I have a feature idea/request and would like comments before attempting anything myself:
When using heatmap to display images that are larger than the pixel area, the data gets decimated by the backend in order to fit into the display. This is fine if your interested in big features.
However the data I frequently work with is small peaks against a large noisy background.
If the data then gets automatically decimated, it ends up being luck whether the peak is sampled or discarded and it changes as the display window gets resized (physically resized and/or adding plot labels etc.).
Thus ideally I want to be able to apply my own decimation to ensure that the peak is not discarded. I can think in other applications, if interested in minimums, the decimation would look different to ensure they are retained.
For a backend that generates an image file, I can try to work around this effect, by first doing my own decimation beforehand. However for an onscreen display, it becomes more problematic. Also if the backend allows pan and zoom, the decimated data is not what you want to display when the user starts zooming.
Thus my feature request would be a way to be able to apply a custom/controlled decimation to the data of a heatmap, with some sort of feedback/call back on viewport changes (only if using a backend with interactivity).
While at some level I feel this should be separate from a package like Plots, working on top of it, however I would still like to be able to have that as a subplot in a larger plot, where the others are not necessarily this type of decimated heatmap. I’m also not sure if the interactive feedback would be achievable if this functionality is in a library running on top of Plots. Is this where Plot recipes come in?
Or is it only Plotly(JS) providing this interactivity and I show work at that level instead of trying to use Plots abstraction?
Would something like this be useful to others?