Frame modes¶

Choose where each spine ends: at the outermost tick, at a round number bracketing the data, at the data itself, or at a mark you place.

apply(ax) ends the spines at the outermost ticks, and so does range_frame(ax), the framing step on its own; the two take the same arguments, so everything below holds for either. Two other modes end the spines at the data itself, or at round numbers bracketing it:

ax.vzs.range_frame(frame="data")
ax.vzs.range_frame(frame="loose")

The ticks and the spine are two separate decisions. The ticks are the locator's: nice numbers inside the data, except under feature, where they're the marks a FixedLocator sets. Where the spine ends is the frame mode's, and nice, loose, and feature all make it follow the ticks:

So under data the spine runs a little past its last tick at each end, and a tick sits at the data's extreme only when a locator you set puts one there, as QuartileLocator does in the Anscombe figure. The same warming record sits under all four modes below. nice, loose and data share their ticks, nice numbers inside the data, and move only where the spine ends; feature instead sets its ticks at the marks you place and carries the spine out to one beyond the data, here the 2 °C target the record has not reached:

Both columns read the data, never the view's padding: matplotlib's autoscale leaves 5% of air around the data, and the frame ignores it. A view you pin inside the data with set_xlim or set_ylim crops the frame to the data left on screen, and the ticks re-fit it as matplotlib's own would. A view wider than the data changes nothing, and is how you make room for a fixed tick outside it, as the resonance figure does for its zero baseline.

A tuple sets the modes per spine, (x, y), so a measurement record can end exactly where the data does while the value axis keeps nice bounds:

ax.vzs.range_frame(frame=("data", "loose"))

Either entry can itself be a pair (low, high) that sets the two ends of one spine apart. A record that begins in 1903 reads better from a round 1900 yet should still stop at its last observation, which is one spine with a loose start and a data end:

ax.vzs.range_frame(frame=(("loose", "data"), "nice"))

The ticks follow the ends: the loose end takes the round number just beyond the data while the other end keeps its ticks inside, so that record gets ticks at 1900, 1950 and 2000 under a spine from 1900 to 2023.

A spine with a loose end also stands off the plot by 8 points: a loose frame rounds outward past the data, so the spine is a detached reference scale rather than the data's own edge, and the gap says so (data and nice sit flush). The offset argument sets that gap yourself, in points, and like frame takes a tuple (x, y) to move the bottom and left spine apart; a None in either slot keeps that spine's mode default:

ax.vzs.range_frame(frame="loose", offset=(8, 2))

feature ends each spine at the outermost mark a FeatureLocator, SummaryLocator, QuartileLocator, or bare FixedLocator sets on that axis, or an AugmentedLocator sets through its .extra side (the marks, not the nice-number ticks it also carries): the peak, the quartiles, a zero baseline. A mark can sit beyond the data, and the spine reaches it anyway, growing the view to keep it on screen. Like data and nice it sits flush, offset 0 by default rather than loose's 8, because the mark is the data's own landmark rather than a detached reference scale. Set the locator before framing:

ax.yaxis.set_major_locator(vzs.FeatureLocator(...))
ax.vzs.apply(frame=("data", "feature"))