[WIP] Sampling points from geometry #184
Conversation
asinghvi17
commented
Jul 21, 2024
asinghvi17
commented
|
|
||
| ```@example sample | ||
| import GeoInterface as GI, GeometryOps as GO | ||
| p1 = GI.Polygon([[[-55965.680060140774, -31588.16072168928], [-55956.50771556479, -31478.09258677756], [-31577.548550575284, -6897.015828572996], [-15286.184961223798, -15386.952072224134], [-9074.387601621409, -27468.20712382156], [-8183.4538916097845, -31040.003969070774], [-27011.85123029944, -38229.02388009402], [-54954.72822634951, -32258.9734800704], [-55965.680060140774, -31588.16072168928]]]) |
There was a problem hiding this comment.
Can we use tuple points in examples? (To teach good habits)
| end | ||
|
|
||
| function sample(alg, geom, n) | ||
| return apply(x -> _sample(alg, GI.trait(x), x, n), application_level(alg), geom) |
There was a problem hiding this comment.
Should we even use apply for this? What happens to a feature collection or DataFrame?
My intuitive understanding would be to sample points from all geometries of all of the features weighted by area.
There was a problem hiding this comment.
At this point each feature will be sampled independently, so it's a dataframe to dataframe transfer. Since everything else also works the same way (centroid etc) I think this is at least consistent?
Am open to discussion though, not sure how people generally use this stuff.
There was a problem hiding this comment.
GeoPandas samples per geometry, and returns a multipoint:
https://geopandas.org/en/stable/docs/reference/api/geopandas.GeoSeries.sample_points.html
GO.centroid does return a single point even if you pass in a feature collection or table. (There's an issue with the Natural Earth north Korea geometry as well :D)
There was a problem hiding this comment.
Aah, yeah apply won't work here since it will try to reconstruct.
I guess this is a good time to figure out the generic apply iterator thing?
Co-authored-by: Rafael Schouten <rafaelschouten@gmail.com>
| end | ||
|
|
||
| function sample(alg, geom, n) | ||
| return apply(x -> _sample(alg, GI.trait(x), x, n), application_level(alg), geom) |
There was a problem hiding this comment.
Aah, yeah apply won't work here since it will try to reconstruct.
I guess this is a good time to figure out the generic apply iterator thing?
| return map(1:n) do _ | ||
| edge_idx = sample(edge_idxs, edge_probabilities) | ||
| x1, y1 = edges[edge_idx][1] | ||
| x2, y2 = edges[edge_idx][2] | ||
| distance = edge_lengths[edge_idx] | ||
| t = rand() * distance | ||
| (x1 + t * (x2 - x1), y1 + t * (y2 - y1)) | ||
| end |
There was a problem hiding this comment.
The better way to do this is probably to define a vector of points, and then set indices in that vector. Could also open the way for multithreading in the future.
|
Some interesting links about triangle sampling:
1D floating point sampling: |
