Interface VegaLiteEncoding

Secondary positional channels — the far end of a mark that spans a range instead of standing on a baseline.

A bar carrying x and x2 draws an interval between two positions on the same axis rather than a magnitude measured from zero, which is what a gantt chart, a ranged bar and a waterfall step all are. Vega-Lite's own y2 accepts a datum constant as well as a field, so the adapter checks for a field before reading either as an interval: a constant baseline is a bar drawn the ordinary way.

Vertical counterpart of x2.

Modern Vega-Lite dodge channel — when paired with a categorical field, Vega-Lite places bars of each subcategory side-by-side within the same x slot. The adapter inspects this to classify a bar mark as DODGED rather than STACKED.

Vertical counterpart of xOffset.

Angular extent of an arc mark — the channel that makes one a pie (or, with mark.innerRadius, a doughnut). It carries the slice magnitudes; the slice labels come from color/fill, since an arc has no x or y.

Radial extent of an arc mark — the channel that turns a pie into a polar area (coxcomb, rose) chart, where the wedge's radius carries the magnitude and the angle only says which category it is.

A pie may also set a radius, but as a constant ({value: 100}) or a scale tweak rather than a data field, so only a bound field marks the mark as radial.

Grouping without a visual channel: it splits one mark into several without colouring or otherwise marking them apart. Vega-Lite's ranged dot plot uses it to draw one connector per category.

Modelled so such a spec type-checks, not dispatched on: the same chart puts its grouping on detail in one layer and on color in the next, so the adapter decides whether the rows really are paired by looking at the rows.

What a text mark writes at each position.

The mark's whole payload: a labelled scatter's point is a country or a gene, and the two coordinates are what the reader can already see the shape of. Read into ScatterPoint.label (#1124).