mdsumner
commented
7 years ago the segment table has a "segment_" id using the same ids machine - but there's a bunch of half-baked sc_edge and NARC and other stuff I left around - it's very much up in the air. I think I've seen all the important use-cases that are needed, but not yet mastered enough to see the big picture.
I've toyed with this "join-ramp" idea where a model has an attribute that tells it what tables it can safely join with a loop, because each table has the next-in-line primary key "up the ramp". This either to build the single big mega-join, or to propagate a subset through the tables. All open to review, certainly I've struggled to find analogous ideas elsewhere, though I'm sure they exist.
extending from this
osmdatadiscussion, rownames are not a solution for naming objects, and more memory-intensive alternatives are almost certainly the only general way. Where and how to store these in a way that still allows for convenientdplyrjoinoperations is the question? Also note that edges insc::PRIMITIVEs should also be considered named/nameable objects in their own right - it seems at the moment that they just get dumped with a generic integer index, right?