crop should be able to read arbitrarily large inputs without running out of memory. Strategy is to
build a spatial of index input files
Stream read the points and accumulate them in the building footprints.
Once all points for a building have been read we can output that building and release the memory it occupied.
this should build further on how the StreamCropper currently already works. Initially implemented on one thread, multithreading can be added later if it makes sense. Output can be written to disk or streamed directly into the reconstruct app. Preferably this is done in a way with minimal overhead due to encoding/decoding.
crop
should be able to read arbitrarily large inputs without running out of memory. Strategy is tothis should build further on how the StreamCropper currently already works. Initially implemented on one thread, multithreading can be added later if it makes sense. Output can be written to disk or streamed directly into the reconstruct app. Preferably this is done in a way with minimal overhead due to encoding/decoding.