voronoi | Animated Fortune 's Algorithm for Voronoi Diagram | Learning library

You need an option clip=box() within voronoi_frames(). Relevant code follows.

So an easy solve is to use scipy instead of pytess.

You could do the following:

I am not sure what's going on with powershell but if the issue is the length of the command-line, you can supply the sparse colour from a file like this:

A legend is created in ggplot2 when one or more columns in your dataset are mapped via aes() to one of the aesthetic modifiers (for example, fill, shape, size, color, alpha...). In your plot code, you are specifying a list of color values, but ggplot2 has no idea what that should be linked to in your data. Consequently, the colors are mapped according to the observations in your data frame tiny_countries line by line, without actually generating a link between your data and the color.

When you put color inside aes(), you indicate this modifier should be mapped to a particular column in your data. ggplot2 will then assign colors to each "different" value in your dataframe. If you just want ggplot2 to specify each tiny country is a different color, then, you need to pick a label for each country. We would naturally want the label to correspond to the name, since... that makes sense, so we would create the mapping via aes(color=name).

Note that this does not specify the color. Be default, ggplot2 will use it's default coloring scheme, but as you have already specified a list of proper number of values for the color... we can tell ggplot2 we want to override the default coloring and use your own specific colors manually. We do that using a scale_*_manual() function, where * corresponds to the particular aesthetic modifier in question. Putting this together you get this:

You can create a ListedColormap listing the desired colors. To decide which number maps to which color, a norm can be used, fixing 4 for the first color and 7 for the last. Both the colormap and norm need to be assigned to the PatchCollection. To position the tick labels, one can divide the range of 4 colored cells into 9 equally-spaced positions and take the ones at odd indexes.

Replace the line

If you are going to change the Figure then you must create a new canvas so you will have to destroy the previous one:

It seems that you are after a fixed-radius near-neighbor count query.

Using a KD-tree (K=2), you can list the points in a given circle, in time roughly O(k + Log n). This is achieved by overlapping the circle with the rectangles of the KD-tree subdivision. When a rectangle is wholly contained in the circle, there is no need to subdivide further, all points inside the rectangle are inside the circle.

So you can enhance the tree with the number of points contained in every node of the tree. This will lower the running time to O(m + Log n) where m is the number of rectangles contained in or split by the circle. It is also possible to stop subdividing when a rectangle contains less than a predefined number of points, and test them by brute force.

These are heuristics. You have to test them on your cases to see if they are worth.

Update:

Just an idea, not really investigated: Vantage-point trees are appropriate for k-nearest-neighbors searches and they partition the plane using circles. That might better blend with circle queries.

https://en.wikipedia.org/wiki/Vantage-point_tree#Searching_through_a_vantage-point_tree

The problem might be that geom_voronoi_tile expects the voronoi polygons to be closed, and your dataset lacks the outer boundaries. A quick alternative is to fall back to ggvoronoi::geom_voronoi().