def test_assert_similar_hists():
""" Test assert on similarity of list of histograms
Check similarity of: type, n-dim, sub-hists, specific type attributes
"""
# dummy dataset with mixed types
# convert timestamp (col D) to nanosec since 1970-1-1
df = pd.util.testing.makeMixedDataFrame()
df['date'] = df['D'].apply(to_ns)
# building 1d-, 2d-, and 3d-histogram (iteratively)
hist0 = hg.Bin(5, 0, 5, unit('A'))
hist1 = hg.Categorize(unit('C'))
hist2 = hg.Bin(5, 0, 5, unit('A'), value=hist1)
hist3 = hg.Categorize(unit('C'), value=hist0)
hist4 = hg.SparselyBin(origin=pd.Timestamp('2009-01-01').value,
binWidth=pd.Timedelta(days=1).value,
quantity=unit('date'),
value=hist2)
hist5 = hg.SparselyBin(origin=pd.Timestamp('2009-01-01').value,
binWidth=pd.Timedelta(days=1).value,
quantity=unit('date'),
value=hist3)
# fill them
for hist in [hist0, hist1, hist2, hist3, hist4, hist5]:
hist.fill.numpy(df)
hc0 = HistogramContainer(hist0)
hc1 = HistogramContainer(hist1)
hc2 = HistogramContainer(hist2)
hc3 = HistogramContainer(hist3)
hc4 = HistogramContainer(hist4)
hc5 = HistogramContainer(hist5)
for hc in [hc0, hc1, hc2, hc3, hc4, hc5]:
assert check_similar_hists([hc, hc])
args01 = ['']
args23 = ['']
args45 = ['']
try:
assert_similar_hists([hc0, hc1])
except AssertionError as e:
args01 = e.args
try:
assert_similar_hists([hc2, hc3])
except AssertionError as e:
args23 = e.args
try:
assert_similar_hists([hc4, hc5])
except AssertionError as e:
args45 = e.args
assert args01[0] == 'Input histograms are not all similar.'
assert args23[0] == 'Input histograms are not all similar.'
assert args45[0] == 'Input histograms are not all similar.'
def test_assert_similar_hists():
"""Test assert on similarity of list of histograms
Check similarity of: type, n-dim, sub-hists, specific type attributes
"""
# dummy dataset with mixed types
# convert timestamp (col D) to nanosec since 1970-1-1
df = pd.util.testing.makeMixedDataFrame()
df["date"] = df["D"].apply(to_ns)
# building 1d-, 2d-, and 3d-histogram (iteratively)
hist0 = hg.Bin(5, 0, 5, unit("A"))
hist1 = hg.Categorize(unit("C"))
hist2 = hg.Bin(5, 0, 5, unit("A"), value=hist1)
hist3 = hg.Categorize(unit("C"), value=hist0)
hist4 = hg.SparselyBin(
origin=pd.Timestamp("2009-01-01").value,
binWidth=pd.Timedelta(days=1).value,
quantity=unit("date"),
value=hist2,
)
hist5 = hg.SparselyBin(
origin=pd.Timestamp("2009-01-01").value,
binWidth=pd.Timedelta(days=1).value,
quantity=unit("date"),
value=hist3,
)
# fill them
for hist in [hist0, hist1, hist2, hist3, hist4, hist5]:
hist.fill.numpy(df)
for hist in [hist0, hist1, hist2, hist3, hist4, hist5]:
assert check_similar_hists([hist, hist])
args01 = [""]
args23 = [""]
args45 = [""]
try:
assert_similar_hists([hist0, hist1])
except ValueError as e:
args01 = e.args
try:
assert_similar_hists([hist2, hist3])
except ValueError as e:
args23 = e.args
try:
assert_similar_hists([hist4, hist5])
except ValueError as e:
args45 = e.args
assert args01[0] == "Input histograms are not all similar."
assert args23[0] == "Input histograms are not all similar."
assert args45[0] == "Input histograms are not all similar."
def test_check_similar_hists():
"""Test similarity of list of histograms
Check similarity of: type, n-dim, sub-hists, specific type attributes
"""
# dummy dataset with mixed types
# convert timestamp (col D) to nanosec since 1970-1-1
df = pd.util.testing.makeMixedDataFrame()
df["date"] = df["D"].apply(to_ns)
# building 1d-, 2d-, and 3d-histogram (iteratively)
hist0 = hg.Bin(5, 0, 5, unit("A"))
hist1 = hg.Categorize(unit("C"))
hist2 = hg.Bin(5, 0, 5, unit("A"), value=hist1)
hist3 = hg.Categorize(unit("C"), value=hist0)
hist4 = hg.SparselyBin(
origin=pd.Timestamp("2009-01-01").value,
binWidth=pd.Timedelta(days=1).value,
quantity=unit("date"),
value=hist2,
)
hist5 = hg.SparselyBin(
origin=pd.Timestamp("2009-01-01").value,
binWidth=pd.Timedelta(days=1).value,
quantity=unit("date"),
value=hist3,
)
# fill them
for hist in [hist0, hist1, hist2, hist3, hist4, hist5]:
hist.fill.numpy(df)
hc0 = HistogramContainer(hist0)
hc1 = HistogramContainer(hist1)
hc2 = HistogramContainer(hist2)
hc3 = HistogramContainer(hist3)
hc4 = HistogramContainer(hist4)
hc5 = HistogramContainer(hist5)
for hc in [hc0, hc1, hc2, hc3, hc4, hc5]:
assert check_similar_hists([hc, hc])
assert not check_similar_hists([hc0, hc1])
assert not check_similar_hists([hc2, hc3])
assert not check_similar_hists([hc4, hc5])
def test_hist_splitter_filter():
""" Test of hist_splitter option filter_empty_split_hists
One of the split histograms of type date:A_score:num_employees is empty and only contains a NaN.
In this test, those empty split-histograms are *not* removed, leading to split-histograms of
inconsistent types.
"""
hist_list = ["date:A_score:num_employees"]
features = ["A_score:num_employees"]
pipeline = Pipeline(modules=[
JsonReader(
file_path=resources.data("example_histogram.json"),
store_key="example_hist",
),
HistSplitter(
read_key="example_hist",
store_key="output_hist",
features=hist_list,
filter_empty_split_hists=False,
),
])
datastore = pipeline.transform(datastore={})
assert "output_hist" in datastore and isinstance(datastore["output_hist"],
dict)
assert len(datastore["output_hist"].keys()) == len(features)
for f in features:
assert f in datastore["output_hist"]
for f in features:
assert isinstance(datastore["output_hist"][f], pd.DataFrame)
for f in features:
df = datastore["output_hist"][f]
split_list = df.reset_index().to_dict("records")
hlist = [s["histogram"] for s in split_list]
check = check_similar_hists(hlist)
assert check is False
