Does PATH$path_link_vertex need an order column?

[krlmlr]

Otherwise the vertex order is determined only by the order of rows in the data frame, and would be lost if copying to a database.

[mdsumner]

yes for sure ... I consider this all up-for-grabs but there's been very little interest or feedback at all - in practice I've found I use PATH0() much more readily (long character keys are too heavy), but the same issue applies in the nested data frames on PATH0()$object$path_.

I didn't really know how to go about any of this, I originally thought there'd be some kind of tricks available but it seems you just have to trade compactness for flexibility as required, and the tools in silicate reflect the different intermediate forms I tend to use as well as some loftier ideas that don't necessarily need to exist.

Thanks!

[mdsumner]

I'd be inclined to write some kind of extension or replace to PATH0() that assumes the use of a database back-end for the vertex indices than modify silicate itself (ala #134 ) and also takes into account the use of {geometries} and/or {interleave} for the conversion/s: dcooley/geometries#4