Gallery¶
Fourteen figures in the order a reader should meet them, from one call to ticks that are the argument.
Every figure comes from examples/gallery.py; regenerate them with uv run --group examples examples/gallery.py. The datasets sit in examples/data, each file naming its source and licence; what is not a measurement says so.
The warming scenarios. The same plotting calls twice. apply trims the frame, line_labels turns the legend into names at the line ends, and one label names the observed record where it starts. The tutorial builds the right panel one call at a time.
One record, four frame modes. nice ends the spines at the outermost ticks, loose at round numbers bracketing the data with the spine standing off, data at the record's own extremes. Those three share their ticks and move only the spine; feature ticks the marks you place instead and runs the left spine out to the 2 °C target the warming has not reached.
Old Faithful. The plainest range frame: a scatter under frame="data", each spine reporting one variable's minimum and maximum.
Waiting times. The same data as a histogram under frame="data", both spines flush with the bars, the geyser's two modes standing clear.
Kepler's third law. The eight planets on log-log axes under frame="loose": offset spines end at the powers of ten bounding the data, and period against semi-major axis falls on the line T = a3/2. Earth alone takes a colour, and label names it in that same colour: one mark picked out of the grey. The tutorial builds it one call at a time.
Tick spacing. The same record at 12 cm and at 2 cm under the same call. The tick count follows the axis's length in tick-label heights, so the wide axes carries four year labels and the narrow one two, and the two y axes agree because they are the same height. The tick spacing how-to has the arguments.
Anscombe's quartet. Four sets built to share their summary statistics, under frame="data" with QuartileLocator ticks. Each panel's spines span its own data and its ticks sit at its own five-number summary, so the axes refuse to repeat the identity; set IV's bottom axis collapses to two marks because ten of its eleven x values are the same number.
A seaborn line plot. vanzelfsprekend reads an axes, not the library that filled it. apply trims seaborn's whitegrid box to two spines and drops the grid; line_labels stands in for the legend, with the labels passed in because seaborn keeps its legend text on proxy artists.
Power profiles. Four peer series, so the figure opts into colour the house way: a muted scheme cycle before the loop, no per-line colour strings, and line_labels naming each curve in its own colour where they crowd at the right. The model is spelled out in the script; the parameters are illustrative.
The resonance peak. After Doumont. FeatureLocator marks the band edges and the peak's own frequency; SummaryLocator sets a minor tick at half power; and the two labels anchor at the summit and the flank. The tutorial builds it one call at a time.
Ticks in π. Phase-shifted voltages on an axis ticked in multiples of π instead of round decimals: TalbotLocator(unit=...) runs the nice-number search in units of π, and secondary_frame mirrors those ticks as whole degrees along the top. V_R is picked out and the other three fade to grey. The radian-axes how-to has the arguments.
The grand tours. Small multiples on one shared scale, each panel ticked at its own race's slowest, median and fastest winner. The wars are holes in every record and the Vuelta's broken start is its own story. The tutorial reads it.
A small-multiples grid. Monthly CO₂ at four stations from the Arctic to the South Pole on one shared scale. Only the left column and bottom row keep furniture, so the seasonal swing shrinking toward the pole reads on equal terms.
The labour-income decomposition. A monthly income split into a permanent step and a transient residual. Small multiples put income and the permanent component on one scale, so the raise reads against income's own swings, and AugmentedLocator unions the frame's nice ticks with the decomposition's own numbers: the two recovered levels, the zero line, and the June and December the bumps recur on. accent colours the two permanent levels and the June and December ticks to match their points, so the figure needs no callouts. The tutorial builds it one call at a time.