Handle whether JPEG compression exists in the TIFF image

Author: catileptic

ingestors/media/tiff.py

@@ -1,12 +1,10 @@
-import logging
 from followthemoney import model
 
 from ingestors.ingestor import Ingestor
 from ingestors.support.pdf import PDFSupport
 from ingestors.support.shell import ShellSupport
 from ingestors.support.temp import TempFileSupport
-
-log = logging.getLogger(__name__)
+from ingestors.exc import ProcessingException
 
 
 class TIFFIngestor(Ingestor, PDFSupport, TempFileSupport, ShellSupport):
@@ -22,9 +20,24 @@ class TIFFIngestor(Ingestor, PDFSupport, TempFileSupport, ShellSupport):
     def ingest(self, file_path, entity):
         entity.schema = model.get("Pages")
         pdf_path = self.make_work_file("tiff.pdf")
-        self.exec_command(
-            "tiff2pdf", file_path, "-n", "-j", "-x", "300", "-y", "300", "-o", pdf_path
-        )
+        try:
+            self.exec_command(
+                "tiff2pdf",
+                file_path,
+                "-n",
+                "-j",
+                "-x",
+                "300",
+                "-y",
+                "300",
+                "-o",
+                pdf_path,
+            )
+        except ProcessingException:
+            self.exec_command(
+                "tiff2pdf", file_path, "-x", "300", "-y", "300", "-o", pdf_path
+            )
+
         self.assert_outfile(pdf_path)
 
         self.pdf_alternative_extract(entity, pdf_path, self.manager)

tests/test_image.py

@@ -63,6 +63,13 @@ def test_tesseract_ocr_regression(self):
             )
             image_entity = emitted_image_entities.pop()
 
+            # Is the mimeType correct?
+            self.assertEqual(
+                image_entity.first("mimeType"),
+                test_data[test_image_type]["mime_type"],
+                f"Test failed for {test_data[test_image_type]['file']}",
+            )
+
             # Is the processing status of the entity == SUCCESS?
             self.assertEqual(
                 image_entity.first("processingStatus"),

tests/test_tiff.py

@@ -5,14 +5,45 @@ class TIFFIngestorTest(TestCase):
     def test_match(self):
         fixture_path, entity = self.fixture("multipage_tiff_example.tif")
         self.manager.ingest(fixture_path, entity)
-        self.assertEqual(entity.first("mimeType"), "image/tiff")
-        self.assertEqual(entity.first("processingStatus"), self.manager.STATUS_SUCCESS)
+
+        emitted_image_entities = [
+            x
+            for x in self.get_emitted()
+            if "mimeType" in x.properties and "image" in x.first("mimeType")
+        ]
+
+        # Have entities been emitted with a mime type that contains "image"?
+        self.assertTrue(
+            len(emitted_image_entities) != 0,
+            f"Test failed for multipage_tiff_example.tif",
+        )
+        image_entity = emitted_image_entities.pop()
+
+        self.assertEqual(image_entity.first("mimeType"), "image/tiff")
+        self.assertEqual(
+            image_entity.first("processingStatus"), self.manager.STATUS_SUCCESS
+        )
         entities = self.get_emitted()
         self.assertEqual(len(entities), 11)
 
     def test_ingest_tiff_format(self):
         fixture_path, entity = self.fixture("hello_world_tiff.tif")
         self.manager.ingest(fixture_path, entity)
-        self.assertEqual(entity.first("processingStatus"), self.manager.STATUS_SUCCESS)
-        entity = self.get_emitted_by_id(entity.id)
-        self.assertEqual(entity.first("indexText"), "HELLO WORLD")
+
+        emitted_image_entities = [
+            x
+            for x in self.get_emitted()
+            if "mimeType" in x.properties and "image" in x.first("mimeType")
+        ]
+
+        # Have entities been emitted with a mime type that contains "image"?
+        self.assertTrue(
+            len(emitted_image_entities) != 0,
+            f"Test failed for multipage_tiff_example.tif",
+        )
+        image_entity = emitted_image_entities.pop()
+
+        self.assertEqual(
+            image_entity.first("processingStatus"), self.manager.STATUS_SUCCESS
+        )
+        self.assertEqual(image_entity.first("indexText"), "HELLO WORLD")