Discussing about the vispy.plot interface.

IMO, this interface should offer a modern, user-friendly, high-level plotting interface for common 2D and 3D visualizations.

Here are a few general principles:

• Forget about pyplot and MATLAB plotting API. vispy.plot doesn't intend to implement any of this. Everything related to pyplot should be implemented in vispy.mpl_plot, which offers a backend for matplotlib (which is already largely sufficient). We might also implement a couple of highly common pyplot operations there.
• Take inspiration from seaborn, bokeh, ggplot2, gadfly for the interface. The idea is to create ready-to-use high-level functions to create common plots like scatter plots, correlation plots, etc.
• Focus on those types of plots that scale to large datasets. For examples, things like that (sometimes called "charts") do not really scale to large datasets, and we shouldn't focus on them at first. Although having bar charts might be useful in some occasions, that's not what Vispy users will want.
• 2D and 3D visualizations should both be first-class citizen, and it should be easy to combine them.
• This interface should also provide user-friendly facilities for common interaction patterns like pan/zoon, linked brushing, turntable camera, arcball, etc.
• The plotting functions should focus on the data: what data structures will people generally have? If we can answer this, we ensure that most users won't have to transform their datasets to match whatever vispy.plot requires.
• vispy.plot should not be restricted to scientific plotting; it should allow users to display arbitrary shapes, 3D models, and most visuals we'll implement.

Questions

• Maps could be accessible here, or in another package vispy.maps or something.
• How customizable should this interface be? (CSS-like styling, per-vertex values for most parameters, possibility to write custom shaders using predefined hooks in the visuals?)