#
# Copyright (C) 2009, 2015, 2016, 2018 - 2024
# Smithsonian Astrophysical Observatory
#
#
# This program is free software; you can redistribute it and/or modify
# it under the terms of the GNU General Public License as published by
# the Free Software Foundation; either version 3 of the License, or
# (at your option) any later version.
# This program is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
# GNU General Public License for more details.
#
# You should have received a copy of the GNU General Public License along
# with this program; if not, write to the Free Software Foundation, Inc.,
# 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
#
"""A visualization interface to Sherpa.
Classes provide access to common plotting tasks, which is done by the
plotting backend defined in the ``options.plot_pkg`` setting of the
Sherpa configuration file. Note that plot objects can be created
and used even when only the `sherpa.plot.backends.BasicBackend` is
available.
"""
from __future__ import annotations
from configparser import ConfigParser
import contextlib
import copy
import logging
import importlib
from typing import Any, Literal, Optional, Sequence, Union
import numpy as np
from sherpa import get_config
from sherpa.estmethods import Covariance
from sherpa.optmethods import LevMar, NelderMead
from sherpa.plot.backends import BaseBackend, BasicBackend, PLOT_BACKENDS
from sherpa.stats import Likelihood, LeastSq, Chi2XspecVar
from sherpa.utils import NoNewAttributesAfterInit, erf, \
bool_cast, parallel_map, dataspace1d, histogram1d, get_error_estimates
from sherpa.utils.err import ArgumentTypeErr, ConfidenceErr, \
IdentifierErr, PlotErr, StatErr
from sherpa.utils.numeric_types import SherpaFloat
# PLOT_BACKENDS only contains backends in modules that are imported successfully
# but modules are not discovered by itself. Entrypoints would solve this problem
# but the current implementation does not have this capability.
# See docstring of sherpa.plot.backends.MetaBaseBackend for details.
#
for name in ["pylab", "pylab_area", "bokeh"]:
try:
importlib.import_module(f"sherpa.plot.{name}_backend")
except ImportError:
pass
config = ConfigParser()
config.read(get_config())
warning = logging.getLogger(__name__).warning
# TODO: why is this module globally changing the invalid mode of NumPy?
_ = np.seterr(invalid='ignore')
basicbackend = BasicBackend()
'''This backend has all the defaults that are backend-independent.
For all the plot classes in this module (e.g. FitPlot, JointPlot, ...)
the default settings are set in code like `plot_prefs = xxx.get_plot_defaults()`
where xxx is some backend.
'''
# This may be over-ridden below, depending on the options.plot_pkg setting.
#
backend: BaseBackend = basicbackend
'''Currently active backend module for plotting.'''
# In the current design, this code is executed when this
# module imported and then the values are shared between all instances of a class.
# That allows one to change those default "globally" (i.e. for all instances of the class)
# but it means that the default values are based on the backend that is active when
# this module is imported. That depends on the sherpa.rc setting and thus `plot_prefs`
# might contain e.g. matplotlib specific defaults that are not applicable when other
# backends are used.
# Thus, we currently initialize them with the BasicBackend that has only the
# backend-independent (i.e. those that works for any backend) defaults set.
plot_opt_str = config.get('options', 'plot_pkg', fallback='BasicBackend')
plot_opt = [o.strip() for o in plot_opt_str.split()]
for plottry in plot_opt:
if plottry in PLOT_BACKENDS:
backend = PLOT_BACKENDS[plottry]()
break
warning("Plotting backend '%s' not found or dependencies missing. Trying next option.", plottry)
else:
# The backend will default to BasicBackend from above
warning(f"Tried the following backends listed in {get_config()}: \n" +
f"{plot_opt}\n" +
"None of these imported correctly, so using the 'BasicBackend'.\n" +
f"List of backends that have loaded and would be available: {[k for k in PLOT_BACKENDS]}")
__all__ = ('Plot', 'Contour', 'Point', 'Histogram',
'MultiPlot',
'HistogramPlot', 'DataHistogramPlot',
'ModelHistogramPlot', 'SourceHistogramPlot',
'PDFPlot', 'CDFPlot', 'LRHistogram',
'SplitPlot', 'JointPlot',
'DataPlot', 'TracePlot', 'ScatterPlot',
'PSFKernelPlot',
'DataContour', 'PSFKernelContour',
'ModelPlot', 'ComponentModelPlot',
'ComponentModelHistogramPlot', 'ComponentTemplateModelPlot',
'SourcePlot', 'ComponentSourcePlot',
'ComponentSourceHistogramPlot', 'ComponentTemplateSourcePlot',
'PSFPlot',
'ModelContour', 'PSFContour', 'SourceContour',
'FitPlot',
'FitContour',
'DelchiPlot', 'ChisqrPlot', 'ResidPlot',
'ResidContour',
'RatioPlot',
'RatioContour',
'Confidence1D', 'Confidence2D',
'IntervalProjection', 'IntervalUncertainty',
'RegionProjection', 'RegionUncertainty',
'TemporaryPlottingBackend',
'set_backend',
)
_stats_noerr = ('cash', 'cstat', 'leastsq', 'wstat')
"""Statistics that do not make use of uncertainties."""
[docs]
def set_backend(new_backend: Union[str, BaseBackend, type[BaseBackend]]) -> None:
'''Set the Sherpa plotting backend.
Plotting backends are registered in Sherpa with a string name.
See the examples below for how to get a list of available backends.
Parameters
----------
new_backend : string, class, or instance
Set a sherpa plotting backend. The backend can be passed in as an
object, or as a convenience, as a simple string
Example
-------
Set the backend to use Pylab from matplotlib for plotting. This is
probably what most users need:
>>> from sherpa.plot import set_backend
>>> set_backend('pylab')
Get a list of registered backends:
>>> from sherpa.plot import PLOT_BACKENDS
>>> PLOT_BACKENDS
{'BaseBackend': <class 'sherpa.plot.backends.BaseBackend'>, ...}
This list shows the names and the class for each backend.
Details for each backend can be found in the Sherpa documentation or by
inspecting the backend classes using normal Python functionality:
>>> from sherpa.plot.backends import BasicBackend
>>> help(BasicBackend)
Help on class BasicBackend in module sherpa.plot.backends:
<BLANKLINE>
class BasicBackend(BaseBackend)
| A dummy backend for plotting.
...
'''
global backend
if isinstance(new_backend, str):
try:
backend = PLOT_BACKENDS[new_backend]()
except KeyError:
raise IdentifierErr('noplotbackend', new_backend,
list(PLOT_BACKENDS.keys())) from None
return
# It's a class and that class is a subclass of BaseBackend
if isinstance(new_backend, type) and issubclass(new_backend, BaseBackend):
backend = new_backend()
return
if isinstance(new_backend, BaseBackend):
backend = new_backend
return
raise ArgumentTypeErr('tempplotbackend', new_backend)
[docs]
class TemporaryPlottingBackend(contextlib.AbstractContextManager):
'''Set the Sherpa plotting backend as a context, e.g. for a single plot
This changes the logging level globally for all modules in sherpa.
Parameters
----------
new_backend : string, class, or instance
Set a sherpa plotting backend. The backend can be passed in as an
instance of a plotting backend class. For simplicity, the user can
also pass in a string naming a loaded backend class or the class
itself; calling this context manager will then create an instance.
Example
-------
>>> from sherpa.plot import TemporaryPlottingBackend, DataPlot
>>> from sherpa.data import Data1D
>>> with TemporaryPlottingBackend('pylab'):
... x1 = [100, 200, 600, 1200]
... y1 = [2000, 2100, 1400, 3050]
... d1 = Data1D('oned', x1, y1)
... plot1 = DataPlot()
... plot1.prepare(d1)
... plot1.plot()
'''
def __init__(self, new_backend: BaseBackend) -> None:
self.backend = new_backend
def __enter__(self) -> TemporaryPlottingBackend:
global backend
self.old = backend
set_backend(self.backend)
return self
# As exc_type/val/tb are not used here we type them as Any, and the
# return value is marked as an explicit False rather than bool since
# it avoids warnings from mypy (as __exit__ is special).
#
def __exit__(self,
exc_type: Any, exc_val: Any, exc_tb: Any) -> Literal[False]:
global backend
backend = self.old
return False
def _make_title(title, name=''):
"""Return the plot title to use.
Parameters
----------
title : str
The main title to use
name : str or None
The identifier for the dataset.
Returns
-------
title : str
If the name is empty or None then use title,
otherwise use title + ' for ' + name.
"""
if name in [None, '']:
return title
return f"{title} for {name}"
def _errorbar_warning(stat):
"""The warning message to display when error bars are being "faked".
Parameters
----------
stat : sherpa.stats.Stat instance
The name attribute is used in the error message.
Returns
-------
msg : str
The warning message
"""
return "The displayed errorbars have been supplied with the " + \
"data or calculated using chi2xspecvar; the errors are not " + \
f"used in fits with {stat.name}"
def calculate_errors(data, stat, yerrorbars=True):
"""Calculate errors from the statistics object."""
if stat is None:
return None
# Do we warn about adding in error values? The logic here isn't
# quite the same as the other times _errorbar_warning is used.
# This suggests that the code really should be re-worked to
# go through a common code path.
#
# Do we require that self.yerr is always set, even if the
# value is not used in a plot? I am going to assume not
# (since the existing code will not change self.yerr if
# the stat.name is not in _stats_noerr but an exception is
# raised by get_yerr).
#
# This assumes that the error bars calculated by data.to_plot are
# over-ridden once a statistic is given. However, how does this
# work if the statistic is a Chi2 variant - e.g. Chi2DataVar -
# but the user has given explicit errors (i.e. they are not to
# be calculated by the "DataVar" part but used as is). Does
# data.get_yerr handle this for us, or are invalid errors
# used here? It appears that the correct answers are being
# returned, but should we only call data.get_yerr if yerr
# is None/empty/whatever is returned by to_plot? This also
# holds for the Resid/RatioPlot classes.
#
# Note that we should probably return a value for yerr if
# we can, even if 'yerrorbars' is set to False, so that
# downstream users can make use of the value even if the
# plot doesn't. This is similar to how labels or xerr
# attributes are created even if they don't get used by the
# plot.
#
msg = _errorbar_warning(stat)
if stat.name in _stats_noerr:
if yerrorbars:
warning(msg)
return data.get_yerr(True, Chi2XspecVar.calc_staterror)
try:
return data.get_yerr(True, stat.calc_staterror)
except Exception:
# TODO: can we report a useful error here?
#
# It is possible that this is actually an unrelated
# error: it's unclear what error class is expected to
# be thrown, but likely ValueError as this is raised
# by Chi2DataVar when sent values < 0
# (note that over time the behavior has changed from
# <= 0 to < 0, but this error message has not been
# changed).
#
if yerrorbars:
warning("%s\nzeros or negative values found", msg)
return None
[docs]
class Plot(NoNewAttributesAfterInit):
"Base class for line plots"
plot_prefs = basicbackend.get_plot_defaults()
"The preferences for the plot."
def __init__(self):
"""
Initialize a Plot object. All 1D line plot
instances utilize Plot, which provides a generic
interface to a backend.
Once an instance of Plot is initialized no new
attributes of the class can be made. (To eliminate
the accidental creation of erroneous attributes)
"""
self.plot_prefs = self.plot_prefs.copy()
super().__init__()
[docs]
@staticmethod
def vline(x, ymin=0, ymax=1,
linecolor=None, linestyle=None, linewidth=None,
overplot=False, clearwindow=True):
"Draw a line at constant x, extending over the plot."
backend.vline(x, ymin=ymin, ymax=ymax, linecolor=linecolor,
linestyle=linestyle, linewidth=linewidth,
overplot=overplot, clearwindow=clearwindow)
[docs]
@staticmethod
def hline(y, xmin=0, xmax=1,
linecolor=None, linestyle=None, linewidth=None,
overplot=False, clearwindow=True):
"Draw a line at constant y, extending over the plot."
backend.hline(y, xmin=xmin, xmax=xmax, linecolor=linecolor,
linestyle=linestyle, linewidth=linewidth,
overplot=overplot, clearwindow=clearwindow)
def _merge_settings(self, kwargs):
"""Return the plot preferences merged with user settings."""
return {**self.plot_prefs, **kwargs}
[docs]
def plot(self, x, y, yerr=None, xerr=None, title=None, xlabel=None,
ylabel=None, overplot=False, clearwindow=True,
**kwargs):
"""Plot the data.
Parameters
----------
x, y
The data values to plot. They should have the same length.
yerr, xerr : optional
The symmetric errors to apply to the y or x values, or `None`
for no errors along the axis. If given, they should match the
length of the data.
title, xlabel, ylabel : optional, string
The optional plot title and axis labels. These are ignored
if overplot is set to `True`.
overplot : bool, optional
If `True` then add the data to an existing plot, otherwise
create a new plot.
clearwindow : bool, optional
Should the existing plot area be cleared before creating this
new plot (e.g. for multi-panel plots)?
**kwargs
These values are passed on to the plot backend, and must
match the names of the keys of the object's
plot_prefs dictionary.
See Also
--------
overplot
"""
opts = self._merge_settings(kwargs)
backend.plot(x, y, yerr=yerr, xerr=xerr,
title=title, xlabel=xlabel, ylabel=ylabel,
overplot=overplot, clearwindow=clearwindow,
**opts)
[docs]
def overplot(self, *args, **kwargs):
"""Add the data to an existing plot.
This is the same as calling the plot method with
overplot set to True.
See Also
--------
plot
"""
kwargs['overplot'] = True
self.plot(*args, **kwargs)
[docs]
class Contour(NoNewAttributesAfterInit):
"Base class for contour plots"
contour_prefs = basicbackend.get_contour_defaults()
"The preferences for the plot."
def __init__(self):
"""
Initialize a Contour object. All 2D contour plot
instances utilize Contour, which provides a generic
interface to a backend.
Once an instance of Contour is initialized no new
attributes of the class can be made. (To eliminate
the accidental creation of erroneous attributes)
"""
self.contour_prefs = self.contour_prefs.copy()
super().__init__()
def _merge_settings(self, kwargs):
"""Return the plot preferences merged with user settings."""
return {**self.contour_prefs, **kwargs}
[docs]
def contour(self, x0, x1, y, title=None, xlabel=None,
ylabel=None, overcontour=False, clearwindow=True,
**kwargs):
opts = self._merge_settings(kwargs)
backend.contour(x0, x1, y, title=title,
xlabel=xlabel, ylabel=ylabel,
overcontour=overcontour,
clearwindow=clearwindow, **opts)
[docs]
def overcontour(self, *args, **kwargs):
kwargs['overcontour'] = True
self.contour(*args, **kwargs)
[docs]
class Point(NoNewAttributesAfterInit):
"Base class for point plots"
point_prefs = basicbackend.get_point_defaults()
"The preferences for the plot."
def __init__(self):
"""
Initialize a Point object. All 1D point plot
instances utilize Point, which provides a generic
interface to a backend.
Once an instance of Point is initialized no new
attributes of the class can be made. (To eliminate
the accidental creation of erroneous attributes)
"""
self.point_prefs = self.point_prefs.copy()
super().__init__()
def _merge_settings(self, kwargs):
"""Return the plot preferences merged with user settings."""
return {**self.point_prefs, **kwargs}
[docs]
def point(self, x, y, overplot=True, clearwindow=False, **kwargs):
"""Draw a point at the given location.
Parameters
----------
x, y
The coordinates of the plot.
overplot : bool, optional
If `True` then add the data to an existing plot, otherwise
create a new plot.
clearwindow : bool, optional
Should the existing plot area be cleared before creating this
new plot (e.g. for multi-panel plots)?
**kwargs
These values are passed on to the plot backend, and must
match the names of the keys of the object's
point_prefs dictionary.
"""
opts = self._merge_settings(kwargs)
backend.plot(x, y,
overplot=overplot, clearwindow=clearwindow,
**opts)
[docs]
class Image(NoNewAttributesAfterInit):
"""Base class for image plots
.. warning::
This class is experimental and subject to change in the future.
Currently, is is only used within _repr_html_ methods.
"""
image_prefs = basicbackend.get_image_defaults()
"The preferences for the plot."
def __init__(self):
"""
Initialize a Image object.
Once an instance of Image is initialized no new
attributes of the class can be made. (To eliminate
the accidental creation of erroneous attributes.)
"""
self.image_prefs = self.image_prefs.copy()
super().__init__()
def _merge_settings(self, kwargs):
"""Return the plot preferences merged with user settings."""
return {**self.image_prefs, **kwargs}
[docs]
def plot(self, x0, x1, y, overplot=True, clearwindow=False, **kwargs):
"""Draw a point at the given location.
Parameters
----------
x0, x1 : array-like
Value for the image axes.
y : array-like, 2D
The coordinates of the plot.
overplot : bool, optional
If `True` then add the data to an existing plot, otherwise
create a new plot.
clearwindow : bool, optional
Should the existing plot area be cleared before creating this
new plot (e.g. for multi-panel plots)?
**kwargs
These values are passed on to the plot backend, and must
match the names of the keys of the object's
image_prefs dictionary.
"""
opts = self._merge_settings(kwargs)
backend.image(x0, x1, y,
overplot=overplot, clearwindow=clearwindow,
**opts)
[docs]
class Histogram(NoNewAttributesAfterInit):
"Base class for histogram plots"
histo_prefs = basicbackend.get_histo_defaults()
"The preferences for the plot."
def __init__(self):
"""
Initialize a Histogram object. All 1D histogram plot
instances utilize Histogram, which provides a generic
interface to a backend.
Once an instance of Histogram is initialized no new
attributes of the class can be made. (To eliminate
the accidental creation of erroneous attributes)
"""
self.histo_prefs = self.histo_prefs.copy()
super().__init__()
def _merge_settings(self, kwargs):
"""Return the plot preferences merged with user settings."""
return {**self.histo_prefs, **kwargs}
[docs]
def plot(self, xlo, xhi, y, yerr=None, title=None, xlabel=None,
ylabel=None, overplot=False, clearwindow=True, **kwargs):
"""Plot the data.
Parameters
----------
xlo, xhi, y
The data values to plot (the bin edges along the X axis and
the bin values for the Y axis). They should have the same length.
yerr : optional
The symmetric errors to apply to the y values, or `None`
for no errors along the axis. If given, they should match the
length of the data.
title, xlabel, ylabel : optional, string
The optional plot title and axis labels. These are ignored
if overplot is set to `True`.
overplot : bool, optional
If `True` then add the data to an existing plot, otherwise
create a new plot.
clearwindow : bool, optional
Should the existing plot area be cleared before creating this
new plot (e.g. for multi-panel plots)?
**kwargs
These values are passed on to the plot backend, and must
match the names of the keys of the object's
plot_prefs dictionary.
See Also
--------
overplot
"""
opts = self._merge_settings(kwargs)
backend.histo(xlo, xhi, y, yerr=yerr, title=title,
xlabel=xlabel, ylabel=ylabel, overplot=overplot,
clearwindow=clearwindow, **opts)
[docs]
def overplot(self, *args, **kwargs):
kwargs['overplot'] = True
self.plot(*args, **kwargs)
[docs]
class HistogramPlot(Histogram):
def __init__(self):
self.xlo = None
self.xhi = None
self.y = None
self.xlabel = None
self.ylabel = None
self.title = None
super().__init__()
def __str__(self):
xlo = self.xlo
if self.xlo is not None:
xlo = np.array2string(np.asarray(self.xlo), separator=',',
precision=4, suppress_small=False)
xhi = self.xhi
if self.xhi is not None:
xhi = np.array2string(np.asarray(self.xhi), separator=',',
precision=4, suppress_small=False)
y = self.y
if self.y is not None:
y = np.array2string(np.asarray(self.y), separator=',',
precision=4, suppress_small=False)
return (('xlo = %s\n' +
'xhi = %s\n' +
'y = %s\n' +
'xlabel = %s\n' +
'ylabel = %s\n' +
'title = %s\n' +
'histo_prefs = %s') %
(xlo,
xhi,
y,
self.xlabel,
self.ylabel,
self.title,
self.histo_prefs))
def _repr_html_(self):
"""Return a HTML (string) representation of the histogram plot."""
return backend.as_html_histogram(self)
@property
def x(self):
"""Return (xlo + xhi) / 2
This is intended to make it easier to swap between
plot- and histogram-style plots by providing access
to an X value.
"""
if self.xlo is None or self.xhi is None:
return None
# As we do not (yet) require NumPy arrays, enforce it.
xlo = np.asarray(self.xlo)
xhi = np.asarray(self.xhi)
return (xlo + xhi) / 2
[docs]
def plot(self, overplot=False, clearwindow=True, **kwargs):
"""Plot the data.
This will plot the data sent to the prepare method.
Parameters
----------
overplot : bool, optional
If `True` then add the data to an existing plot, otherwise
create a new plot.
clearwindow : bool, optional
Should the existing plot area be cleared before creating this
new plot (e.g. for multi-panel plots)?
**kwargs
These values are passed on to the plot backend, and must
match the names of the keys of the object's
plot_prefs dictionary.
See Also
--------
prepare, overplot
"""
Histogram.plot(self, self.xlo, self.xhi, self.y, title=self.title,
xlabel=self.xlabel, ylabel=self.ylabel,
overplot=overplot, clearwindow=clearwindow, **kwargs)
# I think we want slightly different histogram preferences
# than most (mark by point rather than line).
#
def get_data_hist_prefs():
"""Copy the data preferences to the histogram class"""
hprefs = basicbackend.get_model_histo_defaults()
dprefs = basicbackend.get_data_plot_defaults()
for k, v in dprefs.items():
if k in hprefs:
hprefs[k] = v
return hprefs
[docs]
class DataHistogramPlot(HistogramPlot):
"""Create 1D histogram plots of data values."""
histo_prefs = get_data_hist_prefs()
"The preferences for the plot."
def __init__(self):
self.xerr = None
self.yerr = None
super().__init__()
[docs]
def prepare(self, data, stat=None):
"""Create the data to plot
Parameters
----------
data
The Sherpa data object to display (it is assumed to be
one dimensional and represent binned data).
stat : optional
The Sherpa statistics object to use to add Y error bars
if the data has none.
See Also
--------
plot
"""
# Need a better way of accessing the binning of the data.
# Maybe to_plot should return the lo/hi edges as a pair
# here.
#
(_, self.y, self.yerr, self.xerr, self.xlabel,
self.ylabel) = data.to_plot()
self.xlo, self.xhi = data.get_indep(True)
if stat is not None:
yerrorbars = self.histo_prefs.get('yerrorbars', True)
self.yerr = calculate_errors(data, stat, yerrorbars)
self.title = data.name
[docs]
def plot(self, overplot=False, clearwindow=True, **kwargs):
"""Plot the data.
This will plot the data sent to the prepare method.
Parameters
----------
overplot : bool, optional
If `True` then add the data to an existing plot, otherwise
create a new plot.
clearwindow : bool, optional
Should the existing plot area be cleared before creating this
new plot (e.g. for multi-panel plots)?
**kwargs
These values are passed on to the plot backend, and must
match the names of the keys of the object's
plot_prefs dictionary.
See Also
--------
prepare, overplot
"""
Histogram.plot(self, self.xlo, self.xhi, self.y,
yerr=self.yerr, title=self.title,
xlabel=self.xlabel, ylabel=self.ylabel,
overplot=overplot, clearwindow=clearwindow, **kwargs)
[docs]
class ModelHistogramPlot(HistogramPlot):
"""Display a model as a histogram."""
def __init__(self):
super().__init__()
self.title = 'Model'
[docs]
def prepare(self, data, model, stat=None):
"""Create the plot.
The stat parameter is ignored.
"""
plot = data.to_plot(yfunc=model)
(_, y, _, _, self.xlabel, self.ylabel) = plot
# taken from get_x from Data1DInt
indep = data.get_evaluation_indep(filter=True, model=model)
self.xlo = indep[0]
self.xhi = indep[1]
self.y = y[1]
assert self.y.size == self.xlo.size
[docs]
class SourceHistogramPlot(ModelHistogramPlot):
"""Display a source model as a histogram."""
def __init__(self):
super().__init__()
self.title = 'Source'
[docs]
class PDFPlot(HistogramPlot):
"""Display the probability density of an array.
See Also
--------
CDFPlot
"""
def __init__(self):
self.points = None
super().__init__()
def __str__(self):
points = self.points
if self.points is not None:
points = np.array2string(np.asarray(self.points),
separator=',', precision=4,
suppress_small=False)
return ('points = %s\n' % (points) + HistogramPlot.__str__(self))
def _repr_html_(self):
"""Return a HTML (string) representation of the PDF plot."""
return backend.as_html_pdf(self)
[docs]
def prepare(self, points, bins=12, normed=True, xlabel="x", name="x"):
"""Create the data to plot.
Parameters
----------
points
The values to display.
bins : int, optional
The number of bins to use.
normed : bool, optional
See the description of the density parameter in the
numpy.histogram routine.
xlabel : optional, string
The label for the X axis.
name : optional, string
Used to create the plot title.
See Also
--------
plot
"""
self.points = points
self.y, xx = np.histogram(points, bins=bins, density=normed)
self.xlo = xx[:-1]
self.xhi = xx[1:]
self.ylabel = "probability density"
self.xlabel = xlabel
self.title = "PDF: {}".format(name)
[docs]
class CDFPlot(Plot):
"""Display the cumulative distribution of an array.
The cumulative distribution of the data is drawn along with
vertical lines to indicate the median, 15.87%, and 84.13%
percentiles.
See Also
--------
PDFPlot
Examples
--------
Show the cumulative distribution of 1000 randomly-distributed
points from a Gumbel distribution:
>>> import numpy as np
>>> rng = np.random.default_rng() # requires NumPy 1.17 or later
>>> pts = rng.gumbel(loc=100, scale=20, size=1000)
>>> plot = CDFPlot()
>>> plot.prepare(pts)
>>> plot.plot()
"""
median_defaults = {"linestyle": 'dash', "linecolor": 'orange',
"linewidth": 1.5}
"""The options used to draw the median line."""
lower_defaults = {"linestyle": 'dash', "linecolor": 'blue',
"linewidth": 1.5}
"""The options used to draw the 15.87% line."""
upper_defaults = {"linestyle": 'dash', "linecolor": 'blue',
"linewidth": 1.5}
"""The options used to draw the 84.13% line."""
plot_prefs = basicbackend.get_cdf_plot_defaults()
"""The plot options (the CDF and axes)."""
def __init__(self):
self.x = None
self.y = None
self.points = None
self.median = None
self.lower = None
self.upper = None
self.xlabel = None
self.ylabel = None
self.title = None
super().__init__()
def __str__(self):
x = self.x
if self.x is not None:
x = np.array2string(self.x, separator=',', precision=4,
suppress_small=False)
y = self.y
if self.y is not None:
y = np.array2string(self.y, separator=',', precision=4,
suppress_small=False)
points = self.points
if self.points is not None:
points = np.array2string(self.points, separator=',',
precision=4, suppress_small=False)
return f"""points = {points}
x = {x}
y = {y}
median = {self.median}
lower = {self.lower}
upper = {self.upper}
xlabel = {self.xlabel}
ylabel = {self.ylabel}
title = {self.title}
plot_prefs = {self.plot_prefs}"""
def _repr_html_(self):
"""Return a HTML (string) representation of the CDF plot."""
return backend.as_html_cdf(self)
[docs]
def prepare(self, points, xlabel="x", name="x"):
"""Create the data to plot.
Parameters
----------
points
The values to display (they do not have to be sorted).
xlabel : optional, string
The label for the X axis.
name : optional, string
Used to create the Y-axis label and plot title.
See Also
--------
plot
"""
self.points = np.asarray(points)
self.x = np.sort(points)
(self.median, self.lower,
self.upper) = get_error_estimates(self.x, True)
xsize = len(self.x)
self.y = (np.arange(xsize) + 1.0) / xsize
self.xlabel = xlabel
self.ylabel = "p(<={})".format(xlabel)
self.title = "CDF: {}".format(name)
[docs]
def plot(self, overplot=False, clearwindow=True, **kwargs):
"""Plot the data.
This will plot the data sent to the prepare method.
Parameters
----------
overplot : bool, optional
If `True` then add the data to an existing plot, otherwise
create a new plot.
clearwindow : bool, optional
Should the existing plot area be cleared before creating this
new plot (e.g. for multi-panel plots)?
**kwargs
These values are passed on to the plot backend, and must
match the names of the keys of the object's
plot_prefs dictionary.
See Also
--------
prepare, overplot
"""
Plot.plot(self, self.x, self.y, title=self.title,
xlabel=self.xlabel, ylabel=self.ylabel,
overplot=overplot, clearwindow=clearwindow, **kwargs)
# Note: the user arguments are not applied to the vertical lines
#
Plot.vline(self.median, overplot=True, clearwindow=False,
**self.median_defaults)
Plot.vline(self.lower, overplot=True, clearwindow=False,
**self.lower_defaults)
Plot.vline(self.upper, overplot=True, clearwindow=False,
**self.upper_defaults)
[docs]
class LRHistogram(HistogramPlot):
"Derived class for creating 1D likelihood ratio distribution plots"
def __init__(self):
self.ratios = None
self.lr = None
self.ppp = None
super().__init__()
def __str__(self):
ratios = self.ratios
if self.ratios is not None:
ratios = np.array2string(np.asarray(self.ratios),
separator=',', precision=4,
suppress_small=False)
return '\n'.join(['ratios = %s' % ratios,
'lr = %s' % str(self.lr),
HistogramPlot.__str__(self)])
def _repr_html_(self):
"""Return a HTML (string) representation of the LRHistogram plot."""
return backend.as_html_lr(self)
[docs]
def prepare(self, ratios, bins, niter, lr, ppp):
"Create the data to plot"
self.ppp = float(ppp)
self.lr = float(lr)
y = np.asarray(ratios)
self.ratios = y
self.xlo, self.xhi = dataspace1d(y.min(), y.max(),
numbins=bins + 1)[:2]
y = histogram1d(y, self.xlo, self.xhi)
self.y = y / float(niter)
self.title = "Likelihood Ratio Distribution"
self.xlabel = "Likelihood Ratio"
self.ylabel = "Frequency"
[docs]
def plot(self, overplot=False, clearwindow=True, **kwargs):
"""Plot the data.
This will plot the data sent to the prepare method.
Parameters
----------
overplot : bool, optional
If `True` then add the data to an existing plot, otherwise
create a new plot.
clearwindow : bool, optional
Should the existing plot area be cleared before creating this
new plot (e.g. for multi-panel plots)?
**kwargs
These values are passed on to the plot backend, and must
match the names of the keys of the object's
plot_prefs dictionary.
See Also
--------
prepare, overplot
"""
Histogram.plot(self, self.xlo, self.xhi, self.y, title=self.title,
xlabel=self.xlabel, ylabel=self.ylabel,
overplot=overplot, clearwindow=clearwindow, **kwargs)
if self.lr is None:
return
# Note: the user arguments are not applied to the vertical line
#
if self.lr <= self.xhi.max() and self.lr >= self.xlo.min():
Plot.vline(self.lr, linecolor="orange", linestyle="solid",
linewidth=1.5, overplot=True, clearwindow=False)
[docs]
class SplitPlot(Plot, Contour):
"""Create multiple plots.
Attributes
----------
rows : int
Number of rows of plots. The default is 2.
cols : int
Number of columns of plots. The default is 1.
"""
plot_prefs = basicbackend.get_split_plot_defaults()
"The preferences for the plot."
def __init__(self, rows=2, cols=1):
self.reset(rows, cols)
super().__init__()
def __str__(self):
return (('rows = %s\n' +
'cols = %s\n' +
'plot_prefs = %s') %
(self.rows,
self.cols,
self.plot_prefs))
[docs]
def reset(self, rows=2, cols=1):
"Prepare for a new set of plots or contours."
self.rows = rows
self.cols = cols
self._reset_used()
self._current = (0, 0)
self._cleared_window = False
def _clear_window(self):
if not self._cleared_window:
backend.clear_window()
self._cleared_window = True
def _reset_used(self):
self._used = np.zeros((self.rows, self.cols), np.bool_)
def _next_subplot(self):
row, col = np.where(~self._used)
if row.size != 0:
row, col = row[0], col[0]
else:
self._reset_used()
row, col = 0, 0
return row, col
[docs]
def addplot(self, plot, *args, **kwargs):
"Add the plot to the next space."
row, col = self._next_subplot()
self.plot(row, col, plot, *args, **kwargs)
[docs]
def addcontour(self, plot, *args, **kwargs):
"Add the contour plot to the next space."
row, col = self._next_subplot()
self.contour(row, col, plot, *args, **kwargs)
[docs]
def plot(self, row, col, plot, *args, **kwargs):
"Add the plot in the given space."
self._clear_window()
clearaxes = ((not kwargs.get('overplot', False)) and
kwargs.get('clearwindow', True))
backend.set_subplot(row, col, self.rows, self.cols, clearaxes,
**self.plot_prefs)
kwargs['clearwindow'] = False
plot.plot(*args, **kwargs)
self._used[row, col] = True
self._current = (row, col)
[docs]
def contour(self, row, col, plot, *args, **kwargs):
"Add the contour in the given space."
self._clear_window()
clearaxes = ((not kwargs.get('overcontour', False)) and
kwargs.get('clearwindow', True))
backend.set_subplot(row, col, self.rows, self.cols, clearaxes,
**self.plot_prefs)
kwargs['clearwindow'] = False
plot.contour(*args, **kwargs)
self._used[row, col] = True
self._current = (row, col)
[docs]
def overlayplot(self, plot, *args, **kwargs):
"Add the plot to the current space without destroying the contents."
self.overplot(self._current[0], self._current[1], plot, *args,
**kwargs)
[docs]
def overlaycontour(self, plot, *args, **kwargs):
"Add the contour to the current space without destroying the contents."
self.overcontour(self._current[0], self._current[1], plot, *args,
**kwargs)
# FIXME: work on overplot issue
[docs]
def overplot(self, row, col, plot, *args, **kwargs):
"Add a plot to the given space without destroying the contents."
kwargs['overplot'] = True
self.plot(row, col, plot, *args, **kwargs)
[docs]
def overcontour(self, row, col, plot, *args, **kwargs):
"Add a contour plot to the given space without destroying the contents."
kwargs['overcontour'] = True
self.contour(row, col, plot, *args, **kwargs)
# TODO: move logic from sherpa.ui.utils.Session._plot_jointplot
# regarding the X axis here
#
[docs]
class JointPlot(SplitPlot):
"""Multiple plots that share a common axis
This supports two plots, where the top plot is twice as
tall as the bottom plot.
"""
def __init__(self):
super().__init__()
[docs]
def plottop(self, plot, *args, overplot=False, clearwindow=True,
**kwargs):
create = clearwindow and not overplot
if create:
self._clear_window()
backend.set_jointplot(0, 0, self.rows, self.cols,
create=create)
# The code used to check if the plot was an instance of
# FitPlot, which has been updated to check for the presence
# of attributes instead.
#
if hasattr(plot, 'xlabel'):
plot.xlabel = ''
elif hasattr(plot, 'dataplot') and hasattr(plot, 'modelplot'):
dplot = plot.dataplot
mplot = plot.modelplot
if hasattr(dplot, 'xlabel'):
dplot.xlabel = ''
if hasattr(mplot, 'xlabel'):
mplot.xlabel = ''
kwargs['clearwindow'] = False
plot.plot(*args, overplot=overplot, **kwargs)
[docs]
def plotbot(self, plot, *args, overplot=False, **kwargs):
backend.set_jointplot(1, 0, self.rows, self.cols,
create=False)
# FIXME: terrible hack to remove title from bottom
plot.title = ''
kwargs['clearwindow'] = False
plot.plot(*args, overplot=overplot, **kwargs)
[docs]
class DataPlot(Plot):
"""Create 1D plots of data values.
Attributes
----------
x : array_like
The X value for each point (the independent variable).
y : array_like
The Y value for each point (the dependent variable).
xerr : array_like
The half-width of each X "bin", if set.
yerr : array_like
The error on the Y value, if set.
xlabel, ylabel, title : str
Plot labels.
Examples
--------
Plot up an example dataset. The default appearance is to draw
a symbol at each point, but no line connecting the points (the
actual choice depends on the plot backend):
>>> from sherpa.data import Data1D
>>> from sherpa.plot import DataPlot
>>> data = Data1D('a dataset', [10, 20, 25], [2, -7, 4])
>>> dplot = DataPlot()
>>> dplot.prepare(data)
>>> dplot.plot()
The plot attributes can be changed to adjust the appearance of
the plot, and the data re-drawn. The following also shows how
the plot preferences can be over-ridden to turn off the points
and draw a dotted line connecting the points (this assumes that
the Matplotlib backend is in use):
>>> dplot.xlabel = 'length'
>>> dplot.ylabel = 'data values'
>>> dplot.plot(marker=' ', linestyle='dashed')
"""
plot_prefs = basicbackend.get_data_plot_defaults()
"The preferences for the plot."
def __init__(self):
self.x = None
self.y = None
self.yerr = None
self.xerr = None
self.xlabel = None
self.ylabel = None
self.title = None
super().__init__()
def __str__(self):
x = self.x
if self.x is not None:
x = np.array2string(self.x, separator=',', precision=4,
suppress_small=False)
y = self.y
if self.y is not None:
y = np.array2string(self.y, separator=',', precision=4,
suppress_small=False)
yerr = self.yerr
if self.yerr is not None:
yerr = np.array2string(self.yerr, separator=',', precision=4,
suppress_small=False)
xerr = self.xerr
if self.xerr is not None:
xerr = np.array2string(self.xerr, separator=',', precision=4,
suppress_small=False)
return (('x = %s\n' +
'y = %s\n' +
'yerr = %s\n' +
'xerr = %s\n' +
'xlabel = %s\n' +
'ylabel = %s\n' +
'title = %s\n' +
'plot_prefs = %s') %
(x,
y,
yerr,
xerr,
self.xlabel,
self.ylabel,
self.title,
self.plot_prefs))
def _repr_html_(self):
"""Return a HTML (string) representation of the data plot."""
return backend.as_html_data(self)
[docs]
def prepare(self, data, stat=None):
"""Create the data to plot
Parameters
----------
data
The Sherpa data object to display (it is assumed to be
one dimensional).
stat : optional
The Sherpa statistics object to use to add Y error bars
if the data has none.
See Also
--------
plot
"""
(self.x, self.y, self.yerr, self.xerr, self.xlabel,
self.ylabel) = data.to_plot()
if stat is not None:
yerrorbars = self.plot_prefs.get('yerrorbars', True)
self.yerr = calculate_errors(data, stat, yerrorbars)
self.title = data.name
[docs]
def plot(self, overplot=False, clearwindow=True, **kwargs):
"""Plot the data.
This will plot the data sent to the prepare method.
Parameters
----------
overplot : bool, optional
If `True` then add the data to an existing plot, otherwise
create a new plot.
clearwindow : bool, optional
Should the existing plot area be cleared before creating this
new plot (e.g. for multi-panel plots)?
**kwargs
These values are passed on to the plot backend, and must
match the names of the keys of the object's
plot_prefs dictionary.
See Also
--------
prepare, overplot
"""
Plot.plot(self, self.x, self.y, yerr=self.yerr, xerr=self.xerr,
title=self.title, xlabel=self.xlabel, ylabel=self.ylabel,
overplot=overplot, clearwindow=clearwindow, **kwargs)
[docs]
class TracePlot(DataPlot):
plot_prefs = basicbackend.get_model_plot_defaults()
[docs]
def prepare(self, points, xlabel="x", name="x"):
"""The data to plot.
Parameters
----------
points
The trace to plot (one dimensional). The X axis is
set to consecutive integers, starting at 0.
xlabel : optional
This value is ignored.
name : optional, string
Used to create the Y-axis label and plot title.
See Also
--------
plot
"""
self.x = np.arange(len(points), dtype=SherpaFloat)
self.y = points
self.xlabel = "iteration"
self.ylabel = name
self.title = "Trace: {}".format(name)
[docs]
class ScatterPlot(DataPlot):
plot_prefs = basicbackend.get_scatter_plot_defaults()
[docs]
def prepare(self, x, y, xlabel="x", ylabel="y", name="(x,y)"):
"""The data to plot.
Parameters
----------
x, y
The points to plot. The number of points in the two
sequences should be the same.
xlabel, ylabel, name : optional, str
The axis labels and name for the plot title.
See Also
--------
plot
"""
self.x = np.asarray(x, dtype=SherpaFloat)
self.y = np.asarray(y, dtype=SherpaFloat)
self.xlabel = xlabel
self.ylabel = ylabel
self.title = "Scatter: {}".format(name)
[docs]
class PSFKernelPlot(DataPlot):
"Derived class for creating 1D PSF kernel data plots"
[docs]
def prepare(self, psf, data=None, stat=None):
"""Create the data to plot
Parameters
----------
psf
The Sherpa PSF to display
data : optional
The Sherpa data object to pass to the PSF.
stat : optional
The Sherpa statistics object to use to add Y error bars
if the data has none.
See Also
--------
plot
"""
psfdata = psf.get_kernel(data)
DataPlot.prepare(self, psfdata, stat)
# self.ylabel = 'PSF value'
# self.xlabel = 'PSF Kernel size'
self.title = 'PSF Kernel'
[docs]
class DataContour(Contour):
"""Create contours of 2D data.
Attributes
----------
x0, x1 : array_like
The coordinates of each point (the independent variables), as
one-dimensional arrays.
y : array_like
The Y value for each point (the dependent variable), as a
one-dimensional array.
levels : array_like or `None`
The values at which to draw contours.
xlabel, ylabel, title : str
Plot labels.
"""
contour_prefs = basicbackend.get_data_contour_defaults()
"The preferences for the plot."
def __init__(self):
self.x0 = None
self.x1 = None
self.y = None
self.xlabel = None
self.ylabel = None
self.title = None
self.levels = None
super().__init__()
def __str__(self):
x0 = self.x0
if self.x0 is not None:
x0 = np.array2string(self.x0, separator=',', precision=4,
suppress_small=False)
x1 = self.x1
if self.x1 is not None:
x1 = np.array2string(self.x1, separator=',', precision=4,
suppress_small=False)
y = self.y
if self.y is not None:
y = np.array2string(self.y, separator=',', precision=4,
suppress_small=False)
return (('x0 = %s\n' +
'x1 = %s\n' +
'y = %s\n' +
'xlabel = %s\n' +
'ylabel = %s\n' +
'title = %s\n' +
'levels = %s\n' +
'contour_prefs = %s') %
(x0,
x1,
y,
self.xlabel,
self.ylabel,
self.title,
self.levels,
self.contour_prefs))
def _repr_html_(self):
"""Return a HTML (string) representation of the contour plot."""
return backend.as_html_datacontour(self)
[docs]
def prepare(self, data, stat=None):
(self.x0, self.x1, self.y, self.xlabel,
self.ylabel) = data.to_contour()
self.title = data.name
[docs]
def contour(self, overcontour=False, clearwindow=True, **kwargs):
Contour.contour(self, self.x0, self.x1, self.y,
levels=self.levels, title=self.title,
xlabel=self.xlabel, ylabel=self.ylabel,
overcontour=overcontour,
clearwindow=clearwindow, **kwargs)
[docs]
class PSFKernelContour(DataContour):
"Derived class for creating 2D PSF Kernel contours"
[docs]
def prepare(self, psf, data=None, stat=None):
psfdata = psf.get_kernel(data)
DataContour.prepare(self, psfdata)
# self.xlabel = 'PSF Kernel size x0'
# self.ylabel = 'PSF Kernel size x1'
self.title = 'PSF Kernel'
[docs]
class ModelPlot(Plot):
"""Create 1D plots of model values.
Attributes
----------
x : array_like
The X value for each point (the independent variable).
y : array_like
The Y value for each point (the dependent variable).
xerr : array_like
The half-width of each X "bin", if set.
yerr : array_like
The error on the Y value, if set.
xlabel, ylabel, title : str
Plot labels.
Examples
--------
Plot up an example dataset. The default appearance is to draw
a line between each point:
>>> from sherpa.data import Data1D
>>> from sherpa.models.basic import StepLo1D
>>> from sherpa.plot import ModelPlot
>>> data = Data1D('a dataset', [10, 20, 25], [2, -7, 4])
>>> model = StepLo1D()
>>> model.xcut = 19
>>> mplot = ModelPlot()
>>> mplot.prepare(data, model)
>>> mplot.plot()
The plot attributes can be changed to adjust the appearance of
the plot, and the data re-drawn. The following also shows how
the plot preferences can be over-ridden to draw the model as
squares with no line connecting them (this assumes that
the Matplotlib backend is in use):
>>> mplot.xlabel = 'length'
>>> mplot.ylabel = 'model values'
>>> mplot.plot(marker='s', linestyle='none')
"""
plot_prefs = basicbackend.get_model_plot_defaults()
"The preferences for the plot."
def __init__(self):
self.x = None
self.y = None
self.yerr = None
self.xerr = None
self.xlabel = None
self.ylabel = None
self.title = 'Model'
super().__init__()
def __str__(self):
x = self.x
if self.x is not None:
x = np.array2string(self.x, separator=',', precision=4,
suppress_small=False)
y = self.y
if self.y is not None:
y = np.array2string(self.y, separator=',', precision=4,
suppress_small=False)
yerr = self.yerr
if self.yerr is not None:
yerr = np.array2string(self.yerr, separator=',', precision=4,
suppress_small=False)
xerr = self.xerr
if self.xerr is not None:
xerr = np.array2string(self.xerr, separator=',', precision=4,
suppress_small=False)
return (('x = %s\n' +
'y = %s\n' +
'yerr = %s\n' +
'xerr = %s\n' +
'xlabel = %s\n' +
'ylabel = %s\n' +
'title = %s\n' +
'plot_prefs = %s') %
(x,
y,
yerr,
xerr,
self.xlabel,
self.ylabel,
self.title,
self.plot_prefs))
def _repr_html_(self):
"""Return a HTML (string) representation of the model plot."""
return backend.as_html_model(self)
[docs]
def prepare(self, data, model, stat=None):
"""Create the data to plot
Parameters
----------
data
The Sherpa data object to display (it is assumed to be
one dimensional). This defines the grid over which the
model is displayed.
model
The Sherpa model expression to evaluate and display.
stat : optional
This parameter is unused.
See Also
--------
plot
"""
(self.x, self.y, self.yerr, self.xerr,
self.xlabel, self.ylabel) = data.to_plot(yfunc=model)
self.y = self.y[1]
[docs]
def plot(self, overplot=False, clearwindow=True, **kwargs):
"""Plot the data.
This will plot the data sent to the prepare method.
Parameters
----------
overplot : bool, optional
If `True` then add the data to an existing plot, otherwise
create a new plot.
clearwindow : bool, optional
Should the existing plot area be cleared before creating this
new plot (e.g. for multi-panel plots)?
**kwargs
These values are passed on to the plot backend, and must
match the names of the keys of the object's
plot_prefs dictionary.
See Also
--------
prepare, overplot
"""
Plot.plot(self, self.x, self.y, title=self.title, xlabel=self.xlabel,
ylabel=self.ylabel, overplot=overplot,
clearwindow=clearwindow, **kwargs)
[docs]
class ComponentModelPlot(ModelPlot):
plot_prefs = basicbackend.get_component_plot_defaults()
"The preferences for the plot."
[docs]
def prepare(self, data, model, stat=None):
ModelPlot.prepare(self, data, model, stat)
self.title = 'Model component: %s' % model.name
[docs]
class ComponentModelHistogramPlot(ModelHistogramPlot):
# Is this the correct setting?
plot_prefs = basicbackend.get_component_plot_defaults()
"The preferences for the plot."
[docs]
def prepare(self, data, model, stat=None):
super().prepare(data, model, stat)
self.title = 'Model component: {}'.format(model.name)
[docs]
class ComponentTemplateModelPlot(ComponentModelPlot):
[docs]
def prepare(self, data, model, stat=None):
self.x = model.get_x()
self.y = model.get_y()
self.xlabel = data.get_xlabel()
self.ylabel = data.get_ylabel()
self.title = 'Model component: {}'.format(model.name)
[docs]
class SourcePlot(ModelPlot):
"""Create 1D plots of unconvolved model values.
Attributes
----------
x : array_like
The X value for each point (the independent variable).
y : array_like
The Y value for each point (the model value).
xerr : array_like
The half-width of each X "bin", if set.
yerr : array_like
The error on the Y value, if set.
xlabel, ylabel, title : str
Plot labels.
"""
def __init__(self):
super().__init__()
self.title = 'Source'
[docs]
class ComponentSourcePlot(SourcePlot):
plot_prefs = basicbackend.get_component_plot_defaults()
[docs]
def prepare(self, data, model, stat=None):
(self.x, self.y, self.yerr, self.xerr,
self.xlabel, self.ylabel) = data.to_component_plot(yfunc=model)
self.y = self.y[1]
self.title = 'Source model component: {}'.format(model.name)
[docs]
class ComponentSourceHistogramPlot(SourceHistogramPlot):
# Is this the correct setting?
plot_prefs = basicbackend.get_component_plot_defaults()
[docs]
def prepare(self, data, model, stat=None):
plot = data.to_component_plot(yfunc=model)
(_, y, _, _, self.xlabel, self.ylabel) = plot
# taken from get_x from Data1DInt
indep = data.get_evaluation_indep(filter=True, model=model)
self.xlo = indep[0]
self.xhi = indep[1]
self.y = y[1]
assert self.y.size == self.xlo.size
self.title = 'Source model component: {}'.format(model.name)
[docs]
class ComponentTemplateSourcePlot(ComponentSourcePlot):
[docs]
def prepare(self, data, model, stat=None):
self.x = model.get_x()
self.y = model.get_y()
if np.iterable(data.mask):
x = data.to_plot()[0]
mask = (self.x > x.min()) & (self.x <= x.max())
self.x = self.x[mask]
self.y = self.y[mask]
self.xlabel = data.get_xlabel()
self.ylabel = data.get_ylabel()
self.title = 'Source model component: {}'.format(model.name)
[docs]
class PSFPlot(DataPlot):
"Derived class for creating 1D PSF kernel data plots"
[docs]
def prepare(self, psf, data=None, stat=None):
"""Create the data to plot
Parameters
----------
psf
The Sherpa PSF to display
data : optional
The Sherpa data object to pass to the PSF.
stat : optional
The Sherpa statistics object to use to add Y error bars
if the data has none.
See Also
--------
plot
"""
psfdata = psf.get_kernel(data, False)
DataPlot.prepare(self, psfdata, stat)
self.title = psf.kernel.name
[docs]
class ModelContour(Contour):
"Derived class for creating 2D model contours"
contour_prefs = basicbackend.get_model_contour_defaults()
"The preferences for the plot."
def __init__(self):
self.x0 = None
self.x1 = None
self.y = None
self.xlabel = None
self.ylabel = None
self.title = 'Model'
self.levels = None
super().__init__()
def __str__(self):
x0 = self.x0
if self.x0 is not None:
x0 = np.array2string(self.x0, separator=',', precision=4,
suppress_small=False)
x1 = self.x1
if self.x1 is not None:
x1 = np.array2string(self.x1, separator=',', precision=4,
suppress_small=False)
y = self.y
if self.y is not None:
y = np.array2string(self.y, separator=',', precision=4,
suppress_small=False)
return (('x0 = %s\n' +
'x1 = %s\n' +
'y = %s\n' +
'xlabel = %s\n' +
'ylabel = %s\n' +
'title = %s\n' +
'levels = %s\n' +
'contour_prefs = %s') %
(x0,
x1,
y,
self.xlabel,
self.ylabel,
self.title,
self.levels,
self.contour_prefs))
def _repr_html_(self):
"""Return a HTML (string) representation of the model contour plot."""
return backend.as_html_modelcontour(self)
[docs]
def prepare(self, data, model, stat):
(self.x0, self.x1, self.y, self.xlabel,
self.ylabel) = data.to_contour(yfunc=model)
self.y = self.y[1]
[docs]
def contour(self, overcontour=False, clearwindow=True, **kwargs):
Contour.contour(self, self.x0, self.x1, self.y, levels=self.levels,
title=self.title, xlabel=self.xlabel,
ylabel=self.ylabel, overcontour=overcontour,
clearwindow=clearwindow, **kwargs)
[docs]
class PSFContour(DataContour):
"Derived class for creating 2D PSF contours"
[docs]
def prepare(self, psf, data=None, stat=None):
psfdata = psf.get_kernel(data, False)
DataContour.prepare(self, psfdata)
self.title = psf.kernel.name
[docs]
class SourceContour(ModelContour):
"Derived class for creating 2D model contours"
def __init__(self):
super().__init__()
self.title = 'Source'
[docs]
class FitPlot(Plot):
"""Combine data and model plots for 1D data.
Attributes
----------
dataplot
The Sherpa plot object used to display the data.
modelplot
The Sherpa plot object used to display the model.
Examples
--------
If dplot and mplot are data and model plots respectively,
such as those created in the examples for `DataPlot` and `ModelPlot`,
then a combined plot can be created with:
>>> from sherpa.data import Data1D
>>> from sherpa.models.basic import StepLo1D
>>> from sherpa.plot import DataPlot, ModelPlot, FitPlot
>>> data = Data1D('a dataset', [10, 20, 25], [2, -7, 4])
>>> dplot = DataPlot()
>>> fplot = FitPlot()
>>> model = StepLo1D()
>>> model.xcut = 19
>>> mplot = ModelPlot()
>>> dplot.prepare(data)
>>> mplot.prepare(data, model)
>>> fplot.prepare(dplot, mplot)
>>> fplot.plot()
Keyword arguments can be given in the `plot` call, and these
are passed through to both the data and model plots (in the
following example the Matplotlib backend is assumed to be in use):
>>> fplot.plot(color='k')
"""
plot_prefs = basicbackend.get_fit_plot_defaults()
"""The preferences for the plot. Note that the display for the
data and model plots are controlled by the preferences for
the dataplot and modelplot objects, so this is currently
unused.
"""
def __init__(self):
self.dataplot = None
self.modelplot = None
super().__init__()
def __str__(self):
data_title = None
if self.dataplot is not None:
data_title = self.dataplot.title
model_title = None
if self.modelplot is not None:
model_title = self.modelplot.title
return (('dataplot = %s\n%s\n\nmodelplot = %s\n%s') %
(data_title,
self.dataplot,
model_title,
self.modelplot))
def _repr_html_(self):
"""Return a HTML (string) representation of the fit plot."""
return backend.as_html_fit(self)
[docs]
def prepare(self, dataplot, modelplot):
"""Create the data to plot
Parameters
----------
dataplot
The Sherpa plot object representing the data (normally a
DataPlot instance).
modelplot
The Sherpa plot object representing the model (normally a
ModelPlot instance).
See Also
--------
plot
"""
self.dataplot = dataplot
self.modelplot = modelplot
[docs]
def plot(self, overplot=False, clearwindow=True, **kwargs):
"""Plot the data.
This will plot the data sent to the prepare method.
Parameters
----------
overplot : bool, optional
If `True` then add the data to an existing plot, otherwise
create a new plot.
clearwindow : bool, optional
Should the existing plot area be cleared before creating this
new plot (e.g. for multi-panel plots)?
**kwargs
These values are passed on to the plot backend, and must
match the names of the keys of the object's
plot_prefs dictionary. They are sent to both the dataplot
and modelplot instances.
See Also
--------
prepare, overplot
"""
if self.dataplot is None or self.modelplot is None:
raise PlotErr("nodataormodel")
# Note: the user preferences are sent to *both* the data and
# model plot. Is this a good idea?
#
self.dataplot.plot(overplot=overplot, clearwindow=clearwindow,
**kwargs)
self.modelplot.overplot(**kwargs)
[docs]
class FitContour(Contour):
"Derived class for creating 2D combination data and model contours"
contour_prefs = basicbackend.get_fit_contour_defaults()
"The preferences for the plot."
def __init__(self):
self.datacontour = None
self.modelcontour = None
super().__init__()
def __str__(self):
data_title = None
if self.datacontour is not None:
data_title = self.datacontour.title
model_title = None
if self.modelcontour is not None:
model_title = self.modelcontour.title
return (('datacontour = %s\n%s\n\nmodelcontour = %s\n%s') %
(data_title,
self.datacontour,
model_title,
self.modelcontour))
def _repr_html_(self):
"""Return a HTML (string) representation of the fit contour plot."""
return backend.as_html_fitcontour(self)
[docs]
def prepare(self, datacontour, modelcontour):
self.datacontour = datacontour
self.modelcontour = modelcontour
[docs]
def contour(self, overcontour=False, clearwindow=True, **kwargs):
# Note: the user arguments are applied to both plots
self.datacontour.contour(overcontour=overcontour,
clearwindow=clearwindow, **kwargs)
self.modelcontour.overcontour(**kwargs)
[docs]
class DelchiPlot(ModelPlot):
"""Create plots of the delta-chi value per point.
The value of (data-model)/error is plotted for each point.
Attributes
----------
x : array_like
The X value for each point.
y : array_like
The Y value for each point: (data-model)/error
xerr : array_like
The half-width of each X "bin", if set.
yerr : array_like
The error on the Y value (each element is `1`).
xlabel, ylabel, title : str
Plot labels.
Notes
-----
The ylog setting is ignored, whether given as a preference or
a keyword argument, so the Y axis is always drawn with a
linear scale.
"""
plot_prefs = basicbackend.get_resid_plot_defaults()
"The preferences for the plot."
def _calc_delchi(self, ylist, staterr):
return (ylist[0] - ylist[1]) / staterr
[docs]
def prepare(self, data, model, stat):
(self.x, y, staterr, self.xerr,
self.xlabel, self.ylabel) = data.to_plot(model)
if staterr is None:
if stat.name in _stats_noerr:
raise StatErr('badstat', "DelchiPlot", stat.name)
staterr = data.get_yerr(True, stat.calc_staterror)
self.y = self._calc_delchi(y, staterr)
self.yerr = staterr / staterr
self.ylabel = 'Sigma'
self.title = _make_title('Sigma Residuals', data.name)
[docs]
def plot(self, overplot=False, clearwindow=True, **kwargs):
self.plot_prefs['ylog'] = False
kwargs.pop('ylog', True)
Plot.plot(self, self.x, self.y, yerr=self.yerr, xerr=self.xerr,
title=self.title, xlabel=self.xlabel, ylabel=self.ylabel,
overplot=overplot, clearwindow=clearwindow, **kwargs)
super().hline(y=0, xmin=0, xmax=1, linecolor='k',
linewidth=.8, overplot=True)
[docs]
class ChisqrPlot(ModelPlot):
"""Create plots of the chi-square value per point.
The value of ((data-model)/error)^2 is plotted for each point.
Attributes
----------
x : array_like
The X value for each point.
y : array_like
The Y value for each point: ((data-model)/error)^2
xerr : array_like
The half-width of each X "bin", if set.
yerr : array_like
The error on the Y value. Will be `None` here.
xlabel, ylabel, title : str
Plot labels.
"""
plot_prefs = basicbackend.get_model_plot_defaults()
"The preferences for the plot."
def _calc_chisqr(self, ylist, staterr):
dy = ylist[0] - ylist[1]
return dy * dy / (staterr * staterr)
[docs]
def prepare(self, data, model, stat):
(self.x, y, staterr, self.xerr,
self.xlabel, self.ylabel) = data.to_plot(model)
# if staterr is None:
if stat.name in _stats_noerr:
raise StatErr('badstat', "ChisqrPlot", stat.name)
staterr = data.get_yerr(True, stat.calc_staterror)
self.y = self._calc_chisqr(y, staterr)
self.ylabel = backend.get_latex_for_string(r'\chi^2')
self.title = _make_title(
backend.get_latex_for_string(r'\chi^2'), data.name)
[docs]
def plot(self, overplot=False, clearwindow=True, **kwargs):
Plot.plot(self, self.x, self.y, title=self.title,
xlabel=self.xlabel, ylabel=self.ylabel,
overplot=overplot, clearwindow=clearwindow, **kwargs)
[docs]
class ResidPlot(ModelPlot):
"""Create plots of the residuals (data - model) per point.
The value of (data-model) is plotted for each point.
Attributes
----------
x : array_like
The X value for each point.
y : array_like
The Y value for each point: data - model.
xerr : array_like
The half-width of each X "bin", if set.
yerr : array_like
The error on the Y value, if set.
xlabel, ylabel, title : str
Plot labels.
Notes
-----
The ylog setting is ignored, whether given as a preference or
a keyword argument, so the Y axis is always drawn with a
linear scale.
"""
plot_prefs = basicbackend.get_resid_plot_defaults()
"The preferences for the plot."
def _calc_resid(self, ylist):
return ylist[0] - ylist[1]
[docs]
def prepare(self, data, model, stat):
(self.x, y, self.yerr, self.xerr,
self.xlabel, self.ylabel) = data.to_plot(model)
self.y = self._calc_resid(y)
# See the discussion in DataPlot.prepare
try:
yerrorbars = self.plot_prefs['yerrorbars']
except KeyError:
yerrorbars = True
if stat.name in _stats_noerr:
self.yerr = data.get_yerr(True, Chi2XspecVar.calc_staterror)
if yerrorbars:
warning(_errorbar_warning(stat))
else:
self.yerr = data.get_yerr(True, stat.calc_staterror)
# Some data sets (e.g. DataPHA, which shows the units) have a y
# label that could (should?) be displayed (or added to the label).
# To avoid a change in behavior, the label is only changed if
# the "generic" Y axis label is used. To be reviewed.
#
if self.ylabel == 'y':
self.ylabel = 'Data - Model'
self.title = _make_title('Residuals', data.name)
[docs]
def plot(self, overplot=False, clearwindow=True, **kwargs):
self.plot_prefs['ylog'] = False
kwargs.pop('ylog', True)
Plot.plot(self, self.x, self.y, yerr=self.yerr, xerr=self.xerr,
title=self.title, xlabel=self.xlabel, ylabel=self.ylabel,
overplot=overplot, clearwindow=clearwindow, **kwargs)
super().hline(y=0, xmin=0, xmax=1, linecolor='k',
linewidth=.8, overplot=True)
[docs]
class ResidContour(ModelContour):
"Derived class for creating 2D residual contours (data-model)"
contour_prefs = basicbackend.get_resid_contour_defaults()
"The preferences for the plot."
def _calc_resid(self, ylist):
return ylist[0] - ylist[1]
[docs]
def prepare(self, data, model, stat):
(self.x0, self.x1, self.y, self.xlabel,
self.ylabel) = data.to_contour(yfunc=model)
self.y = self._calc_resid(self.y)
self.title = _make_title('Residuals', data.name)
[docs]
def contour(self, overcontour=False, clearwindow=True, **kwargs):
Contour.contour(self, self.x0, self.x1, self.y, levels=self.levels,
title=self.title,
xlabel=self.xlabel, ylabel=self.ylabel,
overcontour=overcontour, clearwindow=clearwindow,
**kwargs)
[docs]
class RatioPlot(ModelPlot):
"""Create plots of the ratio of data to model per point.
The value of data / model is plotted for each point.
Attributes
----------
x : array_like
The X value for each point.
y : array_like
The Y value for each point: data / model.
xerr : array_like
The half-width of each X "bin", if set.
yerr : array_like
The error on the Y value, if set.
xlabel, ylabel, title : str
Plot labels.
Notes
-----
The ylog setting is ignored, whether given as a preference or
a keyword argument, so the Y axis is always drawn with a
linear scale.
"""
plot_prefs = basicbackend.get_ratio_plot_defaults()
"The preferences for the plot."
def _calc_ratio(self, ylist):
data = np.array(ylist[0])
model = np.asarray(ylist[1])
bad = np.where(model == 0.0)
data[bad] = 0.0
model[bad] = 1.0
return data / model
[docs]
def prepare(self, data, model, stat):
(self.x, y, self.yerr, self.xerr,
self.xlabel, self.ylabel) = data.to_plot(model)
self.y = self._calc_ratio(y)
# See the discussion in DataPlot.prepare
try:
yerrorbars = self.plot_prefs['yerrorbars']
except KeyError:
yerrorbars = True
if stat.name in _stats_noerr:
self.yerr = data.get_yerr(True, Chi2XspecVar.calc_staterror)
self.yerr = self.yerr / y[1]
if yerrorbars:
warning(_errorbar_warning(stat))
else:
staterr = data.get_yerr(True, stat.calc_staterror)
self.yerr = staterr / y[1]
self.ylabel = 'Data / Model'
self.title = _make_title('Ratio of Data to Model', data.name)
[docs]
def plot(self, overplot=False, clearwindow=True, **kwargs):
self.plot_prefs['ylog'] = False
kwargs.pop('ylog', True)
Plot.plot(self, self.x, self.y, yerr=self.yerr, xerr=self.xerr,
title=self.title, xlabel=self.xlabel, ylabel=self.ylabel,
overplot=overplot, clearwindow=clearwindow, **kwargs)
super().hline(y=1, xmin=0, xmax=1, linecolor='k',
linewidth=.8, overplot=True)
[docs]
class RatioContour(ModelContour):
"Derived class for creating 2D ratio contours (data divided by model)"
contour_prefs = basicbackend.get_ratio_contour_defaults()
"The preferences for the plot."
def _calc_ratio(self, ylist):
data = np.array(ylist[0])
model = np.asarray(ylist[1])
bad = np.where(model == 0.0)
data[bad] = 0.0
model[bad] = 1.0
return data / model
[docs]
def prepare(self, data, model, stat):
(self.x0, self.x1, self.y, self.xlabel,
self.ylabel) = data.to_contour(yfunc=model)
self.y = self._calc_ratio(self.y)
self.title = _make_title('Ratio of Data to Model', data.name)
[docs]
def contour(self, overcontour=False, clearwindow=True, **kwargs):
Contour.contour(self, self.x0, self.x1, self.y, levels=self.levels,
title=self.title, xlabel=self.xlabel,
ylabel=self.ylabel,
overcontour=overcontour, clearwindow=clearwindow,
**kwargs)
def calc_par_range(minval: float,
maxval: float,
nloop: int,
delv: Optional[float] = None,
log: Optional[bool] = False) -> np.ndarray:
"""Calculate the parameter range to use.
This assumes that the arguments have already been checked for
validity.
Parameters
----------
minval, maxval : number
The requested parameter range, with maxval > minval.
nloop : int
The number of values to create. This is not used if delv is
set.
delv : float or None, optional
If set this is the spacing to use (and over-rides the nloop
parameter). It must be positive and, when `log` is set, the
spacing is 10**delv.
log : bool, optional
Should the spacing be linear or logarithmic?
Returns
-------
pars : ndarray
The parameter values.
Raises
------
ConfidenceErr
If either minval or maxval are not positive definite when the
`log` option is set.
"""
if delv is not None:
# This assumes that delv > eps, but this should be safe.
# The float(eps) term is for mypy.
#
eps = np.finfo(np.float32).eps
maxval = maxval + float(eps)
if log:
if minval <= 0.0 or maxval <= 0.0:
raise ConfidenceErr('badarg', 'Log scale',
'on positive boundaries')
minval = np.log10(minval)
maxval = np.log10(maxval)
if delv is None:
x = np.linspace(minval, maxval, nloop)
else:
x = np.arange(minval, maxval, delv)
if log:
x = 10**x
return x
[docs]
class Confidence1D(DataPlot):
"""The base class for 1D confidence plots.
.. versionchanged:: 4.16.1
Handling of log-scaled axes and use of the delv argument has
been improved, and the string output now includes the parameter
value (if available).
"""
plot_prefs = basicbackend.get_confid_plot_defaults()
"The preferences for the plot."
# This is only used for an error message
conf_type = "unknown"
"The type of confidence analysis."
def __init__(self):
self.min = None
self.max = None
self.nloop = 20
self.delv = None
self.fac = 1
self.log = False
self.parval = None
self.stat = None
self.numcores = None
super().__init__()
def __setstate__(self, state):
self.__dict__.update(state)
if 'stat' not in state:
self.__dict__['stat'] = None
if 'parval' not in state:
self.__dict__['parval'] = None
if 'numcores' not in state:
self.__dict__['numcores'] = None
def __str__(self):
x = self.x
if self.x is not None:
x = np.array2string(self.x, separator=',', precision=4,
suppress_small=False)
y = self.y
if self.y is not None:
y = np.array2string(self.y, separator=',', precision=4,
suppress_small=False)
return (f'x = {x}\n' +
f'y = {y}\n' +
f'min = {self.min}\n' +
f'max = {self.max}\n' +
f'nloop = {self.nloop}\n' +
f'delv = {self.delv}\n' +
f'fac = {self.fac}\n' +
f'log = {self.log}\n' +
f'parval = {self.parval}')
def _repr_html_(self):
"""Return a HTML (string) representation of the confidence 1D plot."""
return backend.as_html_contour1d(self)
[docs]
def prepare(self, min=None, max=None, nloop=20,
delv=None, fac=1, log=False, numcores=None):
"""Set the data to plot.
This defines the range over which the statistic will be
calculated, but does not perform the evaluation.
Parameters
----------
min, max : number or None, optional
The minimum and maximum parameter value to used. If either
is not set then the range is calculated using the fac
parameter.
nloop : int, optional
The number of points at which to evaluate the
statistic. It must be greater than 1. This is used when
delv is set to None.
delv : number or None, optional
The spacing of the parameter grid. This takes precedence
over nloop.
fac : number, optional
Used when either min or max are not set. The parameter
range in this case is taken to be fac times the separation
of the covariance limits for the parameter (unless
explicitly given).
log : bool, optional
Should the parameter be evaluated on a
logarithmically-spaced grid rather than a linearly-spaced
one?
numcores : int or None, optional
Should the parameter evaluation use multiple CPU cores if
available?
See Also
--------
calc
"""
self.min = min
self.max = max
self.nloop = nloop
self.delv = delv
self.fac = fac
self.log = log
self.numcores = numcores
def _interval_init(self, fit, par):
"""Calculate the grid to use for the parameter.
Parameters
----------
fit : sherpa.fit.Fit instance
The current fit.
par : sherpa.models.parameter.Parameter instance
The parameter to analyze.
Returns
-------
x : ndarray
The parameter values to use (also set to self.x)
"""
# Validate the values first (as much as we can).
#
if self.min is not None and not np.isscalar(self.min):
raise ConfidenceErr('badarg', 'Parameter limits', 'scalars')
if self.max is not None and not np.isscalar(self.max):
raise ConfidenceErr('badarg', 'Parameter limits', 'scalars')
if self.nloop <= 1:
raise ConfidenceErr('badarg', 'Nloop parameter', '> 1')
if self.delv is not None and self.delv <= 0:
raise ConfidenceErr('badarg', 'delv parameter', '> 0')
self.stat = fit.calc_stat()
self.parval = par.val
self.xlabel = par.fullname
self.ylabel = 'Statistic Value'
if self.min is None or self.max is None:
oldestmethod = fit.estmethod
fit.estmethod = Covariance()
r = fit.est_errors()
index = list(r.parnames).index(par.fullname)
fit.estmethod = oldestmethod
if self.min is None:
self.min = par.min
minval = r.parmins[index]
if minval is not None and not np.isnan(minval):
self.min = par.val + minval
if self.max is None:
self.max = par.max
maxval = r.parmaxes[index]
if maxval is not None and not np.isnan(maxval):
self.max = par.val + maxval
v = (self.max + self.min) / 2.
dv = np.fabs(v - self.min)
self.min = v - self.fac * dv
self.max = v + self.fac * dv
if self.min < par.min:
self.min = par.min
if self.max > par.max:
self.max = par.max
# check user limits for errors
if self.min >= self.max:
raise ConfidenceErr('badlimits')
self.x = calc_par_range(self.min, self.max, self.nloop,
delv=self.delv, log=self.log)
return self.x
[docs]
def calc(self, fit, par):
"""Evaluate the statistic for the parameter range.
This requires prepare to have been called, and must be
called before plot is called.
Parameters
----------
fit
The Sherpa fit instance to use (defines the statistic
and optimiser to use).
par
The parameter to iterate over.
See Also
--------
plot, prepare
Notes
-----
This method is assumed to be over-ridden in derived classes,
where it will perform the statistic calculations needed to
create the visualization. This version should be called from
these classes as it validates the fit and par arguments.
"""
if isinstance(fit.stat, LeastSq):
raise ConfidenceErr('badargconf', fit.stat.name)
# Check the parameter
# - is thawed
# - is part of the model expression
#
if par.frozen:
raise ConfidenceErr('frozen', par.fullname,
f'interval {self.conf_type}')
if par not in fit.model.pars:
raise ConfidenceErr('thawed', par.fullname, fit.model.name)
[docs]
def plot(self, overplot=False, clearwindow=True, **kwargs):
if self.log:
self.plot_prefs['xlog'] = True
Plot.plot(self, self.x, self.y, title=self.title, xlabel=self.xlabel,
ylabel=self.ylabel, overplot=overplot,
clearwindow=clearwindow, **kwargs)
# Note: the user arguments are not applied to the lines
#
if self.stat is not None:
Plot.hline(self.stat, linecolor="green", linestyle="dash",
linewidth=1.5, overplot=True, clearwindow=False)
if self.parval is not None:
Plot.vline(self.parval, linecolor="orange", linestyle="dash",
linewidth=1.5, overplot=True, clearwindow=False)
if self.log:
self.plot_prefs['xlog'] = False
[docs]
class Confidence2D(DataContour, Point):
"""The base class for 2D confidence contours.
.. versionchanged:: 4.16.1
Handling of log-scaled axes and use of the delv argument has
been improved.
"""
contour_prefs = basicbackend.get_confid_contour_defaults()
point_prefs = basicbackend.get_confid_point_defaults()
# This is only used for an error message
conf_type = "unknown"
"The type of confidence analysis."
def __init__(self):
self.min = None
self.max = None
self.nloop = (10, 10)
self.fac = 4
self.delv = None
self.log = (False, False)
self.sigma = (1, 2, 3)
self.parval0 = None
self.parval1 = None
self.stat = None
self.numcores = None
super().__init__()
def __setstate__(self, state):
self.__dict__.update(state)
if 'stat' not in state:
self.__dict__['stat'] = None
if 'numcores' not in state:
self.__dict__['numcores'] = None
def __str__(self):
x0 = self.x0
if self.x0 is not None:
x0 = np.array2string(self.x0, separator=',', precision=4,
suppress_small=False)
x1 = self.x1
if self.x1 is not None:
x1 = np.array2string(self.x1, separator=',', precision=4,
suppress_small=False)
y = self.y
if self.y is not None:
y = np.array2string(self.y, separator=',', precision=4,
suppress_small=False)
return (f'x0 = {x0}\n' +
f'x1 = {x1}\n' +
f'y = {y}\n' +
f'min = {self.min}\n' +
f'max = {self.max}\n' +
f'nloop = {self.nloop}\n' +
f'fac = {self.fac}\n' +
f'delv = {self.delv}\n' +
f'log = {self.log}\n' +
f'sigma = {self.sigma}\n' +
f'parval0 = {self.parval0}\n' +
f'parval1 = {self.parval1}\n' +
f'levels = {self.levels}')
def _repr_html_(self):
"""Return a HTML (string) representation of the confidence 2D plot."""
return backend.as_html_contour2d(self)
[docs]
def prepare(self, min=None, max=None, nloop=(10, 10),
delv=None, fac=4, log=(False, False),
sigma=(1, 2, 3), levels=None, numcores=None):
"""Set the data to plot.
This defines the ranges over which the statistic will be
calculated, but does not perform the evaluation.
Parameters
----------
min, max : sequence of number or None, optional
The minimum and maximum parameter values to used. If set
then they must contain two elements, and if not then the
range is calculated using the fac parameter.
nloop : sequence of int, optional
The number of points at which to evaluate the statistic,
where each value must be greater than 1. This is used when
delv is set to None.
delv : sequence of number or None, optional
The spacing of the parameter grids, and if set it must
contain two values each greater than 0. This takes
precedence over nloop.
fac : number, optional
Used when either min or max are not set. The parameter
range in this case is taken to be fac times the separation
of the covariance limits for the parameter (unless
explicitly given).
log : sequence of bool, optional
Should each parameter be evaluated on a
logarithmically-spaced grid rather than a linearly-spaced
one?
sigma : sequence of number, optional
The sigma values at which to draw contours. This is only
used if levels is set to None.
levels : sequence of number or None, optional
The levels at which the contours are drawn. This over-rides
the sigma setting.
numcores : int or None, optional
Should the parameter evaluation use multiple CPU cores if
available?
See Also
--------
calc
"""
self.min = min
self.max = max
self.nloop = nloop
self.delv = delv
self.fac = fac
self.log = log
self.sigma = sigma
self.levels = levels
self.parval0 = None
self.parval1 = None
self.numcores = numcores
def _region_init(self, fit, par0, par1):
"""Calculate the grid to use for the parameters.
Parameters
----------
fit : sherpa.fit.Fit instance
The current fit.
par0, par1 : sherpa.models.parameter.Parameter instance
The parameters to analyze.
Returns
-------
x : 2D ndarray
The parameter values to use. Unlike the Confidence1D case
this is not just self.x0 or self.x1, but is a 2D array
where each row represent each pair of parameters, so the
first column is par0 and the second column is par1.
"""
def check2(value, pname, opt=True):
"""Check value is None (when opt set) or has size of 2"""
if opt and value is None:
return None
if np.isscalar(value):
raise ConfidenceErr('badarg', pname, 'a list')
if len(value) != 2:
raise ConfidenceErr('badarg', pname, 'a list of size 2')
# Ensure we return an ndarray
return np.asarray(value)
# Issue #1093 points out that if min or max is a tuple we can
# have a problem, as below the code can assign to an element
# of one of them. So ensure we have a list not a tuple. The
# exact setting of min/max (whether a tuple, list, or ndarray)
# makes the following code a bit messy.
#
# See also #1967 which discusses what type we should accept
# for fields like min and max.
#
if self.min is not None:
try:
self.min = list(self.min)
except TypeError:
raise ConfidenceErr(
'badarg', 'Parameter limits', 'a list') from None
if self.max is not None:
try:
self.max = list(self.max)
except TypeError:
raise ConfidenceErr(
'badarg', 'Parameter limits', 'a list') from None
# We ignore the return value for min/max
check2(self.min, "Parameter limits")
check2(self.max, "Parameter limits")
nloop = check2(self.nloop, "Nloop parameter", opt=False)
delv = check2(self.delv, "delv parameter")
if np.any(nloop <= 1):
raise ConfidenceErr('badarg', 'Nloop parameter',
'a list with elements > 1')
if delv is not None and np.any(delv <= 0):
raise ConfidenceErr('badarg', 'delv parameter',
'a list with elements > 0')
self.stat = fit.calc_stat()
self.xlabel = par0.fullname
self.ylabel = par1.fullname
self.parval0 = par0.val
self.parval1 = par1.val
if self.levels is None:
stat = self.stat
if self.sigma is None or np.isscalar(self.sigma):
raise ConfidenceErr('needlist', 'sigma bounds')
sigma = np.asarray(self.sigma)
lvls = stat - (2. * np.log(1. - erf(sigma / np.sqrt(2.))))
self.levels = np.asarray(lvls, dtype=SherpaFloat)
if self.min is None or self.max is None:
oldestmethod = fit.estmethod
fit.estmethod = Covariance()
r = fit.est_errors()
fit.estmethod = oldestmethod
index0 = list(r.parnames).index(par0.fullname)
index1 = list(r.parnames).index(par1.fullname)
if self.min is None:
self.min = np.array([par0.min, par1.min])
min0 = r.parmins[index0]
min1 = r.parmins[index1]
if min0 is not None and not np.isnan(min0):
self.min[0] = par0.val + min0
if min1 is not None and not np.isnan(min1):
self.min[1] = par1.val + min1
if self.max is None:
self.max = np.array([par0.max, par1.max])
max0 = r.parmaxes[index0]
max1 = r.parmaxes[index1]
if max0 is not None and not np.isnan(max0):
self.max[0] = par0.val + max0
if max1 is not None and not np.isnan(max1):
self.max[1] = par1.val + max1
# This assumes that self.min/max are ndarray
v = (self.max + self.min) / 2.
dv = np.fabs(v - self.min)
self.min = v - self.fac * dv
self.max = v + self.fac * dv
hmin = np.array([par0.min, par1.min])
hmax = np.array([par0.max, par1.max])
for i in [0, 1]:
# check user limits for errors
if self.min[i] >= self.max[i]:
raise ConfidenceErr('badlimits')
if self.min[i] < hmin[i]:
self.min[i] = hmin[i]
if self.max[i] > hmax[i]:
self.max[i] = hmax[i]
if delv is None:
# Just make the following code easier to write
delv = [None, None]
x0 = calc_par_range(self.min[0], self.max[0],
self.nloop[0], delv=delv[0],
log=self.log[0])
x1 = calc_par_range(self.min[1], self.max[1],
self.nloop[1], delv=delv[1],
log=self.log[1])
# Does it matter whether this is x0,x1 or x1,x0? Not for the
# calculation of the parameter values, but the contour
# plotting may care (but at least self.x0 and self.x1 should
# match the ordering of self.y).
#
x0g, x1g = np.meshgrid(x0, x1)
self.x0 = x0g.flatten()
self.x1 = x1g.flatten()
return np.array([self.x0, self.x1]).T
[docs]
def calc(self, fit, par0, par1):
"""Evaluate the statistic for the parameter range.
This requires prepare to have been called, and must be
called before contour is called.
Parameters
----------
fit
The Sherpa fit instance to use (defines the statistic
and optimiser to use).
par0, par1
The parameters to iterate over.
See Also
--------
contour, prepare
Notes
-----
This method is assumed to be over-ridden in derived classes,
where it will perform the statistic calculations needed to
create the visualization. This version should be called from
these classes as it validates the fit and par arguments.
"""
if isinstance(fit.stat, LeastSq):
raise ConfidenceErr('badargconf', fit.stat.name)
# Check that each parameter
# - is thawed
# - is part of the model expression
#
for par in [par0, par1]:
if par.frozen:
raise ConfidenceErr('frozen', par.fullname,
f'region {self.conf_type}')
if par not in fit.model.pars:
raise ConfidenceErr('thawed', par.fullname, fit.model.name)
# TODO: should this be overcontour rather than overplot?
[docs]
def contour(self, overplot=False, clearwindow=True, **kwargs):
if self.log[0]:
self.contour_prefs['xlog'] = True
if self.log[1]:
self.contour_prefs['ylog'] = True
# Work around possible naming issue with overplot/overcontour.
# Set overcontour if either overcontour or overplot are set.
# The overcontour argument must be removed from kwargs if
# sent, hence it us used first here (if second lazy evaluation
# could mean the pop is never executed).
#
overcontour = kwargs.pop('overcontour', False) or overplot
Contour.contour(self, self.x0, self.x1, self.y, levels=self.levels,
title=self.title, xlabel=self.xlabel,
ylabel=self.ylabel, overcontour=overcontour,
clearwindow=clearwindow, **kwargs)
# Note: the user arguments are not applied to the point
#
point = Point()
point.point_prefs = self.point_prefs
point.point(self.parval0, self.parval1,
overplot=True, clearwindow=False)
if self.log[0]:
self.contour_prefs['xlog'] = False
if self.log[1]:
self.contour_prefs['ylog'] = False
class IntervalProjectionWorker():
"""Used to evaluate the model by IntervalProjection.
.. versionchanged:: 4.16.1
The calling convention was changed.
See Also
--------
IntervalUncertaintyWorker
"""
def __init__(self, par, fit, otherpars):
self.par = par
self.fit = fit
self.otherpars = otherpars
def __call__(self, val):
self.par.val = val
# It there are other parameters then we need to fit to get the
# best-fit statistic.
#
if self.otherpars:
r = self.fit.fit()
return r.statval
# If this was the only free parameter we can just calculate
# the statistic value.
#
return self.fit.calc_stat()
[docs]
class IntervalProjection(Confidence1D):
"""The Interval-Projection method.
Evaluate the parameter value on a grid of points, allowing the
other thawed parameters to be fit.
.. versionchanged:: 4.16.1
Handling of log-scaled axes has been improved and the string
output now includes the parameter value (if available).
See Also
--------
IntervalUncertainty
"""
conf_type = "projection"
def __init__(self):
self.fast = True
super().__init__()
[docs]
def prepare(self, fast=True, min=None, max=None, nloop=20,
delv=None, fac=1, log=False, numcores=None):
self.fast = fast
Confidence1D.prepare(self, min, max, nloop, delv, fac, log, numcores)
[docs]
def calc(self, fit, par, methoddict=None, cache=True):
self.title = 'Interval-Projection'
Confidence1D.calc(self, fit, par)
# If "fast" option enabled, set fitting method to
# lmdif if stat is chi-squared,
# else set to neldermead.
# If current method is not LM or NM, warn it is not a good
# method for estimating parameter limits.
#
if not isinstance(fit.method, (NelderMead, LevMar)):
warning("%s is inappropriate for confidence limit "
"estimation", fit.method.name)
oldfitmethod = fit.method
if bool_cast(self.fast) is True and methoddict is not None:
if isinstance(fit.stat, Likelihood):
if not isinstance(fit.method, NelderMead):
fit.method = methoddict['neldermead']
warning("Setting optimization to %s for interval "
"projection plot", fit.method.name)
elif not isinstance(fit.method, LevMar):
fit.method = methoddict['levmar']
warning("Setting optimization to %s for interval "
"projection plot", fit.method.name)
xvals = self._interval_init(fit, par)
oldpars = fit.model.thawedpars
par.freeze()
# We know that par is thawed, so we can check to see whether a fit
# is needed by looking for other parameters.
#
otherpars = len(oldpars) > 1
# Store these before we enter the try block as they are used
# in the finally block.
#
startup = fit.model.startup
teardown = fit.model.teardown
try:
fit.model.startup(cache)
# these calls are unnecessary for every fit
fit.model.startup = return_none
fit.model.teardown = return_none
worker = IntervalProjectionWorker(par, fit, otherpars)
res = parallel_map(worker, xvals, self.numcores)
self.y = np.asarray(res)
finally:
# Set back data that we changed
par.thaw()
fit.model.startup = startup
fit.model.teardown = teardown
fit.model.teardown()
fit.model.thawedpars = oldpars
fit.method = oldfitmethod
class IntervalUncertaintyWorker():
"""Used to evaluate the model by IntervalUncertainty.
.. versionchanged:: 4.16.1
The calling convention was changed.
See Also
--------
IntervalProjectionWorker
"""
def __init__(self, par, fit):
self.par = par
self.fit = fit
def __call__(self, val):
self.par.val = val
return self.fit.calc_stat()
[docs]
class IntervalUncertainty(Confidence1D):
"""The Interval-Projection method.
Evaluate the parameter value on a grid of points, where the other
thawed parameters are *not* changed from their current values.
.. versionchanged:: 4.16.1
Handling of log-scaled axes has been improved and the string
output now includes the parameter value (if available).
See Also
--------
IntervalProjection
"""
conf_type = "uncertainty"
[docs]
def calc(self, fit, par, methoddict=None, cache=True):
self.title = 'Interval-Uncertainty'
Confidence1D.calc(self, fit, par)
thawed = [p for p in fit.model.pars if not p.frozen]
oldpars = fit.model.thawedpars
xvals = self._interval_init(fit, par)
for p in thawed:
p.freeze()
try:
fit.model.startup(cache)
worker = IntervalUncertaintyWorker(par, fit)
res = parallel_map(worker, xvals, self.numcores)
self.y = np.asarray(res)
finally:
# Set back data that we changed
for p in thawed:
p.thaw()
fit.model.teardown()
fit.model.thawedpars = oldpars
class RegionProjectionWorker():
"""Used to evaluate the model by RegionProjection.
.. versionchanged:: 4.16.1
The calling convention was changed.
See Also
--------
RegionUncertaintyWorker
"""
def __init__(self, par0, par1, fit, otherpars):
self.par0 = par0
self.par1 = par1
self.fit = fit
self.otherpars = otherpars
def __call__(self, pars):
(self.par0.val, self.par1.val) = pars
# It there are other parameters then we need to fit to get the
# best-fit statistic.
#
if self.otherpars:
r = self.fit.fit()
return r.statval
# If these were the only two free parameters we can just
# calculate the statistic value.
#
return self.fit.calc_stat()
def return_none(cache=None):
"""dummy implementation of callback for multiprocessing"""
return None
[docs]
class RegionProjection(Confidence2D):
"""The Region-Projection method.
Evaluate the statistic on a grid of points for two parameters,
where the other thawed parameters are fit for each location.
.. versionchanged:: 4.16.1
Support for logarithmically-spaced grids has been improved.
See Also
--------
RegionUncertainty
"""
conf_type = "projection"
def __init__(self):
self.fast = True
super().__init__()
[docs]
def prepare(self, fast=True, min=None, max=None, nloop=(10, 10),
delv=None, fac=4, log=(False, False),
sigma=(1, 2, 3), levels=None, numcores=None):
self.fast = fast
Confidence2D.prepare(self, min, max, nloop, delv, fac, log,
sigma, levels, numcores)
[docs]
def calc(self, fit, par0, par1, methoddict=None, cache=True):
self.title = 'Region-Projection'
Confidence2D.calc(self, fit, par0, par1)
# If "fast" option enabled, set fitting method to
# lmdif if stat is chi-squared,
# else set to neldermead
# If current method is not LM or NM, warn it is not a good
# method for estimating parameter limits.
#
if not isinstance(fit.method, (NelderMead, LevMar)):
warning("%s is inappropriate for confidence limit "
"estimation", fit.method.name)
oldfitmethod = fit.method
if bool_cast(self.fast) is True and methoddict is not None:
if isinstance(fit.stat, Likelihood):
if not isinstance(fit.method, NelderMead):
fit.method = methoddict['neldermead']
warning("Setting optimization to %s for region "
"projection plot", fit.method.name)
elif not isinstance(fit.method, LevMar):
fit.method = methoddict['levmar']
warning("Setting optimization to %s for region "
"projection plot", fit.method.name)
oldpars = fit.model.thawedpars
otherpars = len(oldpars) > 2
# Store these before we enter the try block as they are used
# in the finally block.
#
startup = fit.model.startup
teardown = fit.model.teardown
try:
fit.model.startup(cache)
# these calls are unnecessary for every fit
fit.model.startup = return_none
fit.model.teardown = return_none
grid = self._region_init(fit, par0, par1)
par0.freeze()
par1.freeze()
worker = RegionProjectionWorker(par0, par1, fit, otherpars)
results = parallel_map(worker, grid, self.numcores)
self.y = np.asarray(results)
finally:
# Set back data after we changed it
par0.thaw()
par1.thaw()
fit.model.startup = startup
fit.model.teardown = teardown
fit.model.teardown()
fit.model.thawedpars = oldpars
fit.method = oldfitmethod
class RegionUncertaintyWorker():
"""Used to evaluate the model by RegionUncertainty.
.. versionchanged:: 4.16.1
The calling convention was changed.
See Also
--------
RegionProjectionWorker
"""
def __init__(self, par0, par1, fit):
self.par0 = par0
self.par1 = par1
self.fit = fit
def __call__(self, pars):
(self.par0.val, self.par1.val) = pars
return self.fit.calc_stat()
[docs]
class RegionUncertainty(Confidence2D):
"""The Region-Projection method.
Evaluate the statistic on a grid of points for two parameters,
where the other thawed parameters are *not* changed from their
current values.
.. versionchanged:: 4.16.1
Support for logarithmically-spaced grids has been improved.
See Also
--------
RegionProjection
"""
conf_type = "uncertainty"
[docs]
def calc(self, fit, par0, par1, methoddict=None, cache=True):
self.title = 'Region-Uncertainty'
Confidence2D.calc(self, fit, par0, par1)
thawed = [i for i in fit.model.pars if not i.frozen]
oldpars = fit.model.thawedpars
try:
fit.model.startup(cache)
grid = self._region_init(fit, par0, par1)
for p in thawed:
p.freeze()
worker = RegionUncertaintyWorker(par0, par1, fit)
result = parallel_map(worker, grid, self.numcores)
self.y = np.asarray(result)
finally:
# Set back data after we changed it
for p in thawed:
p.thaw()
fit.model.teardown()
fit.model.thawedpars = oldpars
class MultiPlot(NoNewAttributesAfterInit):
"""Combine multiple line-style plots.
Allow multiple line-style plots - so those plot classes
that use the `plot` method to display - to be drawn in
the same area. Each plot is added with the add method.
"""
def __init__(self) -> None:
self.plots: list[Union[Plot, HistogramPlot]] = []
self.title = ""
# The typing here says Plot but we actually want the sub-classes
# like DataPlot (i.e. those that use the prepare method to set up
# the data to plot so that the plot method requires no data
# arguments) rather than the actual Plot class.
#
def add(self, plot: Union[Plot, HistogramPlot]) -> None:
"""Add the plot to the list of data to plot.
A copy of the plot object is stored, rather than the
input argument. The `title` attribute can be set or
changed.
Parameters
----------
plot : instance
The plot or histogram object to add. It must have
a `plot` method.
"""
# A copy is stored since often the same underlying object is
# returned by the UI routines. It also allows the code to
# change the self.plots array and not worry about changing the
# calling code.
#
self.plots.append(copy.deepcopy(plot))
def plot(self,
overplot: bool = False,
clearwindow: bool = True,
**kwargs) -> None:
"""Plot the data.
Parameters
----------
overplot : bool, optional
If `True` then add the data to an existing plot, otherwise
create a new plot.
clearwindow : bool, optional
Should the existing plot area be cleared before creating this
new plot (e.g. for multi-panel plots)?
**kwargs
These values are passed on to the plot backend, and must
match the names of the keys of the object's
plot_prefs dictionary. Note that the same arguments are
passed to each plot.
"""
for plot in self.plots:
plot.plot(overplot=overplot, clearwindow=clearwindow,
**kwargs)
# To decide whether we draw a title or not we do rely on
# the clearwindow setting, since it is not guaranteed to
# be set (e.g. if this is being displayed within a
# SplitPlot). So the overplot setting is used to decide
# whether to add in the title or not.
#
if not overplot:
backend.set_title(self.title)
overplot = True
clearwindow = False