:
result = converter.convert_local(file_)
self.assertEquals("Lorem Ipsum", result.strip())
def test_convert_docx(self):
converter = TikaConverter()
with open(os.path.join(ASSETS, 'lorem.docx')) as file_:
result = converter.convert_local(file_)
self.assertEquals('Lorem Ipsum', result.strip())
def test_invalid_jar_path_causes_converter_to_raise(self):
tika_converter = TikaConverter(path="/nonexistent")
with self.assertRaises(TikaJarNotFound):
tika_converter.convert('')