clearml/trains/utilities/plotlympl/mplexporter/renderers/base.py
2020-11-25 11:20:49 +02:00

397 lines
15 KiB
Python

import warnings
import itertools
from contextlib import contextmanager
from distutils.version import LooseVersion
import numpy as np
import matplotlib as mpl
from matplotlib import transforms
from .. import utils
from .. import _py3k_compat as py3k
class Renderer(object):
@staticmethod
def ax_zoomable(ax):
return bool(ax and ax.get_navigate())
@staticmethod
def ax_has_xgrid(ax):
if not ax:
return False
_gridOnMajor = ax.xaxis._gridOnMajor if hasattr(ax.xaxis, '_gridOnMajor') \
else ax.xaxis._major_tick_kw['gridOn']
return bool(ax and _gridOnMajor and ax.yaxis.get_gridlines())
@staticmethod
def ax_has_ygrid(ax):
if not ax:
return False
_gridOnMajor = ax.yaxis._gridOnMajor if hasattr(ax.yaxis, '_gridOnMajor') \
else ax.yaxis._major_tick_kw['gridOn']
return bool(ax and _gridOnMajor and ax.yaxis.get_gridlines())
@property
def current_ax_zoomable(self):
return self.ax_zoomable(self._current_ax)
@property
def current_ax_has_xgrid(self):
return self.ax_has_xgrid(self._current_ax)
@property
def current_ax_has_ygrid(self):
return self.ax_has_ygrid(self._current_ax)
@contextmanager
def draw_figure(self, fig, props):
if hasattr(self, "_current_fig") and self._current_fig is not None:
warnings.warn("figure embedded in figure: something is wrong")
self._current_fig = fig
self._fig_props = props
self.open_figure(fig=fig, props=props)
yield
self.close_figure(fig=fig)
self._current_fig = None
self._fig_props = {}
@contextmanager
def draw_axes(self, ax, props):
if hasattr(self, "_current_ax") and self._current_ax is not None:
warnings.warn("axes embedded in axes: something is wrong")
self._current_ax = ax
self._ax_props = props
self.open_axes(ax=ax, props=props)
yield
self.close_axes(ax=ax)
self._current_ax = None
self._ax_props = {}
@contextmanager
def draw_legend(self, legend, props):
self._current_legend = legend
self._legend_props = props
self.open_legend(legend=legend, props=props)
yield
self.close_legend(legend=legend)
self._current_legend = None
self._legend_props = {}
# Following are the functions which should be overloaded in subclasses
def open_figure(self, fig, props):
"""
Begin commands for a particular figure.
Parameters
----------
fig : matplotlib.Figure
The Figure which will contain the ensuing axes and elements
props : dictionary
The dictionary of figure properties
"""
pass
def close_figure(self, fig):
"""
Finish commands for a particular figure.
Parameters
----------
fig : matplotlib.Figure
The figure which is finished being drawn.
"""
pass
def open_axes(self, ax, props):
"""
Begin commands for a particular axes.
Parameters
----------
ax : matplotlib.Axes
The Axes which will contain the ensuing axes and elements
props : dictionary
The dictionary of axes properties
"""
pass
def close_axes(self, ax):
"""
Finish commands for a particular axes.
Parameters
----------
ax : matplotlib.Axes
The Axes which is finished being drawn.
"""
pass
def open_legend(self, legend, props):
"""
Beging commands for a particular legend.
Parameters
----------
legend : matplotlib.legend.Legend
The Legend that will contain the ensuing elements
props : dictionary
The dictionary of legend properties
"""
pass
def close_legend(self, legend):
"""
Finish commands for a particular legend.
Parameters
----------
legend : matplotlib.legend.Legend
The Legend which is finished being drawn
"""
pass
def draw_marked_line(self, data, coordinates, linestyle, markerstyle,
label, mplobj=None):
"""Draw a line that also has markers.
If this isn't reimplemented by a renderer object, by default, it will
make a call to BOTH draw_line and draw_markers when both markerstyle
and linestyle are not None in the same Line2D object.
"""
if linestyle is not None:
self.draw_line(data, coordinates, linestyle, label, mplobj)
if markerstyle is not None:
self.draw_markers(data, coordinates, markerstyle, label, mplobj)
def draw_line(self, data, coordinates, style, label, mplobj=None):
"""
Draw a line. By default, draw the line via the draw_path() command.
Some renderers might wish to override this and provide more
fine-grained behavior.
In matplotlib, lines are generally created via the plt.plot() command,
though this command also can create marker collections.
Parameters
----------
data : array_like
A shape (N, 2) array of datapoints.
coordinates : string
A string code, which should be either 'data' for data coordinates,
or 'figure' for figure (pixel) coordinates.
style : dictionary
a dictionary specifying the appearance of the line.
mplobj : matplotlib object
the matplotlib plot element which generated this line
"""
pathcodes = ['M'] + (data.shape[0] - 1) * ['L']
pathstyle = dict(facecolor='none', **style)
pathstyle['edgecolor'] = pathstyle.pop('color')
pathstyle['edgewidth'] = pathstyle.pop('linewidth')
self.draw_path(data=data, coordinates=coordinates,
pathcodes=pathcodes, style=pathstyle, mplobj=mplobj)
@staticmethod
def _iter_path_collection(paths, path_transforms, offsets, styles):
"""Build an iterator over the elements of the path collection"""
N = max(len(paths), len(offsets))
# Before mpl 1.4.0, path_transform can be a false-y value, not a valid
# transformation matrix.
if LooseVersion(mpl.__version__) < LooseVersion('1.4.0'):
if path_transforms is None:
path_transforms = [np.eye(3)]
edgecolor = styles['edgecolor']
if np.size(edgecolor) == 0:
edgecolor = ['none']
facecolor = styles['facecolor']
if np.size(facecolor) == 0:
facecolor = ['none']
elements = [paths, path_transforms, offsets,
edgecolor, styles['linewidth'], facecolor]
it = itertools
return it.islice(py3k.zip(*py3k.map(it.cycle, elements)), N)
def draw_path_collection(self, paths, path_coordinates, path_transforms,
offsets, offset_coordinates, offset_order,
styles, mplobj=None):
"""
Draw a collection of paths. The paths, offsets, and styles are all
iterables, and the number of paths is max(len(paths), len(offsets)).
By default, this is implemented via multiple calls to the draw_path()
function. For efficiency, Renderers may choose to customize this
implementation.
Examples of path collections created by matplotlib are scatter plots,
histograms, contour plots, and many others.
Parameters
----------
paths : list
list of tuples, where each tuple has two elements:
(data, pathcodes). See draw_path() for a description of these.
path_coordinates: string
the coordinates code for the paths, which should be either
'data' for data coordinates, or 'figure' for figure (pixel)
coordinates.
path_transforms: array_like
an array of shape (*, 3, 3), giving a series of 2D Affine
transforms for the paths. These encode translations, rotations,
and scalings in the standard way.
offsets: array_like
An array of offsets of shape (N, 2)
offset_coordinates : string
the coordinates code for the offsets, which should be either
'data' for data coordinates, or 'figure' for figure (pixel)
coordinates.
offset_order : string
either "before" or "after". This specifies whether the offset
is applied before the path transform, or after. The matplotlib
backend equivalent is "before"->"data", "after"->"screen".
styles: dictionary
A dictionary in which each value is a list of length N, containing
the style(s) for the paths.
mplobj : matplotlib object
the matplotlib plot element which generated this collection
"""
if offset_order == "before":
raise NotImplementedError("offset before transform")
for tup in self._iter_path_collection(paths, path_transforms,
offsets, styles):
(path, path_transform, offset, ec, lw, fc) = tup
vertices, pathcodes = path
path_transform = transforms.Affine2D(path_transform)
vertices = path_transform.transform(vertices)
# This is a hack:
if path_coordinates == "figure":
path_coordinates = "points"
style = {"edgecolor": utils.export_color(ec),
"facecolor": utils.export_color(fc),
"edgewidth": lw,
"dasharray": "10,0",
"alpha": styles['alpha'],
"zorder": styles['zorder']}
self.draw_path(data=vertices, coordinates=path_coordinates,
pathcodes=pathcodes, style=style, offset=offset,
offset_coordinates=offset_coordinates,
mplobj=mplobj)
def draw_markers(self, data, coordinates, style, label, mplobj=None):
"""
Draw a set of markers. By default, this is done by repeatedly
calling draw_path(), but renderers should generally overload
this method to provide a more efficient implementation.
In matplotlib, markers are created using the plt.plot() command.
Parameters
----------
data : array_like
A shape (N, 2) array of datapoints.
coordinates : string
A string code, which should be either 'data' for data coordinates,
or 'figure' for figure (pixel) coordinates.
style : dictionary
a dictionary specifying the appearance of the markers.
mplobj : matplotlib object
the matplotlib plot element which generated this marker collection
"""
vertices, pathcodes = style['markerpath']
pathstyle = dict((key, style[key]) for key in ['alpha', 'edgecolor',
'facecolor', 'zorder',
'edgewidth'])
pathstyle['dasharray'] = "10,0"
for vertex in data:
self.draw_path(data=vertices, coordinates="points",
pathcodes=pathcodes, style=pathstyle,
offset=vertex, offset_coordinates=coordinates,
mplobj=mplobj)
def draw_text(self, text, position, coordinates, style,
text_type=None, mplobj=None):
"""
Draw text on the image.
Parameters
----------
text : string
The text to draw
position : tuple
The (x, y) position of the text
coordinates : string
A string code, which should be either 'data' for data coordinates,
or 'figure' for figure (pixel) coordinates.
style : dictionary
a dictionary specifying the appearance of the text.
text_type : string or None
if specified, a type of text such as "xlabel", "ylabel", "title"
mplobj : matplotlib object
the matplotlib plot element which generated this text
"""
raise NotImplementedError()
def draw_path(self, data, coordinates, pathcodes, style,
offset=None, offset_coordinates="data", mplobj=None):
"""
Draw a path.
In matplotlib, paths are created by filled regions, histograms,
contour plots, patches, etc.
Parameters
----------
data : array_like
A shape (N, 2) array of datapoints.
coordinates : string
A string code, which should be either 'data' for data coordinates,
'figure' for figure (pixel) coordinates, or "points" for raw
point coordinates (useful in conjunction with offsets, below).
pathcodes : list
A list of single-character SVG pathcodes associated with the data.
Path codes are one of ['M', 'm', 'L', 'l', 'Q', 'q', 'T', 't',
'S', 's', 'C', 'c', 'Z', 'z']
See the SVG specification for details. Note that some path codes
consume more than one datapoint (while 'Z' consumes none), so
in general, the length of the pathcodes list will not be the same
as that of the data array.
style : dictionary
a dictionary specifying the appearance of the line.
offset : list (optional)
the (x, y) offset of the path. If not given, no offset will
be used.
offset_coordinates : string (optional)
A string code, which should be either 'data' for data coordinates,
or 'figure' for figure (pixel) coordinates.
mplobj : matplotlib object
the matplotlib plot element which generated this path
"""
raise NotImplementedError()
def draw_image(self, imdata, extent, coordinates, style, mplobj=None):
"""
Draw an image.
Parameters
----------
imdata : string
base64 encoded png representation of the image
extent : list
the axes extent of the image: [xmin, xmax, ymin, ymax]
coordinates: string
A string code, which should be either 'data' for data coordinates,
or 'figure' for figure (pixel) coordinates.
style : dictionary
a dictionary specifying the appearance of the image
mplobj : matplotlib object
the matplotlib plot object which generated this image
"""
raise NotImplementedError()