aw_interpolate() fails when attempting to cast to MULTIPOLYGON about areal HOT 5 CLOSED

Also, just wanted to give an early thanks for this package! I've done some areal interpolation in R before, but it makes a lot of sense to turn it into a package.

from areal.

Thanks @mfherman!

This might be a bit of an edge case - the only reference I can find to the error from sf::st_cast(), which was the root case, is here.

Your intersection is creating geometry collections, which our test data has not done. Reading between the lines of the sf issue, it seems like you might be creating slivers with the intersection but I didn't verify that. I've added a line to aw_intersect() that gets a version of your code working correctly.

I'm going to leave this open until we merge the PR - we've got a backlog since the package is under CRAN review. What we'll do is wait until we hear back from the initial review. If we have to resubmit, will merge the multipolygon branch before we do.

Otherwise, will merge into master after CRAN acceptance and then submit a patch to CRAN in January.

from areal.

Thanks, @chris-prener! aw_intersect() on the multipolygon branch is working for my sf objects now.

from areal.

Glad to hear it @mfherman!

from areal.

@chris-prener -- I think the fix to this issue might be causing some performance issues when it isn't really needed.

A little back of the envelope profiling of my areal function vs a non-areal function gives me ~9 minutes for areal and ~30 seconds for non-areal. I dug into the function calls I'm making and found that this line is the killer. I also get the warning:

Warning message:
In st_collection_extract.sf(intersection) : x is already of type POLYGON.

Reproduce with: big_test.rda.zip

load("big_test.rda")
int <- areal::aw_intersect(target_polygons, data_source_cells, "area")

I'm not really sure what the consequences of not calling that are or what the best check would be to see if you don't need it... will do some research and get back to you.

from areal.