Closed daniel-j-h closed 7 years ago
Yeah we don't bundle it because we don't want to depend on libgdal
(of which we require a super outdated version that doesn't even ship with Debian anymore as far as I recall).
One way around this would be to let osrm-components
emit geojson and use tippecanoe
to convert it to vector tiles we can use.
What I can see on the osrm side is components not being supported by the mason build? Is this because of libgdal? @springmeyer any insights here - what's your take on this ticket?
Yes, we could support libgdal in the mason build (there are mason packages for gdal) but @TheMarex and I decided not to. I'm still in support of this because gdal is a large download and would slow down the mason fetch. I also agree that the best route forward is to target geojson in osrm-components
to avoid a gdal dep.
Another alternative:
I've used this, it works. The shapefile format is significantly smaller than a GeoJSON output would be. For our use case, we would only need a very simple wrapper.
Let's keep the discussion here free of upstream decisions, we already have
which outlines pros and cons of going the GeoJSON way.
https://github.com/Project-OSRM/osrm-backend/pull/3570 just landed upstream unblocking us from bundling osrm-components
with node-osrm
in the 5.6 release. Let's ship this!
The use case came up to quickly iterate on components and a modified profile making use of the shapefile generated by
osrm-components
. At the moment we do not bundleosrm-components
mainly because it is an internal tool, we run it on the demo server anyway, and there is the debug map rendering components in real-time.Can we bundle
osrm-components
easily with this package for npm?What I can see on the osrm side is components not being supported by the mason build? Is this because of libgdal? @springmeyer any insights here - what's your take on this ticket?
cc @geohacker @srividyacb @oini