def test_analysis_layout_build(self, small_df, populated_config):
ds = DataSource(small_df, None)
layout = AnalysisLayout([["Scatter"], ["Scatter"]], 400, ds)
root_widget = layout.build()
assert isinstance(root_widget, widgets.VBox)
children = root_widget.children
assert (
len(children) == 4
) # 2 + 2 for HTML(css) and selection types (standard, additive, subtractive)
assert isinstance(children[1], widgets.HBox)
assert isinstance(children[2], widgets.HBox)
assert len(children[1].children) == 1
assert len(children[2].children) == 1
def test_build_row_height_list(self, small_df, populated_config):
ds = DataSource(small_df, None)
build_return = widgets.HBox()
scatterInstance = Mock()
scatterInstance.build = Mock(return_value=build_return)
mockScatterClass = Mock(return_value=scatterInstance)
wcr = WidgetClassRegistry()
wcr.registry["Scatter"] = mockScatterClass
layout = AnalysisLayout([["Scatter"], ["Scatter"]], [300, 200], ds)
layout.build()
call1 = call(ANY, ANY, ANY, ANY, 300)
call2 = call(ANY, ANY, ANY, ANY, 200)
mockScatterClass.assert_has_calls([call1, call2], any_order=False)
def test_change_to_std(self, small_df):
ds = DataSource(small_df, None)
layout = AnalysisLayout([["Scatter"], ["Scatter"]], 400, ds)
layout.selection_type_widget.value = (
"add" # change first so observe gets triggered
)
layout.selection_type_widget.value = "std"
assert layout.data_source.selection_type == SelectionType.STANDARD
def __init__(
self,
data: typing.Union[DataFrame, DataSource],
layout: typing.Union[str, typing.List[typing.List[str]]] = "default",
categorical_columns: typing.Union[typing.List[str], None] = None,
row_height: typing.Union[int, typing.List[int]] = 400,
sample: typing.Union[float, int, None] = None,
select_color: typing.Union[str, typing.Tuple[int, int, int]] = "#323EEC",
deselect_color: typing.Union[str, typing.Tuple[int, int, int]] = "#8A8C93",
alpha: float = 0.75,
seed: typing.Union[int, None] = None,
):
"""
:param data: A pandas.DataFrame object or a :class:`DataSource`.
:param layout: Layout specification name or explicit definition of widget names in rows.
Those columns have to include all columns of the DataFrame
which have type `object`, `str`, `bool` or `category`.
This means it can only add columns which do not have the aforementioned types.
Defaults to 'default'.
:param categorical_columns: If given, specifies which columns are to be interpreted as categorical.
Defaults to None.
:param row_height: Height in pixels each row should have. If given an integer, each row has the height
specified by that value, if given a list of integers, each value in the list specifies the height of
the corresponding row.
Defaults to 400.
:param sample: Int or float value specifying if the DataFrame should be sub-sampled.
When an int is given, the DataFrame will be limited to that number of rows given by the value.
When a float is given, the DataFrame will include the fraction of rows given by the value.
Defaults to None.
:param select_color: RGB tuple or hex color specifying the color display selected data points.
Values in the tuple have to be between 0 and 255 inclusive or a hex string that converts to
such RGB values.
Defaults to '#323EEC'.
:param deselect_color: RGB tuple or hex color specifying the color display deselected data points.
Values in the tuple have to be between 0 and 255 inclusive or a hex string that converts to
such RGB values.
Defaults to '#8A8C93'.
:param alpha: Opacity of data points when applicable ranging from 0.0 to 1.0 inclusive. Defaults to 0.75.
:param seed: Random seed used for sampling the data.
Values can be any integer between 0 and 2**32 - 1 inclusive or None.
Defaults to None.
"""
super().__init__()
validate.validate_data(data)
validate.validate_alpha(alpha)
validate.validate_color(select_color)
validate.validate_color(deselect_color)
if isinstance(select_color, str):
self.select_color: typing.Tuple[int, int, int] = hex_to_rgb(select_color)
elif isinstance(select_color, tuple):
self.select_color: typing.Tuple[int, int, int] = select_color
if isinstance(deselect_color, str):
self.deselect_color: typing.Tuple[int, int, int] = hex_to_rgb(
deselect_color
)
elif isinstance(deselect_color, tuple):
self.deselect_color: typing.Tuple[int, int, int] = deselect_color
self.alpha = alpha
self.color_scale = [
[0, "rgb(%d,%d,%d)" % self.deselect_color],
[1, "rgb(%d,%d,%d)" % self.select_color],
]
config = Config()
config["alpha"] = self.alpha
config["select_color"] = self.select_color
config["deselect_color"] = self.deselect_color
config["color_scale"] = self.color_scale
if isinstance(data, DataFrame):
self.data_source = DataSource(
df=data,
categorical_columns=categorical_columns,
sample=sample,
seed=seed,
)
elif isinstance(data, DataSource):
self.data_source = data
self.layout = AnalysisLayout(
layout=layout, row_height=row_height, data_source=self.data_source
)
if self.data_source.few_num_cols and len(self._check_numerical_plots()) != 0:
warnings.warn(
"The passed DataFrame only has %d NUMERICAL column, which is insufficient for some plots "
"like Parallel Coordinates. These plots will not be displayed."
% len(self.data_source.numerical_columns)
)
if self.data_source.few_cat_cols and len(self._check_categorical_plots()) != 0:
warnings.warn(
"The passed DataFrame only has %d CATEGORICAL column, which is insufficient for some plots "
"like Parallel Categories. These plots will not be displayed."
% len(self.data_source.numerical_columns)
)
class VisualAnalysis:
"""
Generate plots that support linked-brushing from a pandas `DataFrame` and display them in Jupyter notebooks.
"""
def __init__(
self,
data: typing.Union[DataFrame, DataSource],
layout: typing.Union[str, typing.List[typing.List[str]]] = "default",
categorical_columns: typing.Union[typing.List[str], None] = None,
row_height: typing.Union[int, typing.List[int]] = 400,
sample: typing.Union[float, int, None] = None,
select_color: typing.Union[str, typing.Tuple[int, int, int]] = "#323EEC",
deselect_color: typing.Union[str, typing.Tuple[int, int, int]] = "#8A8C93",
alpha: float = 0.75,
seed: typing.Union[int, None] = None,
):
"""
:param data: A pandas.DataFrame object or a :class:`DataSource`.
:param layout: Layout specification name or explicit definition of widget names in rows.
Those columns have to include all columns of the DataFrame
which have type `object`, `str`, `bool` or `category`.
This means it can only add columns which do not have the aforementioned types.
Defaults to 'default'.
:param categorical_columns: If given, specifies which columns are to be interpreted as categorical.
Defaults to None.
:param row_height: Height in pixels each row should have. If given an integer, each row has the height
specified by that value, if given a list of integers, each value in the list specifies the height of
the corresponding row.
Defaults to 400.
:param sample: Int or float value specifying if the DataFrame should be sub-sampled.
When an int is given, the DataFrame will be limited to that number of rows given by the value.
When a float is given, the DataFrame will include the fraction of rows given by the value.
Defaults to None.
:param select_color: RGB tuple or hex color specifying the color display selected data points.
Values in the tuple have to be between 0 and 255 inclusive or a hex string that converts to
such RGB values.
Defaults to '#323EEC'.
:param deselect_color: RGB tuple or hex color specifying the color display deselected data points.
Values in the tuple have to be between 0 and 255 inclusive or a hex string that converts to
such RGB values.
Defaults to '#8A8C93'.
:param alpha: Opacity of data points when applicable ranging from 0.0 to 1.0 inclusive. Defaults to 0.75.
:param seed: Random seed used for sampling the data.
Values can be any integer between 0 and 2**32 - 1 inclusive or None.
Defaults to None.
"""
super().__init__()
validate.validate_data(data)
validate.validate_alpha(alpha)
validate.validate_color(select_color)
validate.validate_color(deselect_color)
if isinstance(select_color, str):
self.select_color: typing.Tuple[int, int, int] = hex_to_rgb(select_color)
elif isinstance(select_color, tuple):
self.select_color: typing.Tuple[int, int, int] = select_color
if isinstance(deselect_color, str):
self.deselect_color: typing.Tuple[int, int, int] = hex_to_rgb(
deselect_color
)
elif isinstance(deselect_color, tuple):
self.deselect_color: typing.Tuple[int, int, int] = deselect_color
self.alpha = alpha
self.color_scale = [
[0, "rgb(%d,%d,%d)" % self.deselect_color],
[1, "rgb(%d,%d,%d)" % self.select_color],
]
config = Config()
config["alpha"] = self.alpha
config["select_color"] = self.select_color
config["deselect_color"] = self.deselect_color
config["color_scale"] = self.color_scale
if isinstance(data, DataFrame):
self.data_source = DataSource(
df=data,
categorical_columns=categorical_columns,
sample=sample,
seed=seed,
)
elif isinstance(data, DataSource):
self.data_source = data
self.layout = AnalysisLayout(
layout=layout, row_height=row_height, data_source=self.data_source
)
if self.data_source.few_num_cols and len(self._check_numerical_plots()) != 0:
warnings.warn(
"The passed DataFrame only has %d NUMERICAL column, which is insufficient for some plots "
"like Parallel Coordinates. These plots will not be displayed."
% len(self.data_source.numerical_columns)
)
if self.data_source.few_cat_cols and len(self._check_categorical_plots()) != 0:
warnings.warn(
"The passed DataFrame only has %d CATEGORICAL column, which is insufficient for some plots "
"like Parallel Categories. These plots will not be displayed."
% len(self.data_source.numerical_columns)
)
def _ipython_display_(self):
"""
Builds the layout and calls :func:`IPython.core.display.display`
:return:
"""
from IPython.core.display import display
from ipywidgets import widgets
root_widget: widgets.Widget = self.layout.build()
# noinspection PyTypeChecker
display(root_widget)
def _check_numerical_plots(self) -> typing.List[str]:
"""
Checks if the layout contains widgets that can only display numerical data.
:return: Set of widgets in the layout that are strictly numerical. Empty set otherwise.
"""
numerical_plots = {"ParallelCoordinates"}
found_plots = set()
for row in self.layout.layout_spec:
for el in row:
if el in numerical_plots:
found_plots.add(el)
return list(found_plots)
def _check_categorical_plots(self) -> typing.List[str]:
"""
Checks if the layout contains widgets that can only display categorical data.
:return: Set of widgets in the layout that are strictly categorical. Empty set otherwise.
"""
numerical_plots = {"ParallelCategories"}
found_plots = set()
for row in self.layout.layout_spec:
for el in row:
if el in numerical_plots:
found_plots.add(el)
return list(found_plots)
@staticmethod
def widgets():
"""
:return: All the widget names that are available as input to a layout.
"""
return WidgetClassRegistry().widget_list
def test_type_error(self, small_df):
with pytest.raises(TypeError):
AnalysisLayout("default", "400px", DataSource(small_df))
def test_int_negative(self, small_df):
with pytest.raises(ValueError):
AnalysisLayout("default", -200, DataSource(small_df))
def test_int_normal(self, small_df):
AnalysisLayout("default", 400, DataSource(small_df))
def test_analysis_layout_incorrect_widget_name(self, small_df):
with pytest.raises(ValueError):
AnalysisLayout([["asdfasdf"]], 400, DataSource(small_df, None))
def test_analysis_layout_wrong_layout_type_error(self, small_df):
with pytest.raises(TypeError):
AnalysisLayout(set(["A", "B"]), 400, DataSource(small_df, None))
def test_analysis_layout_wrong_predefined_layout_error(self, small_df):
with pytest.raises(ValueError):
AnalysisLayout("some_unknown_layout", 400,
DataSource(small_df, None))
def test_analysis_layout_one_num_col_no_warning(small_df):
assert AnalysisLayout([["Scatter"]], 400, DataSource(small_df, None))
def test_change_to_sub(self, small_df):
ds = DataSource(small_df, None)
layout = AnalysisLayout([["Scatter"], ["Scatter"]], 400, ds)
layout.selection_type_widget.value = "sub"
assert layout.data_source.selection_type == SelectionType.SUBTRACTIVE
def test_std_standard(self, small_df):
ds = DataSource(small_df, None)
layout = AnalysisLayout([["Scatter"], ["Scatter"]], 400, ds)
assert layout.data_source.selection_type == SelectionType.STANDARD
def test_list_negative(self, small_df):
with pytest.raises(ValueError):
AnalysisLayout([["Scatter"], ["Scatter"]], [200, -300],
DataSource(small_df))
def test_list_normal(self, small_df):
AnalysisLayout([["Scatter"], ["Scatter"]], [200, 300],
DataSource(small_df))