Band support for the time-based scale

A very common graph is a bar chart with the temporal X axis. Yet, Vega has relatively little support for that usecase. Graph designers must hardcode the bar width, like the wheat & wages example, and know ahead of time the “bar unit” of dates, e.g. monthly vs daily. This goes against the whole idea of the data driver graphs.

Goals

• time scale should calculate the smallest gap size between time elements, and use it as the band width.
• there should be a way to restrict that size with min/max as percentages of the width
• time scale should work similar to the band scale: width: {scale: 'xscale', band: 1} (consistent, simpler for novice users)
• the band size should account for any padding/rounding that time scale adds
• it should be possible for the user to add in-between bar padding, same as in band scale

Note that there was a similar request in 2013, but there was no example for it, so there might be a workaround.

CC: @domoritz @dnmiller @gerwitz @mkscrg

[kanitw]

@josesaranda Im not sure what you mean “the approach interaction occurs”.

In any case, one can make timeUnit “band” on a continuous scale by calculating two new fields, one for the start of the time unit and another for the end of time unit. Then we can draw rect from start to the end like this.

This is also the approach that we take in Vega-Lite 4.0.

[nyurik]

@kanitw your example highlights the issue I was talking about – you have to “know” your data range when designing the graph – by specifying yearmonth. You cannot design a truly data-driven graph, where the data decides the time units - months or minutes, and the graph can change on the fly depending on the data given.

The bands here are not discrete, but are also continuous. Bars do not have to align - in fact bars could appear at any place anywhere on the scale. It is imputing that makes them descrete. Using minimum gap as the bar width will make both the continuous and “pre-descretized” (data generated on the server as buckets) data work the same way.