Cartesian axes

This section documents features used for modifying Cartesian x and y axes, including axis scales, tick locations, tick label formatting, and several twin and dual axes commands.

Tick locations

Matplotlib tick locators select sensible tick locations based on the axis data limits. In proplot, you can change the tick locator using the format keyword arguments xlocator, ylocator, xminorlocator, and yminorlocator (or their aliases, xticks, yticks, xminorticks, and yminorticks). This is powered by the Locator constructor function.

You can use these keyword arguments to apply built-in matplotlib Locators by their “registered” names (e.g., xlocator='log'), to draw ticks every N data values with MultipleLocator (e.g., xlocator=2), or to tick the specific locations in a list using FixedLocator (just like set_xticks and set_yticks). If you want to work with the locator classes directly, they are available in the top-level namespace (e.g., xlocator=pplt.MultipleLocator(...) is allowed).

To generate lists of tick locations, we recommend using proplot’s arange function – it’s basically an endpoint-inclusive version of numpy.arange, which is usually what you’ll want in this context.

[1]:
import proplot as pplt
import numpy as np
state = np.random.RandomState(51423)
pplt.rc.update(
    metawidth=1, fontsize=10,
    metacolor='dark blue', suptitlecolor='dark blue',
    titleloc='upper center', titlecolor='dark blue', titleborder=False,
    axesfacecolor=pplt.scale_luminance('powderblue', 1.15),
)
fig = pplt.figure(share=False, refwidth=5, refaspect=(8, 1))
fig.format(suptitle='Tick locators demo')

# Step size for tick locations
ax = fig.subplot(711, title='MultipleLocator')
ax.format(xlim=(0, 200), xminorlocator=10, xlocator=30)

# Specific list of locations
ax = fig.subplot(712, title='FixedLocator')
ax.format(xlim=(0, 10), xminorlocator=0.1, xlocator=[0, 0.3, 0.8, 1.6, 4.4, 8, 8.8])

# Ticks at numpy.linspace(xmin, xmax, N)
ax = fig.subplot(713, title='LinearLocator')
ax.format(xlim=(0, 10), xlocator=('linear', 21))

# Logarithmic locator, used automatically for log scale plots
ax = fig.subplot(714, title='LogLocator')
ax.format(xlim=(1, 100), xlocator='log', xminorlocator='logminor')

# Maximum number of ticks, but at "nice" locations
ax = fig.subplot(715, title='MaxNLocator')
ax.format(xlim=(1, 7), xlocator=('maxn', 11))

# Hide all ticks
ax = fig.subplot(716, title='NullLocator')
ax.format(xlim=(-10, 10), xlocator='null')

# Tick locations that cleanly divide 60 minute/60 second intervals
ax = fig.subplot(717, title='Degree-Minute-Second Locator (requires cartopy)')
ax.format(xlim=(0, 2), xlocator='dms', xformatter='dms')

pplt.rc.reset()
_images/cartesian_2_0.svg

Tick formatting

Matplotlib tick formatters convert floating point numbers to nicely-formatted tick labels. In proplot, you can change the tick formatter using the format keyword arguments xformatter and yformatter (or their aliases, xticklabels and yticklabels). This is powered by the Formatter constructor function.

You can use these keyword arguments to apply built-in matplotlib Formatters by their “registered” names (e.g., xformatter='log'), to apply a %-style format directive with FormatStrFormatter (e.g., xformatter='%.0f'), or to apply custom tick labels with FixedFormatter (just like set_xticklabels). You can also apply one of proplot’s new tick formatters – for example, xformatter='deglat' to label ticks as geographic latitude coordinates, xformatter='pi' to label ticks as fractions of \(\pi\), or xformatter='sci' to label ticks with scientific notation. If you want to work with the formatter classes directly, they are available in the top-level namespace (e.g., xformatter=pplt.SciFormatter(...) is allowed).

Proplot also changes the default tick formatter to AutoFormatter. This class trims trailing zeros by default, can optionally omit or wrap tick values within particular number ranges, and can add prefixes and suffixes to each label. See AutoFormatter for details. To disable the trailing zero-trimming feature, set rc['formatter.zerotrim'] to False.

[2]:
import proplot as pplt
pplt.rc.fontsize = 11
pplt.rc.metawidth = 1.5
pplt.rc.gridwidth = 1

# Create the figure
fig, axs = pplt.subplots(ncols=2, nrows=2, refwidth=1.5, share=False)
axs.format(
    ytickloc='both', yticklabelloc='both',
    titlepad='0.5em', suptitle='Default formatters demo'
)

# Formatter comparison
locator = [0, 0.25, 0.5, 0.75, 1]
axs[0].format(xformatter='scalar', yformatter='scalar', title='Matplotlib formatter')
axs[1].format(title='Proplot formatter')
axs[:2].format(xlocator=locator, ylocator=locator)

# Limiting the tick range
axs[2].format(
    title='Omitting tick labels', ticklen=5, xlim=(0, 5), ylim=(0, 5),
    xtickrange=(0, 2), ytickrange=(0, 2), xlocator=1, ylocator=1
)

# Setting the wrap range
axs[3].format(
    title='Wrapping the tick range', ticklen=5, xlim=(0, 7), ylim=(0, 6),
    xwraprange=(0, 5), ywraprange=(0, 3), xlocator=1, ylocator=1
)
pplt.rc.reset()
_images/cartesian_4_0.svg
[3]:
import proplot as pplt
import numpy as np
pplt.rc.update(
    metawidth=1.2, fontsize=10, axesfacecolor='gray0', figurefacecolor='gray2',
    metacolor='gray8', gridcolor='gray8', titlecolor='gray8', suptitlecolor='gray8',
    titleloc='upper center', titleborder=False,
)
fig = pplt.figure(refwidth=5, refaspect=(8, 1), share=False)

# Scientific notation
ax = fig.subplot(911, title='SciFormatter')
ax.format(xlim=(0, 1e20), xformatter='sci')

# N significant figures for ticks at specific values
ax = fig.subplot(912, title='SigFigFormatter')
ax.format(
    xlim=(0, 20), xlocator=(0.0034, 3.233, 9.2, 15.2344, 7.2343, 19.58),
    xformatter=('sigfig', 2),  # 2 significant digits
)

# Fraction formatters
ax = fig.subplot(913, title='FracFormatter')
ax.format(xlim=(0, 3 * np.pi), xlocator=np.pi / 4, xformatter='pi')
ax = fig.subplot(914, title='FracFormatter')
ax.format(xlim=(0, 2 * np.e), xlocator=np.e / 2, xticklabels='e')

# Geographic formatters
ax = fig.subplot(915, title='Latitude Formatter')
ax.format(xlim=(-90, 90), xlocator=30, xformatter='deglat')
ax = fig.subplot(916, title='Longitude Formatter')
ax.format(xlim=(0, 360), xlocator=60, xformatter='deglon')

# User input labels
ax = fig.subplot(917, title='FixedFormatter')
ax.format(
    xlim=(0, 5), xlocator=np.arange(5),
    xticklabels=['a', 'b', 'c', 'd', 'e'],
)

# Custom style labels
ax = fig.subplot(918, title='FormatStrFormatter')
ax.format(xlim=(0, 0.001), xlocator=0.0001, xformatter='%.E')
ax = fig.subplot(919, title='StrMethodFormatter')
ax.format(xlim=(0, 100), xtickminor=False, xlocator=20, xformatter='{x:.1f}')
fig.format(ylocator='null', suptitle='Tick formatters demo')
pplt.rc.reset()
_images/cartesian_5_0.svg

Datetime ticks

The above examples all assumed typical “numeric” axes. However format can also modify the tick locations and tick labels for “datetime” axes. To draw ticks on each occurence of some particular time unit, use a unit string (e.g., xlocator='month'). To draw ticks every N time units, use a (unit, N) tuple (e.g., xlocator=('day', 5)). For % style formatting of datetime tick labels with strftime, you can use a string containing '%' (e.g. xformatter='%Y-%m-%d'). By default, x axis datetime axis labels are rotated 90 degrees, like in pandas. This can be disabled by passing xrotation=0 to format or by setting rc['formatter.timerotation'] to 0. See Locator and Formatter for details.

[4]:
import proplot as pplt
import numpy as np
pplt.rc.update(
    metawidth=1.2, fontsize=10, ticklenratio=0.7,
    figurefacecolor='w', axesfacecolor='pastel blue',
    titleloc='upper center', titleborder=False,
)
fig, axs = pplt.subplots(nrows=5, refwidth=6, refaspect=(8, 1), share=False)

# Default date locator
# This is enabled if you plot datetime data or set datetime limits
ax = axs[0]
ax.format(
    xlim=(np.datetime64('2000-01-01'), np.datetime64('2001-01-02')),
    title='Auto date locator and formatter'
)

# Concise date formatter introduced in matplotlib 3.1
ax = axs[1]
ax.format(
    xlim=(np.datetime64('2000-01-01'), np.datetime64('2001-01-01')),
    xformatter='concise', title='Concise date formatter',
)

# Minor ticks every year, major every 10 years
ax = axs[2]
ax.format(
    xlim=(np.datetime64('2000-01-01'), np.datetime64('2050-01-01')),
    xlocator=('year', 10), xformatter='\'%y', title='Ticks every N units',
)

# Minor ticks every 10 minutes, major every 2 minutes
ax = axs[3]
ax.format(
    xlim=(np.datetime64('2000-01-01T00:00:00'), np.datetime64('2000-01-01T12:00:00')),
    xlocator=('hour', range(0, 24, 2)), xminorlocator=('minute', range(0, 60, 10)),
    xformatter='T%H:%M:%S', title='Ticks at specific intervals',
)

# Month and year labels, with default tick label rotation
ax = axs[4]
ax.format(
    xlim=(np.datetime64('2000-01-01'), np.datetime64('2008-01-01')),
    xlocator='year', xminorlocator='month',  # minor ticks every month
    xformatter='%b %Y', title='Ticks with default rotation',
)
axs[:4].format(xrotation=0)  # no rotation for the first four examples
fig.format(ylocator='null', suptitle='Datetime locators and formatters demo')
pplt.rc.reset()
_images/cartesian_7_0.svg

Axis positions

The locations of axis spines, tick marks, tick labels, and axis labels can be controlled with proplot.axes.CartesianAxes.format keyword arguments like xspineloc (shorthand xloc), xtickloc, xticklabelloc, and xlabelloc. Valid locations include 'left', 'right', 'top', 'bottom', 'neither', 'none', or 'both'. Spine locations can also be set to a valid set_position value, e.g. 'zero' or ('axes', 1.5). The top or right spine is used when the coordinate is more than halfway across the axes. This is often convenient when passing e.g. loc to “alternate” axes commands. These keywords provide the functionality of matplotlib’s tick_left, tick_right, tick_top, and tick_bottom, and set_position, but with additional flexibility.

[5]:
import proplot as pplt
pplt.rc.update(
    metawidth=1.2, fontsize=10, gridcolor='coral',
    axesedgecolor='deep orange', figurefacecolor='white',
)
fig = pplt.figure(share=False, refwidth=2, suptitle='Axis locations demo')

# Spine location demonstration
ax = fig.subplot(121, title='Various locations')
ax.format(xloc='top', xlabel='original axis')
ax.twiny(xloc='bottom', xcolor='black', xlabel='locked twin')
ax.twiny(xloc=('axes', 1.25), xcolor='black', xlabel='offset twin')
ax.twiny(xloc=('axes', -0.25), xcolor='black', xlabel='offset twin')
ax.format(ytickloc='both', yticklabelloc='both')
ax.format(ylabel='labels on both sides')

# Other locations locations
ax = fig.subplot(122, title='Zero-centered spines', titlepad='1em')
ax.format(xlim=(-10, 10), ylim=(-3, 3), yticks=1)
ax.format(xloc='zero', yloc='zero')
pplt.rc.reset()
_images/cartesian_9_0.svg

Axis scales

“Axis scales” like 'linear' and 'log' control the x and y axis coordinate system. To change the axis scale, pass e.g. xscale='log' or yscale='log' to format. This is powered by the Scale constructor function. Proplot makes several changes to the axis scale API:

  • The AutoFormatter formatter is now used for all axis scales by default, including 'log' and 'symlog'. Matplotlib’s behavior can be restored by passing e.g. xformatter='log' or yformatter='log' to format.

  • To make its behavior consistent with Locator and Formatter, the Scale constructor function returns instances of ScaleBase, and set_xscale and set_yscale now accept these class instances in addition to “registered” names like 'log'.

  • While matplotlib axis scales must be instantiated with an Axis instance (for backwards compatibility reasons), proplot axis scales can be instantiated without the axis instance (e.g., pplt.LogScale() instead of pplt.LogScale(ax.xaxis)).

  • The default subs for the 'symlog' axis scale is now np.arange(1, 10), and the default linthresh is now 1. Also the 'log' and 'symlog' axis scales now accept the keywords base, linthresh, linscale, and subs rather than keywords with trailing x or y.

Proplot also includes a few new axis scales. The 'cutoff' scale CutoffScale is useful when the statistical distribution of your data is very unusual. The 'sine' scale SineLatitudeScale scales the axis with a sine function (resulting in an area-weighted spherical latitude coordinate) and the 'mercator' scale MercatorLatitudeScale scales the axis with the Mercator projection latitude coordinate. The 'inverse' scale InverseScale can be useful when working with spectral data, especially with “dual” unit axes. If you want to work with the axis scale classes directly, they are available in the top-level namespace (e.g., xscale=pplt.CutoffScale(...) is allowed).

[6]:
import proplot as pplt
import numpy as np
N = 200
lw = 3
pplt.rc.update({'meta.width': 1, 'label.weight': 'bold', 'tick.labelweight': 'bold'})
fig = pplt.figure(refwidth=1.8, share=False)

# Linear and log scales
ax1 = fig.subplot(221)
ax1.format(yscale='linear', ylabel='linear scale')
ax2 = fig.subplot(222)
ax2.format(ylim=(1e-3, 1e3), yscale='log', ylabel='log scale')
for ax in (ax1, ax2):
    ax.plot(np.linspace(0, 1, N), np.linspace(0, 1000, N), lw=lw)

# Symlog scale
ax = fig.subplot(223)
ax.format(yscale='symlog', ylabel='symlog scale')
ax.plot(np.linspace(0, 1, N), np.linspace(-1000, 1000, N), lw=lw)

# Logit scale
ax = fig.subplot(224)
ax.format(yscale='logit', ylabel='logit scale')
ax.plot(np.linspace(0, 1, N), np.linspace(0.01, 0.99, N), lw=lw)

fig.format(suptitle='Axis scales demo', ytickminor=True)
pplt.rc.reset()
_images/cartesian_11_0.svg
[7]:
import proplot as pplt
import numpy as np

# Create figure
x = np.linspace(0, 4 * np.pi, 100)
dy = np.linspace(-1, 1, 5)
ys = (np.sin(x), np.cos(x))
state = np.random.RandomState(51423)
data = state.rand(len(dy) - 1, len(x) - 1)
colors = ('coral', 'sky blue')
cmap = pplt.Colormap('grays', right=0.8)
fig, axs = pplt.subplots(nrows=4, refaspect=(5, 1), figwidth=5.5, sharex=False)

# Loop through various cutoff scale options
titles = ('Zoom out of left', 'Zoom into left', 'Discrete jump', 'Fast jump')
args = (
    (np.pi, 3),  # speed up
    (3 * np.pi, 1 / 3),  # slow down
    (np.pi, np.inf, 3 * np.pi),  # discrete jump
    (np.pi, 5, 3 * np.pi)  # fast jump
)
locators = (
    np.pi / 3,
    np.pi / 3,
    np.pi * np.append(np.linspace(0, 1, 4), np.linspace(3, 4, 4)),
    np.pi * np.append(np.linspace(0, 1, 4), np.linspace(3, 4, 4)),
)
for ax, iargs, title, locator in zip(axs, args, titles, locators):
    ax.pcolormesh(x, dy, data, cmap=cmap)
    for y, color in zip(ys, colors):
        ax.plot(x, y, lw=4, color=color)
    ax.format(
        xscale=('cutoff', *iargs), xlim=(0, 4 * np.pi),
        xlocator=locator, xformatter='pi', xtickminor=False,
        ygrid=False, ylabel='wave amplitude',
        title=title, suptitle='Cutoff axis scales demo'
    )
_images/cartesian_12_0.svg
[8]:
import proplot as pplt
import numpy as np

# Create figure
n = 30
state = np.random.RandomState(51423)
data = state.rand(n - 1, n - 1)
colors = ('coral', 'sky blue')
cmap = pplt.Colormap('grays', right=0.8)
gs = pplt.GridSpec(nrows=2, ncols=2)
fig = pplt.figure(refwidth=2.3, share=False)
fig.format(grid=False, suptitle='Other axis scales demo')

# Geographic scales
x = np.linspace(-180, 180, n)
y = np.linspace(-85, 85, n)
for i, scale in enumerate(('sine', 'mercator')):
    ax = fig.subplot(gs[i, 0])
    ax.plot(x, y, '-', color=colors[i], lw=4)
    ax.pcolormesh(x, y, data, cmap='grays', cmap_kw={'right': 0.8})
    ax.format(
        yscale=scale, title=scale.title() + ' scale',
        ylim=(-85, 85), ylocator=20, yformatter='deg',
    )

# Exponential scale
n = 50
x = np.linspace(0, 1, n)
y = 3 * np.linspace(0, 1, n)
data = state.rand(len(y) - 1, len(x) - 1)
ax = fig.subplot(gs[0, 1])
title = 'Exponential $e^x$ scale'
ax.pcolormesh(x, y, data, cmap='grays', cmap_kw={'right': 0.8})
ax.plot(x, y, lw=4, color=colors[0])
ax.format(ymin=0.05, yscale=('exp', np.e), title=title)

# Power scale
ax = fig.subplot(gs[1, 1])
title = 'Power $x^{0.5}$ scale'
ax.pcolormesh(x, y, data, cmap='grays', cmap_kw={'right': 0.8})
ax.plot(x, y, lw=4, color=colors[1])
ax.format(ymin=0.05, yscale=('power', 0.5), title=title)
_images/cartesian_13_0.svg

Alternate axes

The matplotlib.axes.Axes class includes twinx and twiny commands for drawing “twin” x and y axes in the same subplot. Proplot expands on these commands and adds the arguably more intuitive altx and alty options. Here altx is equivalent to twiny (makes an alternate x axes and an identical twin y axes) and alty is equivalent to twinx (makes an alternate y axes and an identical twin x axes). The proplot versions can be quickly formatted by passing proplot.axes.CartesianAxes.format keyword arguments to the commands (e.g., ax.alty(ycolor='red') or, since the y prefix in this context is redundant, just ax.alty(color='red')). They also enforce sensible default locations for the spines, ticks, and labels, and disable the twin axes background patch and gridlines by default.

Note

Unlike matplotlib, proplot adds alternate axes as children of the original axes. This helps simplify the tight layout algorithm but means that the drawing order is controlled by the difference between the zorders of the alternate axes and the content inside the original axes rather than the zorder of the original axes itself (see this issue page for details).

[9]:
import proplot as pplt
import numpy as np
state = np.random.RandomState(51423)
c0 = 'gray5'
c1 = 'red8'
c2 = 'blue8'
N, M = 50, 10

# Alternate y axis
data = state.rand(M) + (state.rand(N, M) - 0.48).cumsum(axis=0)
altdata = 5 * (state.rand(N) - 0.45).cumsum(axis=0)
fig = pplt.figure(share=False)
ax = fig.subplot(121, title='Alternate y twin x')
ax.line(data, color=c0, ls='--')
ox = ax.alty(color=c2, label='alternate ylabel', linewidth=1)
ox.line(altdata, color=c2)

# Alternate x axis
data = state.rand(M) + (state.rand(N, M) - 0.48).cumsum(axis=0)
altdata = 5 * (state.rand(N) - 0.45).cumsum(axis=0)
ax = fig.subplot(122, title='Alternate x twin y')
ax.linex(data, color=c0, ls='--')
ox = ax.altx(color=c1, label='alternate xlabel', linewidth=1)
ox.linex(altdata, color=c1)
fig.format(xlabel='xlabel', ylabel='ylabel', suptitle='Alternate axes demo')
_images/cartesian_15_0.svg

Dual unit axes

The dualx and dualy methods can be used to draw duplicate x and y axes meant to represent alternate units in the same coordinate range as the “parent” axis. This feature is powered by the FuncScale class. dualx and dualy accept the same axis formatting keyword arguments as altx and alty. The alternate units are specified with either of the following three positional arguments:

  1. A single linear forward function.

  2. A 2-tuple of arbitrary forward and inverse functions.

  3. An axis scale name or class instance.

In the third case, the axis scale transforms are used for the forward and inverse functions, and the default axis scale locators and formatters are used for the default dual axis locators and formatters. In the below examples, we generate dual axes with each of these three methods. Note that the “parent” axis scale is arbitrary – in the first example, we create a dualx axis for a symlog-scaled axis.

[10]:
import proplot as pplt
pplt.rc.update({'grid.alpha': 0.4, 'meta.width': 1, 'grid.linewidth': 1})
c1 = pplt.scale_luminance('cerulean', 0.5)
c2 = pplt.scale_luminance('red', 0.5)
fig = pplt.figure(refaspect=2.2, refwidth=3, share=False)
axs = fig.subplots(
    [[1, 1, 2, 2], [0, 3, 3, 0]],
    suptitle='Duplicate axes with simple transformations',
    ylocator=[], yformatter=[], xcolor=c1, gridcolor=c1,
)

# Meters and kilometers
ax = axs[0]
ax.format(xlim=(0, 5000), xlabel='meters')
ax.dualx(
    lambda x: x * 1e-3,
    label='kilometers', grid=True, color=c2, gridcolor=c2
)

# Kelvin and Celsius
ax = axs[1]
ax.format(xlim=(200, 300), xlabel='temperature (K)')
ax.dualx(
    lambda x: x - 273.15,
    label='temperature (\N{DEGREE SIGN}C)', grid=True, color=c2, gridcolor=c2
)

# With symlog parent
ax = axs[2]
ax.format(xlim=(-100, 100), xscale='symlog', xlabel='MegaJoules')
ax.dualx(
    lambda x: x * 1e6,
    label='Joules', formatter='log', grid=True, color=c2, gridcolor=c2
)
pplt.rc.reset()
_images/cartesian_17_0.svg
[11]:
import proplot as pplt
pplt.rc.update({'grid.alpha': 0.4, 'meta.width': 1, 'grid.linewidth': 1})
c1 = pplt.scale_luminance('cerulean', 0.5)
c2 = pplt.scale_luminance('red', 0.5)
fig = pplt.figure(
    share=False, refaspect=0.4, refwidth=1.8,
    suptitle='Duplicate axes with pressure and height'
)

# Pressure as the linear scale, height on opposite axis (scale height 7km)
ax = fig.subplot(121)
ax.format(
    xformatter='null', ylabel='pressure (hPa)',
    ylim=(1000, 10), xlocator=[], ycolor=c1, gridcolor=c1
)
ax.dualy(
    'height', label='height (km)', ticks=2.5, color=c2, gridcolor=c2, grid=True
)

# Height as the linear scale, pressure on opposite axis (scale height 7km)
ax = fig.subplot(122)
ax.format(
    xformatter='null', ylabel='height (km)', ylim=(0, 20), xlocator='null',
    grid=True, gridcolor=c2, ycolor=c2
)
ax.dualy(
    'pressure', label='pressure (hPa)', locator=100, color=c1, gridcolor=c1, grid=True
)
pplt.rc.reset()
_images/cartesian_18_0.svg
[12]:
import proplot as pplt
import numpy as np
pplt.rc.margin = 0
c1 = pplt.scale_luminance('cerulean', 0.5)
c2 = pplt.scale_luminance('red', 0.5)
fig, ax = pplt.subplots(refaspect=(3, 1), figwidth=6)

# Sample data
cutoff = 1 / 5
x = np.linspace(0.01, 0.5, 1000)  # in wavenumber days
response = (np.tanh(-((x - cutoff) / 0.03)) + 1) / 2  # response func
ax.axvline(cutoff, lw=2, ls='-', color=c2)
ax.fill_between([cutoff - 0.03, cutoff + 0.03], 0, 1, color=c2, alpha=0.3)
ax.plot(x, response, color=c1, lw=2)

# Add inverse scale to top
ax.format(
    title='Imaginary response function',
    suptitle='Duplicate axes with wavenumber and period',
    xlabel='wavenumber (days$^{-1}$)', ylabel='response', grid=False,
)
ax = ax.dualx(
    'inverse', locator='log', locator_kw={'subs': (1, 2, 5)}, label='period (days)'
)
pplt.rc.reset()
_images/cartesian_19_0.svg