def test_adjusted_expected_tile_some_nans_and_square_tiling():
print("Running tile some nans la_exp test + square tiling")
# first, generate that locally-adjusted expected:
nnans = 1
band_idx = int(band / b)
res_df = pd.DataFrame([])
for tilei, tilej in square_matrix_tiling(start,
stop,
tile_size=40,
edge=w,
square=False):
# define origin:
origin = (tilei[0], tilej[0])
# RAW observed matrix slice:
observed = mock_M_raw[slice(*tilei), slice(*tilej)]
# trying new expected function:
expected = tile_of_expected(start, tilei, tilej, get_mock_exp)
# for diagonal chuynking/tiling tilei==tilej:
ice_weight_i = mock_v_ice[slice(*tilei)]
ice_weight_j = mock_v_ice[slice(*tilej)]
# that's the main working function from loopify:
res = get_adjusted_expected_tile_some_nans(
origin=origin,
observed=observed,
expected=expected,
bal_weight=(ice_weight_i, ice_weight_j),
kernels={
"donut": kernel,
"footprint": np.ones_like(kernel)
},
# nan_threshold=1,
verbose=False)
is_inside_band = (res["row"] > (res["col"] - band_idx))
# new style, selecting good guys:
does_comply_nans = (res["la_exp." + "footprint" + ".nnans"] < nnans)
# so, select inside band and nNaNs compliant results and append:
res_df = res_df.append(res[is_inside_band & does_comply_nans],
ignore_index=True)
# drop dups (from overlaping tiles), sort and reset index:
res_df = res_df \
.drop_duplicates() \
.sort_values(by=['row','col']) \
.reset_index(drop=True)
# prepare mock_data for comparison:
# apparently sorting is needed in this case:
mock_res_sorted = mock_res.sort_values(by=['row', 'col']).reset_index(
drop=True)
# ACTUAL TESTS:
# integer part of DataFrame must equals exactly:
assert (res_df[['row', 'col']].equals(mock_res_sorted[['row', 'col']]))
# compare floating point part separately:
assert (np.isclose(res_df["la_exp." + "donut" + ".value"],
mock_res_sorted['la_expected'],
equal_nan=True).all())
def test_tile_exp_diag_rectangle():
# ####################
# requested tile is on the diag
# rectangular tile!
# ####################
i0, i1 = 50, 100
j0, j1 = 50, 80
#
E_tile_ref = mock_E_ice[i0:i1, j0:j1]
# now, let's try to reconstruct it
# from diag-indexed vector using
expected = tile_of_expected(start, (i0, i1), (j0, j1), get_mock_exp)
##########################
assert np.isclose(E_tile_ref, expected, equal_nan=True).all()
def test_tile_exp_lower():
# ####################
# requested tile is in the lower
# triangle of the matrix region
# ####################
i0, i1 = 150, 200
j0, j1 = 5, 100
#
E_tile_ref = mock_E_ice[i0:i1, j0:j1]
# now, let's try to reconstruct it
# from diag-indexed vector using
expected = tile_of_expected(start, (i0, i1), (j0, j1), get_mock_exp)
##########################
assert np.isclose(E_tile_ref, expected, equal_nan=True).all()