def test_Instance(self):
p = Instance(HasProps)
with pytest.raises(ValueError) as e:
p.validate("junk")
assert not str(e).endswith("ValueError")
class Mixin(HasProps):
mixin_num = Int(12)
mixin_container = List(String)
mixin_child = Instance(HasProps)
class SomeModelInTestClientServer(Model):
foo = Int(2)
child = Instance(Model)
class DrawTool(Tool):
__implementation__ = JS_CODE
source = Instance(ColumnDataSource)
class AttrSpec(HasProps):
"""A container for assigning attributes to values and retrieving them as needed.
A special function this provides is automatically handling cases where the provided
iterator is too short compared to the distinct values provided.
Once created as attr_spec, you can do attr_spec[data_label], where data_label must
be a one dimensional tuple of values, representing the unique group in the data.
See the :meth:`AttrSpec.setup` method for the primary way to provide an existing
AttrSpec with data and column values and update all derived property values.
"""
data = Instance(ColumnDataSource)
iterable = List(Any, default=None)
attrname = String(help='Name of the attribute the spec provides.')
columns = Either(ColumnLabel,
List(ColumnLabel),
help="""
The label or list of column labels that correspond to the columns that will be
used to find all distinct values (single column) or combination of values (
multiple columns) to then assign a unique attribute to. If not enough unique
attribute values are found, then the attribute values will be cycled.
""")
default = Any(default=None,
help="""
The default value for the attribute, which is used if no column is assigned to
the attribute for plotting. If the default value is not provided, the first
value in the `iterable` property is used.
""")
attr_map = Dict(Any,
Any,
help="""
Created by the attribute specification when `iterable` and `data` are
available. The `attr_map` will include a mapping between the distinct value(s)
found in `columns` and the attribute value that has been assigned.
""")
items = Any(default=None,
help="""
The attribute specification calculates this list of distinct values that are
found in `columns` of `data`.
""")
sort = Bool(default=True,
help="""
A boolean flag to tell the attribute specification to sort `items`, when it is
calculated. This affects which value of `iterable` is assigned to each distinct
value in `items`.
""")
ascending = Bool(default=True,
help="""
A boolean flag to tell the attribute specification how to sort `items` if the
`sort` property is set to `True`. The default setting for `ascending` is `True`.
""")
bins = Instance(Bins,
help="""
If an attribute spec is binning data, so that we can map one value in the
`iterable` to one value in `items`, then this attribute will contain an instance
of the Bins stat. This is used to create unique labels for each bin, which is
then used for `items` instead of the actual unique values in `columns`.
""")
def __init__(self,
columns=None,
df=None,
iterable=None,
default=None,
items=None,
**properties):
"""Create a lazy evaluated attribute specification.
Args:
columns: a list of column labels
df(:class:`~pandas.DataFrame`): the data source for the attribute spec.
iterable: an iterable of distinct attribute values
default: a value to use as the default attribute when no columns are passed
items: the distinct values in columns. If items is provided as input,
then the values provided are used instead of being calculated. This can
be used to force a specific order for assignment.
**properties: other properties to pass to parent :class:`HasProps`
"""
properties['columns'] = self._ensure_list(columns)
if df is not None:
properties['data'] = ColumnDataSource(df)
if default is None and iterable is not None:
default_iter = copy(iterable)
properties['default'] = next(iter(default_iter))
elif default is not None:
properties['default'] = default
if iterable is not None:
properties['iterable'] = iterable
if items is not None:
properties['items'] = items
super(AttrSpec, self).__init__(**properties)
if self.default is None and self.iterable is not None:
self.default = next(iter(copy(self.iterable)))
if self.data is not None and self.columns is not None:
if df is None:
df = self.data.to_df()
self._generate_items(df, columns=self.columns)
if self.items is not None and self.iterable is not None:
self.attr_map = self._create_attr_map()
@staticmethod
def _ensure_list(attr):
"""Always returns a list with the provided value. Returns the value if a list."""
if isinstance(attr, str):
return [attr]
elif isinstance(attr, tuple):
return list(attr)
else:
return attr
@staticmethod
def _ensure_tuple(attr):
"""Return tuple with the provided value. Returns the value if a tuple."""
if not isinstance(attr, tuple):
return (attr, )
else:
return attr
def _setup_default(self):
"""Stores the first value of iterable into `default` property."""
self.default = next(self._setup_iterable())
def _setup_iterable(self):
"""Default behavior is to copy and cycle the provided iterable."""
return cycle(copy(self.iterable))
def _generate_items(self, df, columns):
"""Produce list of unique tuples that identify each item."""
if self.sort:
# TODO (fpliger): this handles pandas API change so users do not experience
# the related annoying deprecation warning. This is probably worth
# removing when pandas deprecated version (0.16) is "old" enough
try:
df = df.sort_values(by=columns, ascending=self.ascending)
except AttributeError:
df = df.sort(columns=columns, ascending=self.ascending)
items = df[columns].drop_duplicates()
self.items = [tuple(x) for x in items.to_records(index=False)]
def _create_attr_map(self, df=None, columns=None):
"""Creates map between unique values and available attributes."""
if df is not None and columns is not None:
self._generate_items(df, columns)
iterable = self._setup_iterable()
return {item: next(iterable) for item in self._item_tuples()}
def _item_tuples(self):
return [self._ensure_tuple(item) for item in self.items]
def set_columns(self, columns):
"""Set columns property and update derived properties as needed."""
columns = self._ensure_list(columns)
if all([col in self.data.column_names for col in columns]):
self.columns = columns
else:
# we have input values other than columns
# assume this is now the iterable at this point
self.iterable = columns
self._setup_default()
def setup(self, data=None, columns=None):
"""Set the data and update derived properties as needed."""
if data is not None:
self.data = data
if columns is not None and self.data is not None:
self.set_columns(columns)
if self.columns is not None and self.data is not None:
self.attr_map = self._create_attr_map(self.data.to_df(),
self.columns)
def update_data(self, data):
self.setup(data=data, columns=self.columns)
def __getitem__(self, item):
"""Lookup the attribute to use for the given unique group label."""
if not self.attr_map:
return self.default
elif self._ensure_tuple(item) not in self.attr_map.keys():
# make sure we have attr map
self.setup()
return self.attr_map[self._ensure_tuple(item)]
@property
def series(self):
if not self.attr_map:
return pd.Series()
else:
index = pd.MultiIndex.from_tuples(self._item_tuples(),
names=self.columns)
return pd.Series(list(self.attr_map.values()), index=index)
class Interval(AggregateGlyph):
"""A rectangle representing aggregated values.
The interval is a rect glyph where two of the parallel sides represent a
summary of values. Each of the two sides is derived from a separate aggregation of
the values provided to the interval.
.. note::
A bar is a special case interval where one side is pinned and used to
communicate a value relative to it.
"""
width = Float(default=0.8)
start_agg = Either(Instance(Stat),
Enum(*list(stats.keys())),
default=Min(),
help="""
The stat used to derive the starting point of the composite glyph.""")
end_agg = Either(Instance(Stat),
Enum(*list(stats.keys())),
default=Max(),
help="""
The stat used to derive the end point of the composite glyph.""")
start = Float(default=0.0)
end = Float()
def __init__(self, label, values, **kwargs):
kwargs['label'] = label
kwargs['values'] = values
super(Interval, self).__init__(**kwargs)
self.setup()
def get_start(self):
"""Get the value for the start of the glyph."""
if len(self.values.index) == 1:
self.start_agg = None
return self.values[0]
elif isinstance(self.start_agg, str):
self.start_agg = stats[self.start_agg]()
self.start_agg.set_data(self.values)
return self.start_agg.value
def get_end(self):
"""Get the value for the end of the glyph."""
if len(self.values.index
) == 1 and not self.values.dtype.name == 'object':
self.end_agg = None
return self.values[0]
elif isinstance(self.end_agg, str):
self.end_agg = stats[self.end_agg]()
self.end_agg.set_data(self.values)
return self.end_agg.value
def get_span(self):
"""The total range between the start and end."""
return self.end - self.start
def build_source(self):
# ToDo: Handle rotation
self.start = self.get_start()
self.end = self.get_end()
self.span = self.get_span()
width = [self.width]
if self.dodge_shift is not None:
x = [self.get_dodge_label()]
else:
x = [self.x_label]
height = [self.span]
y = [self.stack_shift + (self.span / 2.0) + self.start]
color = [self.color]
fill_alpha = [self.fill_alpha]
line_color = [self.line_color]
line_alpha = [self.line_alpha]
return dict(x=x,
y=y,
width=width,
height=height,
color=color,
fill_alpha=fill_alpha,
line_color=line_color,
line_alpha=line_alpha)
@property
def x_max(self):
"""The maximum extent of the glyph in x.
.. note::
Dodging the glyph can affect the value.
"""
return (self.dodge_shift or self.x_label_value) + (self.width / 2.0)
@property
def x_min(self):
"""The maximum extent of the glyph in y.
.. note::
Dodging the glyph can affect the value.
"""
return (self.dodge_shift or self.x_label_value) - (self.width / 2.0)
@property
def y_max(self):
"""Maximum extent of all `Glyph`s.
How much we are stacking + the height of the interval + the base of the interval
.. note::
the start and end of the glyph can swap between being associated with the
min and max when the glyph end represents a negative value.
"""
return max(self.bottom, self.top)
@property
def y_min(self):
"""The minimum extent of all `Glyph`s in y.
.. note::
the start and end of the glyph can swap between being associated with the
min and max when the glyph end represents a negative value.
"""
return min(self.bottom, self.top)
@property
def bottom(self):
"""The value associated with the start of the stacked glyph."""
return self.stack_shift + self.start
@property
def top(self):
"""The value associated with the end of the stacked glyph."""
return self.stack_shift + self.span + self.start
def build_renderers(self):
"""Yields a `GlyphRenderer` associated with a `Rect` glyph."""
glyph = Rect(x='x',
y='y',
width='width',
height='height',
fill_color='color',
fill_alpha='fill_alpha',
line_color='line_color')
yield GlyphRenderer(glyph=glyph)
class HistogramGlyph(AggregateGlyph):
"""Depicts the distribution of values using rectangles created by binning.
The histogram represents a distribution, so will likely include other
options for displaying it, such as KDE and cumulative density.
"""
# input properties
bin_width = Float()
bin_count = Float(
help="""Provide a manually specified number of bins to use.""")
# derived models
bins = Instance(Bins,
help="""A stat used to calculate the bins. The bins stat
includes attributes about each composite bin.""")
bars = List(Instance(BarGlyph),
help="""The histogram is comprised of many
BarGlyphs that are derived from the values.""")
def __init__(self,
values,
label=None,
color=None,
bin_count=None,
**kwargs):
if label is not None:
kwargs['label'] = label
kwargs['values'] = values
kwargs['bin_count'] = bin_count
if color is not None:
kwargs['color'] = color
# remove width, since this is handled automatically
kwargs.pop('width', None)
super(HistogramGlyph, self).__init__(**kwargs)
self.setup()
def _set_sources(self):
# No need to set sources, since composite glyphs handle this
pass
def build_source(self):
# No need to build source, since composite glyphs handle this
return None
def build_renderers(self):
"""Yield a bar glyph for each bin."""
self.bins = Bins(values=self.values, bin_count=self.bin_count)
centers = [bin.center for bin in self.bins.bins]
self.bin_width = centers[1] - centers[0]
bars = []
for bin in self.bins.bins:
bars.append(
BarGlyph(label=self.label,
x_label=bin.center,
values=bin.values,
color=self.color,
fill_alpha=self.fill_alpha,
agg=bin.stat,
width=self.bin_width))
# provide access to bars as children for bounds properties
self.bars = bars
self.children = self.bars
for comp_glyph in self.bars:
for renderer in comp_glyph.renderers:
yield renderer
@property
def y_min(self):
return 0.0
class SomeModelInTestPullDoc(Model):
foo = Int(2)
child = Instance(Model)
class LuminoDock(HTMLBox):
children = List(Either(Tuple(String, Instance(LayoutDOM)),
Tuple(String, Instance(LayoutDOM), InsertMode),
Tuple(String, Instance(LayoutDOM), InsertMode,
Int)),
default=[])
def test_Instance(self):
p = Instance(HasProps)
with pytest.raises(ValueError) as e:
p.validate("junk")
assert not str(e).endswith("ValueError")
def test_Instance(self, detail):
p = Instance(HasProps)
with pytest.raises(ValueError) as e:
p.validate("junk", detail)
assert str(e).endswith("ValueError") == (not detail)
class HistogramGlyph(AggregateGlyph):
"""Depicts the distribution of values using rectangles created by binning.
The histogram represents a distribution, so will likely include other
options for displaying it, such as KDE and cumulative density.
"""
# derived models
bins = Instance(BinnedStat,
help="""A stat used to calculate the bins. The bins stat
includes attributes about each composite bin.""")
bars = List(Instance(BarGlyph),
help="""The histogram is comprised of many
BarGlyphs that are derived from the values.""")
density = Bool(False,
help="""
Whether to normalize the histogram.
If True, the result is the value of the probability *density* function
at the bin, normalized such that the *integral* over the range is 1. If
False, the result will contain the number of samples in each bin.
For more info check :class:`~bokeh.charts.stats.Histogram` documentation.
(default: False)
""")
def __init__(self, values, label=None, color=None, bins=None, **kwargs):
if label is not None:
kwargs['label'] = label
kwargs['values'] = values
if color is not None:
kwargs['color'] = color
# remove width, since this is handled automatically
kwargs.pop('width', None)
# keep original bins setting private since it just needs to be
# delegated to the Histogram stat
self._bins = bins
super(HistogramGlyph, self).__init__(**kwargs)
self.setup()
def _set_sources(self):
# No need to set sources, since composite glyphs handle this
pass
def build_source(self):
# No need to build source, since composite glyphs handle this
return None
def build_renderers(self):
"""Yield a bar glyph for each bin."""
# TODO(fpliger): We should expose the bin stat class so we could let
# users specify other bins other the Histogram Stat
self.bins = Histogram(values=self.values,
bins=self._bins,
density=self.density)
bars = []
for bin in self.bins.bins:
bars.append(
BarGlyph(label=bin.label[0],
x_label=bin.center,
values=bin.values,
color=self.color,
fill_alpha=self.fill_alpha,
agg=bin.stat,
width=bin.width))
# provide access to bars as children for bounds properties
self.bars = self.children = bars
for comp_glyph in self.bars:
for renderer in comp_glyph.renderers:
yield renderer
@property
def y_min(self):
return 0.0
class Builder(HasProps):
""" A prototype class to inherit each new chart Builder type.
It provides useful methods to be used by the inherited builder classes,
in order to automate most of the charts creation tasks and leave the
core customization to specialized builder classes. In that pattern
inherited builders just need to provide the following methods:
Required:
* :meth:`~bokeh.charts.builder.Builder.yield_renderers`: yields the glyphs to be
rendered into the plot. Here you should call the
:meth:`~bokeh.charts.builder.Builder.add_glyph` method so that the builder can
setup the legend for you.
* :meth:`~bokeh.charts.builder.Builder.set_ranges`: setup the ranges for the
glyphs. This is called after glyph creation, so you are able to inspect the
comp_glyphs for their minimum and maximum values. See the
:meth:`~bokeh.charts.builder.Builder.create` method for more information on
when this is called and how the builder provides the ranges to the containing
:class:`Chart` using the :meth:`Chart.add_ranges` method.
Optional:
* :meth:`~bokeh.charts.builder.Builder.setup`: provides an area
where subclasses of builder can introspect properties, setup attributes, or change
property values. This is called before
:meth:`~bokeh.charts.builder.Builder.process_data`.
* :meth:`~bokeh.charts.builder.Builder.process_data`: provides an area
where subclasses of builder can manipulate the source data before renderers are
created.
"""
# Optional Inputs
x_range = Instance(Range)
y_range = Instance(Range)
xlabel = String()
ylabel = String()
xscale = String()
yscale = String()
palette = List(Color,
help="""Optional input to override the default palette used
by any color attribute.
""")
# Dimension Configuration
"""
The dimension labels that drive the position of the
glyphs. Subclasses should implement this so that the Builder
base class knows which dimensions it needs to operate on.
An example for a builder working with cartesian x and y
coordinates would be dimensions = ['x', 'y']. You should
then instantiate the x and y dimensions as attributes of the
subclass of builder using the :class:`Dimension
<bokeh.charts.properties.Dimension>` class. One for x, as x
= Dimension(...), and one as y = Dimension(...).
"""
dimensions = None # None because it MUST be overridden
"""
The dimension labels that must exist to produce the
glyphs. This specifies what are the valid configurations for
the chart, with the option of specifying the type of the
columns. The
:class:`~bokeh.charts.data_source.ChartDataSource` will
inspect this property of your subclass of Builder and use
this to fill in any required dimensions if no keyword
arguments are used.
"""
req_dimensions = []
# Attribute Configuration
attributes = Dict(String,
Instance(AttrSpec),
help="""
The attribute specs used to group data. This is a mapping between the role of
the attribute spec (e.g. 'color') and the
:class:`~bokeh.charts.attributes.AttrSpec` class (e.g.,
:class:`~bokeh.charts.attributes.ColorAttr`). The Builder will use this
attributes property during runtime, which will consist of any attribute specs
that are passed into the chart creation function (e.g.,
:class:`~bokeh.charts.Bar`), ones that are created for the user from simple
input types (e.g. `Bar(..., color='red')` or `Bar(..., color=<column_name>)`),
or lastly, the attribute spec found in the default_attributes configured for
the subclass of :class:`~bokeh.charts.builder.Builder`.
""")
"""
The default attribute specs used to group data. This is
where the subclass of Builder should specify what the
default attributes are that will yield attribute values to
each group of data, and any specific configuration. For
example, the :class:`ColorAttr` utilizes a default palette
for assigning color based on groups of data. If the user
doesn't assign a column of the data to the associated
attribute spec, then the default attrspec is used, which
will yield a constant color value for each group of
data. This is by default the first color in the default
palette, but can be customized by setting the default color
in the ColorAttr.
"""
default_attributes = None # None because it MUST be overridden
# Derived properties (created by Builder at runtime)
attribute_columns = List(ColumnLabel,
help="""
All columns used for specifying attributes for the Chart. The Builder will set
this value on creation so that the subclasses can know the distinct set of columns
that are being used to assign attributes.
""")
comp_glyphs = List(Instance(CompositeGlyph),
help="""
A list of composite glyphs, where each represents a unique subset of data. The
composite glyph is a helper class that encapsulates all low level
:class:`~bokeh.models.glyphs.Glyph`, that represent a higher level group of
data. For example, the :class:`BoxGlyph` is a single class that yields
each :class:`GlyphRenderer` needed to produce a Box on a :class:`BoxPlot`. The
single Box represents a full array of values that are aggregated, and is made
up of multiple :class:`~bokeh.models.glyphs.Rect` and
:class:`~bokeh.models.glyphs.Segment` glyphs.
""")
labels = List(
String,
help="""Represents the unique labels to be used for legends.""")
"""List of attributes to use for legends."""
label_attributes = []
"""
Used to assign columns to dimensions when no selections have been provided. The
default behavior is provided by the :class:`OrderedAssigner`, which assigns
a single column to each dimension available in the `Builder`'s `dims` property.
"""
column_selector = OrderedAssigner
comp_glyph_types = List(Instance(CompositeGlyph))
sort_dim = Dict(String, Bool, default={})
sort_legend = List(Tuple(String, Bool),
help="""
List of tuples to use for sorting the legend, in order that they should be
used for sorting. This sorting can be different than the sorting used for the
rest of the chart. For example, you might want to sort only on the column
assigned to the color attribute, or sort it descending. The order of each tuple
is (Column, Ascending).
""")
legend_sort_field = String(help="""
Attribute that should be used to sort the legend, for example: color,
dash, maker, etc. Valid values for this property depend on the type
of chart.
""")
legend_sort_direction = Enum(SortDirection,
help="""
Sort direction to apply to :attr:`~bokeh.charts.builder.Builder.sort_legend`.
Valid values are: `ascending` or `descending`.
""")
source = Instance(ColumnDataSource)
tooltips = Either(List(Tuple(String, String)),
List(String),
Bool,
default=None,
help="""
Tells the builder to add tooltips to the chart by either using the columns
specified to the chart attributes (True), or by generating tooltips for each
column specified (list(str)), or by explicit specification of the tooltips
using the valid input for the `HoverTool` tooltips kwarg.
""")
__deprecated_attributes__ = ('sort_legend', )
def __init__(self, *args, **kws):
"""Common arguments to be used by all the inherited classes.
Args:
data (:ref:`userguide_charts_data_types`): source data for the chart
legend (str, bool): the legend of your plot. The legend content is
inferred from incoming input.It can be ``top_left``,
``top_right``, ``bottom_left``, ``bottom_right``.
It is ``top_right`` is you set it as True.
Attributes:
source (obj): datasource object for your plot,
initialized as a dummy None.
x_range (obj): x-associated datarange object for you plot,
initialized as a dummy None.
y_range (obj): y-associated datarange object for you plot,
initialized as a dummy None.
groups (list): to be filled with the incoming groups of data.
Useful for legend construction.
data (dict): to be filled with the incoming data and be passed
to the ChartDataSource for each Builder class.
attr (list(AttrSpec)): to be filled with the new attributes created after
loading the data dict.
"""
data = None
if len(args) != 0 or len(kws) != 0:
# chart dimensions can be literal dimensions or attributes
attrs = list(self.default_attributes.keys())
dims = self.dimensions + attrs
# pop the dimension inputs from kwargs
data_args = {}
for dim in dims:
if dim in kws.keys():
data_args[dim] = kws[dim]
# build chart data source from inputs, given the dimension configuration
data_args['dims'] = tuple(dims)
data_args['required_dims'] = tuple(self.req_dimensions)
data_args['attrs'] = attrs
data_args['column_assigner'] = self.column_selector
data = ChartDataSource.from_data(*args, **data_args)
# make sure that the builder dimensions have access to the chart data source
for dim in self.dimensions:
getattr(getattr(self, dim), 'set_data')(data)
# handle input attrs and ensure attrs have access to data
attributes = self._setup_attrs(data, kws)
# remove inputs handled by dimensions and chart attributes
for dim in dims:
kws.pop(dim, None)
else:
attributes = dict()
kws['attributes'] = attributes
super(Builder, self).__init__(**kws)
# collect unique columns used for attributes
self.attribute_columns = collect_attribute_columns(**self.attributes)
for k in self.__deprecated_attributes__:
if k in kws:
setattr(self, k, kws[k])
self._data = data
self._legends = []
def _setup_attrs(self, data, kws):
"""Handle overridden attributes and initialize them with data.
Makes sure that all attributes have access to the data
source, which is used for mapping attributes to groups
of data.
Returns:
None
"""
source = ColumnDataSource(data.df)
attr_names = self.default_attributes.keys()
custom_palette = kws.get('palette')
attributes = dict()
for attr_name in attr_names:
attr = kws.pop(attr_name, None)
# if given an attribute use it
if isinstance(attr, AttrSpec):
attributes[attr_name] = attr
# if we are given columns, use those
elif isinstance(attr, (str, list)):
attributes[attr_name] = self.default_attributes[
attr_name]._clone()
# override palette if available
if isinstance(attributes[attr_name], ColorAttr):
if custom_palette is not None:
attributes[attr_name].iterable = custom_palette
attributes[attr_name].setup(data=source, columns=attr)
else:
# override palette if available
if (isinstance(self.default_attributes[attr_name], ColorAttr)
and custom_palette is not None):
attributes[attr_name] = self.default_attributes[
attr_name]._clone()
attributes[attr_name].iterable = custom_palette
else:
attributes[attr_name] = self.default_attributes[
attr_name]._clone()
# make sure all have access to data source
for attr_name in attr_names:
attributes[attr_name].update_data(data=source)
return attributes
def setup(self):
"""Perform any initial pre-processing, attribute config.
Returns:
None
"""
pass
def process_data(self):
"""Make any global data manipulations before grouping.
It has to be implemented by any of the inherited class
representing each different chart type. It is the place
where we make specific calculations for each chart.
Returns:
None
"""
pass
def yield_renderers(self):
""" Generator that yields the glyphs to be draw on the plot
It has to be implemented by any of the inherited class
representing each different chart type.
Yields:
:class:`GlyphRenderer`
"""
raise NotImplementedError(
'Subclasses of %s must implement _yield_renderers.' %
self.__class__.__name__)
def set_ranges(self):
"""Calculate and set the x and y ranges.
It has to be implemented by any of the subclasses of builder
representing each different chart type, and is called after
:meth:`yield_renderers`.
Returns:
None
"""
raise NotImplementedError(
'Subclasses of %s must implement _set_ranges.' %
self.__class__.__name__)
def get_dim_extents(self):
"""Helper method to retrieve maximum extents of all the renderers.
Returns:
a dict mapping between dimension and value for x_max, y_max, x_min, y_min
"""
return {
'x_max': max([renderer.x_max for renderer in self.comp_glyphs]),
'y_max': max([renderer.y_max for renderer in self.comp_glyphs]),
'x_min': min([renderer.x_min for renderer in self.comp_glyphs]),
'y_min': min([renderer.y_min for renderer in self.comp_glyphs])
}
def add_glyph(self, group, glyph):
"""Add a composite glyph.
Manages the legend, since the builder might not want all attribute types
used for the legend.
Args:
group (:class:`DataGroup`): the data the `glyph` is associated with
glyph (:class:`CompositeGlyph`): the glyph associated with the `group`
Returns:
None
"""
if isinstance(glyph, list):
for sub_glyph in glyph:
self.comp_glyphs.append(sub_glyph)
else:
self.comp_glyphs.append(glyph)
# handle cases where builders have specified which attributes to use for labels
label = None
if len(self.label_attributes) > 0:
for attr in self.label_attributes:
# this will get the last attribute group label for now
if self.attributes[attr].columns is not None:
label = self._get_group_label(group, attr=attr)
# if no special case for labeling, just use the group label
if label is None:
label = self._get_group_label(group, attr='label')
# add to legend if new and unique label
if str(label) not in self.labels and label is not None:
self._legends.append((label, glyph.renderers))
self.labels.append(label)
def _get_group_label(self, group, attr='label'):
"""Get the label of the group by the attribute name.
Args:
group (:attr:`DataGroup`: the group of data
attr (str, optional): the attribute name containing the label, defaults to
'label'.
Returns:
str: the label for the group
"""
if attr is 'label':
label = group.label
else:
label = group[attr]
if isinstance(label, dict):
label = tuple(label.values())
return self._get_label(label)
@staticmethod
def _get_label(raw_label):
"""Converts a label by string or tuple to a string representation.
Args:
raw_label (str or tuple(any, any)): a unique identifier for the data group
Returns:
str: a label that is usable in charts
"""
# don't convert None type to string so we can test for it later
if raw_label is None:
return None
if (isinstance(raw_label, tuple) or isinstance(raw_label, list)) and \
len(raw_label) == 1:
raw_label = raw_label[0]
elif isinstance(raw_label, dict):
raw_label = label_from_index_dict(raw_label)
return str(raw_label)
def collect_attr_kwargs(self):
if hasattr(super(self.__class__, self), 'default_attributes'):
attrs = set(self.default_attributes.keys()) - set(
(super(self.__class__, self).default_attributes or {}).keys())
else:
attrs = set()
return attrs
def get_group_kwargs(self, group, attrs):
return {attr: group[attr] for attr in attrs}
def create(self, chart=None):
"""Builds the renderers, adding them and other components to the chart.
Args:
chart (:class:`Chart`, optional): the chart that will contain the glyph
renderers that the `Builder` produces.
Returns:
:class:`Chart`
"""
# call methods that allow customized setup by subclasses
self.setup()
self.process_data()
# create and add renderers to chart
renderers = self.yield_renderers()
if chart is None:
chart = Chart()
chart.add_renderers(self, renderers)
# handle ranges after renders, since ranges depend on aggregations
# ToDo: should reconsider where this occurs
self.set_ranges()
chart.add_ranges('x', self.x_range)
chart.add_ranges('y', self.y_range)
# sort the legend if we are told to
self._legends = self._sort_legend(self.legend_sort_field,
self.legend_sort_direction,
self._legends, self.attributes)
# always contribute legends, let Chart sort it out
chart.add_legend(self._legends)
chart.add_labels('x', self.xlabel)
chart.add_labels('y', self.ylabel)
chart.add_scales('x', self.xscale)
chart.add_scales('y', self.yscale)
if self.tooltips is not None:
tooltips = build_hover_tooltips(hover_spec=self.tooltips,
chart_cols=self.attribute_columns)
chart.add_tooltips(tooltips)
return chart
@classmethod
def generate_help(cls):
help_str = ''
for comp_glyph in cls.comp_glyph_types:
help_str += str(comp_glyph.glyph_properties())
return help_str
@staticmethod
def _sort_legend(legend_sort_field, legend_sort_direction, legends,
attributes):
"""Sort legends sorted by looping though sort_legend items (
see :attr:`Builder.sort_legend` for more details)
"""
if legend_sort_field:
if len(attributes[legend_sort_field].columns) > 0:
# TODO(fpliger): attributes should be consistent and not
# need any type checking but for
# the moment it is not, specially when going
# though a process like binning or when data
# is built for HeatMap, Scatter, etc...
item_order = [
x[0] if isinstance(x, tuple) else x
for x in attributes[legend_sort_field].items
]
item_order = [
str(x) if not isinstance(x, string_types) else x
for x in item_order
]
def foo(leg):
return item_order.index(leg[0])
reverse = legend_sort_direction == 'descending'
return list(sorted(legends, key=foo, reverse=reverse))
return legends
@property
def sort_legend(self):
deprecated((0, 12, 0), 'Chart.sort_legend', 'Chart.legend_sort_field')
return [(self.legend_sort_field, self.legend_sort_direction)]
@sort_legend.setter
def sort_legend(self, value):
deprecated((0, 12, 0), 'Chart.sort_legend', 'Chart.legend_sort_field')
self.legend_sort_field, direction = value[0]
if direction:
self.legend_sort_direction = "ascending"
else:
self.legend_sort_direction = "descending"
class BinnedStat(Stat):
""" Base class for shared functionality accross bins and aggregates
dimensions for plotting.
"""
bin_stat = Instance(BinStats,
help="""
A mapping between each dimension and associated binning calculations.
""")
bins = List(Instance(Bin),
help="""
A list of the `Bin` instances that were produced as result of the inputs.
Iterating over `Bins` will iterate over this list. Each `Bin` can be inspected
for metadata about the bin and the values associated with it.
""")
stat = Instance(Stat,
default=Count(),
help="""
The statistical operation to be used on the values in each bin.
""")
bin_column = String()
centers_column = String()
aggregate = Bool(default=True)
bin_values = Bool(default=False)
bin_width = Float()
def __init__(self,
values=None,
column=None,
bins=None,
stat='count',
source=None,
**properties):
if isinstance(stat, str):
stat = stats[stat]()
properties['column'] = column or 'vals'
properties['stat'] = stat
properties['values'] = values
properties['source'] = source
self._bins = bins
super(BinnedStat, self).__init__(**properties)
def _get_stat(self):
stat_kwargs = {}
if self.source is not None:
stat_kwargs['source'] = self.source
stat_kwargs['column'] = self.column
elif self.values is not None:
stat_kwargs['values'] = self.values
stat_kwargs['bins'] = self._bins
return BinStats(**stat_kwargs)
def update(self):
self.bin_stat = self._get_stat()
self.bin_stat.update()
class Stat(HasProps):
"""Represents a statistical operation to summarize a column of data.
Can be computed from either a ColumnLabel with a ColumnDataSource, *or*, a
discrete column of data.
"""
# inputs
column = ColumnLabel(
help="""A column to use for the stat calculation. Required
when providing a ColumnDataSource as input.""")
source = Instance(ColumnDataSource,
help="""One option for providing the data
source for stat calculation.""")
values = EitherColumn(Column(Float),
Column(Int),
Column(String),
Column(Date),
Column(Datetime),
Column(Bool),
default=None,
help="""
Second option for providing values for stat calculation is by
passing the actual column of data.""")
# output
value = Float(
help="""The value calculated for the stat. Some stats could use
multiple properties to provide the calculation if required.""")
def __init__(self, **properties):
source = properties.pop('source', None)
if source is not None:
if isinstance(source, pd.DataFrame):
source = ColumnDataSource(source)
properties['source'] = source
super(Stat, self).__init__(**properties)
self._refresh()
def _refresh(self):
"""Lazy update of properties, used for initial transform init."""
if self.get_data() is not None:
self.update()
self.calculate()
def set_data(self, data, column=None):
"""Set data properties and update all dependent properties."""
if isinstance(data, pd.DataFrame):
data = ColumnDataSource(data)
if isinstance(data, ColumnDataSource):
self.source = data
if column is not None:
self.column = column
else:
self.values = data
self.update()
self.calculate()
def get_data(self, column=None):
"""Returns the available columnlabel/source values or column values."""
if self.source is not None and (self.column is not None
or column is not None):
if column is not None:
col = column
else:
col = self.column
return pd.Series(self.source.data[col])
elif self.values is None and self.source is not None:
return pd.Series(self.source.to_df().index)
elif self.values is not None:
return self.values
else:
return None
def calculate(self):
"""Return transformed value from column label/source or column-like data."""
raise NotImplementedError('You must implement the calculate method '
'for each stat type.')
def update(self):
"""Perform any initial work before the actual calculation is performed."""
pass
class DrawTool(Tool):
__implementation__ = TypeScript(JS_CODE)
source = Instance(ColumnDataSource)
class AggregateGlyph(NestedCompositeGlyph):
"""A base composite glyph for aggregating an array.
Implements default stacking and dodging behavior that other composite
glyphs can inherit.
"""
x_label = String()
x_label_value = Any()
stack_label = String()
stack_shift = Float(default=0.0)
dodge_label = String(
help="""Where on the scale the glyph should be placed.""")
dodge_shift = Float(default=None)
agg = Instance(Stat, default=Sum())
span = Float(help="""The range of values represented by the aggregate.""")
def __init__(self, x_label=None, **kwargs):
label = kwargs.get('label')
if x_label is not None:
kwargs['x_label_value'] = x_label
if not isinstance(x_label, str):
x_label = str(x_label)
kwargs['x_label'] = x_label
elif label is not None:
kwargs['x_label'] = str(label)
super(AggregateGlyph, self).__init__(**kwargs)
def get_dodge_label(self, shift=0.0):
"""Generate the label defining an offset in relation to a position on a scale."""
if self.dodge_shift is None:
shift_str = ':' + str(0.5 + shift)
elif self.dodge_shift is not None:
shift_str = ':' + str(self.dodge_shift + shift)
else:
shift_str = ''
return str(label_from_index_dict(self.x_label)) + shift_str
def filter_glyphs(self, glyphs):
"""Return only the glyphs that are of the same class."""
return [glyph for glyph in glyphs if isinstance(glyph, self.__class__)]
@staticmethod
def groupby(glyphs, prop):
"""Returns a dict of `CompositeGlyph`s, grouped by unique values of prop.
For example, if all glyphs had a value of 'a' or 'b' for glyph.prop, the dict
would contain two keys, 'a' and 'b', where each value is a list of the glyphs
that had each of the values.
"""
grouped = defaultdict(list)
labels = [getattr(glyph, prop) for glyph in glyphs]
labels = [
tuple(label.values()) if isinstance(label, dict) else label
for label in labels
]
[grouped[label].append(glyph) for label, glyph in zip(labels, glyphs)]
labels = pd.Series(labels).drop_duplicates().values
return labels, grouped
def __stack__(self, glyphs):
"""Apply relative shifts to the composite glyphs for stacking."""
filtered_glyphs = self.filter_glyphs(glyphs)
labels, grouped = self.groupby(filtered_glyphs, 'x_label')
for label in labels:
group = grouped[label]
# separate the negative and positive aggregates into separate groups
neg_group = [glyph for glyph in group if glyph.span < 0]
pos_group = [glyph for glyph in group if glyph.span >= 0]
# apply stacking to each group separately
for group in [neg_group, pos_group]:
shift = []
for i, glyph in enumerate(group):
# save off the top of each rect's height
shift.append(glyph.span)
if i > 0:
glyph.stack_shift = sum(shift[0:i])
glyph.refresh()
def __dodge__(self, glyphs):
"""Apply relative shifts to the composite glyphs for dodging."""
if self.dodge_label is not None:
filtered_glyphs = self.filter_glyphs(glyphs)
labels, grouped = self.groupby(filtered_glyphs, 'dodge_label')
# calculate transformations
step = np.linspace(0, 1.0, len(grouped.keys()) + 1, endpoint=False)
width = min(0.2, (1. / len(grouped.keys()))**1.1)
# set bar attributes and re-aggregate
for i, label in enumerate(labels):
group = grouped[label]
for glyph in group:
glyph.dodge_shift = step[i + 1]
glyph.width = width
glyph.refresh()
def test_has_ref(self) -> None:
prop = bcpc.Seq(Int)
assert not prop.has_ref
prop = bcpc.Seq(Instance(_TestModel))
assert prop.has_ref
class BoxGlyph(AggregateGlyph):
"""Summarizes the distribution with a collection of glyphs.
A box glyph produces one "box" for a given array of vales. The box
is made up of multiple other child composite glyphs (intervals,
scatter) and directly produces glyph renderers for the whiskers,
as well.
"""
q1 = Float(help="""Derived value for 25% of all values.""")
q2 = Float(help="""Derived value for 50% of all values.""")
q3 = Float(help="""Derived value for 75% of all values.""")
iqr = Float()
w0 = Float(help='Lower whisker')
w1 = Float(help='Upper whisker')
q2_glyph = Instance(QuartileGlyph)
q3_glyph = Instance(QuartileGlyph)
whisker_glyph = Instance(GlyphRenderer)
outliers = Either(Bool, Instance(PointGlyph))
marker = String(default='circle')
whisker_width = Float(default=0.3)
whisker_line_width = Float(default=2)
whisker_span_line_width = Float(default=2)
whisker_color = String(default='black')
outlier_fill_color = String(default='red')
outlier_line_color = String(default='red')
outlier_size = Float(default=5)
bar_color = String(default='DimGrey')
def __init__(self, label, values, outliers=True, **kwargs):
width = kwargs.pop('width', None)
bar_color = kwargs.pop('color', None) or kwargs.get(
'bar_color', None) or self.lookup('bar_color').class_default()
kwargs['outliers'] = kwargs.pop('outliers', None) or outliers
kwargs['label'] = label
kwargs['values'] = values
x_label = kwargs.get('x_label')
kwargs['q2_glyph'] = QuartileGlyph(label=label,
x_label=x_label,
values=values,
interval1=0.25,
interval2=0.5,
width=width,
color=bar_color)
kwargs['q3_glyph'] = QuartileGlyph(label=label,
x_label=x_label,
values=values,
interval1=0.5,
interval2=0.75,
width=width,
color=bar_color)
super(BoxGlyph, self).__init__(**kwargs)
self.setup()
def build_renderers(self):
"""Yields all renderers that make up the BoxGlyph."""
self.calc_quartiles()
outlier_values = self.values[((self.values < self.w0) |
(self.values > self.w1))]
self.whisker_glyph = GlyphRenderer(
glyph=Segment(x0='x0s',
y0='y0s',
x1='x1s',
y1='y1s',
line_width=self.whisker_line_width,
line_color=self.whisker_color))
if len(outlier_values) > 0 and self.outliers:
self.outliers = PointGlyph(label=self.label,
y=outlier_values,
x=[self.get_dodge_label()] *
len(outlier_values),
line_color=self.outlier_line_color,
fill_color=self.outlier_fill_color,
size=self.outlier_size,
marker=self.marker)
for comp_glyph in self.composite_glyphs:
for renderer in comp_glyph.renderers:
yield renderer
yield self.whisker_glyph
def calc_quartiles(self):
"""Sets all derived stat properties of the BoxGlyph."""
self.q1 = self.q2_glyph.start
self.q2 = self.q2_glyph.end
self.q3 = self.q3_glyph.end
self.iqr = self.q3 - self.q1
mx = Max()
mx.set_data(self.values)
mn = Min()
mn.set_data(self.values)
self.w0 = max(self.q1 - (1.5 * self.iqr), mn.value)
self.w1 = min(self.q3 + (1.5 * self.iqr), mx.value)
def build_source(self):
"""Calculate stats and builds and returns source for whiskers."""
self.calc_quartiles()
x_label = self.get_dodge_label()
x_w0_label = self.get_dodge_label(shift=(self.whisker_width / 2.0))
x_w1_label = self.get_dodge_label(shift=-(self.whisker_width / 2.0))
# span0, whisker bar0, span1, whisker bar1
x0s = [x_label, x_w0_label, x_label, x_w0_label]
y0s = [self.w0, self.w0, self.q3, self.w1]
x1s = [x_label, x_w1_label, x_label, x_w1_label]
y1s = [self.q1, self.w0, self.w1, self.w1]
return dict(x0s=x0s, y0s=y0s, x1s=x1s, y1s=y1s)
def _set_sources(self):
"""Set the column data source on the whisker glyphs."""
self.whisker_glyph.data_source = self.source
def get_extent(self, func, prop_name):
return func([
getattr(renderer, prop_name) for renderer in self.composite_glyphs
])
@property
def composite_glyphs(self):
"""Returns list of composite glyphs, excluding the regular glyph renderers."""
comp_glyphs = [self.q2_glyph, self.q3_glyph]
if isinstance(self.outliers, PointGlyph):
comp_glyphs.append(self.outliers)
return comp_glyphs
@property
def x_max(self):
return self.get_extent(max, 'x_max') + self.right_buffer
@property
def x_min(self):
return self.get_extent(min, 'x_min') - self.left_buffer
@property
def y_max(self):
return max(self.w1, self.get_extent(max, 'y_max')) + self.top_buffer
@property
def y_min(self):
return min(self.w0, self.get_extent(min, 'y_min')) - self.bottom_buffer
class OtherModel(Model):
child = Instance(Model)
class SomeModel(Model):
data = ColumnData(Any, Any, default={})
ref1 = Instance(OtherModel, default=lambda: OtherModel())
ref2 = Instance(OtherModel, default=lambda: OtherModel())
class SomeModelInTestDocument(Model):
foo = Int(2)
child = Nullable(Instance(Model))
def test_Instance(self) -> None:
p = Instance(HasProps)
with pytest.raises(ValueError) as e:
p.validate("junk")
assert matches(str(e.value), r"expected an instance of type HasProps, got junk of type str")
def test_has_ref(self) -> None:
prop0 = bcpn.Nullable(Int)
assert not prop0.has_ref
prop1 = bcpn.Nullable(Instance(_TestModel))
assert prop1.has_ref
class Base(HasProps):
num = Int(12)
container = List(String)
child = Instance(HasProps)
def test_has_ref(self):
prop = bcpc.Dict(String, Int)
assert not prop.has_ref
prop = bcpc.Dict(String, Instance(_TestModel))
assert prop.has_ref
class Sub(Base, Mixin):
sub_num = Int(12)
sub_container = List(String)
sub_child = Instance(HasProps)
def test_has_ref(self):
prop = bcpc.Tuple(Int, Int)
assert not prop.has_ref
prop = bcpc.Tuple(Int, Instance(_TestModel))
assert prop.has_ref
def test_Instance(self, detail) -> None:
p = Instance(HasProps)
with pytest.raises(ValueError) as e:
p.validate("junk", detail)
assert (str(e.value) == "") == (not detail)
class BinGlyph(XyGlyph):
"""Represents a group of data that was aggregated and is represented by a glyph.
"""
bins = Instance(Bins)
column = String()
stat = String()
glyph_name = String()
width = Float()
height = Float()
def __init__(self,
x,
y,
values,
column=None,
stat='count',
glyph='rect',
width=1,
height=1,
**kwargs):
df = pd.DataFrame(dict(x_vals=x, y_vals=y, values_vals=values))
df.drop_duplicates(inplace=True)
kwargs['x'] = df.x_vals
kwargs['y'] = df.y_vals
kwargs['values'] = df.values_vals
kwargs['column'] = column
kwargs['stat'] = stat
kwargs['glyph_name'] = glyph
kwargs['height'] = height
kwargs['width'] = width
if 'glyphs' not in kwargs:
kwargs['glyphs'] = {'rect': Rect}
super(XyGlyph, self).__init__(**kwargs)
self.setup()
def build_source(self):
return {'x': self.x, 'y': self.y, 'values': self.values}
def build_renderers(self):
glyph_class = self.glyphs[self.glyph_name]
glyph = glyph_class(x='x',
y='y',
height=self.height,
width=self.width,
fill_color=self.fill_color,
line_color=self.line_color,
dilate=True)
yield GlyphRenderer(glyph=glyph)
@property
def x_max(self):
return self.get_data_range('x')[1] + self.width / 2.0
@property
def x_min(self):
return self.get_data_range('x')[0] - self.width / 2.0
@property
def y_max(self):
return self.get_data_range('y')[1] + self.height / 2.0
@property
def y_min(self):
return self.get_data_range('y')[0] - self.height / 2.0
def get_data_range(self, col):
data = self.source.data[col]
if ChartDataSource.is_number(data):
return min(data), max(data)
else:
return 1, len(data.drop_duplicates())
def test_Instance(self, detail):
p = Instance(HasProps)
with pytest.raises(ValueError) as e:
p.validate("junk", detail)
assert str(e).endswith("ValueError") == (not detail)
class Bin(Stat):
"""Represents a single bin of data values and attributes of the bin."""
label = Either(String, List(String))
start = Either(Float, List(Float))
stop = Either(Float, List(Float))
start_label = String()
stop_label = String()
center = Either(Float, List(Float))
stat = Instance(Stat, default=Count())
width = Float()
def __init__(self, bin_label, values=None, source=None, **properties):
if isinstance(bin_label, tuple):
bin_label = list(bin_label)
else:
bin_label = [bin_label]
properties['label'] = bin_label
bounds = self.process_bounds(bin_label)
starts, stops = zip(*bounds)
centers = [(start + stop) / 2.0 for start, stop in zip(starts, stops)]
if len(starts) == 1:
starts = starts[0]
stops = stops[0]
centers = centers[0]
else:
starts = list(starts)
stops = list(stops)
centers = list(centers)
properties['start'] = starts
properties['stop'] = stops
properties['center'] = centers
properties['values'] = values
super(Bin, self).__init__(**properties)
@staticmethod
def binstr_to_list(bins):
"""Produce a consistent display of a bin of data."""
value_chunks = bins.split(',')
value_chunks = [
val.replace('[', '').replace(']', '').replace('(',
'').replace(')', '')
for val in value_chunks
]
bin_values = [float(value) for value in value_chunks]
return bin_values[0], bin_values[1]
def process_bounds(self, bin_label):
if isinstance(bin_label, list):
return [self.binstr_to_list(dim) for dim in bin_label]
else:
return [self.binstr_to_list(bin_label)]
def update(self):
self.stat.set_data(self.values)
def calculate(self):
self.value = self.stat.value
