You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
geom_raster would be handy for heatmaps with many tiles. The current approach of drawing a single rectangle for each tile is extremely wasteful for large heatmaps, which essentially represent bitmaps.
We should attempt to utilize cairo rastering to implement this. As far as I understand it, cairo raster gets two callbacks to determine the color for each pixel.
Essentially we provide a proc, which calculates the color of each pixel in the main plot viewport in (I assume) global coordinates.
If this works out, we can think about implementing geom_hexbin as well. However, in order to support hex binning two dimensional data, we need a histogram proc for hex bins. But essentially if the hex size is known the computation of that is rather simple.
The text was updated successfully, but these errors were encountered:
What still is broken is using geom_raster with bin / count statistics. This is almost a requisite for geom_hexbin, because I assume almost no one would calculate a hex bin statistics manually. Drawing the hex bin is reasonably easy. Might be a good idea at that point to move some of the raster logic from ginger back to ggplotnim, so that a ginger backend only has to draw a finished bitmap without and logic attached.
geom_raster
would be handy for heatmaps with many tiles. The current approach of drawing a single rectangle for each tile is extremely wasteful for large heatmaps, which essentially represent bitmaps.We should attempt to utilize cairo rastering to implement this. As far as I understand it, cairo raster gets two callbacks to determine the color for each pixel.
Essentially we provide a proc, which calculates the color of each pixel in the main plot viewport in (I assume) global coordinates.
If this works out, we can think about implementing
geom_hexbin
as well. However, in order to support hex binning two dimensional data, we need a histogram proc for hex bins. But essentially if the hex size is known the computation of that is rather simple.The text was updated successfully, but these errors were encountered: