Simplify conditional logic for gale local ocr

ppa/archive/gale.py

@@ -238,20 +238,19 @@ def get_item_pages(self, item_id, gale_record=None):
             # OCR. Set a tag to indicate when local OCR is present.
             tags = []
             ocr_text = None
-            if local_ocr_text:
+            if local_ocr_text and page_number in local_ocr_text:
                 ocr_text = local_ocr_text.get(page_number)
                 # if we have content for this page, set tag to indicate local ocr.
                 # If page is present but content is the empty string
                 # (i.e., blank page), still set tag since it was the local OCR
                 # that determined the page was blank
-                if ocr_text is not None:
-                    tags = ["local_ocr"]
-                # If None, page is not present in the data;
-                else:
-                    logger.warning(f"No local OCR for {item_id} {page_number}")
-                    # try getting the ocr from the gale api result
-                    # (may be empty, since some pages have no text)
-                    ocr_text = page.get("ocrText")
+                tags = ["local_ocr"]
+            # If page is not present in the data, use Gale OCR as fallback
+            else:
+                logger.warning(f"No local OCR for {item_id} {page_number}")
+                # try getting the ocr from the gale api result
+                # (may be empty, since some pages have no text)
+                ocr_text = page.get("ocrText")
 
             info = {
                 "page_id": page_number,

ppa/archive/tests/test_gale.py

@@ -304,7 +304,7 @@ def test_get_item_pages(self, mock_get_item, mock_get_local_ocr, mockrequests):
             f"http://example.com/img/{i+1}" for i in range(3)
         ]
 
-        # page present but content empty should still return a local ocr tag
+        # page present but content empty should set local ocr tag
         mock_get_local_ocr.return_value = {"0001": "", "0002": "", "0003": ""}
         page_data = list(gale_api.get_item_pages(item_id))
         assert [p["tags"] for p in page_data] == [
@@ -313,6 +313,15 @@ def test_get_item_pages(self, mock_get_item, mock_get_local_ocr, mockrequests):
             ["local_ocr"],
         ]
 
+        # no local ocr file - should use gale ocr
+        mock_get_local_ocr.return_value = None
+        page_data = list(gale_api.get_item_pages(item_id))
+        assert [p["content"] for p in page_data] == [
+            None,
+            "more test content",
+            "ignored text",
+        ]
+
         # skip api call if record is provided
         mock_get_item.reset_mock()
         mock_get_local_ocr.reset_mock()