Support for columnar datasets: arquero and quarto - #1324

Hi @Fil .

Just a note to say I'm very interested in the feature. Is it still planned?

Context:
I'm writing a visualizer for large sets of timeseries data that I've gone to a great deal of trouble to transform into Apache-Arrow IPC on the backend, so it's a shame that Observable plot has to transpose it to rows for rendering (though I was very pleased to discover it will accept an arrow Table as a data object!)

To be clear what this PR is trying to do is autodetect (by duck-typing) a few columnar formats and transform them into iterators that generate objects with key:value pairs. I don't think we should go this way, and I'll close this PR.

Plot supports channels passed directly as columns, saving one copy of the values (a copy will have to be made as soon as you have a scale or a data transform, anyway). I think that is what you want?

For example with apache-arrow@14, if C is a column:

const C = Arrow.makeVector(Float64Array.from({ length: 100 }, d3.randomLcg(42)));
Plot.lineY({length: C.length}, {y: C}).plot()