Harden Dataflow preview behavior

This commit is contained in:
2026-07-28 01:30:36 +02:00
parent 223c008ff3
commit df468a2bd8
4 changed files with 122 additions and 6 deletions

View File

@@ -2,7 +2,7 @@ from __future__ import annotations
import unittest
from govoplan_dataflow.backend.executor import execute_preview
from govoplan_dataflow.backend.executor import PipelineExecutionError, execute_preview
from govoplan_dataflow.backend.schemas import (
GraphEdge,
GraphNode,
@@ -120,6 +120,89 @@ class DataflowGraphAndSqlTests(unittest.TestCase):
self.assertEqual(3, result.total_rows)
self.assertTrue(result.truncated)
def test_empty_aggregate_and_null_sort_are_deterministic(self) -> None:
graph, _, _ = compile_sql(
"""
SELECT status, COUNT(*) AS records
FROM monthly_files
GROUP BY status
ORDER BY status DESC
""",
source_nodes=[
inline_source().model_copy(
update={
"config": {
"source_name": "monthly_files",
"rows": [
{"status": None},
{"status": "open"},
{"status": "closed"},
],
}
}
)
],
)
first = execute_preview(graph, row_limit=100)
second = execute_preview(graph, row_limit=100)
self.assertEqual(first.rows, second.rows)
self.assertEqual(
[
{"status": "open", "records": 1},
{"status": "closed", "records": 1},
{"status": None, "records": 1},
],
first.rows,
)
self.assertEqual(
[
diagnostic.model_dump(exclude={"duration_ms"})
for diagnostic in first.node_diagnostics
],
[
diagnostic.model_dump(exclude={"duration_ms"})
for diagnostic in second.node_diagnostics
],
)
def test_empty_global_aggregate_returns_a_count_row(self) -> None:
graph, _, _ = compile_sql(
"SELECT COUNT(*) AS records FROM monthly_files",
source_nodes=[
inline_source().model_copy(
update={"config": {"source_name": "monthly_files", "rows": []}}
)
],
)
result = execute_preview(graph, row_limit=100)
self.assertEqual([{"records": 0}], result.rows)
self.assertEqual(0, result.input_row_count)
def test_invalid_runtime_comparison_is_scoped_to_the_node(self) -> None:
graph, _, _ = compile_sql(
"SELECT * FROM monthly_files WHERE amount > 10",
source_nodes=[
inline_source().model_copy(
update={
"config": {
"source_name": "monthly_files",
"rows": [{"amount": "not-a-number"}],
}
}
)
],
)
with self.assertRaises(PipelineExecutionError) as raised:
execute_preview(graph, row_limit=100)
self.assertEqual("filter-1", raised.exception.node_id)
self.assertIn("Cannot apply", str(raised.exception))
if __name__ == "__main__":
unittest.main()