def process_ocr(self, page, index=True):
_logger.debug("extracting ocr text and word coords for %s" % page.url)
url = urlparse.urljoin(self.current_batch.storage_url, page.ocr_filename)
lang_text, coords = ocr_extractor(url)
if self.PROCESS_COORDINATES:
self._process_coordinates(page, coords)
ocr = OCR()
ocr.page = page
ocr.save()
for lang, text in lang_text.iteritems():
try:
language = models.Language.objects.get(Q(code=lang) | Q(lingvoj__iendswith=lang))
except models.Language.DoesNotExist:
# default to english as per requirement
language = models.Language.objects.get(code="eng")
ocr.language_texts.create(language=language, text=text)
page.ocr = ocr
if index:
_logger.debug("indexing ocr for: %s" % page.url)
self.solr.add(**page.solr_doc)
page.indexed = True
page.save()
def test_extractor(self):
dir = join(dirname(dirname(__file__)), 'test-data')
ocr_file = join(dir, 'ocr.xml')
text, coord_info = ocr_extractor(ocr_file)
coords = coord_info["coords"]
expected_text = {"eng": file(join(dir, 'ocr.txt')).read().decode('utf-8')}
self.assertEqual(text, expected_text)
self.assertEqual(len(coords.keys()), 2150)
self.assertEqual(len(coords['place']), 3)
# Craft. should be normalized to Craft
# since Solr's highlighting will not include
# trailing punctuation in highlighted text
self.assertTrue(coords.has_key('Craft'))
self.assertTrue(not coords.has_key('Craft.'))
def process_coordinates(self, batch_path):
logging.info("process word coordinates for batch at %s", batch_path)
dirname, batch_name = os.path.split(batch_path.rstrip("/"))
if dirname:
batch_source = None
else:
batch_source = urlparse.urljoin(settings.BATCH_STORAGE, batch_name)
if not batch_source.endswith("/"):
batch_source += "/"
batch_name = _normalize_batch_name(batch_name)
try:
batch = self._get_batch(batch_name, batch_source, create=False)
self.current_batch = batch
for issue in batch.issues.all():
for page in issue.pages.all():
url = urlparse.urljoin(self.current_batch.storage_url, page.ocr_filename)
lang_text, coords = ocr_extractor(url)
self._process_coordinates(page, coords)
except Exception, e:
msg = "unable to process coordinates for batch: %s" % e
_logger.error(msg)
_logger.exception(e)
raise BatchLoaderException(msg)
