test_geodataframe.py's test_memory_usage is failing due to size mismatch

[jakirkham]

As part of the Python 3.11 bringup, we noticed that test_geodataframe.py's test_memory_usage fails

Namely in Python 3.9 and 3.10 it uses 224 bytes, but in Python 3.11 it uses 228 bytes

cuspatial/python/cuspatial/cuspatial/tests/test_geodataframe.py

Lines 323 to 330 in 59972f6

	def test_memory_usage(gs):
	assert gs.memory_usage() == 224
	host_dataframe = gpd.read_file(
	gpd.datasets.get_path("naturalearth_lowres")
	)
	gpu_dataframe = cuspatial.from_geopandas(host_dataframe)
	# The df size is 8kb of cudf rows and 217kb of the geometry column
	assert gpu_dataframe.memory_usage().sum() == 224945

Unfortunately in Python these kinds of tests can be pretty flaky as things can change in under-the-hood dependencies (perhaps in Python itself) or do to platform differences, so it might be worth rethinking this test altogether

For now this test is being skipped. Filing this issue to track

This test was originally added in PR ( #604 )

[harrism]

Seems a bit problematic to have tests that expect exact data sizes for data structures that might change.