intersection analysis is missleading

[javisantana]

We have a intersection analysis that works with any geometry but no all the intersecions make sense:

• polygon - point <- perfect sense and the typical use case
• point - polygon <- what is the use case?
• line - point <- this does not work because the change a line a point intersects is really low
• line - line <- what is the use case?
• polygon - polygon <- what is the use case?
• point - point <- wont work

So my proposal is to simplify and have a polygon - point and then some specific like "aggregate by line" where we find closest point.

cc @saleiva

[saleiva]

+1

[zingbot]

Having done work with tons of students on this, here's the answer to a few:

point - polygon <- what is the use case?

• same as polygon-point

line - point <- this does not work because the change a line a point intersects is really low

• I can't think of one

line - line <- what is the use case?

• Actual intersections, like streets, or ship paths, trying to find the points where lines cross. Very useful in transportation work.

polygon - polygon <- what is the use case?

• Sales territories and metropolitan regions, or watersheds and states, or zip codes and census blocks

You are missing:
line- polygon
*find all roads in this provice, give me the routes paths of all runners in this park

[talos]

If you consider "intersection" more broadly as a spatial join, where the two geometries do not necessarily exist in the same space but have a relationship that the user might be concerned about, I've listed the use cases below. However, they might not really fit under the name "intersection" and should live in a different analysis:

• point - polygon <- This is the inverse of polygon-point, is the expectation that these are not symmetrical?
• line - point <- Finding the closest PoIs to a road. This probably dissolves to a line->buffer->point in polygon equation
• line - line <- Finding intersections (points where the lines cross)
• polygon - polygon <- Boolean operations. For example, subtracting one shape from another
• point - point <- Closest N points to another point. For example, I am thinking of building a fruit stand and I want to know the 3 nearest fruit stands. This doesn't dissolve to a buffer issue, it's nearest neighbors.
• polygon - line (not mentioned above) <- splitting a polygon into N other polygons based off of lines.

[talos]

Also, I'd say that I don't think any of the additional cases I outlined above are at all high priority, just that they do exist. And to reiterate, "intersection" is not the right term for many of them.

The only one I think that could be useful sooner rather than later would be point - point, just plugging into KNN. And also to make sure that point-polygon and polygon-point are interchangeable, if possible.

[stale]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.