Fix the image build, and stop the fallback refusing Clusters

Dockerfile

@@ -25,18 +25,22 @@ RUN pip install --no-cache-dir --no-binary numpy \
 # copies the schema's src into src/vfb and puts the checkout on PYTHONPATH.
 # Do exactly that, so what runs here is what runs in the bulk job.
 #
-# Pin by passing --build-arg INDEXER_REF=<sha>; the default tracks master the
-# way the Jenkins job does. Both resolved SHAs are baked into the image so
-# /status can report which schema built a given document.
-ARG INDEXER_REF=master
-ARG JSON_SCHEMA_REF=master
+# Pin by passing --build-arg INDEXER_REF=<sha>; the default, `default`, keeps
+# whatever branch the clone checks out, the way the Jenkins job does. Do not
+# name a branch here: the two repositories disagree (the indexer's default is
+# `main`, the schema's is `master`), and hardcoding either breaks the other --
+# and would break again silently if a default branch were ever renamed. Both
+# resolved SHAs are baked into the image so a rebuilt document can say which
+# schema produced it.
+ARG INDEXER_REF=default
+ARG JSON_SCHEMA_REF=default
 RUN apt-get update && \
     apt-get install -y --no-install-recommends git && \
     rm -rf /var/lib/apt/lists/*
 RUN git clone --quiet https://github.com/VirtualFlyBrain/VFB_json_schema_indexer.git /opt/vfb_indexer && \
-    git -C /opt/vfb_indexer checkout --quiet "${INDEXER_REF}" && \
+    { [ "${INDEXER_REF}" = default ] || git -C /opt/vfb_indexer checkout --quiet "${INDEXER_REF}"; } && \
     git clone --quiet https://github.com/VirtualFlyBrain/VFB_json_schema.git /opt/vfb_json_schema && \
-    git -C /opt/vfb_json_schema checkout --quiet "${JSON_SCHEMA_REF}" && \
+    { [ "${JSON_SCHEMA_REF}" = default ] || git -C /opt/vfb_json_schema checkout --quiet "${JSON_SCHEMA_REF}"; } && \
     mkdir -p /opt/vfb_indexer/src/vfb && \
     cp -r /opt/vfb_json_schema/src/* /opt/vfb_indexer/src/vfb/ && \
     printf 'VFB_INDEXER_SHA=%s\nVFB_JSON_SCHEMA_SHA=%s\n' \

src/test/test_term_info_fallback.py

@@ -52,15 +52,16 @@ def test_template_wins_over_individual():
 
 
 # ---------------------------------------------------------------------------
-# Exclusions -- ids the bulk indexer never writes a document for
+# Exclusions -- anatomical-branch only
 # ---------------------------------------------------------------------------
 
-@pytest.mark.parametrize("short_form", [
-    "VFBc_00000001", "FBlc0006125", "SAMN12345678", "VFB_internal_thing",
-])
-def test_excluded_ids_are_not_built(short_form):
-    payload, doc = tif.build_term_info(short_form)
-    assert payload is None and doc is None
+def test_exclusions_are_scoped_to_the_anatomical_branch():
+    """FBlc ids are Clusters, which have their own indexer. The anatomical
+    indexer's parameter query excludes them; the term_info index as a whole
+    does not. Applying the list globally would refuse every Cluster."""
+    assert "FBlc" in tif.ANATOMICAL_EXCLUDED_ID_PREFIXES
+    assert tif.choose_indexer(["Individual", "Cluster"]) == "cluster"
+    assert tif.choose_indexer(["Individual", "Anatomy"]) == "anatomical_ind"
 
 
 # ---------------------------------------------------------------------------

src/vfbquery/term_info_fallback.py

@@ -244,9 +244,12 @@ def fallback_unavailable_reason():
 # Type dispatch
 # --------------------------------------------------------------------------
 
-#: Ids the indexer's parameter queries exclude, so we exclude them too rather
-#: than writing a document the bulk job would never have written.
-EXCLUDED_ID_PREFIXES = ("VFBc_", "FBlc", "SAMN", "VFB_internal")
+#: Id prefixes the *anatomical individual* indexer's parameter query excludes.
+#: They are not excluded from the term_info index as a whole: FBlc ids are
+#: Clusters, which ClusterTermInfoQueryIndexer owns, so this list only applies
+#: on the anatomical branch. Applying it globally -- as the first version of
+#: this module did -- would have refused to rebuild any Cluster document.
+ANATOMICAL_EXCLUDED_ID_PREFIXES = ("VFBc_", "FBlc", "SAMN", "VFB_internal")
 
 
 def choose_indexer(labels):
@@ -308,10 +311,6 @@ def build_term_info(short_form, neo=None):
     if not fallback_available():
         _warn_unavailable_once()
         return None, None
-    if short_form.startswith(EXCLUDED_ID_PREFIXES):
-        print("term_info fallback: %s is excluded from the term_info index"
-              % short_form)
-        return None, None
 
     from .vfb_queries import vc, get_dict_cursor
     neo = neo or vc.nc
@@ -325,6 +324,13 @@ def build_term_info(short_form, neo=None):
         print("term_info fallback: no term_info indexer covers labels %s (%s)"
               % (sorted(labels), short_form))
         return None, None
+    if (key == "anatomical_ind"
+            and short_form.startswith(ANATOMICAL_EXCLUDED_ID_PREFIXES)):
+        # The anatomical indexer's parameter query skips these, so the bulk
+        # job would never have written a document for one either.
+        print("term_info fallback: %s is excluded from the anatomical "
+              "individual index" % short_form)
+        return None, None
 
     indexer = _INDEXERS[key]()
     rows = get_dict_cursor()(neo.commit_list([