You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
We found some issues with pickling/unpickling the RTree during multiprocessing that was causing incorrect results and so we ended up pulling the method out to prevent confusion. We've moved to a new RTree data structure and so could investigate reintroducing the method with new checks around the safety of serializing the RTree.
As a sidebar, we've also just introduced map matching into our RouteE Compass application. That application is written in Rust and natively supports parallelism over batches. Since we just introduced it we don't have much documentation but it could be a good solution if you need higher performance over a large number of traces.
Thanks I lot!
In case it is useful, I meant that even with a single process I'm getting faster results (3-4x). I'm batching ~15 traces at a time and using the same matcher (before, I was initializing one geofence and matcher per trace), and this method is handy.
Activity
nreinicke commentedon Feb 26, 2026
Thanks for flagging! I've updated the API docs in #219 that should remove this.
jpherrerap commentedon Feb 27, 2026
Thanks! Any reason for removing
match_trace_batch? I've been using it on my project and it is a lot faster.nreinicke commentedon Feb 27, 2026
We found some issues with pickling/unpickling the RTree during multiprocessing that was causing incorrect results and so we ended up pulling the method out to prevent confusion. We've moved to a new RTree data structure and so could investigate reintroducing the method with new checks around the safety of serializing the RTree.
nreinicke commentedon Feb 27, 2026
As a sidebar, we've also just introduced map matching into our RouteE Compass application. That application is written in Rust and natively supports parallelism over batches. Since we just introduced it we don't have much documentation but it could be a good solution if you need higher performance over a large number of traces.
jpherrerap commentedon Feb 27, 2026
Thanks I lot!
In case it is useful, I meant that even with a single process I'm getting faster results (3-4x). I'm batching ~15 traces at a time and using the same matcher (before, I was initializing one geofence and matcher per trace), and this method is handy.