Backport hybrid search filter and CI fixes to release/1.4.0

.github/scripts/check-pytest-summary.sh

@@ -0,0 +1,27 @@
+#!/usr/bin/env bash
+
+# Importing pylibseekdb can mask pytest's non-zero process status, so CI must
+# verify the final pytest summary instead. Fail closed if the summary reports
+# failures/errors or does not confirm that at least one test passed.
+set -euo pipefail
+
+pytest_log="${1:-pytest.log}"
+
+if [[ ! -s "${pytest_log}" ]]; then
+  echo "pytest log is missing or empty: ${pytest_log}" >&2
+  exit 1
+fi
+
+summary="$(awk 'NF { last = $0 } END { print last }' "${pytest_log}")"
+failure_pattern='(^|[[:space:]])[1-9][0-9]*[[:space:]]+(failed|errors?)([,=[:space:]]|$)'
+success_pattern='(^|[[:space:]])[1-9][0-9]*[[:space:]]+passed([,=[:space:]]|$)'
+
+if printf '%s\n' "${summary}" | grep -Eiq "${failure_pattern}"; then
+  echo "pytest reported failures: ${summary}" >&2
+  exit 1
+fi
+
+if ! printf '%s\n' "${summary}" | grep -Eiq "${success_pattern}"; then
+  echo "could not verify a successful pytest summary: ${summary}" >&2
+  exit 1
+fi

.github/workflows/ci.yml

@@ -85,7 +85,7 @@ jobs:
           fi
           set -o pipefail
           uv run pytest tests/unit_tests/ -v --log-cli-level=${log_level} | tee pytest.log
-          tail -n 1 pytest.log | grep '=======' | grep 'passed' |grep -q 'failed' && exit 1 || exit 0
+          bash .github/scripts/check-pytest-summary.sh pytest.log
 
       - name: Run integration tests without any test mode
         run: |
@@ -97,7 +97,7 @@ jobs:
           set -o pipefail
           uv run pytest tests/integration_tests/ -v --log-cli-level=${log_level} \
             -k "not server and not embedded and not oceanbase" | tee pytest.log
-          tail -n 1 pytest.log | grep '=======' | grep 'passed' |grep -q 'failed' && exit 1 || exit 0
+          bash .github/scripts/check-pytest-summary.sh pytest.log
 
   integration-test:
     runs-on: ubuntu-latest
@@ -161,4 +161,4 @@ jobs:
           set -o pipefail
           uv run pytest tests/integration_tests/ -v --log-cli-level=${log_level} \
             -k "${{ matrix.test_mode }}" | tee pytest.log
-          tail -n 1 pytest.log | grep '=======' | grep 'passed' | grep -q 'failed' && exit 1 || exit 0
+          bash .github/scripts/check-pytest-summary.sh pytest.log

src/pyseekdb/client/client_base.py

@@ -77,6 +77,14 @@
 
 _COLLECTION_NAME_PATTERN = re.compile(r"^[A-Za-z0-9_]+$")
 
+# DBMS_HYBRID_SEARCH.GET_SQL may quote a JSON_EXTRACT expression as though it
+# were a column identifier.  Require JSON_EXTRACT to start the quoted content
+# so the match cannot span from one ordinary quoted identifier to another.
+_QUOTED_JSON_EXTRACT_EXPRESSION_PATTERN = re.compile(
+    r"`(?P<expression>\s*\(*\s*JSON_EXTRACT\s*\([^`]*\)\s*\)*\s*)`",
+    re.IGNORECASE,
+)
+
 # Maximum allowed length for user-facing collection names.
 _MAX_COLLECTION_NAME_LENGTH = 512
 
@@ -92,6 +100,11 @@
 from .types import _NOT_PROVIDED, _NotProvided  # noqa: E402, F401
 
 
+def _unquote_json_extract_expressions(query_sql: str) -> str:
+    """Unquote JSON_EXTRACT expressions without touching adjacent SQL identifiers."""
+    return _QUOTED_JSON_EXTRACT_EXPRESSION_PATTERN.sub(r"\g<expression>", query_sql)
+
+
 def is_lakebase_version_string(version_str: str) -> bool:
     """Return whether a ``SELECT version()`` string identifies a LakeBase cluster."""
     return _LAKEBASE_VERSION_MARKER in version_str.lower()
@@ -4507,16 +4520,10 @@ def _collection_hybrid_search(
             # Remove any surrounding quotes if present
             query_sql = query_sql.strip().strip("'\"")
 
-        # OB's GET_SQL wraps field names in backticks, which turns
-        # `JSON_EXTRACT(metadata, '$.key')` (with or without outer
-        # parentheses) into a literal column name instead of a
-        # function call. Strip the backticks so OB evaluates the
-        # expression as a function call.
-        query_sql = re.sub(
-            r"`([^`]*JSON_EXTRACT[^`]*)`",
-            r"\1",
-            query_sql,
-        )
+        # OB's GET_SQL can wrap JSON_EXTRACT expressions in backticks, which
+        # turns them into literal column names. Unquote only those complete
+        # expressions; ordinary identifiers around them must remain quoted.
+        query_sql = _unquote_json_extract_expressions(query_sql)
 
         # Add query hint to the generated SQL
         hint_sql = _query_hint_to_sql(query_hint, table_name=table_name)

tests/unit_tests/test_ci_pytest_summary.py

@@ -0,0 +1,44 @@
+"""Tests for the CI fallback that validates pytest's final summary line."""
+
+import subprocess
+from pathlib import Path
+
+import pytest
+
+CHECK_SCRIPT = Path(__file__).parents[2] / ".github" / "scripts" / "check-pytest-summary.sh"
+
+
+@pytest.mark.parametrize(
+    ("summary", "expected_returncode"),
+    [
+        ("================ 92 passed, 1 xfailed in 1.23s ================", 0),
+        ("================ 92 passed in 1.23s ================\n\n \t", 0),
+        ("= 12 failed, 92 passed, 311 skipped, 1 xpassed in 135.47s =", 1),
+        ("================ 1 error, 2 passed in 0.42s ================", 1),
+        ("================ 5 skipped in 0.10s ================", 1),
+        ("test session interrupted", 1),
+    ],
+)
+def test_check_pytest_summary(tmp_path: Path, summary: str, expected_returncode: int) -> None:
+    pytest_log = tmp_path / "pytest.log"
+    pytest_log.write_text(f"test output\n{summary}\n", encoding="utf-8")
+
+    result = subprocess.run(  # noqa: S603 - execute the repository's fixed CI script
+        ["/bin/bash", str(CHECK_SCRIPT), str(pytest_log)],
+        check=False,
+        capture_output=True,
+        text=True,
+    )
+
+    assert result.returncode == expected_returncode, result.stderr
+
+
+def test_check_pytest_summary_rejects_missing_log(tmp_path: Path) -> None:
+    result = subprocess.run(  # noqa: S603 - execute the repository's fixed CI script
+        ["/bin/bash", str(CHECK_SCRIPT), str(tmp_path / "missing.log")],
+        check=False,
+        capture_output=True,
+        text=True,
+    )
+
+    assert result.returncode == 1

tests/unit_tests/test_hybrid_search_sql_rewrite.py

@@ -0,0 +1,41 @@
+"""Regression tests for SQL returned by DBMS_HYBRID_SEARCH.GET_SQL."""
+
+import pytest
+
+from pyseekdb.client.client_base import _unquote_json_extract_expressions
+
+
+@pytest.mark.parametrize(
+    "query_sql",
+    [
+        (
+            "SELECT MATCH(`document`) AGAINST ('machine') "
+            "WHERE (JSON_EXTRACT(metadata, '$.category')) = 'AI' "
+            "ORDER BY `_score` DESC, `__pk_increment`"
+        ),
+        ("SELECT * FROM `c$v1$test` WHERE JSON_EXTRACT(metadata, '$.score') >= 90 ORDER BY `_distance`"),
+    ],
+)
+def test_unquoted_json_extract_does_not_remove_adjacent_identifier_quotes(query_sql: str) -> None:
+    assert _unquote_json_extract_expressions(query_sql) == query_sql
+
+
+@pytest.mark.parametrize(
+    ("query_sql", "expected"),
+    [
+        (
+            "WHERE `(JSON_EXTRACT(metadata, '$.category'))` = 'AI' ORDER BY `_score`",
+            "WHERE (JSON_EXTRACT(metadata, '$.category')) = 'AI' ORDER BY `_score`",
+        ),
+        (
+            "WHERE (`JSON_EXTRACT(metadata, '$.score')`) >= 90 ORDER BY `_distance`",
+            "WHERE (JSON_EXTRACT(metadata, '$.score')) >= 90 ORDER BY `_distance`",
+        ),
+        (
+            "WHERE `json_extract(metadata, '$.tag')` = 'ml'",
+            "WHERE json_extract(metadata, '$.tag') = 'ml'",
+        ),
+    ],
+)
+def test_quoted_json_extract_expression_is_unquoted(query_sql: str, expected: str) -> None:
+    assert _unquote_json_extract_expressions(query_sql) == expected