Fix TIFF processing. Add tests to prevent regression in OCR for gif, jpg, jp2, tiff, webp

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):
@@ -22,8 +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, "-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

@@ -14,10 +14,80 @@ def test_ingest_on_svg(self):
         self.assertIn("TEST", entity.first("bodyText"))
         self.assertEqual(entity.first("processingStatus"), self.manager.STATUS_SUCCESS)
 
-    def test_ingest_on_jpeg(self):
-        fixture_path, entity = self.fixture("jpegtest.jpg")
-        self.manager.ingest(fixture_path, entity)
-        self.assertIn("Debian", entity.first("bodyText"))
-        self.assertEqual(entity.first("mimeType"), "image/jpeg")
+    def test_tesseract_ocr_regression(self):
+        """This test is meant to catch a regression in the OCR behaviour
+        described in this PR: https://github.com/alephdata/ingest-file/pull/585"""
 
-        self.assertEqual(entity.first("processingStatus"), self.manager.STATUS_SUCCESS)
+        test_data = {
+            "jpeg": {
+                "file": "regression_jpg.jpg",
+                "content": "Debian -- Packages",
+                "mime_type": "image/jpeg",
+            },
+            "gif": {
+                "file": "regression_gif.gif",
+                "content": "This is text inside a GIF image",
+                "mime_type": "image/gif",
+            },
+            "tiff": {
+                "file": "regression_tiff.tiff",
+                "content": "Debian -- Packages",
+                "mime_type": "image/tiff",
+            },
+            "webp": {
+                "file": "regression_webp.webp",
+                "content": "Debian -- Packages",
+                "mime_type": "image/webp",
+            },
+            "openjpeg": {
+                "file": "regression_openjpeg.jp2",
+                "content": "Debian -- Packages",
+                "mime_type": "image/jp2",
+            },
+        }
+
+        for test_image_type in test_data:
+            fixture_path, entity = self.fixture(test_data[test_image_type]["file"])
+            self.manager.ingest(fixture_path, entity)
+
+            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 {test_data[test_image_type]['file']}",
+            )
+            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"),
+                self.manager.STATUS_SUCCESS,
+                f"Test failed for {test_data[test_image_type]['file']}",
+            )
+
+            # Does either the bodyText prop or the indexText prop contain
+            # the text resulted from OCR?
+            try:
+                self.assertIn(
+                    test_data[test_image_type]["content"],
+                    image_entity.first("bodyText"),
+                    f"Test failed for {test_data[test_image_type]['file']}",
+                )
+            except TypeError:
+                self.assertIn(
+                    test_data[test_image_type]["content"],
+                    image_entity.first("indexText"),
+                    f"Test failed for {test_data[test_image_type]['file']}",
+                )

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,
+            "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,
+            "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")