A reactive notebook for Python — run reproducible experiments, query with SQL, execute as a script, deploy as an app, and version with git. Stored as pure Python. All in a modern, AI-native editor.
Python artificial-intelligencedagdata-sciencedata-visualizationdataflowdeveloper-toolsmachine-learningnotebookspipelinepythonreactivesqlweb-app
from typing import cast
import pytest
from marimo._runtime.app.kernel_runner import AppKernelRunner
from marimo._types.ids import CellId_t
def _make_runner():
# Bypass __init__ since we only test caching logic
runner = AppKernelRunner.__new__(AppKernelRunner)
runner._previously_seen_defs = None
runner._outputs = {cast(CellId_t, "cell-1"): "dummy"}
return runner
@pytest.mark.requires("numpy")
def test_numpy_defs_do_not_crash_and_invalidate_cache():
runner = _make_runner()
import numpy as np
defs1 = {"arr": np.ones(1)}
defs2 = {"arr": np.zeros(2)}
runner._previously_seen_defs = defs1
# Must not raise ValueError
cached = runner.are_outputs_cached(defs2)
assert cached is False
@pytest.mark.requires("numpy")
def test_numpy_defs_equal_use_cache():
runner = _make_runner()
import numpy as np
arr = np.array([1, 2, 3])
defs1 = {"arr": arr}
defs2 = {"arr": arr.copy()} # different object, same values
runner._previously_seen_defs = defs1
cached = runner.are_outputs_cached(defs2)
assert cached is True