Better big data support for `scattermapbox`

[jackparmer]

We're getting Dash apps through the roof that use scattermapbox. Would be nice to see how far we can push the # of points with a WebGL overlay instead of an SVG one to improve big data support for scattermapbox traces.

Maybe we could investigate: https://github.com/vicapow/react-map-gl-stack-gl-overlay-example

This plot of 🌳 s in Manhattan for example is starting to push the mapbox-gl limits:

https://plot.ly/~jackluo/444

/cc @jackwluo @dfcreative

[etpinard]

I think you mean the mapbox-gl limits? scattermapbox renders all things in using mapbox-gl which uses purely WebGL. No SVG whatsoever here. Editing ⤴️ accordingly.

I think we should investigate other pure mapbox-gl solutions before doing this:

1. find what are those pure mapbox-gl limits i.e. try to plot the as many as possible using only mapbox-gl (not via the plotly.js wrapper)
2. plotly.js uses mapbox-gl@0.22.1, which more than 10 minor versions behind the mapbox-gl latest, so we should test bleeding edge mapbox-gl to see if the rendering better
3. try mapbox-gl clustering.

[immotus]

Are there any plans to update mapbox-gl dependency to the latest version?

[rreusser]

@PavelGolodoniuc See #2029 to track progress on that.

[etpinard]

Performance for

Plotly.d3.json('https://plot.ly/~jackluo/444.json', (err, fig) => {
    delete fig.layout.mapbox.accesstoken
    Plotly.newPlot(gd, fig)
})

is pretty solid using mapbox-gl@0.44.0 🎉

[etpinard]

#2361 will improve scattermapbox performance significantly.

Closing this thing (for now).