Add a scatter (2d) plot?

[nvaytet]

Would look like:

[nvaytet]

What kind of API?

Would be nice to be able to call pp.scatter(x, y, z) where z is for the color.

In scatter3d, x, and y can only be strings. Should we consider allowing either strings or variables?
-> if they are strings, look up the coordinates in z, if they are variables, use as is?

This would mean that z could be optional if both x and y are variables.

[nvaytet]

What kind of API?

Would be nice to be able to call pp.scatter(x, y, z) where z is for the color.

In scatter3d, x, and y can only be strings. Should we consider allowing either strings or variables? -> if they are strings, look up the coordinates in z, if they are variables, use as is?

This would mean that z could be optional if both x and y are variables.

I don't think this works when we want to update data on an existing scatter plot.
If a view receives new data, it will be a data array where potentially both color values and x,y positions have changed.
So the x and y must be strings (set once and for all when the plot is created) that pick coordinates in the data array it receives upon update.

[nvaytet]

So it would have to be like the scatter3d plot:

pp.scatter(da, x='time', y='cost')

[nvaytet]

If you want to make a scatter plot using just two variables, and no color, this is covered by the new xyplot.

[nvaytet]

So it would have to be like the scatter3d plot:

I am no longer so sure about this. Maybe we can allow supplying colors and positions as separate variables, instead of all being bundled up in a single data array.
You can pass nodes to the xyplot and then update only the y values by changing the data in the corresponding node.
So we could do the same for the scatter plots.

Then maybe the xyplot is not needed, and it is just a scatter plot?
Or maybe it is needed because right now it supports making line plots, and also if the x coordinate is bin edges.