Source code for bokeh.models.transforms
''' Represent transformations of data to happen on the client (browser) side.
'''
from __future__ import absolute_import
from textwrap import dedent
from types import FunctionType
from ..core.enums import StepMode, JitterRandomDistribution
from ..core.has_props import abstract
from ..core.properties import Bool, Dict, Either, Enum, Float, Instance, Seq, String, AnyRef
from ..model import Model
from ..util.compiler import nodejs_compile, CompilationError
from ..util.dependencies import import_required
from ..util.future import get_param_info, signature
from .sources import ColumnarDataSource
[docs]@abstract
class Transform(Model):
''' Base class for ``Transform`` models that represent a computation
to be carried out on the client-side.
JavaScript implementations should implement the following methods:
.. code-block:: coffeescript
compute: (x) ->
# compute the transform of a single value
v_compute: (xs) ->
# compute the transform of an array of values
'''
pass
[docs]class CustomJSTransform(Transform):
''' Apply a custom defined transform to data.
.. warning::
The explicit purpose of this Bokeh Model is to embed *raw JavaScript
code* for a browser to execute. If any part of the code is derived
from untrusted user inputs, then you must take appropriate care to
sanitize the user input prior to passing to Bokeh.
'''
[docs] @classmethod
def from_py_func(cls, func, v_func):
''' Create a ``CustomJSTransform`` instance from a pair of Python
functions. The function is translated to JavaScript using PScript.
The python functions must have no positional arguments. It's
possible to pass Bokeh models (e.g. a ``ColumnDataSource``) as keyword
arguments to the functions.
The ``func`` function namespace will contain the variable ``x`` (the
untransformed value) at render time. The ``v_func`` function namespace
will contain the variable ``xs`` (the untransformed vector) at render
time.
.. warning::
The vectorized function, ``v_func``, must return an array of the
same length as the input ``xs`` array.
Example:
.. code-block:: python
def transform():
from pscript.stubs import Math
return Math.cos(x)
def v_transform():
from pscript.stubs import Math
return [Math.cos(x) for x in xs]
customjs_transform = CustomJSTransform.from_py_func(transform, v_transform)
Args:
func (function) : a scalar function to transform a single ``x`` value
v_func (function) : a vectorized function to transform a vector ``xs``
Returns:
CustomJSTransform
'''
if not isinstance(func, FunctionType) or not isinstance(v_func, FunctionType):
raise ValueError('CustomJSTransform.from_py_func only accepts function objects.')
pscript = import_required(
'pscript',
dedent("""\
To use Python functions for CustomJSTransform, you need PScript
'("conda install -c conda-forge pscript" or "pip install pscript")""")
)
def pscript_compile(func):
sig = signature(func)
all_names, default_values = get_param_info(sig)
if len(all_names) - len(default_values) != 0:
raise ValueError("Function may only contain keyword arguments.")
if default_values and not any(isinstance(value, Model) for value in default_values):
raise ValueError("Default value must be a Bokeh Model.")
func_kwargs = dict(zip(all_names, default_values))
# Wrap the code attr in a function named `formatter` and call it
# with arguments that match the `args` attr
code = pscript.py2js(func, 'transformer') + 'return transformer(%s);\n' % ', '.join(all_names)
return code, func_kwargs
jsfunc, func_kwargs = pscript_compile(func)
v_jsfunc, v_func_kwargs = pscript_compile(v_func)
# Have to merge the function arguments
func_kwargs.update(v_func_kwargs)
return cls(func=jsfunc, v_func=v_jsfunc, args=func_kwargs)
[docs] @classmethod
def from_coffeescript(cls, func, v_func, args={}):
''' Create a ``CustomJSTransform`` instance from a pair of CoffeeScript
snippets. The function bodies are translated to JavaScript functions
using node and therefore require return statements.
The ``func`` snippet namespace will contain the variable ``x`` (the
untransformed value) at render time. The ``v_func`` snippet namespace
will contain the variable ``xs`` (the untransformed vector) at render
time.
Example:
.. code-block:: coffeescript
func = "return Math.cos(x)"
v_func = "return [Math.cos(x) for x in xs]"
transform = CustomJSTransform.from_coffeescript(func, v_func)
Args:
func (str) : a coffeescript snippet to transform a single ``x`` value
v_func (str) : a coffeescript snippet function to transform a vector ``xs``
Returns:
CustomJSTransform
'''
compiled = nodejs_compile(func, lang="coffeescript", file="???")
if "error" in compiled:
raise CompilationError(compiled.error)
v_compiled = nodejs_compile(v_func, lang="coffeescript", file="???")
if "error" in v_compiled:
raise CompilationError(v_compiled.error)
return cls(func=compiled.code, v_func=v_compiled.code, args=args)
args = Dict(String, AnyRef, help="""
A mapping of names to Python objects. In particular those can be bokeh's models.
These objects are made available to the transform' code snippet as the values of
named parameters to the callback.
""")
func = String(default="", help="""
A snippet of JavaScript code to transform a single value. The variable
``x`` will contain the untransformed value and can be expected to be
present in the function namespace at render time. The snippet will be
into the body of a function and therefore requires a return statement.
Example:
.. code-block:: javascript
func = '''
return Math.floor(x) + 0.5
'''
""")
v_func = String(default="", help="""
A snippet of JavaScript code to transform an array of values. The variable
``xs`` will contain the untransformed array and can be expected to be
present in the function namespace at render time. The snippet will be
into the body of a function and therefore requires a return statement.
Example:
.. code-block:: javascript
v_func = '''
var new_xs = new Array(xs.length)
for(var i = 0; i < xs.length; i++) {
new_xs[i] = xs[i] + 0.5
}
return new_xs
'''
.. warning::
The vectorized function, ``v_func``, must return an array of the
same length as the input ``xs`` array.
""")
use_strict = Bool(default=False, help="""
Enables or disables automatic insertion of ``"use strict";`` into ``func`` or ``v_func``.
""")
[docs]class Dodge(Transform):
''' Apply either fixed dodge amount to data.
'''
value = Float(default=0, help="""
The amount to dodge the input data.
""")
range = Instance("bokeh.models.ranges.Range", help="""
When applying ``Dodge`` to categorical data values, the corresponding
``FactorRange`` must be supplied as the ``range`` property.
""")
[docs]class Jitter(Transform):
''' Apply either a uniform or normally sampled random jitter to data.
'''
mean = Float(default=0, help="""
The central value for the random sample
""")
width = Float(default=1, help="""
The width (absolute for uniform distribution and sigma for the normal
distribution) of the random sample.
""")
distribution = Enum(JitterRandomDistribution, default='uniform', help="""
The random distribution upon which to pull the random scatter
""")
range = Instance("bokeh.models.ranges.Range", help="""
When applying Jitter to Categorical data values, the corresponding
``FactorRange`` must be supplied as the ``range`` property.
""")
[docs]@abstract
class Interpolator(Transform):
''' Base class for interpolator transforms.
Interpolators return the value of a function which has been evaluated
between specified (x, y) pairs of data. As an example, if two control
point pairs were provided to the interpolator, a linear interpolaction
at a specific value of 'x' would result in the value of 'y' which existed
on the line connecting the two control points.
The control point pairs for the interpolators can be specified through either
* A literal sequence of values:
.. code-block:: python
interp = Interpolator(x=[1, 2, 3, 4, 5], y=[2, 5, 10, 12, 16])
* or a pair of columns defined in a ``ColumnDataSource`` object:
.. code-block:: python
interp = Interpolator(x="year", y="earnings", data=jewlery_prices))
This is the base class and is not intended to end use. Please see the
documentation for the final derived classes (``Jitter``, ``LineraInterpolator``,
``StepInterpolator``) for more information on their specific methods of
interpolation.
'''
x = Either(String, Seq(Float), help="""
Independent coordinate denoting the location of a point.
""")
y = Either(String, Seq(Float), help="""
Dependant coordinate denoting the value of a point at a location.
""")
data = Instance(ColumnarDataSource, help="""
Data which defines the source for the named columns if a string is passed to either the ``x`` or ``y`` parameters.
""")
clip = Bool(True, help="""
Determine if the interpolation should clip the result to include only values inside its predefined range.
If this is set to False, it will return the most value of the closest point.
""")
# Define an initialization routine to do some cross checking of input values
def __init__(self, **kwargs):
super(Interpolator, self).__init__(**kwargs)
[docs]class LinearInterpolator(Interpolator):
''' Compute a linear interpolation between the control points provided through
the ``x``, ``y``, and ``data`` parameters.
'''
pass
[docs]class StepInterpolator(Interpolator):
''' Compute a step-wise interpolation between the points provided through
the ``x``, ``y``, and ``data`` parameters.
'''
mode = Enum(StepMode, default="after", help="""
Adjust the behavior of the returned value in relation to the control points. The parameter can assume one of three values:
* ``after`` (default): Assume the y-value associated with the nearest x-value which is less than or equal to the point to transform.
* ``before``: Assume the y-value associated with the nearest x-value which is greater than the point to transform.
* ``center``: Assume the y-value associated with the nearest x-value to the point to transform.
""")