Made hocr parse more robust to errors, added test for preprocess flag

virantha · virantha · commit 623cf32fe2b3 · 2016-02-27T22:55:03.000-05:00
diff --git a/.gitignore b/.gitignore
@@ -1,3 +1,5 @@
 *.pyc
 .*
 *~
+*.hocr
+*.jpg
diff --git a/pypdfocr/pypdfocr_pdf.py b/pypdfocr/pypdfocr_pdf.py
@@ -273,7 +273,13 @@ def add_text_layer(self,pdf, hocrfile, page_num,height, dpi):
         
       """
       hocr = ElementTree()
-      hocr.parse(hocrfile)
+      try: 
+        # It's possible tesseract has failed and written garbage to this hocr file, so we need to catch any exceptions
+          hocr.parse(hocrfile)
+      except Exception:
+          logging.info("Error loading hocr, not adding any text")
+          return 
+
       logging.debug(xml.etree.ElementTree.tostring(hocr.getroot()))
       for c in hocr.getroot():  # Find the <body> tag
           if c.tag != 'body':
diff --git a/test/pdfs/test_recipe.pdf b/test/pdfs/test_recipe.pdf
diff --git a/test/test_option_parsing.py b/test/test_option_parsing.py
@@ -21,6 +21,10 @@ def test_standalone(self):
         self.p.get_options(opts)
         assert(self.p.verbose)
 
+        opts.append('--preprocess')
+        self.p.get_options(opts)
+        assert(not self.p.skip_preprocess)
+
         assert(not self.p.enable_filing)
         assert(self.p.config == {})
 
diff --git a/test/test_pypdfocr.py b/test/test_pypdfocr.py
@@ -42,6 +42,8 @@ def _iter_pdf(self, filename):
                            ]),
             (".", os.path.join("temp","target","recipe"), os.path.join("..","test", "pdfs", "1.pdf"), [ ["Simply","Recipes"],
                                  ]),
+            (".", os.path.join("temp","target","recipe"), os.path.join("..","test", "pdfs", "test_recipe_sideways.pdf"), [ ["Simply","Recipes", 'spinach'],
+                                 ]),
         ]
 
     #@pytest.mark.skipif(True, reason="Just testing")
@@ -93,7 +95,7 @@ def test_standalone_email(self, dirname, tgt_folder, filename, expected):
         with patch("smtplib.SMTP") as mock_smtp:
             cwd = os.getcwd()
             os.chdir(dirname)
-            opts = [filename, "--config=test_pypdfocr_config.yaml", "-m"]
+            opts = [filename, "--preprocess", "--config=test_pypdfocr_config.yaml", "-m"]
             self.p.go(opts)
 
             out_filename = filename.replace(".pdf", "_ocr.pdf")

-Original file line number
+Diff line change
@@ @@ -1,3 +1,5 @@ @@
 *.pyc
 .*
 *~
 +*.hocr
 +*.jpg