def scatter(*args, **kwargs):
"""
This will plot a scatterplot of x and y, iterating over the ColorBrewer
"Set2" color cycle unless a color is specified. The symbols produced are
empty circles, with the outline in the color specified by either 'color'
or 'edgecolor'. If you want to fill the circle, specify 'facecolor'.
Besides the matplotlib scatter(), will also take the parameter
@param show_ticks: Whether or not to show the x and y axis ticks
"""
# Force 'color' to indicate the edge color, so the middle of the
# scatter patches are empty. Can specify
ax, args, kwargs = utils.maybe_get_ax(*args, **kwargs)
if 'color' not in kwargs:
# Assume that color means the edge color. You can assign the
color_cycle = ax._get_lines.color_cycle
kwargs['color'] = next(color_cycle)
if 'edgecolor' not in kwargs:
kwargs['edgecolor'] = almost_black
if 'alpha' not in kwargs:
kwargs['alpha'] = 0.5
lw = utils.maybe_get_linewidth(**kwargs)
kwargs['lw'] = lw
show_ticks = kwargs.pop('show_ticks', False)
scatterpoints = ax.scatter(*args, **kwargs)
utils.remove_chartjunk(ax, ['top', 'right'], show_ticks=show_ticks)
return ax
def plot(*args, **kwargs):
ax, args, kwargs = maybe_get_ax(*args, **kwargs)
show_ticks = kwargs.pop('show_ticks', False)
lines = ax.plot(*args, **kwargs)
remove_chartjunk(ax, ['top', 'right'], show_ticks=show_ticks)
return lines
def scatter(*args, **kwargs):
"""
This will plot a scatterplot of x and y, iterating over the ColorBrewer
"Set2" color cycle unless a color is specified. The symbols produced are
empty circles, with the outline in the color specified by either 'color'
or 'edgecolor'. If you want to fill the circle, specify 'facecolor'.
Besides the matplotlib scatter(), will also take the parameter
@param show_ticks: Whether or not to show the x and y axis ticks
"""
# Force 'color' to indicate the edge color, so the middle of the
# scatter patches are empty. Can specify
ax, args, kwargs = utils.maybe_get_ax(*args, **kwargs)
if 'color' not in kwargs:
# Assume that color means the edge color. You can assign the
color_cycle = cycle(mpl.rcParams['axes.color_cycle'])
kwargs['color'] = next(color_cycle)
kwargs.setdefault('edgecolor', almost_black)
kwargs.setdefault('alpha', 0.5)
lw = utils.maybe_get_linewidth(**kwargs)
kwargs['lw'] = lw
show_ticks = kwargs.pop('show_ticks', False)
scatterpoints = ax.scatter(*args, **kwargs)
utils.remove_chartjunk(ax, ['top', 'right'], show_ticks=show_ticks)
return ax
def hist(*args, **kwargs):
"""
Plots a histogram of the provided data. Can provide optional argument
"grid='x'" or "grid='y'" to draw a white grid over the histogram. Almost like "erasing" some of the plot,
but it adds more information!
"""
ax, args, kwargs = maybe_get_ax(*args, **kwargs)
color_cycle = ax._get_lines.color_cycle
color = next(color_cycle)
# Reassign the default colors to Set2 by Colorbrewer
if 'color' not in kwargs:
kwargs['color'] = color
if 'facecolor' not in kwargs:
kwargs['facecolor'] = color
if 'edgecolor' not in kwargs:
kwargs['edgecolor'] = 'white'
show_ticks = kwargs.pop('show_ticks', False)
# If no grid specified, don't draw one.
grid = kwargs.pop('grid', None)
# print 'hist kwargs', kwargs
patches = ax.hist(*args, **kwargs)
remove_chartjunk(ax, ['top', 'right'], grid=grid, show_ticks=show_ticks)
return ax
def hist(*args, **kwargs):
"""
Plots a histogram of the provided data. Can provide optional argument
"grid='x'" or "grid='y'" to draw a white grid over the histogram. Almost like "erasing" some of the plot,
but it adds more information!
"""
ax, args, kwargs = maybe_get_ax(*args, **kwargs)
color_cycle = ax._get_lines.color_cycle
# Reassign the default colors to Set2 by Colorbrewer
if iterable(args[0]):
if isinstance(args[0], list):
ncolors = len(args[0])
else:
if len(args[0].shape) == 2:
ncolors = args[0].shape[1]
else:
ncolors = 1
kwargs.setdefault('color', [next(color_cycle) for _ in range(ncolors)])
else:
kwargs.setdefault('color', next(color_cycle))
kwargs.setdefault('edgecolor', 'white')
show_ticks = kwargs.pop('show_ticks', False)
# If no grid specified, don't draw one.
grid = kwargs.pop('grid', None)
# print 'hist kwargs', kwargs
patches = ax.hist(*args, **kwargs)
remove_chartjunk(ax, ['top', 'right'], grid=grid, show_ticks=show_ticks)
return ax
def stackplot(*args, **kwargs):
ax, args, kwargs = maybe_get_ax(*args, **kwargs)
lw = maybe_get_linewidth(**kwargs)
kwargs['linewidths'] = lw
kwargs.setdefault('edgecolor', almost_black)
show_ticks = kwargs.pop('show_ticks', False)
lines = ax.stackplot(*args, **kwargs)
remove_chartjunk(ax, ['top', 'right'], show_ticks=show_ticks)
return lines
def stackplot(*args, **kwargs):
ax, args, kwargs = maybe_get_ax(*args, **kwargs)
lw = maybe_get_linewidth(**kwargs)
kwargs['linewidths'] = lw
kwargs.setdefault('edgecolor', almost_black)
show_ticks = kwargs.pop('show_ticks', False)
lines = ax.stackplot(*args, **kwargs)
remove_chartjunk(ax, ['top', 'right'], show_ticks=show_ticks)
return lines
def trendlines(queries, norm=False):
"""Plot query results"""
for q in queries:
counts = queries[q]
x = counts[:, 0]
y = np.copy(counts[:, 1])
if norm:
y = y / counts[:, 2]
plt.plot(x, y * 100, label=q, lw=4, alpha=0.8)
plt.xlim(np.min(x), np.max(x))
plt.xlabel('Year')
plt.ylabel('Percent of Refereed\nPublications Mentioning')
plt.legend(loc='upper left', frameon=False)
remove_chartjunk(plt.gca(), ['top', 'right'])
def plot(*args, **kwargs):
ax, args, kwargs = maybe_get_ax(*args, **kwargs)
if 'color' not in kwargs:
# if no color is specified, cycle over the ones in this axis
color_cycle = ax._get_lines.color_cycle
kwargs['color'] = next(color_cycle)
if 'linewidth' not in kwargs:
kwargs['linewidth'] = 0.75
show_ticks = kwargs.pop('show_ticks', False)
lines = ax.plot(*args, **kwargs)
remove_chartjunk(ax, ['top', 'right'], show_ticks=show_ticks)
return lines
def boxplot(*args, **kwargs):
"""
Create a box-and-whisker plot showing the mean, 25th percentile, and 75th
percentile. The difference from matplotlib is only the left axis line is
shown, and ticklabels labeling each category of data can be added.
@param ax:
@param x:
@param kwargs: Besides xticklabels, which is a prettyplotlib-specific
argument which will label each individual boxplot, any argument for
matplotlib.pyplot.boxplot will be accepted:
http://matplotlib.org/api/axes_api.html#matplotlib.axes.Axes.boxplot
@return:
"""
ax, args, kwargs = maybe_get_ax(*args, **kwargs)
# If no ticklabels are specified, don't draw any
xticklabels = kwargs.pop('xticklabels', None)
fontsize = kwargs.pop('fontsize', 10)
kwargs.setdefault('widths', 0.15)
bp = ax.boxplot(*args, **kwargs)
if xticklabels:
ax.xaxis.set_ticklabels(xticklabels, fontsize=fontsize)
show_caps = kwargs.pop('show_caps', True)
show_ticks = kwargs.pop('show_ticks', False)
remove_chartjunk(ax, ['top', 'right', 'bottom'], show_ticks=show_ticks)
linewidth = 0.75
blue = colors.set1[1]
red = colors.set1[0]
plt.setp(bp['boxes'], color=blue, linewidth=linewidth)
plt.setp(bp['medians'], color=red)
plt.setp(bp['whiskers'],
color=blue,
linestyle='solid',
linewidth=linewidth)
plt.setp(bp['fliers'], color=blue)
if show_caps:
plt.setp(bp['caps'], color=blue, linewidth=linewidth)
else:
plt.setp(bp['caps'], color='none')
ax.spines['left']._linewidth = 0.5
return ax
def fill_betweenx(*args, **kwargs):
ax, args, kwargs = maybe_get_ax(*args, **kwargs)
lw = maybe_get_linewidth(**kwargs)
kwargs['linewidths'] = lw
if 'color' not in kwargs:
# if no color is specified, cycle over the ones in this axis
color_cycle = cycle(mpl.rcParams['axes.color_cycle'])
kwargs['color'] = next(color_cycle)
kwargs.setdefault('edgecolor', almost_black)
show_ticks = kwargs.pop('show_ticks', False)
lines = ax.fill_betweenx(*args, **kwargs)
remove_chartjunk(ax, ['top', 'right'], show_ticks=show_ticks)
return ax
def fill_between(*args, **kwargs):
ax, args, kwargs = maybe_get_ax(*args, **kwargs)
lw = maybe_get_linewidth(**kwargs)
kwargs['linewidths'] = lw
if 'color' not in kwargs:
# if no color is specified, cycle over the ones in this axis
color_cycle = ax._get_lines.color_cycle
kwargs['color'] = next(color_cycle)
kwargs.setdefault('edgecolor', almost_black)
show_ticks = kwargs.pop('show_ticks', False)
lines = ax.fill_between(*args, **kwargs)
remove_chartjunk(ax, ['top', 'right'], show_ticks=show_ticks)
return lines
def boxplot(*args, **kwargs):
"""
Create a box-and-whisker plot showing the mean, 25th percentile, and 75th
percentile. The difference from matplotlib is only the left axis line is
shown, and ticklabels labeling each category of data can be added.
@param ax:
@param x:
@param kwargs: Besides xticklabels, which is a prettyplotlib-specific
argument which will label each individual boxplot, any argument for
matplotlib.pyplot.boxplot will be accepted:
http://matplotlib.org/api/axes_api.html#matplotlib.axes.Axes.boxplot
@return:
"""
ax, args, kwargs = maybe_get_ax(*args, **kwargs)
# If no ticklabels are specified, don't draw any
xticklabels = kwargs.pop('xticklabels', None)
fontsize = kwargs.pop('fontsize', 10)
if 'widths' not in kwargs:
kwargs['widths'] = 0.15
bp = ax.boxplot(*args, **kwargs)
if xticklabels:
ax.xaxis.set_ticklabels(xticklabels, fontsize=fontsize)
show_caps = kwargs.pop('show_caps', True)
show_ticks = kwargs.pop('show_ticks', False)
remove_chartjunk(ax, ['top', 'right', 'bottom'], show_ticks=show_ticks)
linewidth = 0.75
blue = colors.set1[1]
red = colors.set1[0]
plt.setp(bp['boxes'], color=blue, linewidth=linewidth)
plt.setp(bp['medians'], color=red)
plt.setp(bp['whiskers'], color=blue, linestyle='solid',
linewidth=linewidth)
plt.setp(bp['fliers'], color=blue)
if show_caps:
plt.setp(bp['caps'], color=blue, linewidth=linewidth)
else:
plt.setp(bp['caps'], color='none')
ax.spines['left']._linewidth = 0.5
return ax
def fill_between(*args, **kwargs):
ax, args, kwargs = maybe_get_ax(*args, **kwargs)
lw = maybe_get_linewidth(**kwargs)
kwargs["linewidths"] = lw
if "color" not in kwargs:
# if no color is specified, cycle over the ones in this axis
color_cycle = ax._get_lines.color_cycle
kwargs["color"] = next(color_cycle)
if "edgecolor" not in kwargs:
kwargs["edgecolor"] = almost_black
show_ticks = kwargs.pop("show_ticks", False)
lines = ax.fill_between(*args, **kwargs)
remove_chartjunk(ax, ["top", "right"], show_ticks=show_ticks)
return ax
def boxplot(*args, **kwargs):
"""
Create a box-and-whisker plot showing the mean, 25th percentile, and 75th
percentile. The difference from matplotlib is only the left axis line is
shown, and ticklabels labeling each category of data can be added.
@param ax:
@param x:
@param kwargs: Besides xticklabels, which is a prettyplotlib-specific
argument which will label each individual boxplot, any argument for
matplotlib.pyplot.boxplot will be accepted:
http://matplotlib.org/api/axes_api.html#matplotlib.axes.Axes.boxplot
@return:
"""
ax, args, kwargs = maybe_get_ax(*args, **kwargs)
# If no ticklabels are specified, don't draw any
xticklabels = kwargs.pop("xticklabels", None)
fontsize = kwargs.pop("fontsize", 10)
kwargs.setdefault("widths", 0.15)
bp = ax.boxplot(*args, **kwargs)
if xticklabels:
ax.xaxis.set_ticklabels(xticklabels, fontsize=fontsize)
show_caps = kwargs.pop("show_caps", True)
show_ticks = kwargs.pop("show_ticks", False)
remove_chartjunk(ax, ["top", "right", "bottom"], show_ticks=show_ticks)
linewidth = 0.75
blue = colors.set1[1]
red = colors.set1[0]
plt.setp(bp["boxes"], color=blue, linewidth=linewidth)
plt.setp(bp["medians"], color=red)
plt.setp(bp["whiskers"], color=blue, linestyle="solid", linewidth=linewidth)
plt.setp(bp["fliers"], color=blue)
if show_caps:
plt.setp(bp["caps"], color=blue, linewidth=linewidth)
else:
plt.setp(bp["caps"], color="none")
ax.spines["left"]._linewidth = 0.5
return bp
def eventplot(*args, **kwargs):
ax, args, kwargs = maybe_get_ax(*args, **kwargs)
show_ticks = kwargs.pop('show_ticks', False)
alpha = kwargs.pop('alpha', 1.0)
if len(args) > 0:
positions = args[0]
else:
positions = kwargs['positions']
if any(iterable(p) for p in positions):
size = len(positions)
else:
size = 1
kwargs.setdefault('colors', [c + (alpha, ) for c in set2[:size]])
event_collections = ax.eventplot(*args, **kwargs)
remove_chartjunk(ax, ['top', 'right'], show_ticks=show_ticks)
return event_collections
def eventplot(*args, **kwargs):
ax, args, kwargs = maybe_get_ax(*args, **kwargs)
show_ticks = kwargs.pop('show_ticks', False)
alpha = kwargs.pop('alpha', 1.0)
if len(args) > 0:
positions = args[0]
else:
positions = kwargs['positions']
if any(iterable(p) for p in positions):
size = len(positions)
else:
size = 1
kwargs.setdefault('colors', [c + (alpha,) for c in set2[:size]])
event_collections = ax.eventplot(*args, **kwargs)
remove_chartjunk(ax, ['top', 'right'], show_ticks=show_ticks)
return event_collections
def pcolormesh(*args, **kwargs):
"""
Use for large datasets
Non-traditional `pcolormesh` kwargs are:
- xticklabels, which will put x tick labels exactly in the center of the
heatmap block
- yticklables, which will put y tick labels exactly aligned in the center
of the heatmap block
- xticklabels_rotation, which can be either 'horizontal' or 'vertical'
depending on how you want the xticklabels rotated. The default is
'horizontal', but if you have xticklabels that are longer, you may want
to do 'vertical' so they don't overlap.
- yticklabels_rotation, which can also be either 'horizontal' or
'vertical'. The default is 'horizontal' and in most cases,
that's what you'll want to stick with. But the option is there if you
want.
- center_value, which will be the centered value for a divergent
colormap, for example if you have data above and below zero, but you want
the white part of the colormap to be equal to 10 rather than 0,
then specify 'center_value=10'.
"""
# Deal with arguments in kwargs that should be there, or need to be taken
# out
fig, ax, args, kwargs = maybe_get_fig_ax(*args, **kwargs)
# If x and y axis are passed in arguments, gets correct data
# Ticks will work with x and y data, although it would be pointless to use
# both x/y and custom ticks
if len(args) == 3:
x = args[0]
y = args[1]
data = args[2]
elif len(args) == 1:
data = args[0]
kwargs.setdefault('vmax', data.max())
kwargs.setdefault('vmin', data.min())
center_value = kwargs.pop('center_value', 0)
# If
divergent_data = False
if kwargs['vmax'] > 0 and kwargs['vmin'] < 0:
divergent_data = True
kwargs['vmax'] += center_value
kwargs['vmin'] += center_value
# If we have both negative and positive values, use a divergent colormap
if 'cmap' not in kwargs:
# Check if this is divergent
if divergent_data:
kwargs['cmap'] = blue_red
elif kwargs['vmax'] <= 0:
kwargs['cmap'] = blues_r
elif kwargs['vmax'] > 0:
kwargs['cmap'] = reds
if 'xticklabels' in kwargs:
xticklabels = kwargs['xticklabels']
kwargs.pop('xticklabels')
else:
xticklabels = None
if 'yticklabels' in kwargs:
yticklabels = kwargs['yticklabels']
kwargs.pop('yticklabels')
else:
yticklabels = None
if 'xticklabels_rotation' in kwargs:
xticklabels_rotation = kwargs['xticklabels_rotation']
kwargs.pop('xticklabels_rotation')
else:
xticklabels_rotation = 'horizontal'
if 'yticklabels_rotation' in kwargs:
yticklabels_rotation = kwargs['yticklabels_rotation']
kwargs.pop('yticklabels_rotation')
else:
yticklabels_rotation = 'horizontal'
ax_colorbar = kwargs.pop('ax_colorbar', None)
orientation_colorbar = kwargs.pop('orientation_colorbar', 'vertical')
p = ax.pcolormesh(*args, **kwargs)
# ax.set_ylim(0, x.shape[0])
# ax.set_xlim(0, x.shape[1])
# Get rid of ALL axes
remove_chartjunk(ax, ['top', 'right', 'left', 'bottom'])
if xticklabels is not None and any(xticklabels):
if len(args) == 1:
xticks = np.arange(1, data.shape[1] + 1)
else:
xticks = []
for i in np.arange(len(x) - 1):
half = float(x[i + 1] - x[i]) / 2. + x[i]
xticks.append(half)
xticks = np.array(xticks)
ax.set_xticks(xticks)
ax.set_xticklabels(xticklabels, rotation=xticklabels_rotation)
if yticklabels is not None and any(yticklabels):
if len(args) == 1:
yticks = np.arange(1, data.shape[2] + 1)
else:
yticks = []
for i in np.arange(len(y) - 1):
half = float(y[i + 1] - y[i]) / 2. + y[i]
yticks.append(half)
yticks = np.array(yticks)
ax.set_yticks(yticks)
ax.set_yticklabels(yticklabels, rotation=yticklabels_rotation)
# Show the scale of the colorbar
cbar = fig.colorbar(p,
cax=ax_colorbar,
use_gridspec=True,
orientation=orientation_colorbar)
return p, cbar
def bar(*args, **kwargs):
"""
Creates a bar plot, with white outlines and a fill color that defaults to
the first teal-ish green in ColorBrewer's Set2. Optionally accepts
grid='y' or grid='x' to draw a white grid over the bars,
to show the scale. Almost like "erasing" some of the plot,
but it adds more information!
Can also add an annotation of the height of the barplots directly onto
the bars with the `annotate` parameter, which can either be True,
which will annotate the values, or a list of strings, which will annotate
with the supplied strings.
Can support stacked bars with the value of each stack shown on the stack
(Added by Salil Banerjee)
@param ax: matplotlib.axes instance
@param left: Vector of values of where to put the left side of the bar
@param height: Vector of values of the bar heights
@param kwargs: Besides xticklabels, which is a prettyplotlib-specific
argument, any additional arguments to matplotlib.bar(): http://matplotlib
.org/api/axes_api.html#matplotlib.axes.Axes.bar is accepted.
"""
ax, args, kwargs = maybe_get_ax(*args, **kwargs)
kwargs.setdefault('color', set2[0])
kwargs.setdefault('edgecolor', 'white')
middle = 0.4 if 'width' not in kwargs else kwargs['width']/2.0
# Check if data contains stacks
stacked = kwargs.pop('stacked',False)
# Check if stack text should be included
stack_text = kwargs.pop('stack_text',False)
# Get legend if available
legend = kwargs.pop('legend',False)
left = args[0]
height = np.array(args[1])
# Label each individual bar, if xticklabels is provided
xtickabels = kwargs.pop('xticklabels', None)
# left+0.4 is the center of the bar
xticks = np.array(left) + middle
# Whether or not to annotate each bar with the height value
annotate = kwargs.pop('annotate', False)
show_ticks = kwargs.pop('show_ticks', False)
# If no grid specified, don't draw one.
grid = kwargs.pop('grid', None)
cmap = kwargs.pop('cmap', False)
if cmap:
kwargs['edgecolor'] = almost_black
if not stacked:
kwargs['color'] = getcolors(cmap, height, 0)
# Check if stacked and plot data accordingly
color = kwargs.get('color', None)
if stacked:
num_stacks, num_data = height.shape
bottom = np.zeros(num_data)
for i in np.arange(num_stacks):
lst = list(args)
lst[1] = height[i]
args = tuple(lst)
# make sure number of user specified colors equals to the stacks
if not color or len(color) != num_stacks:
if cmap:
kwargs['color'] = getcolors(cmap, height[i], i)
else:
kwargs['color'] = set2[i]
else:
kwargs['color'] = color[i]
kwargs['bottom'] = bottom
rectangles = ax.bar(*args, **kwargs)
bottom += height[i]
else:
rectangles = ax.bar(*args, **kwargs)
# add legend
if isinstance(legend, collections.Iterable):
ax.legend(legend,loc='upper center',bbox_to_anchor=(0.5,1.11), ncol=5)
# add whitespace padding on left
xmin, xmax = ax.get_xlim()
xmin -= 0.2
if stacked:
xmax = num_data
ax.set_xlim(xmin, xmax)
# If the user is only plotting one bar, make it an iterable
if not isinstance(height, collections.Iterable):
height = [height]
# If there are negative counts, remove the bottom axes
# and add a line at y=0
if any(h < 0 for h in height.tolist()):
axes_to_remove = ['top', 'right', 'bottom']
ax.hlines(y=0, xmin=xmin, xmax=xmax,
linewidths=0.75)
else:
axes_to_remove = ['top', 'right']
# Remove excess axes
remove_chartjunk(ax, axes_to_remove, grid=grid, show_ticks=show_ticks)
if stacked:
data = height
height = height.sum(axis=0)
# Add the xticklabels if they are there
if xtickabels is not None:
ax.set_xticks(xticks)
ax.set_xticklabels(xtickabels)
if annotate or isinstance(annotate, collections.Iterable):
annotate_yrange_factor = 0.025
ymin, ymax = ax.get_ylim()
yrange = ymax - ymin
# Reset ymax and ymin so there's enough room to see the annotation of
# the top-most
if ymax > 0:
ymax += yrange * 0.1
if ymin < 0:
ymin -= yrange * 0.1
ax.set_ylim(ymin, ymax)
yrange = ymax - ymin
offset_ = yrange * annotate_yrange_factor
if isinstance(annotate, collections.Iterable):
annotations = map(str, annotate)
else:
annotations = ['%.3f' % h if type(h) is np.float_ else str(h)
for h in height]
for x, h, annotation in zip(xticks, height, annotations):
# Adjust the offset to account for negative bars
offset = offset_ if h >= 0 else -1 * offset_
verticalalignment = 'bottom' if h >= 0 else 'top'
# Finally, add the text to the axes
ax.annotate(annotation, (x, h + offset),
verticalalignment=verticalalignment,
horizontalalignment='center',
color=almost_black)
# Text for each block of stack
# This was partially inspired by the following article by Tableau software
# http://www.tableausoftware.com/about/blog/2014/1/new-whitepaper-survey-data-less-ugly-more-understandable-27812
if stack_text:
bottom = np.zeros(num_data)
max_h = max(height)
for i in np.arange(num_stacks):
for x, d, b in zip(xticks, data[i], bottom):
if (d*100.0/max_h) > 4.0:
ax.text(x,b+d/2.0,d, ha='center', va='center', color=almost_black)
bottom += data[i]
return rectangles
def test_remove_chartjunk():
fig, ax = plt.subplots(1)
np.random.seed(14)
ax.bar(np.arange(10), np.abs(np.random.randn(10)), color=set2[0],
edgecolor='white')
remove_chartjunk(ax, ['top', 'right'], grid='y', ticklabels='x')
def pcolormesh(*args, **kwargs):
"""
Use for large datasets
Non-traditional `pcolormesh` kwargs are:
- xticklabels, which will put x tick labels exactly in the center of the
heatmap block
- yticklables, which will put y tick labels exactly aligned in the center
of the heatmap block
- xticklabels_rotation, which can be either 'horizontal' or 'vertical'
depending on how you want the xticklabels rotated. The default is
'horizontal', but if you have xticklabels that are longer, you may want
to do 'vertical' so they don't overlap.
- yticklabels_rotation, which can also be either 'horizontal' or
'vertical'. The default is 'horizontal' and in most cases,
that's what you'll want to stick with. But the option is there if you
want.
- center_value, which will be the centered value for a divergent
colormap, for example if you have data above and below zero, but you want
the white part of the colormap to be equal to 10 rather than 0,
then specify 'center_value=10'.
"""
# Deal with arguments in kwargs that should be there, or need to be taken
# out
fig, ax, args, kwargs = maybe_get_fig_ax(*args, **kwargs)
# If x and y axis are passed in arguments, gets correct data
# Ticks will work with x and y data, although it would be pointless to use
# both x/y and custom ticks
if len(args) == 3:
x = args[0]
y = args[1]
data = args[2]
elif len(args) == 1:
data = args[0]
kwargs.setdefault('vmax', data.max())
kwargs.setdefault('vmin', data.min())
center_value = kwargs.pop('center_value', 0)
# If
divergent_data = False
if kwargs['vmax'] > 0 and kwargs['vmin'] < 0:
divergent_data = True
kwargs['vmax'] += center_value
kwargs['vmin'] += center_value
# If we have both negative and positive values, use a divergent colormap
if 'cmap' not in kwargs:
# Check if this is divergent
if divergent_data:
kwargs['cmap'] = blue_red
elif kwargs['vmax'] <= 0:
kwargs['cmap'] = blues_r
elif kwargs['vmax'] > 0:
kwargs['cmap'] = reds
if 'xticklabels' in kwargs:
xticklabels = kwargs['xticklabels']
kwargs.pop('xticklabels')
else:
xticklabels = None
if 'yticklabels' in kwargs:
yticklabels = kwargs['yticklabels']
kwargs.pop('yticklabels')
else:
yticklabels = None
if 'xticklabels_rotation' in kwargs:
xticklabels_rotation = kwargs['xticklabels_rotation']
kwargs.pop('xticklabels_rotation')
else:
xticklabels_rotation = 'horizontal'
if 'yticklabels_rotation' in kwargs:
yticklabels_rotation = kwargs['yticklabels_rotation']
kwargs.pop('yticklabels_rotation')
else:
yticklabels_rotation = 'horizontal'
ax_colorbar = kwargs.pop('ax_colorbar', None)
orientation_colorbar = kwargs.pop('orientation_colorbar', 'vertical')
p = ax.pcolormesh(*args, **kwargs)
# ax.set_ylim(0, x.shape[0])
# ax.set_xlim(0, x.shape[1])
# Get rid of ALL axes
remove_chartjunk(ax, ['top', 'right', 'left', 'bottom'])
if xticklabels is not None and any(xticklabels):
if len(args) == 1:
xticks = np.arange(0.5, data.shape[1] + 0.5)
else:
xticks = []
for i in np.arange(len(x) - 1):
half = float(x[i + 1] - x[i]) / 2. + x[i]
xticks.append(half)
xticks = np.array(xticks)
ax.set_xticks(xticks)
ax.set_xticklabels(xticklabels, rotation=xticklabels_rotation)
if yticklabels is not None and any(yticklabels):
if len(args) == 1:
yticks = np.arange(0.5, data.shape[1] + 0.5)
else:
yticks = []
for i in np.arange(len(y) - 1):
half = float(y[i + 1] - y[i]) / 2. + y[i]
yticks.append(half)
yticks = np.array(yticks)
ax.set_yticks(yticks)
ax.set_yticklabels(yticklabels, rotation=yticklabels_rotation)
# Show the scale of the colorbar
cbar = fig.colorbar(p, cax=ax_colorbar, use_gridspec=True,
orientation=orientation_colorbar)
return p, cbar
def beeswarm(*args, **kwargs):
"""
Create a R-like beeswarm plot showing the mean and datapoints.
The difference from matplotlib is only the left axis line is
shown, and ticklabels labeling each category of data can be added.
@param ax:
@param x:
@param kwargs: Besides xticklabels, which is a prettyplotlib-specific
argument which will label each individual beeswarm, many arguments for
matplotlib.pyplot.boxplot will be accepted:
http://matplotlib.org/api/axes_api.html#matplotlib.axes.Axes.boxplot
Additional arguments include:
*median_color* : (default gray)
The color of median lines
*median_width* : (default 2)
Median line width
*colors* : (default None)
Colors to use when painting a dataseries, for example
list1 = [1,2,3]
list2 = [5,6,7]
ppl.beeswarm([list1, list2], colors=["red", "blue"], xticklabels=["data1", "data2"])
@return:
"""
ax, args, kwargs = maybe_get_ax(*args, **kwargs)
# If no ticklabels are specified, don't draw any
xticklabels = kwargs.pop('xticklabels', None)
colors = kwargs.pop('colors', None)
fontsize = kwargs.pop('fontsize', 10)
gray = _colors.set1[8]
red = _colors.set1[0]
blue = kwargs.pop('color', _colors.set1[1])
kwargs.setdefault('widths', 0.25)
kwargs.setdefault('sym', "o")
bp = _beeswarm(ax, *args, **kwargs)
kwargs.setdefault("median_color", gray)
kwargs.setdefault("median_linewidth", 2)
if xticklabels:
ax.xaxis.set_ticklabels(xticklabels, fontsize=fontsize)
show_caps = kwargs.pop('show_caps', True)
show_ticks = kwargs.pop('show_ticks', False)
remove_chartjunk(ax, ['top', 'right', 'bottom'], show_ticks=show_ticks)
linewidth = 0.75
plt.setp(bp['boxes'], color=blue, linewidth=linewidth)
plt.setp(bp['medians'], color=kwargs.pop("median_color"), linewidth=kwargs.pop("median_linewidth"))
#plt.setp(bp['whiskers'], color=blue, linestyle='solid',
# linewidth=linewidth)
for color, flier in zip(colors, bp['fliers']):
plt.setp(flier, color=color)
#if show_caps:
# plt.setp(bp['caps'], color=blue, linewidth=linewidth)
#else:
# plt.setp(bp['caps'], color='none')
ax.spines['left']._linewidth = 0.5
return bp
def beeswarm(*args, **kwargs):
"""
Create a R-like beeswarm plot showing the mean and datapoints.
The difference from matplotlib is only the left axis line is
shown, and ticklabels labeling each category of data can be added.
@param ax:
@param x:
@param kwargs: Besides xticklabels, which is a prettyplotlib-specific
argument which will label each individual beeswarm, many arguments for
matplotlib.pyplot.boxplot will be accepted:
http://matplotlib.org/api/axes_api.html#matplotlib.axes.Axes.boxplot
Additional arguments include:
*median_color* : (default gray)
The color of median lines
*median_width* : (default 2)
Median line width
*colors* : (default None)
Colors to use when painting a dataseries, for example
list1 = [1,2,3]
list2 = [5,6,7]
ppl.beeswarm([list1, list2], colors=["red", "blue"], xticklabels=["data1", "data2"])
@return:
"""
ax, args, kwargs = maybe_get_ax(*args, **kwargs)
# If no ticklabels are specified, don't draw any
xticklabels = kwargs.pop('xticklabels', None)
colors = kwargs.pop('colors', None)
fontsize = kwargs.pop('fontsize', 10)
gray = _colors.set1[8]
red = _colors.set1[0]
blue = kwargs.pop('color', _colors.set1[1])
kwargs.setdefault('widths', 0.25)
kwargs.setdefault('sym', "o")
bp = _beeswarm(ax, *args, **kwargs)
kwargs.setdefault("median_color", gray)
kwargs.setdefault("median_linewidth", 2)
if xticklabels:
ax.xaxis.set_ticklabels(xticklabels, fontsize=fontsize)
show_caps = kwargs.pop('show_caps', True)
show_ticks = kwargs.pop('show_ticks', False)
remove_chartjunk(ax, ['top', 'right', 'bottom'], show_ticks=show_ticks)
linewidth = 0.75
plt.setp(bp['boxes'], color=blue, linewidth=linewidth)
plt.setp(bp['medians'],
color=kwargs.pop("median_color"),
linewidth=kwargs.pop("median_linewidth"))
#plt.setp(bp['whiskers'], color=blue, linestyle='solid',
# linewidth=linewidth)
for color, flier in zip(colors, bp['fliers']):
plt.setp(flier, color=color)
#if show_caps:
# plt.setp(bp['caps'], color=blue, linewidth=linewidth)
#else:
# plt.setp(bp['caps'], color='none')
ax.spines['left']._linewidth = 0.5
return bp
def pcolormesh(*args, **kwargs):
"""
Use for large datasets
Non-traditional `pcolormesh` kwargs are:
- xticklabels, which will put x tick labels exactly in the center of the
heatmap block
- yticklables, which will put y tick labels exactly aligned in the center
of the heatmap block
- xticklabels_rotation, which can be either 'horizontal' or 'vertical'
depending on how you want the xticklabels rotated. The default is
'horizontal', but if you have xticklabels that are longer, you may want
to do 'vertical' so they don't overlap.
- yticklabels_rotation, which can also be either 'horizontal' or
'vertical'. The default is 'horizontal' and in most cases,
that's what you'll want to stick with. But the option is there if you
want.
- center_value, which will be the centered value for a divergent
colormap, for example if you have data above and below zero, but you want
the white part of the colormap to be equal to 10 rather than 0,
then specify 'center_value=10'.
"""
# Deal with arguments in kwargs that should be there, or need to be taken
# out
fig, ax, args, kwargs = maybe_get_fig_ax(*args, **kwargs)
x = args[0]
kwargs.setdefault('vmax', x.max())
kwargs.setdefault('vmin', x.min())
center_value = kwargs.pop('center_value', 0)
# If
divergent_data = False
if kwargs['vmax'] > 0 and kwargs['vmin'] < 0:
divergent_data = True
kwargs['vmax'] += center_value
kwargs['vmin'] += center_value
# If we have both negative and positive values, use a divergent colormap
if 'cmap' not in kwargs:
# Check if this is divergent
if divergent_data:
kwargs['cmap'] = blue_red
elif kwargs['vmax'] <= 0:
kwargs['cmap'] = blues_r
elif kwargs['vmax'] > 0:
kwargs['cmap'] = reds
if 'xticklabels' in kwargs:
xticklabels = kwargs['xticklabels']
kwargs.pop('xticklabels')
else:
xticklabels = None
if 'yticklabels' in kwargs:
yticklabels = kwargs['yticklabels']
kwargs.pop('yticklabels')
else:
yticklabels = None
if 'xticklabels_rotation' in kwargs:
xticklabels_rotation = kwargs['xticklabels_rotation']
kwargs.pop('xticklabels_rotation')
else:
xticklabels_rotation = 'horizontal'
if 'yticklabels_rotation' in kwargs:
yticklabels_rotation = kwargs['yticklabels_rotation']
kwargs.pop('yticklabels_rotation')
else:
yticklabels_rotation = 'horizontal'
ax_colorbar = kwargs.pop('ax_colorbar', None)
orientation_colorbar = kwargs.pop('orientation_colorbar', 'vertical')
p = ax.pcolormesh(*args, **kwargs)
ax.set_ylim(0, x.shape[0])
# Get rid of ALL axes
remove_chartjunk(ax, ['top', 'right', 'left', 'bottom'])
if any(xticklabels):
xticks = np.arange(0.5, x.shape[1] + 0.5)
ax.set_xticks(xticks)
ax.set_xticklabels(xticklabels, rotation=xticklabels_rotation)
if any(yticklabels):
yticks = np.arange(0.5, x.shape[0] + 0.5)
ax.set_yticks(yticks)
ax.set_yticklabels(yticklabels, rotation=yticklabels_rotation)
# Show the scale of the colorbar
fig.colorbar(p,
cax=ax_colorbar,
use_gridspec=True,
orientation=orientation_colorbar)
return fig
def barh(*args, **kwargs):
"""
Creates a bar plot, with white outlines and a fill color that defaults to
the first teal-ish green in ColorBrewer's Set2. Optionally accepts
grid='y' or grid='x' to draw a white grid over the bars,
to show the scale. Almost like "erasing" some of the plot,
but it adds more information!
Can also add an annotation of the width of the barplots directly onto
the bars with the `annotate` parameter, which can either be True,
which will annotate the values, or a list of strings, which will annotate
with the supplied strings.
Can support stacked bars with the value of each stack shown on the stack
(Added by Salil Banerjee)
@param ax: matplotlib.axes instance
@param top: Vector of values of where to put the top side of the bar
@param width: Vector of values of the bar widths
@param ytickabels: Vector of labels of the bar widths
@param kwargs: Any additional arguments to matplotlib.bar()
"""
ax, args, kwargs = maybe_get_ax(*args, **kwargs)
if 'color' not in kwargs:
kwargs['color'] = set2[0]
if 'edgecolor' not in kwargs:
kwargs['edgecolor'] = 'white'
if 'width' in kwargs:
# Find the middle of the bar
middle = kwargs['width']/2.0
else:
middle = 0.4
# Check if data contains stacks
stacked = kwargs.pop('stacked',False)
# Check if stack text should be included
stack_text = kwargs.pop('stack_text',False)
# Get legend if available
legend = kwargs.pop('legend',False)
top = args[0]
width = np.array(args[1])
# Label each individual bar, if xticklabels is provided
ytickabels = kwargs.pop('yticklabels', None)
# left+0.4 is the center of the bar
yticks = np.array(top) + middle
# Whether or not to annotate each bar with the width value
annotate = kwargs.pop('annotate', False)
# If no grid specified, don't draw one.
grid = kwargs.pop('grid', None)
# Check if stacked and plot data accordingly
if stacked:
num_stacks, num_data = width.shape
left = np.zeros(num_data)
for i in np.arange(num_stacks):
lst = list(args)
lst[1] = width[i]
args = tuple(lst)
kwargs['color'] = set2[i]
kwargs['left'] = left
rectangles = ax.barh(*args, **kwargs)
left += width[i]
else:
rectangles = ax.barh(*args, **kwargs)
# add legend
if isinstance(legend, collections.Iterable):
ax.legend(legend,loc='upper center',bbox_to_anchor=(0.5,1.11), ncol=5)
# add whitespace padding on left
ymin, ymax = ax.get_ylim()
ymin -= 0.2
if stacked:
ymax = num_data
ax.set_ylim(ymin, ymax)
# If there are negative counts, remove the bottom axes
# and add a line at y=0
if any(w < 0 for w in width.tolist()):
axes_to_remove = ['top', 'right', 'bottom']
ax.vlines(x=0, ymin=ymin, ymax=ymax,
linewidths=0.75)
#ax.hlines(y=0, xmin=xmin, xmax=xmax,
# linewidths=0.75)
else:
axes_to_remove = ['top', 'right']
#Remove excess axes
remove_chartjunk(ax, axes_to_remove, grid=grid)
if stacked:
data = width
width = width.sum(axis=0)
# Add the yticklabels if they are there
if ytickabels is not None:
ax.set_yticks(yticks)
ax.set_yticklabels(ytickabels)
if annotate or isinstance(annotate, collections.Iterable):
annotate_xrange_factor = 0.050
xmin, xmax = ax.get_xlim()
xrange = xmax - xmin
# Reset ymax and ymin so there's enough room to see the annotation of
# the top-most
if xmax > 0:
xmax += xrange * 0.1
if xmin < 0:
xmin -= xrange * 0.1
ax.set_xlim(xmin, xmax)
xrange = xmax - xmin
offset_ = xrange * annotate_xrange_factor
if isinstance(annotate, collections.Iterable):
annotations = map(str, annotate)
else:
annotations = ['%.3f' % w if type(w) is np.float_ else str(w)
for w in width]
for y, w, annotation in zip(yticks, width, annotations):
# Adjust the offset to account for negative bars
offset = offset_ if w >= 0 else -1 * offset_
# Finally, add the text to the axes
ax.annotate(annotation, (w + offset, y),
verticalalignment='center',
horizontalalignment='center',
color=almost_black)
# Text for each block of stack
# This was partially inspired by the following article by Tableau software
# http://www.tableausoftware.com/about/blog/2014/1/new-whitepaper-survey-data-less-ugly-more-understandable-27812
if stack_text:
left = np.zeros(num_data)
max_w = max(width)
for i in np.arange(num_stacks):
for y, d, l in zip(yticks, data[i], left):
if (d*100.0/max_w) > 2.0:
ax.text(l+d/2.0,y,d, ha='center', va='center', color=almost_black)
left += data[i]
return rectangles
def pcolormesh(*args, **kwargs):
"""
Use for large datasets
Non-traditional `pcolormesh` kwargs are:
- xticklabels, which will put x tick labels exactly in the center of the
heatmap block
- yticklables, which will put y tick labels exactly aligned in the center
of the heatmap block
- xticklabels_rotation, which can be either 'horizontal' or 'vertical'
depending on how you want the xticklabels rotated. The default is
'horizontal', but if you have xticklabels that are longer, you may want
to do 'vertical' so they don't overlap.
- yticklabels_rotation, which can also be either 'horizontal' or
'vertical'. The default is 'horizontal' and in most cases,
that's what you'll want to stick with. But the option is there if you
want.
- center_value, which will be the centered value for a divergent
colormap, for example if you have data above and below zero, but you want
the white part of the colormap to be equal to 10 rather than 0,
then specify 'center_value=10'.
"""
# Deal with arguments in kwargs that should be there, or need to be taken
# out
fig, ax, args, kwargs = maybe_get_fig_ax(*args, **kwargs)
x = args[0]
if 'vmax' not in kwargs:
kwargs['vmax'] = x.max()
if 'vmin' not in kwargs:
kwargs['vmin'] = x.min()
center_value = kwargs.pop('center_value', 0)
# If
divergent_data = False
if kwargs['vmax'] > 0 and kwargs['vmin'] < 0:
divergent_data = True
kwargs['vmax'] += center_value
kwargs['vmin'] += center_value
# If we have both negative and positive values, use a divergent colormap
if 'cmap' not in kwargs:
# Check if this is divergent
if divergent_data:
kwargs['cmap'] = blue_red
elif kwargs['vmax'] <= 0:
kwargs['cmap'] = blues_r
elif kwargs['vmax'] > 0:
kwargs['cmap'] = reds
if 'xticklabels' in kwargs:
xticklabels = kwargs['xticklabels']
kwargs.pop('xticklabels')
else:
xticklabels = None
if 'yticklabels' in kwargs:
yticklabels = kwargs['yticklabels']
kwargs.pop('yticklabels')
else:
yticklabels = None
if 'xticklabels_rotation' in kwargs:
xticklabels_rotation = kwargs['xticklabels_rotation']
kwargs.pop('xticklabels_rotation')
else:
xticklabels_rotation = 'horizontal'
if 'yticklabels_rotation' in kwargs:
yticklabels_rotation = kwargs['yticklabels_rotation']
kwargs.pop('yticklabels_rotation')
else:
yticklabels_rotation = 'horizontal'
ax_colorbar = kwargs.pop('ax_colorbar', None)
orientation_colorbar = kwargs.pop('orientation_colorbar', 'vertical')
p = ax.pcolormesh(*args, **kwargs)
ax.set_ylim(0, x.shape[0])
# Get rid of ALL axes
remove_chartjunk(ax, ['top', 'right', 'left', 'bottom'])
if xticklabels:
xticks = np.arange(0.5, x.shape[1] + 0.5)
ax.set_xticks(xticks)
ax.set_xticklabels(xticklabels, rotation=xticklabels_rotation)
if yticklabels:
yticks = np.arange(0.5, x.shape[0] + 0.5)
ax.set_yticks(yticks)
ax.set_yticklabels(yticklabels, rotation=yticklabels_rotation)
# Show the scale of the colorbar
fig.colorbar(p, cax=ax_colorbar, use_gridspec=True,
orientation=orientation_colorbar)
return fig
def barh(*args, **kwargs):
"""
Creates a bar plot, with white outlines and a fill color that defaults to
the first teal-ish green in ColorBrewer's Set2. Optionally accepts
grid='y' or grid='x' to draw a white grid over the bars,
to show the scale. Almost like "erasing" some of the plot,
but it adds more information!
Can also add an annotation of the width of the barplots directly onto
the bars with the `annotate` parameter, which can either be True,
which will annotate the values, or a list of strings, which will annotate
with the supplied strings.
Can support stacked bars with the value of each stack shown on the stack
(Added by Salil Banerjee)
@param ax: matplotlib.axes instance
@param top: Vector of values of where to put the top side of the bar
@param width: Vector of values of the bar widths
@param ytickabels: Vector of labels of the bar widths
@param kwargs: Any additional arguments to matplotlib.bar()
"""
ax, args, kwargs = maybe_get_ax(*args, **kwargs)
if 'color' not in kwargs:
kwargs['color'] = set2[0]
if 'edgecolor' not in kwargs:
kwargs['edgecolor'] = 'white'
if 'width' in kwargs:
# Find the middle of the bar
middle = kwargs['width'] / 2.0
else:
middle = 0.4
# Check if data contains stacks
stacked = kwargs.pop('stacked', False)
# Check if stack text should be included
stack_text = kwargs.pop('stack_text', False)
# Get legend if available
legend = kwargs.pop('legend', False)
top = args[0]
width = np.array(args[1])
# Label each individual bar, if xticklabels is provided
ytickabels = kwargs.pop('yticklabels', None)
# left+0.4 is the center of the bar
yticks = np.array(top) + middle
# Whether or not to annotate each bar with the width value
annotate = kwargs.pop('annotate', False)
# If no grid specified, don't draw one.
grid = kwargs.pop('grid', None)
# Check if stacked and plot data accordingly
if stacked:
num_stacks, num_data = width.shape
left = np.zeros(num_data)
for i in np.arange(num_stacks):
lst = list(args)
lst[1] = width[i]
args = tuple(lst)
kwargs['color'] = set2[i]
kwargs['left'] = left
rectangles = ax.barh(*args, **kwargs)
left += width[i]
else:
rectangles = ax.barh(*args, **kwargs)
# add legend
if isinstance(legend, collections.Iterable):
ax.legend(legend,
loc='upper center',
bbox_to_anchor=(0.5, 1.11),
ncol=5)
# add whitespace padding on left
ymin, ymax = ax.get_ylim()
ymin -= 0.2
if stacked:
ymax = num_data
ax.set_ylim(ymin, ymax)
# If there are negative counts, remove the bottom axes
# and add a line at y=0
if any(w < 0 for w in width.tolist()):
axes_to_remove = ['top', 'right', 'bottom']
ax.vlines(x=0, ymin=ymin, ymax=ymax, linewidths=0.75)
#ax.hlines(y=0, xmin=xmin, xmax=xmax,
# linewidths=0.75)
else:
axes_to_remove = ['top', 'right']
#Remove excess axes
remove_chartjunk(ax, axes_to_remove, grid=grid)
if stacked:
data = width
width = width.sum(axis=0)
# Add the yticklabels if they are there
if ytickabels is not None:
ax.set_yticks(yticks)
ax.set_yticklabels(ytickabels)
if annotate or isinstance(annotate, collections.Iterable):
annotate_xrange_factor = 0.050
xmin, xmax = ax.get_xlim()
xrange = xmax - xmin
# Reset ymax and ymin so there's enough room to see the annotation of
# the top-most
if xmax > 0:
xmax += xrange * 0.1
if xmin < 0:
xmin -= xrange * 0.1
ax.set_xlim(xmin, xmax)
xrange = xmax - xmin
offset_ = xrange * annotate_xrange_factor
if isinstance(annotate, collections.Iterable):
annotations = map(str, annotate)
else:
annotations = [
'%.3f' % w if type(w) is np.float_ else str(w) for w in width
]
for y, w, annotation in zip(yticks, width, annotations):
# Adjust the offset to account for negative bars
offset = offset_ if w >= 0 else -1 * offset_
# Finally, add the text to the axes
ax.annotate(annotation, (w + offset, y),
verticalalignment='center',
horizontalalignment='center',
color=almost_black)
# Text for each block of stack
# This was partially inspired by the following article by Tableau software
# http://www.tableausoftware.com/about/blog/2014/1/new-whitepaper-survey-data-less-ugly-more-understandable-27812
if stack_text:
left = np.zeros(num_data)
max_w = max(width)
for i in np.arange(num_stacks):
for y, d, l in zip(yticks, data[i], left):
if (d * 100.0 / max_w) > 2.0:
ax.text(l + d / 2.0,
y,
d,
ha='center',
va='center',
color=almost_black)
left += data[i]
return rectangles
