About the usability of distances and CRS

I’d like to start by thanking you all for your great work. Also, that I’m presenting this issue as a way to better understand the problems underneath so I can help and solve the issues. I’m willing to do the necessary pull requests.

Anyway, I mentor some students of data science and a recurrent problem I saw them (and myself) having is import data with latitude and longitude in geopandas and expecting the method .distance() will deliver distances in meters. Most of them end up using geopy, which, in their mind, “does the right thing by default”.

So, I’d like to improve the usability in GeoPandas regarding that. Have some suggestions:

• Create a “great circle” function.
• Add in the docs a section about projections and CRS and how to make the distances behave in a known manner.
• Add, in the distance function of the API docs, a link about “this will not necessarily wield a result in meters, please see the FAQ”

Thanks for your attention!

[jwilson8767]

For all my projects I end up projecting data to the state plane (units=meters) before using .distance() or .buffer() I do this so often that I have written my own buffer wrapper function that takes feet, meters, or degrees and attempts to rectify crs issues before applying the buffer to the geometries. I would be happy to contribute this to geopandas if there is interest (perhaps under something like gpd.safe_buffer(gdf, distance, projected_crs=crs) method)

Regardless I agree that documentation should be better and maybe any usage of .distance(), .buffer() or even .area should show a warning if the crs is not projected.

[jwilson8767]

Out of curiosity, beyond the basic geodesic functions, what others would you imagine someday wanting to utilize yourself?