Huge datasets unacceptably slow (250,000 datapoints)

[phildow]

Hypothesis: drawing the overview is the problem. 250,000 points shouldn't be an issue in memory, and the main area only draws the currently visible points instead of everything and then clipping. But the overview area draws all the points.

Proposed solution: if num(datapoints) > some x, downsample for drawing the overview, e.g. take every fifth to downsample to 10m intervals.

Need dataset to test