Clean up marshmallow warnings + count pytest tests in perf report

Three layered cleanups on the now-green perf-test workflow:

1. vfb_queries.py: migrate marshmallow 3→4 deprecations in our own
   code so RemovedInMarshmallow4Warning / ChangedInMarshmallow4Warning
   stop firing for lines 138, 308, 309.
     - `missing=` → `load_default=` on four schema fields at
       preview_columns, preview_results, output_format, count (136-139).
     - `fields.Field()` → `fields.Raw()` on Publications and Synonyms
       (308-309).
   No behavioural change; both renames are marshmallow's recommended
   migration path.

2. pyproject.toml [tool.pytest.ini_options]: filterwarnings to silence
   the marshmallow library-internal deprecation warnings
   (marshmallow/fields.py:582,776,986,1218) that bubble up from
   transitive deps (vfb_connect, dataclasses-json) using legacy
   patterns. Our own code's warnings remain visible because they live
   in our source tree, not marshmallow's.

3. performance-test.yml step #9: report counts now sum both unittest-
   and pytest-format test lines. Was previously showing "Total
   Tests: 1" — only the legacy term_info_performance unittest line —
   because grep "^test_" doesn't match pytest's
   `src/test/...::TestX::test_y PASSED`. Parser now sums:
     unittest: ^test_  AND  ^(FAIL|ERROR):
     pytest:   " (PASSED|FAILED|ERROR)( |$)"
   Totals reflect every test the workflow actually ran across all
   three test steps.

Author: Robbie1977

.github/workflows/performance-test.yml

@@ -264,20 +264,28 @@ jobs:
             fi
             echo "" >> performance.md
             
-            # Count successes and failures. `grep -c` already prints `0` and
-            # exits 1 on no-match, so combining it with `|| echo "0"` gave
-            # variables containing literal `"0\n0"` and broke the arithmetic
-            # below ("syntax error in expression (error token is "0")").
-            # Swallow the exit code with `|| true` and let grep's own output
-            # stand; default to 0 only if the variable is empty (file
-            # missing).
-            TOTAL_TESTS=$(grep -c "^test_" performance_test_output.log 2>/dev/null || true)
-            FAILED_TESTS=$(grep -c "FAIL:" performance_test_output.log 2>/dev/null || true)
-            ERROR_TESTS=$(grep -c "ERROR:" performance_test_output.log 2>/dev/null || true)
-            TOTAL_TESTS=${TOTAL_TESTS:-0}
-            FAILED_TESTS=${FAILED_TESTS:-0}
-            ERROR_TESTS=${ERROR_TESTS:-0}
-            PASSED_TESTS=$((TOTAL_TESTS - FAILED_TESTS - ERROR_TESTS))
+            # Count successes and failures. The log mixes both unittest
+            # (`test_xxx ... ok`/`FAIL:`/`ERROR:`) and pytest
+            # (`... PASSED`/`FAILED`/`ERROR` lines) outputs depending on
+            # which step wrote it. Sum both formats so the report shows a
+            # meaningful total instead of "Total: 1" (which is what we got
+            # when only the unittest-format legacy step matched).
+            #
+            # `grep -c` already prints `0` and exits 1 on no-match, so
+            # `|| true` is needed to swallow the exit code (otherwise
+            # set -e aborts the step). Default empty captures to 0.
+            UNITTEST_TESTS=$(grep -cE "^test_" performance_test_output.log 2>/dev/null || true)
+            PYTEST_TESTS=$(grep -cE " (PASSED|FAILED|ERROR)( |$)" performance_test_output.log 2>/dev/null || true)
+            UNITTEST_FAIL=$(grep -cE "^(FAIL|ERROR):" performance_test_output.log 2>/dev/null || true)
+            PYTEST_FAIL=$(grep -cE " (FAILED|ERROR)( |$)" performance_test_output.log 2>/dev/null || true)
+            UNITTEST_TESTS=${UNITTEST_TESTS:-0}
+            PYTEST_TESTS=${PYTEST_TESTS:-0}
+            UNITTEST_FAIL=${UNITTEST_FAIL:-0}
+            PYTEST_FAIL=${PYTEST_FAIL:-0}
+            TOTAL_TESTS=$((UNITTEST_TESTS + PYTEST_TESTS))
+            FAILED_TESTS=$((UNITTEST_FAIL + PYTEST_FAIL))
+            ERROR_TESTS=0  # ERROR counts already folded into FAILED above
+            PASSED_TESTS=$((TOTAL_TESTS - FAILED_TESTS))
             
             echo "### Test Statistics" >> performance.md
             echo "" >> performance.md

pyproject.toml

@@ -12,4 +12,16 @@ build-backend = "setuptools.build_meta"
 # blocks in performance.md — pure noise.
 markers = [
     "integration: integration tests that exercise the live VFB upstream (Neo4j, SOLR, owlery). Skip via -m 'not integration'.",
+]
+
+# Silence the marshmallow deprecation warnings that bubble up from inside
+# the library itself (marshmallow/fields.py:582,776,986,1218). Our own
+# code calls to `missing=`/`fields.Field()` have been migrated to
+# `load_default=`/`fields.Raw()`, but transitive deps (e.g. via
+# vfb_connect / dataclasses-json) still trigger them. Hiding the
+# library-internal noise — but NOT user-code DeprecationWarnings —
+# keeps performance.md readable.
+filterwarnings = [
+    "ignore::marshmallow.warnings.RemovedInMarshmallow4Warning",
+    "ignore::marshmallow.warnings.ChangedInMarshmallow4Warning",
 ]

src/vfbquery/vfb_queries.py

@@ -133,10 +133,10 @@ class QuerySchema(Schema):
     function = fields.String(required=True)
     takes = fields.Nested(TakesSchema(), required=False, missing={})
     preview = fields.Integer(required=False, missing=0)
-    preview_columns = fields.List(fields.String(), required=False, missing=[])
-    preview_results = fields.List(fields.Dict(), required=False, missing=[])
-    output_format = fields.String(required=False, missing='table')
-    count = fields.Integer(required=False, missing=-1)
+    preview_columns = fields.List(fields.String(), required=False, load_default=[])
+    preview_results = fields.List(fields.Dict(), required=False, load_default=[])
+    output_format = fields.String(required=False, load_default='table')
+    count = fields.Integer(required=False, load_default=-1)
 
 class License:
     def __init__(self, iri, short_form, label, icon, source, source_iri):
@@ -305,8 +305,8 @@ class TermInfoOutputSchema(Schema):
     IsPaintedDomain = fields.Bool(missing=False, required=False)
     Domains = fields.Dict(keys=fields.Integer(), values=fields.Nested(ImageSchema()), required=False, allow_none=True)
     Licenses = fields.Dict(keys=fields.Integer(), values=fields.Nested(LicenseSchema()), required=False, allow_none=True)
-    Publications = fields.List(fields.Dict(keys=fields.String(), values=fields.Field()), required=False)
-    Synonyms = fields.List(fields.Dict(keys=fields.String(), values=fields.Field()), required=False, allow_none=True)
+    Publications = fields.List(fields.Dict(keys=fields.String(), values=fields.Raw()), required=False)
+    Synonyms = fields.List(fields.Dict(keys=fields.String(), values=fields.Raw()), required=False, allow_none=True)
     Technique = fields.List(fields.String(), required=False, allow_none=True)
 
     @post_load