Handle TIFFs in PDFs by converting to PNG (#419)

Fixes #418 and thus alephdata/aleph#2810

Author: stchris

ingestors/support/pdf.py

@@ -178,6 +178,9 @@ def _extract_images(
             if isinstance(image, Image.Image):
                 image.save(filepath_prefix + ".png", "PNG")
             else:
+                pil_image = image.as_pil_image()
+                if pil_image.format == "TIFF":
+                    pil_image.save(filepath_prefix + ".png", "PNG")
                 image.extract_to(fileprefix=filepath_prefix)
 
     def pdf_extract_page(

tests/fixtures/greek.pdf

@@ -195,3 +195,14 @@ def test_ingest_index_text(self):
         assert (
             "Erklärung verkündeten Rechte und Freiheiten zum Ziel hat." in pages_index
         )
+
+    def test_ingest_pdf_ocr_greek(self):
+        fixture_path, entity = self.fixture("greek.pdf")
+        self.manager.ingest(fixture_path, entity)
+
+        emitted = self.get_emitted()
+        assert len(emitted) == 3
+
+        page = emitted[1]
+        assert page.schema.name == "Page"
+        assert "IRIDECEA HOLDINGS LIMITED" in "\n".join(page.get("bodyText"))