Fix: Subgraph stitching query crashes BQ emulator (#5847)

UPGRADING.rst

@@ -20,6 +20,22 @@ reverted. This is all fairly informal and loosely defined. Hopefully we won't
 have too many entries in this file.
 
 
+ #5847 Subgraph stitching query crashes BQ emulator
+===================================================
+
+Operator
+~~~~~~~~
+
+Manually perform a two-phase deployment of the ``shared`` component of every
+main deployment. In a lower deployment, perform the first phase using the
+``apply_keep_unused`` Makefile target just before pushing the PR branch to the
+GitLab instance in that deployment. In a stable deployment (``prod``), perform
+the first phase before pushing the merge commit to the GitLab instance in that
+deployment. In lower and stable deployments, perform the second phase using the
+``apply`` Makefile target after the merge commit was successfully built on the
+GitLab instance in that deployment.
+
+
 #5046 Replace tinyquery with bigquery-emulator
 ==============================================
 

src/azul/__init__.py

@@ -1483,7 +1483,7 @@ def docker_images(self) -> dict[str, str]:
             pycharm=f'docker.io/ucscgi/azul-pycharm:{self.docker_pycharm_version}',
             elasticsearch=f'docker.io/ucscgi/azul-elasticsearch'
                           f':{self.docker_elasticsearch_version}',
-            bigquery_emulator='ghcr.io/goccy/bigquery-emulator:0.4.4',
+            bigquery_emulator='ghcr.io/hannes-ucsc/bigquery-emulator:azul',
             # Updating any of the four images below additionally requires
             # redeploying the `gitlab` TF component.
             clamav='docker.io/clamav/clamav:1.2.1-24',

test/indexer/test_tdr.py

@@ -6,6 +6,7 @@
     BytesIO,
 )
 import json
+import logging
 from operator import (
     attrgetter,
 )
@@ -75,6 +76,7 @@
 from azul.plugins.repository.tdr_hca import (
     TDRBundleFQID,
     TDRHCABundle,
+    log as plugin_log,
 )
 from azul.terra import (
     TDRClient,
@@ -308,28 +310,21 @@ def _test_invalid_links(self, bundle: TDRHCABundle):
         with self.assertRaises(RequirementError):
             self._test_fetch_bundle(bundle, load_tables=False)
 
-    @patch('azul.plugins.repository.tdr_hca.Plugin._find_upstream_bundles')
-    def test_subgraph_stitching(self, _mock_find_upstream_bundles):
+    def test_subgraph_stitching(self):
         downstream_uuid = '4426adc5-b3c5-5aab-ab86-51d8ce44dfbe'
         upstream_uuids = [
             'b0c2c714-45ee-4759-a32b-8ccbbcf911d4',
             'bd4939c1-a078-43bd-8477-99ae59ceb555',
         ]
-        # FIXME: Fix the crash in bigquery-emulator and remove the mock
-        #        https://github.com/DataBiosphere/azul/issues/5847
-        _mock_find_upstream_bundles.side_effect = [
-            {SourcedBundleFQID(source=self.source,
-                               uuid=uuid,
-                               version='2020-08-10T21:24:26.174274Z')}
-            for uuid in upstream_uuids
-        ]
         bundle = self._load_canned_bundle(SourcedBundleFQID(source=self.source,
                                                             uuid=downstream_uuid,
                                                             version='2020-08-10T21:24:26.174274Z'))
         assert any(e['is_stitched'] for e in bundle.manifest)
-        self._test_fetch_bundle(bundle, load_tables=True)
-        self.assertEqual(_mock_find_upstream_bundles.call_count,
-                         len(upstream_uuids))
+        with self.assertLogs(plugin_log, level=logging.DEBUG) as cm:
+            self._test_fetch_bundle(bundle, load_tables=True)
+        record = one(r for r in cm.records if 'Stitched 2 bundle(s): ' in r.message)
+        for upstream_uuid in upstream_uuids:
+            self.assertIn("uuid='" + upstream_uuid, record.message)
 
     def _test_fetch_bundle(self,
                            test_bundle: TDRHCABundle,