## Summary
Two bugs in `agnes describe` surfaced from a real analyst session following the CLAUDE.md agent-rails discovery workflow. Together they break `agnes describe` end-to-end for any analyst (or analyst-AI) who follows the documented form.
### A) CLI parsing
`agnes describe TABLE -n 5` failed with `Missing argument 'TABLE_ID'`. Root cause: the command was registered as a `Typer.Typer` subcommand group via `app.add_typer(describe_app, name="describe")` + `@describe_app.callback(invoke_without_command=True)`, and that pattern mis-parses positional + short-int option in some orderings. Same pattern in `cli/commands/schema.py` works only because schema has no INTEGER short option. Fix: switch to flat `@app.command("describe")`.
### B) Server NaN
`/api/v2/sample/<id>` (called by `agnes describe`) returned HTTP 500 with `ValueError: Out of range float values are not JSON compliant: nan` whenever a row contained NaN. Fix: sanitize NaN/±inf to None before JSON serialization.
## Test plan
- [x] `pytest tests/test_cli_describe*.py` — added regression tests pinning `-n` parsing on either side of the positional.
- [x] `pytest tests/test_api_v2_sample*.py` — added regression test for NaN row → JSON `null` (not 500).
<!-- devin-review-badge-begin -->
---
<a href="https://app.devin.ai/review/keboola/agnes-the-ai-analyst/pull/224" target="_blank">
<picture>
<source media="(prefers-color-scheme: dark)" srcset="https://static.devin.ai/assets/gh-open-in-devin-review-dark.svg?v=1">
<img src="https://static.devin.ai/assets/gh-open-in-devin-review-light.svg?v=1" alt="Open in Devin Review">
</picture>
</a>
<!-- devin-review-badge-end -->
315 lines
12 KiB
Python
315 lines
12 KiB
Python
# tests/test_v2_sample.py
|
|
import asyncio
|
|
import importlib
|
|
from unittest.mock import MagicMock, patch
|
|
import pytest
|
|
from fastapi import HTTPException
|
|
|
|
|
|
@pytest.fixture
|
|
def reload_db(tmp_path, monkeypatch):
|
|
monkeypatch.setenv("DATA_DIR", str(tmp_path))
|
|
import src.db as db_module
|
|
importlib.reload(db_module)
|
|
yield db_module
|
|
|
|
|
|
def _seed(conn):
|
|
_ensure_admin1(conn)
|
|
from src.repositories.table_registry import TableRegistryRepository
|
|
TableRegistryRepository(conn).register(
|
|
id="bq_view", name="bq_view", source_type="bigquery",
|
|
bucket="ds", source_table="bq_view", query_mode="remote",
|
|
)
|
|
|
|
|
|
def _ensure_admin1(conn):
|
|
"""Seed an admin user with id='admin1' + Admin group membership so
|
|
{"id": "admin1", ...} dicts pass the can_access admin shortcut."""
|
|
from src.db import SYSTEM_ADMIN_GROUP
|
|
from src.repositories.users import UserRepository
|
|
from src.repositories.user_group_members import UserGroupMembersRepository
|
|
if UserRepository(conn).get_by_id('admin1') is None:
|
|
UserRepository(conn).create(id='admin1', email='admin1@test.com', name='Admin')
|
|
admin_gid = conn.execute(
|
|
'SELECT id FROM user_groups WHERE name = ?', [SYSTEM_ADMIN_GROUP]
|
|
).fetchone()
|
|
if admin_gid:
|
|
UserGroupMembersRepository(conn).add_member(
|
|
'admin1', admin_gid[0], source='system_seed',
|
|
)
|
|
|
|
|
|
def _bq(billing="billing-proj", data="data-proj"):
|
|
"""Build a BqAccess wired to default factories. For tests that monkeypatch
|
|
`_fetch_bq_sample` whole, the inner factories are never called."""
|
|
from connectors.bigquery.access import BqAccess, BqProjects
|
|
return BqAccess(BqProjects(billing=billing, data=data))
|
|
|
|
|
|
class TestSampleEndpoint:
|
|
def test_returns_n_rows_for_bq_table(self, reload_db, monkeypatch):
|
|
from app.api import v2_sample
|
|
monkeypatch.setattr(
|
|
v2_sample, "_fetch_bq_sample",
|
|
lambda bq, dataset, table, n: [
|
|
{"event_date": "2026-04-27", "country_code": "CZ"},
|
|
{"event_date": "2026-04-26", "country_code": "SK"},
|
|
],
|
|
)
|
|
conn = reload_db.get_system_db()
|
|
try:
|
|
_seed(conn)
|
|
user = {"id": "admin1", "email": "a@x.com"}
|
|
data = v2_sample.build_sample(conn, user, "bq_view", n=2, bq=_bq())
|
|
finally:
|
|
conn.close()
|
|
assert data["table_id"] == "bq_view"
|
|
assert len(data["rows"]) == 2
|
|
|
|
def test_caps_n_at_100(self, reload_db, monkeypatch):
|
|
from app.api import v2_sample
|
|
captured = {}
|
|
def fake_fetch(bq, dataset, table, n):
|
|
captured["n"] = n
|
|
return []
|
|
monkeypatch.setattr(v2_sample, "_fetch_bq_sample", fake_fetch)
|
|
conn = reload_db.get_system_db()
|
|
try:
|
|
_seed(conn)
|
|
user = {"id": "admin1", "email": "a@x.com"}
|
|
v2_sample.build_sample(conn, user, "bq_view", n=999, bq=_bq())
|
|
finally:
|
|
conn.close()
|
|
assert captured["n"] == 100
|
|
|
|
def test_sample_handles_nan_values_in_rows(self, reload_db, monkeypatch):
|
|
"""Regression: rows containing NaN floats from a DuckDB / BigQuery
|
|
scan used to crash the response with `ValueError: Out of range
|
|
float values are not JSON compliant: nan`. The endpoint now
|
|
sanitizes NaN/±inf to None before returning the payload."""
|
|
import math
|
|
from app.api import v2_sample
|
|
v2_sample._sample_cache.clear()
|
|
monkeypatch.setattr(
|
|
v2_sample, "_fetch_bq_sample",
|
|
lambda bq, dataset, table, n: [
|
|
{"col": float("nan"), "ok": 1.0},
|
|
{"col": float("inf"), "ok": 2.0},
|
|
{"col": float("-inf"), "ok": 3.0},
|
|
],
|
|
)
|
|
conn = reload_db.get_system_db()
|
|
try:
|
|
_seed(conn)
|
|
user = {"id": "admin1", "email": "a@x.com"}
|
|
data = v2_sample.build_sample(conn, user, "bq_view", n=3, bq=_bq())
|
|
finally:
|
|
conn.close()
|
|
assert data["rows"] == [
|
|
{"col": None, "ok": 1.0},
|
|
{"col": None, "ok": 2.0},
|
|
{"col": None, "ok": 3.0},
|
|
]
|
|
# Belt-and-braces: payload must round-trip through stdlib json
|
|
# in strict mode (allow_nan=False) — that's what FastAPI's
|
|
# serializer enforces internally.
|
|
import json as _json
|
|
_json.dumps(data, allow_nan=False) # must not raise
|
|
|
|
def test_sample_handles_nested_nan_in_arrays(self, reload_db, monkeypatch):
|
|
"""Sanitizer recurses into nested lists/dicts — array-typed BQ
|
|
cells with NaN inside also serialize cleanly."""
|
|
from app.api import v2_sample
|
|
v2_sample._sample_cache.clear()
|
|
monkeypatch.setattr(
|
|
v2_sample, "_fetch_bq_sample",
|
|
lambda *a, **kw: [{"arr": [1.0, float("nan"), 3.0],
|
|
"nested": {"x": float("inf")}}],
|
|
)
|
|
conn = reload_db.get_system_db()
|
|
try:
|
|
_seed(conn)
|
|
user = {"id": "admin1", "email": "a@x.com"}
|
|
data = v2_sample.build_sample(conn, user, "bq_view", n=1, bq=_bq())
|
|
finally:
|
|
conn.close()
|
|
assert data["rows"][0]["arr"] == [1.0, None, 3.0]
|
|
assert data["rows"][0]["nested"] == {"x": None}
|
|
import json as _json
|
|
_json.dumps(data, allow_nan=False)
|
|
|
|
def test_rbac_check_runs_before_cache(self, reload_db, monkeypatch):
|
|
"""Regression: cache check used to come before RBAC, leaking sample rows
|
|
cached by an authorized user to subsequent unauthorized callers."""
|
|
from app.api import v2_sample
|
|
monkeypatch.setattr(
|
|
v2_sample, "_fetch_bq_sample",
|
|
lambda *a, **kw: [{"col": "secret"}],
|
|
)
|
|
monkeypatch.setattr(
|
|
"app.api.v2_sample.can_access_table",
|
|
lambda user, tid, conn: user.get("id") == "admin1",
|
|
)
|
|
conn = reload_db.get_system_db()
|
|
try:
|
|
_seed(conn)
|
|
admin = {"id": "admin1", "email": "admin@x.com"}
|
|
v2_sample.build_sample(conn, admin, "bq_view", n=2, bq=_bq())
|
|
other = {"id": "viewer1", "email": "viewer@x.com"}
|
|
with pytest.raises(PermissionError):
|
|
v2_sample.build_sample(conn, other, "bq_view", n=2, bq=_bq())
|
|
finally:
|
|
conn.close()
|
|
|
|
|
|
class TestBqAccessErrors:
|
|
"""Issue #134: structured 502 translation on BQ errors in sample path.
|
|
|
|
These tests exercise the REAL translation path through `BqAccess` +
|
|
`translate_bq_error` by injecting a duckdb_session whose execute() raises
|
|
the Google API exception. That's the production path — Phase 1
|
|
monkeypatches of `_fetch_bq_sample` whole would skip the translation logic
|
|
and only test the outer wrap (which has been removed in Phase 2).
|
|
|
|
Key difference from /scan: /sample SQL is server-constructed (validated
|
|
identifiers + LIMIT n), so a BadRequest from BQ means registry corruption,
|
|
NOT user input → translates to `bq_upstream_error` (HTTP 502), not 400.
|
|
"""
|
|
|
|
@pytest.fixture(autouse=True)
|
|
def _clear_sample_cache(self):
|
|
"""The sample-result TTL cache is module-level; clear it between
|
|
tests so cached payloads from a sibling test don't mask call paths."""
|
|
from app.api import v2_sample
|
|
v2_sample._sample_cache.clear()
|
|
yield
|
|
v2_sample._sample_cache.clear()
|
|
|
|
def test_sample_returns_502_on_bq_forbidden_serviceusage(self, reload_db, bq_access):
|
|
"""When the BQ extension raises Forbidden mentioning serviceusage,
|
|
the endpoint must translate to HTTP 502 with a structured body
|
|
whose `error` is `cross_project_forbidden` and whose hint mentions
|
|
`billing_project`."""
|
|
from app.api import v2_sample
|
|
from google.api_core.exceptions import Forbidden
|
|
|
|
mock_conn = MagicMock()
|
|
mock_conn.execute.side_effect = Forbidden(
|
|
"Permission denied: serviceusage.services.use on project foo"
|
|
)
|
|
bq = bq_access(duckdb_conn=mock_conn, billing="billing-proj", data="data-proj")
|
|
|
|
conn = reload_db.get_system_db()
|
|
try:
|
|
_seed(conn)
|
|
user = {"id": "admin1", "email": "a@x.com"}
|
|
|
|
# Endpoint is async — drive it directly. dependency_overrides only
|
|
# fires through TestClient/HTTP, so pass `bq=bq` explicitly.
|
|
with pytest.raises(HTTPException) as exc_info:
|
|
(v2_sample.sample(
|
|
table_id="bq_view", n=5, user=user, conn=conn, bq=bq,
|
|
))
|
|
finally:
|
|
conn.close()
|
|
|
|
assert exc_info.value.status_code == 502
|
|
detail = exc_info.value.detail
|
|
assert isinstance(detail, dict)
|
|
assert detail["error"] == "cross_project_forbidden"
|
|
assert "billing_project" in detail["details"]["hint"].lower()
|
|
|
|
def test_sample_returns_502_on_bq_forbidden_non_serviceusage(self, reload_db, bq_access):
|
|
"""A Forbidden that is NOT about serviceusage (e.g. dataset-level ACL)
|
|
still becomes a 502, but with `bq_forbidden` (no billing_project hint)."""
|
|
from app.api import v2_sample
|
|
from google.api_core.exceptions import Forbidden
|
|
|
|
mock_conn = MagicMock()
|
|
mock_conn.execute.side_effect = Forbidden(
|
|
"Access Denied: Table foo.bar.baz: User does not have permission"
|
|
)
|
|
bq = bq_access(duckdb_conn=mock_conn, billing="billing-proj", data="data-proj")
|
|
|
|
conn = reload_db.get_system_db()
|
|
try:
|
|
_seed(conn)
|
|
user = {"id": "admin1", "email": "a@x.com"}
|
|
|
|
with pytest.raises(HTTPException) as exc_info:
|
|
(v2_sample.sample(
|
|
table_id="bq_view", n=5, user=user, conn=conn, bq=bq,
|
|
))
|
|
finally:
|
|
conn.close()
|
|
|
|
assert exc_info.value.status_code == 502
|
|
assert exc_info.value.detail["error"] == "bq_forbidden"
|
|
|
|
def test_sample_returns_502_on_bq_bad_request(self, reload_db, bq_access):
|
|
"""`/sample` SQL is server-constructed (validated identifiers + LIMIT n),
|
|
so a BQ BadRequest means registry corruption, not user input. Must
|
|
surface as HTTP 502 with `bq_upstream_error` (NOT 400 / `bq_bad_request`
|
|
like /scan does — that's the key difference from Task 2.7)."""
|
|
from app.api import v2_sample
|
|
from google.api_core.exceptions import BadRequest
|
|
|
|
mock_conn = MagicMock()
|
|
mock_conn.execute.side_effect = BadRequest(
|
|
"Syntax error: unexpected token at line 1, column 5"
|
|
)
|
|
bq = bq_access(duckdb_conn=mock_conn, billing="billing-proj", data="data-proj")
|
|
|
|
conn = reload_db.get_system_db()
|
|
try:
|
|
_seed(conn)
|
|
user = {"id": "admin1", "email": "a@x.com"}
|
|
|
|
with pytest.raises(HTTPException) as exc_info:
|
|
(v2_sample.sample(
|
|
table_id="bq_view", n=5, user=user, conn=conn, bq=bq,
|
|
))
|
|
finally:
|
|
conn.close()
|
|
|
|
assert exc_info.value.status_code == 502
|
|
detail = exc_info.value.detail
|
|
assert isinstance(detail, dict)
|
|
assert detail["error"] == "bq_upstream_error"
|
|
assert "Syntax error" in detail["message"]
|
|
|
|
def test_sample_passes_billing_project_to_bigquery_query(self, reload_db, bq_access):
|
|
"""Regression guard: bq.projects.billing must be passed to bigquery_query()
|
|
as the billing project (positional arg 0). Verifies the migration didn't
|
|
regress the original cross-project bug fix."""
|
|
from app.api import v2_sample
|
|
|
|
captured = {}
|
|
|
|
def _fake_execute(sql, params):
|
|
# Capture the bigquery_query() call args.
|
|
if "bigquery_query" in sql:
|
|
captured["billing_project"] = params[0]
|
|
captured["bq_sql"] = params[1]
|
|
result = MagicMock()
|
|
result.fetchdf.return_value.to_dict.return_value = []
|
|
return result
|
|
|
|
mock_conn = MagicMock()
|
|
mock_conn.execute.side_effect = _fake_execute
|
|
bq = bq_access(duckdb_conn=mock_conn, billing="billing-proj", data="data-proj")
|
|
|
|
conn = reload_db.get_system_db()
|
|
try:
|
|
_seed(conn)
|
|
user = {"id": "admin1", "email": "a@x.com"}
|
|
(v2_sample.sample(
|
|
table_id="bq_view", n=5, user=user, conn=conn, bq=bq,
|
|
))
|
|
finally:
|
|
conn.close()
|
|
|
|
assert captured["billing_project"] == "billing-proj"
|
|
# FROM clause uses data project (where the table actually lives)
|
|
assert "`data-proj.ds.bq_view`" in captured["bq_sql"]
|