Add MeanCenter trait #1166
Add MeanCenter trait #1166
Conversation
|
I think this is normally called a centroid. |
|
There are subtle differences between centroids and the mean center of point sets. I've added some tests that compare the centroid for each geometry with the mean center. There are differences, notably for a polygons. Here's a blog post that (briefly) describes its utility. I was planning on adding a |
It's obviously an automatic approval from me because the blog post uses Irish CSO data, but I'd love to see the weighted mean centre method too. |
|
🤣 heard. I've already pushed the weighted_mean_center() method and doc changes. I just need to add tests which I'll find time for tonight or tomorrow. 🫡 |
There was a problem hiding this comment.
Can you add an entry for this in the algorithm index? https://docs.rs/geo/latest/geo/#algorithms
Yup, you're right! OOPS! 🙈 The mean center for This technique is most useful for feature collections. For example, finding the weighted mean center of 311 reports grouped by category (MultiPoint per category) and the weight would be days since report or estimate cost to fix etc. Related doc from arcgis implementation I've updated the trait doc and implementation to reflect this. I will need to rewrite the tests, though! Thank you for your help :) |
| let centroids = self | ||
| .iter() | ||
| .map(|g| g.centroid()) | ||
| .filter(|c| c.is_some()) |
There was a problem hiding this comment.
You can replace the map/filter/map(unwrap) with flat_map(|g| g.centroid())
(this occurs in several places in this PR)
| /// | ||
| /// The `weights` argument takes a slice which is used as an iterator which is cycled. | ||
| /// If there are fewer weights than coordinates, the weights will be recycled until the | ||
| /// calculation has completed. |
There was a problem hiding this comment.
If there are fewer weights than coordinates, the weights will be recycled until the calculation has completed.
Is there a useful reason for this cycle behavior or some precedent?
Or is this more like "do something, anything, so long as we don't explode"?
| return None; | ||
| } | ||
|
|
||
| for ((i, coord), weight) in self.coords_iter().enumerate().zip(weights.iter().cycle()) { |
There was a problem hiding this comment.
Hm. The blog post uses centroids here as well right?:
xy_arr[i] = (ft_geom.Centroid().GetX() * weight, ft_geom.Centroid().GetY() * weight)
| }; | ||
| } | ||
|
|
||
| impl_mean_center_option!(LineString); |
There was a problem hiding this comment.
It looks like you're computing centroid here.
Two things:
-
The blog uses one weight per geometry, not per coord in the geometry.
-
For linear features like a LineString, the blog post uses midpoint instead of centroid. Though I can't think of an application of this algorithm where using a midpoint would be more useful than a centroid. I know there are applications where you want the "middle" of the geometry to be within a geometry, which might not happen with centroid (which is why people love polylabel), but in this case, we're eventually averaging a bunch of points together, so there's no guarantee that the output point will be inside of any geometry anyway.
CHANGES.mdif knowledge of this change could be valuable to users.This PR adds a new trait
MeanCenterwhich calculates the euclidean mean center of a set of coordinates. It is implemented for all geometry types.