Nice! For an orthographic projection like this you could determine the currently viewable extent of the scene similar to how axes ranges are set dynamically: https://github.com/fastplotlib/fastplotlib/blob/main/fastplotlib%2Fgraphics%2F_axes.py#L401

Zoom isn't a reliable way to do it I think.

Once you have the extent, you could use different subsample levels for a few different scales of extent ranges. This would be some kind of basic LOD system I think. @almarklein might have a better idea.

Would need more thought to extend this to perspective projections.

I think with this kind of data we assume an orthographic projection (i.e. fov==0).

Zoom isn't a reliable way to do it I think.

The camera.zoom property is indeed not a good measure; that one is intended for quick temporary zooming. I admit this is a bit confusing...

The current "zoom level" of the camera is defined by its .width and .height properties. In time-series-like data like this, just using .width makes the most sense for this use-case. The .width together with the camera's .world.position.x determines what part of the x-axis is in view.

Super-cool use-case BTW 🤘

I ended up essentially replicating the Clikchouse demo. The SQL from the demo outputs an image as a result, so it is very easy to render using fastplotlib, and it creates some nice-looking graphs:

I also turned it into an animation, using the Simple Image Update example, and made a post on it here (trying to promote fastplotlib a bit🙂).

The animation function will render continuously which will have a performance impact by constantly updating the buffer, unless you put in condition to update only when there's significant changes in the current extent. If you post your animation function I can take a look.

Cool post, thanks for sharing! 😄