"""Offline tests for CSV validation and input constraints.""" from __future__ import annotations import io from pathlib import Path import numpy as np import pandas as pd import pytest from lab.validation import ( MAX_FILE_BYTES, MAX_HORIZON, MAX_POINTS, ValidationError, parse_series_csv, validate_run_params, ) def _write_csv(tmp_path: Path, name: str, df: pd.DataFrame) -> Path: path = tmp_path / name df.to_csv(path, index=False) return path def test_parse_numeric_column_and_optional_date(tmp_path: Path) -> None: df = pd.DataFrame( { "fecha": pd.date_range("2024-01-01", periods=40, freq="D"), "demanda": np.arange(40, dtype=float), "nota": ["x"] * 40, } ) path = _write_csv(tmp_path, "ok.csv", df) series = parse_series_csv(path, value_column="demanda", date_column="fecha") assert series.n_points == 40 assert series.value_column == "demanda" assert series.timestamps is not None assert len(series.values) == 40 assert np.isfinite(series.values).all() def test_reject_non_csv_extension(tmp_path: Path) -> None: path = tmp_path / "data.txt" path.write_text("a,b\n1,2\n", encoding="utf-8") with pytest.raises(ValidationError, match="CSV"): parse_series_csv(path) def test_reject_path_traversal_style_name(tmp_path: Path) -> None: # Symlink / relative escape should be rejected when unsafe name is detected. victim = tmp_path / "secret.csv" victim.write_text("v\n1\n2\n", encoding="utf-8") evil = tmp_path / ".." / victim.name with pytest.raises(ValidationError): parse_series_csv(evil) def test_reject_oversized_file(tmp_path: Path) -> None: path = tmp_path / "big.csv" # Build a file just over the limit without allocating huge RAM for dataframes. header = b"value\n" row = b"1.0\n" need = MAX_FILE_BYTES - len(header) + 1 with path.open("wb") as f: f.write(header) chunk = row * 1000 written = 0 while written < need: f.write(chunk) written += len(chunk) assert path.stat().st_size > MAX_FILE_BYTES with pytest.raises(ValidationError, match="1 ?MB|tamaƱo|size"): parse_series_csv(path) def test_reject_too_many_points(tmp_path: Path) -> None: n = MAX_POINTS + 1 df = pd.DataFrame({"value": np.ones(n, dtype=float)}) path = _write_csv(tmp_path, "many.csv", df) with pytest.raises(ValidationError, match="10000|puntos|points"): parse_series_csv(path) def test_reject_nan_and_nonfinite(tmp_path: Path) -> None: df = pd.DataFrame({"value": [1.0, 2.0, np.nan, 4.0]}) path = _write_csv(tmp_path, "nan.csv", df) with pytest.raises(ValidationError, match="NaN|finite|finit"): parse_series_csv(path) df2 = pd.DataFrame({"value": [1.0, np.inf, 3.0]}) path2 = _write_csv(tmp_path, "inf.csv", df2) with pytest.raises(ValidationError, match="finite|finit|Inf"): parse_series_csv(path2) def test_corrupt_csv_raises(tmp_path: Path) -> None: path = tmp_path / "bad.csv" path.write_bytes(b"\xff\xfe not,csv\x00\x01") with pytest.raises(ValidationError): parse_series_csv(path) def test_minimum_context_and_horizon_limits() -> None: with pytest.raises(ValidationError): validate_run_params(n_points=10, holdout=5, future_horizon=24, seasonality=7) with pytest.raises(ValidationError): validate_run_params(n_points=200, holdout=MAX_HORIZON + 1, future_horizon=24, seasonality=7) params = validate_run_params(n_points=200, holdout=24, future_horizon=24, seasonality=7) assert params.holdout == 24 assert params.future_horizon == 24 def test_auto_pick_last_numeric_column(tmp_path: Path) -> None: df = pd.DataFrame({"a": [1, 2, 3, 4] * 20, "b": [10.0] * 80}) path = _write_csv(tmp_path, "auto.csv", df) series = parse_series_csv(path, value_column=None) assert series.value_column == "b"