Old Faithful¶

Four calls turn a default scatter into a range frame that reports both variables' extremes for free.

Old Faithful erupts for between one and five minutes, and the wait until the next eruption is between forty and a hundred minutes. Plotted against each other the two form two clusters: short eruptions are followed by short waits, long by long. The data are 272 observations from Azzalini and Bowman's 1990 paper, shipped with the examples.

Start with the scatter as matplotlib draws it, then add the frame and the labels one call at a time.

Draw the data¶

import matplotlib.pyplot as plt
import numpy as np

import vanzelfsprekend as vzs

table = np.genfromtxt(
    "examples/data/old_faithful.csv", delimiter=",", names=True, skip_header=4
)
fig, ax = plt.subplots(figsize=(5, 3.5))
ax.scatter(table["eruptions"], table["waiting"], s=10, color=vzs.palettes.DATA_INK)

The file's first four lines name its source, so skip_header steps over them to the column names. The colour is yours to set: vanzelfsprekend never recolours a mark, so a scatter drawn before apply keeps whatever you gave it. Draw it after apply instead and the neutral cycle hands you the same near-black.

This is the figure every matplotlib user knows: a box, ticks at round numbers whether or not the data reaches them, and the axis labels still to come.

Apply the range frame¶

vzs.apply(ax, frame="data")

The box becomes two spines. Each runs from that variable's minimum to its maximum and no further, so the frame now says that the shortest eruption was a little under two minutes and the longest over five, and that nobody waited less than about forty-five minutes. The ticks stayed on round numbers but only the ones inside the data survive, and the tick marks, tick labels and spines turned grey. The points are untouched.

The frame="data" argument is what puts the spine ends exactly at the extremes. The default, nice, ends them at the outermost round tick instead. The frame modes how-to has the three modes side by side.

The same call is also a method on the axes, ax.vzs.apply(frame="data"). Every entry point is, with matplotlib's spelling where it has one, so the vzs.xlabel(ax, ...) below is also ax.vzs.set_xlabel(...). The how-to pages use that form; the registration reference has the rule.

Name the axes¶

vzs.xlabel(ax, "eruption length (min)")
vzs.ylabel(ax, "minutes to the next")

The x label sits under the right end of the bottom spine and the y label sits horizontal at the top of the left spine, where the eye arrives after reading the last tick. Neither is rotated.

The result¶

All four calls, as the gallery script draws it:

ax.scatter(table["eruptions"], table["waiting"], s=10, color=vzs.palettes.DATA_INK)
vzs.apply(ax, frame="data")
vzs.xlabel(ax, "eruption length (min)")
vzs.ylabel(ax, "minutes to the next")

Save it with fig.savefig("old_faithful.svg", bbox_inches="tight"). The frame is installed as a draw hook, so it follows any later change to the limits or the ticks, and vzs.restore(ax) undoes it exactly.

Next: Kepler's third law, where one planet is picked out of eight and named in its colour.