pyworkflow-engine 0.3.2__tar.gz → 0.3.4__tar.gz

{pyworkflow_engine-0.3.2 → pyworkflow_engine-0.3.4}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: pyworkflow-engine
-Version: 0.3.2
+Version: 0.3.4
 Summary: A Python implementation of durable, event-sourced workflows inspired by Vercel Workflow
 Author: PyWorkflow Contributors
 License: MIT
@@ -26,11 +26,11 @@ Requires-Dist: celery<6.0.0,>=5.3.0
 Requires-Dist: cloudpickle>=3.0.0
 Requires-Dist: pydantic<3.0.0,>=2.0.0
 Requires-Dist: loguru>=0.7.0
-Requires-Dist: click>=8.3.3
+Requires-Dist: click>=8.4.1
 Requires-Dist: inquirerpy>=0.3.4; python_version < "4.0"
 Requires-Dist: httpx>=0.25.0
 Requires-Dist: python-dateutil>=2.8.0
-Requires-Dist: filelock>=3.12.0
+Requires-Dist: filelock>=3.29.3
 Requires-Dist: pyyaml>=6.0.0
 Requires-Dist: croniter>=2.0.0
 Provides-Extra: redis
@@ -42,7 +42,7 @@ Requires-Dist: asyncpg>=0.29.0; extra == "postgres"
 Provides-Extra: aws
 Requires-Dist: aws-durable-execution-sdk-python>=0.1.0; extra == "aws"
 Provides-Extra: dynamodb
-Requires-Dist: aiobotocore>=3.6.0; extra == "dynamodb"
+Requires-Dist: aiobotocore>=3.7.0; extra == "dynamodb"
 Provides-Extra: cassandra
 Requires-Dist: cassandra-driver>=3.29.0; extra == "cassandra"
 Provides-Extra: mysql
@@ -60,10 +60,10 @@ Requires-Dist: pytest-asyncio>=0.21.0; extra == "dev"
 Requires-Dist: pytest-cov>=4.1.0; extra == "dev"
 Requires-Dist: pytest-celery>=0.0.0; extra == "dev"
 Requires-Dist: pytest-mock>=3.12.0; extra == "dev"
-Requires-Dist: moto[dynamodb]>=5.2.0; extra == "dev"
-Requires-Dist: black>=23.0.0; extra == "dev"
-Requires-Dist: ruff>=0.15.12; extra == "dev"
-Requires-Dist: mypy>=1.7.0; extra == "dev"
+Requires-Dist: moto[dynamodb]>=5.2.2; extra == "dev"
+Requires-Dist: black>=26.5.1; extra == "dev"
+Requires-Dist: ruff>=0.15.17; extra == "dev"
+Requires-Dist: mypy>=2.1.0; extra == "dev"
 Requires-Dist: pre-commit>=4.6.0; extra == "dev"
 Requires-Dist: types-redis>=4.6.0; extra == "dev"
 Requires-Dist: types-python-dateutil>=2.8.0; extra == "dev"

{pyworkflow_engine-0.3.2 → pyworkflow_engine-0.3.4}/pyproject.toml

@@ -7,7 +7,7 @@ packages = [{include = "pyworkflow"}]
 
 [project]
 name = "pyworkflow-engine"
-version = "0.3.2"
+version = "0.3.4"
 description = "A Python implementation of durable, event-sourced workflows inspired by Vercel Workflow"
 readme = "README.md"
 requires-python = ">=3.11"
@@ -37,13 +37,13 @@ dependencies = [
     "loguru>=0.7.0",
 
     # CLI dependencies
-    "click>=8.3.3",
+    "click>=8.4.1",
     "inquirerpy>=0.3.4; python_version < '4.0'",
 
     # Utilities
     "httpx>=0.25.0",
     "python-dateutil>=2.8.0",
-    "filelock>=3.12.0",
+    "filelock>=3.29.3",
     "pyyaml>=6.0.0",
     "croniter>=2.0.0",
 ]
@@ -67,7 +67,7 @@ aws = [
     "aws-durable-execution-sdk-python>=0.1.0",
 ]
 dynamodb = [
-    "aiobotocore>=3.6.0",
+    "aiobotocore>=3.7.0",
 ]
 
 cassandra = [
@@ -96,12 +96,12 @@ dev = [
     "pytest-cov>=4.1.0",
     "pytest-celery>=0.0.0",
     "pytest-mock>=3.12.0",
-    "moto[dynamodb]>=5.2.0",  # AWS mocking for DynamoDB tests
+    "moto[dynamodb]>=5.2.2",  # AWS mocking for DynamoDB tests
 
     # Code quality
-    "black>=23.0.0",
-    "ruff>=0.15.12",
-    "mypy>=1.7.0",
+    "black>=26.5.1",
+    "ruff>=0.15.17",
+    "mypy>=2.1.0",
     "pre-commit>=4.6.0",
 
     # Type stubs

{pyworkflow_engine-0.3.2 → pyworkflow_engine-0.3.4}/pyworkflow/__init__.py

@@ -29,7 +29,7 @@ Quick Start:
     >>> run_id = await start(my_workflow, "Alice")
 """
 
-__version__ = "0.3.2"
+__version__ = "0.3.4"
 
 # Configuration
 from pyworkflow.config import (

{pyworkflow_engine-0.3.2 → pyworkflow_engine-0.3.4}/pyworkflow/cli/utils/interactive.py

@@ -76,15 +76,12 @@ def select(
     try:
         # Convert string list to Choice objects
         choice_objects: list[Choice] = []
-        if choices and isinstance(choices[0], str):
-            for c in choices:
-                if isinstance(c, str):
-                    choice_objects.append(Choice(value=c, name=c))
-        else:
-            # Dict format: {"name": "...", "value": "..."}
-            for c in choices:
-                if isinstance(c, dict):
-                    choice_objects.append(Choice(value=c.get("value", c["name"]), name=c["name"]))
+        for c in choices:
+            if isinstance(c, str):
+                choice_objects.append(Choice(value=c, name=c))
+            else:
+                # Dict format: {"name": "...", "value": "..."}
+                choice_objects.append(Choice(value=c.get("value", c["name"]), name=c["name"]))
 
         return inquirer.select(
             message=message,
@@ -222,22 +219,19 @@ def multiselect(
     try:
         # Convert string list to Choice objects
         choice_objects: list[Choice] = []
-        if choices and isinstance(choices[0], str):
-            for c in choices:
-                if isinstance(c, str):
-                    choice_objects.append(Choice(value=c, name=c, enabled=(c in (default or []))))
-        else:
-            # Dict format: {"name": "...", "value": "..."}
-            for c in choices:
-                if isinstance(c, dict):
-                    value = c.get("value", c["name"])
-                    choice_objects.append(
-                        Choice(
-                            value=value,
-                            name=c["name"],
-                            enabled=(value in (default or [])),
-                        )
+        for c in choices:
+            if isinstance(c, str):
+                choice_objects.append(Choice(value=c, name=c, enabled=(c in (default or []))))
+            else:
+                # Dict format: {"name": "...", "value": "..."}
+                value = c.get("value", c["name"])
+                choice_objects.append(
+                    Choice(
+                        value=value,
+                        name=c["name"],
+                        enabled=(value in (default or [])),
                     )
+                )
 
         return inquirer.checkbox(
             message=message,

{pyworkflow_engine-0.3.2 → pyworkflow_engine-0.3.4}/pyworkflow/storage/citus.py

@@ -308,6 +308,14 @@ class CitusStorageBackend(PostgresStorageBackend):
             await conn.execute(
                 "CREATE INDEX IF NOT EXISTS idx_runs_created_at ON workflow_runs(created_at DESC)"
             )
+            # Composite index backing keyset pagination + ORDER BY created_at DESC, run_id DESC
+            # for the unfiltered Runs list (issue #482). The summary column projection
+            # (RUN_SUMMARY_COLUMNS) already drops the heavy payload columns from the scatter-gather;
+            # no INCLUDE covering columns, to avoid duplicating the summary columns into the index.
+            await conn.execute(
+                "CREATE INDEX IF NOT EXISTS idx_runs_created_at_run_id "
+                "ON workflow_runs(created_at DESC, run_id DESC)"
+            )
             # Non-unique index: Citus cannot enforce global uniqueness on non-distribution columns
             await conn.execute(
                 "CREATE INDEX IF NOT EXISTS idx_runs_idempotency_key ON workflow_runs(idempotency_key) WHERE idempotency_key IS NOT NULL"

{pyworkflow_engine-0.3.2 → pyworkflow_engine-0.3.4}/pyworkflow/storage/mysql.py

@@ -31,6 +31,15 @@ from pyworkflow.storage.schemas import (
     WorkflowRun,
 )
 
+# Columns needed by list/summary views. Excludes the heavy payload columns
+# (input_args, input_kwargs, result, metadata) which a list view never renders. See issue #482.
+RUN_SUMMARY_COLUMNS = (
+    "run_id, workflow_name, status, created_at, updated_at, started_at, completed_at, "
+    "error, idempotency_key, max_duration, recovery_attempts, max_recovery_attempts, "
+    "recover_on_worker_loss, parent_run_id, nesting_depth, continued_from_run_id, "
+    "continued_to_run_id"
+)
+
 
 class MySQLMigrationRunner(MigrationRunner):
     """MySQL-specific migration runner."""
@@ -286,6 +295,7 @@ class MySQLStorageBackend(StorageBackend):
                         INDEX idx_runs_status (status),
                         INDEX idx_runs_workflow_name (workflow_name),
                         INDEX idx_runs_created_at (created_at DESC),
+                        INDEX idx_runs_created_at_run_id (created_at DESC, run_id DESC),
                         UNIQUE INDEX idx_runs_idempotency_key (idempotency_key),
                         INDEX idx_runs_parent_run_id (parent_run_id),
                         FOREIGN KEY (parent_run_id) REFERENCES workflow_runs(run_id)
@@ -607,15 +617,23 @@ class MySQLStorageBackend(StorageBackend):
         limit: int = 100,
         cursor: str | None = None,
     ) -> tuple[list[WorkflowRun], str | None]:
-        """List workflow runs with optional filtering and pagination."""
+        """List workflow runs with optional filtering and pagination.
+
+        Note: the returned WorkflowRun objects are *summaries* — the heavy payload columns
+        (input_args, input_kwargs, result, metadata) are NOT fetched and carry placeholder
+        defaults. Use get_run() when the full payload is required. See issue #482.
+        """
         pool = self._ensure_connected()
 
         conditions = []
         params: list[Any] = []
 
         if cursor:
+            # Keyset pagination on (created_at, run_id) — created_at is not unique, so paginating on
+            # it alone would skip/duplicate rows sharing a timestamp. See issue #482.
             conditions.append(
-                "created_at < (SELECT created_at FROM workflow_runs WHERE run_id = %s)"
+                "(created_at, run_id) < "
+                "(SELECT created_at, run_id FROM workflow_runs WHERE run_id = %s)"
             )
             params.append(cursor)
 
@@ -640,9 +658,9 @@ class MySQLStorageBackend(StorageBackend):
         params.append(limit + 1)  # Fetch one extra to determine if there are more
 
         sql = f"""
-            SELECT * FROM workflow_runs
+            SELECT {RUN_SUMMARY_COLUMNS} FROM workflow_runs
             {where_clause}
-            ORDER BY created_at DESC
+            ORDER BY created_at DESC, run_id DESC
             LIMIT %s
         """
 
@@ -654,7 +672,7 @@ class MySQLStorageBackend(StorageBackend):
         if has_more:
             rows = rows[:limit]
 
-        runs = [self._row_to_workflow_run(row) for row in rows]
+        runs = [self._row_to_workflow_run_summary(row) for row in rows]
         next_cursor = runs[-1].run_id if runs and has_more else None
 
         return runs, next_cursor
@@ -1486,6 +1504,36 @@ class MySQLStorageBackend(StorageBackend):
             continued_to_run_id=row["continued_to_run_id"],
         )
 
+    def _row_to_workflow_run_summary(self, row: dict) -> WorkflowRun:
+        """Convert a summary row (RUN_SUMMARY_COLUMNS) to a WorkflowRun.
+
+        Heavy payload columns are not selected by list_runs, so they are filled with their schema
+        defaults. Callers needing the real payload must use get_run(). See issue #482.
+        """
+        return WorkflowRun(
+            run_id=row["run_id"],
+            workflow_name=row["workflow_name"],
+            status=RunStatus(row["status"]),
+            created_at=row["created_at"],
+            updated_at=row["updated_at"],
+            started_at=row["started_at"],
+            completed_at=row["completed_at"],
+            input_args="[]",
+            input_kwargs="{}",
+            result=None,
+            error=row["error"],
+            idempotency_key=row["idempotency_key"],
+            max_duration=row["max_duration"],
+            context={},
+            recovery_attempts=row["recovery_attempts"],
+            max_recovery_attempts=row["max_recovery_attempts"],
+            recover_on_worker_loss=bool(row["recover_on_worker_loss"]),
+            parent_run_id=row["parent_run_id"],
+            nesting_depth=row["nesting_depth"],
+            continued_from_run_id=row["continued_from_run_id"],
+            continued_to_run_id=row["continued_to_run_id"],
+        )
+
     def _row_to_event(self, row: dict) -> Event:
         """Convert database row to Event object."""
         return Event(

{pyworkflow_engine-0.3.2 → pyworkflow_engine-0.3.4}/pyworkflow/storage/postgres.py

@@ -39,6 +39,17 @@ from pyworkflow.storage.schemas import (
     WorkflowRun,
 )
 
+# Columns needed by list/summary views (the Runs list, CLI table). Deliberately excludes the heavy
+# unbounded payload columns (input_args, input_kwargs, result, metadata) which a list view never
+# renders. Selecting only these avoids scatter-gathering wide rows across every Citus shard.
+# See issue #482. Keep `error` — it is shown in the list view and is normally short.
+RUN_SUMMARY_COLUMNS = (
+    "run_id, workflow_name, status, created_at, updated_at, started_at, completed_at, "
+    "error, idempotency_key, max_duration, recovery_attempts, max_recovery_attempts, "
+    "recover_on_worker_loss, parent_run_id, nesting_depth, continued_from_run_id, "
+    "continued_to_run_id"
+)
+
 
 class PostgresMigrationRunner(MigrationRunner):
     """PostgreSQL-specific migration runner."""
@@ -429,6 +440,12 @@ class PostgresStorageBackend(StorageBackend):
             await conn.execute(
                 "CREATE INDEX IF NOT EXISTS idx_runs_created_at ON workflow_runs(created_at DESC)"
             )
+            # Composite index backing keyset pagination + ORDER BY created_at DESC, run_id DESC
+            # (see list_runs / issue #482).
+            await conn.execute(
+                "CREATE INDEX IF NOT EXISTS idx_runs_created_at_run_id "
+                "ON workflow_runs(created_at DESC, run_id DESC)"
+            )
             await conn.execute(
                 "CREATE UNIQUE INDEX IF NOT EXISTS idx_runs_idempotency_key ON workflow_runs(idempotency_key) WHERE idempotency_key IS NOT NULL"
             )
@@ -852,7 +869,12 @@ class PostgresStorageBackend(StorageBackend):
         limit: int = 100,
         cursor: str | None = None,
     ) -> tuple[list[WorkflowRun], str | None]:
-        """List workflow runs with optional filtering and pagination."""
+        """List workflow runs with optional filtering and pagination.
+
+        Note: the returned WorkflowRun objects are *summaries* — the heavy payload columns
+        (input_args, input_kwargs, result, metadata) are NOT fetched and carry placeholder
+        defaults. Use get_run() when the full payload is required. See issue #482.
+        """
         pool = await self._get_pool()
 
         conditions = []
@@ -860,8 +882,12 @@ class PostgresStorageBackend(StorageBackend):
         param_idx = 1
 
         if cursor:
+            # Keyset pagination on (created_at, run_id). created_at is not unique, so paginating on
+            # it alone would skip/duplicate rows that share a timestamp. The composite tuple compare
+            # is a stable total order matching ORDER BY created_at DESC, run_id DESC.
             conditions.append(
-                f"created_at < (SELECT created_at FROM workflow_runs WHERE run_id = ${param_idx})"
+                f"(created_at, run_id) < "
+                f"(SELECT created_at, run_id FROM workflow_runs WHERE run_id = ${param_idx})"
             )
             params.append(cursor)
             param_idx += 1
@@ -893,9 +919,9 @@ class PostgresStorageBackend(StorageBackend):
         params.append(limit + 1)  # Fetch one extra to determine if there are more
 
         sql = f"""
-            SELECT * FROM workflow_runs
+            SELECT {RUN_SUMMARY_COLUMNS} FROM workflow_runs
             {where_clause}
-            ORDER BY created_at DESC
+            ORDER BY created_at DESC, run_id DESC
             LIMIT ${param_idx}
         """
 
@@ -906,7 +932,7 @@ class PostgresStorageBackend(StorageBackend):
         if has_more:
             rows = rows[:limit]
 
-        runs = [self._row_to_workflow_run(row) for row in rows]
+        runs = [self._row_to_workflow_run_summary(row) for row in rows]
         next_cursor = runs[-1].run_id if runs and has_more else None
 
         return runs, next_cursor
@@ -1747,6 +1773,36 @@ class PostgresStorageBackend(StorageBackend):
             continued_to_run_id=row["continued_to_run_id"],
         )
 
+    def _row_to_workflow_run_summary(self, row: asyncpg.Record) -> WorkflowRun:
+        """Convert a summary row (RUN_SUMMARY_COLUMNS) to a WorkflowRun.
+
+        The heavy payload columns are not selected by list_runs, so they are filled with their
+        schema defaults. Callers needing the real payload must use get_run(). See issue #482.
+        """
+        return WorkflowRun(
+            run_id=row["run_id"],
+            workflow_name=row["workflow_name"],
+            status=RunStatus(row["status"]),
+            created_at=row["created_at"],
+            updated_at=row["updated_at"],
+            started_at=row["started_at"],
+            completed_at=row["completed_at"],
+            input_args="[]",
+            input_kwargs="{}",
+            result=None,
+            error=row["error"],
+            idempotency_key=row["idempotency_key"],
+            max_duration=row["max_duration"],
+            context={},
+            recovery_attempts=row["recovery_attempts"],
+            max_recovery_attempts=row["max_recovery_attempts"],
+            recover_on_worker_loss=row["recover_on_worker_loss"],
+            parent_run_id=row["parent_run_id"],
+            nesting_depth=row["nesting_depth"],
+            continued_from_run_id=row["continued_from_run_id"],
+            continued_to_run_id=row["continued_to_run_id"],
+        )
+
     def _row_to_event(self, row: asyncpg.Record) -> Event:
         """Convert database row to Event object."""
         return Event(

{pyworkflow_engine-0.3.2 → pyworkflow_engine-0.3.4}/pyworkflow/storage/sqlite.py

@@ -31,6 +31,16 @@ from pyworkflow.storage.schemas import (
     WorkflowRun,
 )
 
+# Columns needed by list/summary views, in the exact order consumed by
+# _row_to_workflow_run_summary (positional). Excludes the heavy payload columns
+# (input_args, input_kwargs, result, metadata) which a list view never renders. See issue #482.
+RUN_SUMMARY_COLUMNS = (
+    "run_id, workflow_name, status, created_at, updated_at, started_at, completed_at, "
+    "error, idempotency_key, max_duration, recovery_attempts, max_recovery_attempts, "
+    "recover_on_worker_loss, parent_run_id, nesting_depth, continued_from_run_id, "
+    "continued_to_run_id"
+)
+
 
 class SQLiteMigrationRunner(MigrationRunner):
     """SQLite-specific migration runner."""
@@ -249,6 +259,12 @@ class SQLiteStorageBackend(StorageBackend):
         await db.execute(
             "CREATE INDEX IF NOT EXISTS idx_runs_created_at ON workflow_runs(created_at DESC)"
         )
+        # Composite index backing keyset pagination + ORDER BY created_at DESC, run_id DESC
+        # (see list_runs / issue #482).
+        await db.execute(
+            "CREATE INDEX IF NOT EXISTS idx_runs_created_at_run_id "
+            "ON workflow_runs(created_at DESC, run_id DESC)"
+        )
         await db.execute(
             "CREATE UNIQUE INDEX IF NOT EXISTS idx_runs_idempotency_key ON workflow_runs(idempotency_key) WHERE idempotency_key IS NOT NULL"
         )
@@ -597,15 +613,24 @@ class SQLiteStorageBackend(StorageBackend):
         limit: int = 100,
         cursor: str | None = None,
     ) -> tuple[list[WorkflowRun], str | None]:
-        """List workflow runs with optional filtering and pagination."""
+        """List workflow runs with optional filtering and pagination.
+
+        Note: the returned WorkflowRun objects are *summaries* — the heavy payload columns
+        (input_args, input_kwargs, result, metadata) are NOT fetched and carry placeholder
+        defaults. Use get_run() when the full payload is required. See issue #482.
+        """
         db = self._ensure_connected()
 
         conditions = []
         params: list[Any] = []
 
         if cursor:
+            # Keyset pagination on (created_at, run_id) — created_at is not unique, so paginating on
+            # it alone would skip/duplicate rows sharing a timestamp. SQLite supports row-value
+            # comparison (>= 3.15). See issue #482.
             conditions.append(
-                "created_at < (SELECT created_at FROM workflow_runs WHERE run_id = ?)"
+                "(created_at, run_id) < "
+                "(SELECT created_at, run_id FROM workflow_runs WHERE run_id = ?)"
             )
             params.append(cursor)
 
@@ -630,9 +655,9 @@ class SQLiteStorageBackend(StorageBackend):
         params.append(limit + 1)  # Fetch one extra to determine if there are more
 
         sql = f"""
-            SELECT * FROM workflow_runs
+            SELECT {RUN_SUMMARY_COLUMNS} FROM workflow_runs
             {where_clause}
-            ORDER BY created_at DESC
+            ORDER BY created_at DESC, run_id DESC
             LIMIT ?
         """
 
@@ -643,7 +668,7 @@ class SQLiteStorageBackend(StorageBackend):
         if has_more:
             rows = rows[:limit]
 
-        runs = [self._row_to_workflow_run(row) for row in rows]
+        runs = [self._row_to_workflow_run_summary(row) for row in rows]
         next_cursor = runs[-1].run_id if runs and has_more else None
 
         return runs, next_cursor
@@ -1801,6 +1826,37 @@ class SQLiteStorageBackend(StorageBackend):
             continued_to_run_id=row[20],
         )
 
+    def _row_to_workflow_run_summary(self, row: Any) -> WorkflowRun:
+        """Convert a summary row to a WorkflowRun.
+
+        Positional mapping MUST match the order of RUN_SUMMARY_COLUMNS. The heavy payload columns
+        are not selected by list_runs, so they are filled with their schema defaults. Callers
+        needing the real payload must use get_run(). See issue #482.
+        """
+        return WorkflowRun(
+            run_id=row[0],
+            workflow_name=row[1],
+            status=RunStatus(row[2]),
+            created_at=datetime.fromisoformat(row[3]),
+            updated_at=datetime.fromisoformat(row[4]),
+            started_at=datetime.fromisoformat(row[5]) if row[5] else None,
+            completed_at=datetime.fromisoformat(row[6]) if row[6] else None,
+            input_args="[]",
+            input_kwargs="{}",
+            result=None,
+            error=row[7],
+            idempotency_key=row[8],
+            max_duration=row[9],
+            context={},
+            recovery_attempts=row[10],
+            max_recovery_attempts=row[11],
+            recover_on_worker_loss=bool(row[12]),
+            parent_run_id=row[13],
+            nesting_depth=row[14],
+            continued_from_run_id=row[15],
+            continued_to_run_id=row[16],
+        )
+
     def _row_to_event(self, row: Any) -> Event:
         """Convert database row to Event object."""
         # Column order: event_id[0], run_id[1], sequence[2], type[3], timestamp[4], data[5], step_id[6]

{pyworkflow_engine-0.3.2 → pyworkflow_engine-0.3.4}/pyworkflow_engine.egg-info/SOURCES.txt

@@ -188,6 +188,7 @@ tests/integration/test_child_workflows.py
 tests/integration/test_continue_as_new.py
 tests/integration/test_dynamodb_storage.py
 tests/integration/test_fault_tolerance.py
+tests/integration/test_list_runs_pagination.py
 tests/integration/test_schedule_storage.py
 tests/integration/test_schema_migrations.py
 tests/integration/test_singleton.py

pyworkflow_engine-0.3.4/tests/integration/test_list_runs_pagination.py

@@ -0,0 +1,94 @@
+"""
+Integration tests for list_runs keyset pagination and summary projection (issue #482).
+
+Run against a real SQLite database so we exercise the actual SQL (row-value cursor comparison,
+ORDER BY created_at DESC, run_id DESC, and the summary column projection) rather than mocks.
+"""
+
+from datetime import UTC, datetime
+
+import pytest
+
+from pyworkflow.storage.schemas import RunStatus, WorkflowRun
+from pyworkflow.storage.sqlite import SQLiteStorageBackend
+
+
+@pytest.fixture
+async def storage(tmp_path):
+    """Real SQLite backend backed by a temp file."""
+    backend = SQLiteStorageBackend(db_path=str(tmp_path / "runs.db"))
+    await backend.connect()
+    yield backend
+    await backend.disconnect()
+
+
+async def _make_run(storage, run_id: str, created_at: datetime, **kwargs) -> None:
+    run = WorkflowRun(
+        run_id=run_id,
+        workflow_name="wf",
+        status=RunStatus.COMPLETED,
+        created_at=created_at,
+        updated_at=created_at,
+        **kwargs,
+    )
+    await storage.create_run(run)
+
+
+@pytest.mark.asyncio
+async def test_list_runs_keyset_pagination_handles_tied_created_at(storage):
+    """Rows sharing one created_at must each appear exactly once across cursor pages.
+
+    The old `created_at < cursor.created_at` predicate skipped/duplicated rows on ties; the keyset
+    `(created_at, run_id)` predicate must page through all of them.
+    """
+    ts = datetime(2024, 1, 1, 12, 0, 0, tzinfo=UTC)
+    # Five runs, all with the SAME created_at — the pathological tie case.
+    run_ids = {f"run_{i}" for i in range(5)}
+    for run_id in run_ids:
+        await _make_run(storage, run_id, ts)
+
+    seen: list[str] = []
+    cursor = None
+    for _ in range(10):  # generous upper bound; should terminate well before this
+        runs, cursor = await storage.list_runs(limit=2, cursor=cursor)
+        seen.extend(r.run_id for r in runs)
+        if cursor is None:
+            break
+
+    assert sorted(seen) == sorted(run_ids)  # every row exactly once, no skips/dupes
+    assert len(seen) == len(set(seen))
+
+
+@pytest.mark.asyncio
+async def test_list_runs_returns_summary_without_heavy_payload(storage):
+    """list_runs omits the heavy payload columns; they come back as defaults, error survives."""
+    ts = datetime(2024, 1, 1, 12, 0, 0, tzinfo=UTC)
+    await _make_run(
+        storage,
+        "run_heavy",
+        ts,
+        input_args='["big"]',
+        input_kwargs='{"k": "v"}',
+        result='{"r": 1}',
+        error="boom",
+        context={"meta": "data"},
+    )
+
+    runs, _ = await storage.list_runs(limit=10)
+
+    assert len(runs) == 1
+    summary = runs[0]
+    # Heavy payload columns are not fetched -> placeholder defaults.
+    assert summary.input_args == "[]"
+    assert summary.input_kwargs == "{}"
+    assert summary.result is None
+    assert summary.context == {}
+    # Lightweight fields used by the list view survive.
+    assert summary.error == "boom"
+    assert summary.status == RunStatus.COMPLETED
+
+    # get_run still returns the full payload.
+    full = await storage.get_run("run_heavy")
+    assert full.input_args == '["big"]'
+    assert full.result == '{"r": 1}'
+    assert full.context == {"meta": "data"}

{pyworkflow_engine-0.3.2 → pyworkflow_engine-0.3.4}/tests/unit/backends/test_postgres_storage.py

@@ -531,6 +531,31 @@ class TestWorkflowRunOperations:
 
         assert len(runs) == 1
         assert runs[0].run_id == "run_1"
+        # list_runs returns summaries: heavy payload columns are not fetched and carry defaults,
+        # while lightweight fields (incl. error) survive. See issue #482.
+        assert runs[0].input_args == "[]"
+        assert runs[0].input_kwargs == "{}"
+        assert runs[0].result is None
+        assert runs[0].context == {}
+        assert runs[0].error is None
+
+        # The query must not SELECT * (avoids scatter-gathering wide rows across Citus shards) and
+        # must order by the keyset tuple (created_at, run_id).
+        sql = mock_conn.fetch.call_args[0][0]
+        assert "SELECT *" not in sql
+        assert "input_args" not in sql
+        assert "ORDER BY created_at DESC, run_id DESC" in sql
+
+    @pytest.mark.asyncio
+    async def test_list_runs_cursor_uses_keyset(self, mock_backend):
+        """Cursor pagination compares the (created_at, run_id) tuple, not created_at alone."""
+        backend, mock_conn = mock_backend
+        mock_conn.fetch.return_value = []
+
+        await backend.list_runs(limit=10, cursor="run_1")
+
+        sql = mock_conn.fetch.call_args[0][0]
+        assert "(created_at, run_id) <" in sql
 
 
 class TestEventOperations: