A declarative chart in Vega-Lite's grammar — a list of dicts, a mark, and channels — rendered to SVG by plain Python. Every figure below was drawn when this page was built, so the page carries the picture and no runtime at all. Hover the last one — the tooltips are markup too.
A colour channel, and the stacking Vega-Lite implies
nothing says `stack`; a bar with a colour channel stacks anyway
xOffset groups instead of stacking
the same spec with an xOffset channel — and it stops the stacking
timeUnit carries its own order
the two charts above needed sort=MONTHS; a timeUnit knows a weekday runs Sun–Sat
bin: a histogram
the bars fill their bucket rather than being hairlines at its start
transform: filter and calculate
filter "datum.weather === 'rain'", then calculate the daily range
layer: two marks, one pair of scales
a line and its points, in one hue, because that is one series
boxplot: five statistics per group
whiskers reach 1.5×IQR; what lies outside is drawn, not hidden
errorbar: how well the mean is known
extent=stderr — narrow. Compare the next one, which is the same data
errorbar: how spread the data is
extent=stdev — the same rows, an order of magnitude wider. They are not the same question
errorband with borders, extent=ci
a bootstrapped 95% interval, seeded so the server and the browser agree
facet: small multiples, on one pair of scales
panels scaled apart look like a comparison and are not one
hconcat: different charts, independent scales
so each panel keeps its own axes — without them it cannot be read
one axis only
a spec may encode just x: the other axis is one slot, with no furniture
tooltip: hover a point
a <title> inside each mark — the browser's own, and free on a static page
16 charts, every one of them markup. Nothing on this page fetches JavaScript — there is no <script src> anywhere in it, and the only script it carries at all is the chrome's inline theme toggle, which has to run before first paint. The data is the 1,461 daily observations of Seattle weather that ship in the repository, and the same figures build standalone with python tools/chart_gallery.py.