In our open-trails.zip, the trailheads.geojson file does not currently contain the trail_ids property, or any values.
In the test data, those fields exist as many discrete columns, "Trail1", "Trail2" etc.
This is likely because I had left those functions commented out, because I wasn't sure about the logic of grabbing strings from those fields and identifying a match in named_trails.csv's "name" column, then grabbing the "id" value of each match, and writing it to an array.
We are 90% to consensus about changing the specification on this matter. By tomorrow at noon, we should be able to make a final call. The new logic would be associating trailheads with segments, via segment_ids.
For the lake-point-ohio data, that would mean simply discarding the columns with the names of named_trails.csv written in the many columns. This is the one dataset we have on hand that is less well served by this change.
cc @migurski