Create missing tables in parents caches before merge

arkindex_worker/cache.py

@@ -188,13 +188,19 @@ def merge_parents_cache(parent_ids, current_database, data_dir="/data", chunk=No
 
     # Merge each table into the local database
     for idx, path in enumerate(paths):
+        with SqliteDatabase(path) as source:
+            with source.bind_ctx(MODELS):
+                source.create_tables(MODELS)
+
         logger.info(f"Merging parent db {path} into {current_database}")
         statements = [
             "PRAGMA page_size=80000;",
             "PRAGMA synchronous=OFF;",
             f"ATTACH DATABASE '{path}' AS source_{idx};",
+            f"REPLACE INTO images SELECT * FROM source_{idx}.images;",
             f"REPLACE INTO elements SELECT * FROM source_{idx}.elements;",
             f"REPLACE INTO transcriptions SELECT * FROM source_{idx}.transcriptions;",
+            f"REPLACE INTO classifications SELECT * FROM source_{idx}.classifications;",
         ]
 
         for statement in statements:

tests/test_merge.py

@@ -5,7 +5,9 @@ import pytest
 
 from arkindex_worker.cache import (
     MODELS,
+    CachedClassification,
     CachedElement,
+    CachedImage,
     CachedTranscription,
     merge_parents_cache,
 )
@@ -76,8 +78,10 @@ def test_merge_databases(
 
     # We always start with an empty database
     with mock_databases["target"]["db"].bind_ctx(MODELS):
+        assert CachedImage.select().count() == 0
         assert CachedElement.select().count() == 0
         assert CachedTranscription.select().count() == 0
+        assert CachedClassification.select().count() == 0
 
     # Merge all requested parents databases into our target
     merge_parents_cache(
@@ -88,8 +92,10 @@ def test_merge_databases(
 
     # The target now should have the expected elements and transcriptions
     with mock_databases["target"]["db"].bind_ctx(MODELS):
+        assert CachedImage.select().count() == 0
         assert CachedElement.select().count() == len(expected_elements)
         assert CachedTranscription.select().count() == len(expected_transcriptions)
+        assert CachedClassification.select().count() == 0
         assert [
             e.id for e in CachedElement.select().order_by("id")
         ] == expected_elements
@@ -124,8 +130,10 @@ def test_merge_chunk(mock_databases, tmpdir, monkeypatch):
 
     # The target should now have 3 elements and 0 transcription
     with mock_databases["target"]["db"].bind_ctx(MODELS):
+        assert CachedImage.select().count() == 0
         assert CachedElement.select().count() == 3
         assert CachedTranscription.select().count() == 0
+        assert CachedClassification.select().count() == 0
         assert [e.id for e in CachedElement.select().order_by("id")] == [
             UUID("42424242-4242-4242-4242-424242424242"),
             UUID("12341234-1234-1234-1234-123412341234"),
@@ -147,8 +155,10 @@ def test_merge_from_worker(
     )
 
     # At first we have no data in our main database
+    assert CachedImage.select().count() == 0
     assert CachedElement.select().count() == 0
     assert CachedTranscription.select().count() == 0
+    assert CachedClassification.select().count() == 0
 
     # Configure worker with a specific data directory
     monkeypatch.setenv("PONOS_DATA", str(tmpdir))
@@ -157,8 +167,10 @@ def test_merge_from_worker(
     mock_base_worker_with_cache.configure()
 
     # Then we have 2 elements and a transcription
+    assert CachedImage.select().count() == 0
     assert CachedElement.select().count() == 3
     assert CachedTranscription.select().count() == 1
+    assert CachedClassification.select().count() == 0
     assert [e.id for e in CachedElement.select().order_by("id")] == [
         UUID("12341234-1234-1234-1234-123412341234"),
         UUID("56785678-5678-5678-5678-567856785678"),