geojson or geopackage as option to shapefile outputs

[williamlidberg]

I find myself pushing the limits of shapefiles due to overly ambitious projects and ridiculously sized study areas. Alternative file types such as geopackage or geojson have been useful alternatives to shapefiles, and I am wondering if it would be technically possible/feasible to include an option to shapefiles in Whitebox, at least for tools such as individual_tree_detection that outputs very large shapefiles.

[jblindsay]

This is an issue for sure and I can see why you want an alternative to the Shapefile. In the past, for tools that can have a large number of point outputs, I have preferred using MultiPoint Shapefiles instead, e.g., see the LasToMultipointShapefile tool. The issue with that is that you then lose the ability to store attributes for individual points. I've looked into the geopackage format and don't think that it's a good candidate for Whitebox. The dependency on SQLite is a real problem for me. GeoJson has more appeal for me, but I think that is an ASCII format and therefore could be quite large and slow to parse when there are a great many shapes, although it wouldn't suffer from the same limitations as Shapefiles.

[brownag]

This is something that I think is worth further consideration as the Shapefile can be an awkward format.

I love Geopackage, and think it would have benefits for whitebox workflows as it can be used to store several grids or vector layers in the same file (for example)... but full support for it could require a lot of effort to integrate. And I understand @jblindsay's concerns w/ an added SQLite dependency in general.

GeoJSON would be a happy medium that most everything could make use of. Perhaps something like geobuf could be used for a much leaner, binary representation of that GeoJSON?

[giswqs]

It seems the geozero package has some functionality for reading and writing GeoPackgae. Would that be useful for this? https://github.com/georust/geozero

[jblindsay]

I'm aware, but it's a binding and requires a local install of SQLite. This is somewhat counter to a development principle of Whitebox.

[brownag]

Looking at dependencies (but not actually trying to implement), geobuf support would require the protobuf crate (and its dependencies) be added; serde-json crate is already used by whitebox-common. Would adding a dependency on a rust crate implementing protocol buffers be out of the question? Or would the whitebox approach (if a priority) be to write your own encoder/decoder?