We can use get_help() to retrieve the config information. Could we change the configuration? For example, I want to increase the number of points allowed in trace.

WIP

this will allow the python bindings to link to a static protobuf library which will circumvent problems with the qgis python console. it's a little patchy still, at some point we might want to include tile building and depend on spatialite which will cause the same problem. there's a good chance though that spatialite static is in debian's package manager and is compiled with position-independent-code :crossed_fingers:

still needs:

• [x] choose a protobuf tag. probably latest, cmake wasn't introduced until very recently
• [x] smth is messed up with pre-commit
• [x] update the manylinux image to have the image built protobuf from source
• [x] adapt and test windows in qgis (comment: couldn't make static linking of protobuf work on windows, maybe bcs libvalhalla is linking dynamically.. actually it took the protobuf-lite.lib but still the resulting .pyd would, maybe additionally, link to the .dll)
• [x] adapt and test mac os in qgis 🤞 (didn't actually test it works in qgis, but only built the lib on mac)

it appears that mvsc does not copy the dependency .dll's to the python distribution (it does do that with valhalla's cmake). so we need another tool similar to auditwheel & delocate. seems that https://github.com/adang1345/delvewheel is a good option. even cibuildwheel considered using it by default.

new, other PR didn't build anymore for some reason..

branch to collect all things to be done before releasing v3.0.0:

• ~~all main OSs are integrated~~
• improve documentation/tutorials with some binder stuff

branch to collect all things to be done before releasing v3.0.0:

• ~~all main OSs are integrated~~
• improve documentation/tutorials with some blender package

fixes #2, fixes #3

finally worked for both linux & windows. eventually the process became:

• we checked in all needed libraries, including valhalla, for all platforms (mac still missing)
• so we have to build any binary dependency locally and also update all headers etc. easy for win (simply copy the vcpkg libs & headers), but it's harder for linux of course (script provided), mac we'll see, expected to be easier than linux
• valhalla is registered as submodule but is solely used locally when we want to update valhalla
• added __version__ and __valhalla_commit to __init__.py, guess that'll become important
• releasing scheme will be .., but only after we're out of 0.x

Package and system information: pyvalhalla 3.0.3 with python 3.10.6 in the Debian Linux 11.

My task was to map match several trips to the roads. The issue was that, after trace_route returned a RuntimeError with "Exact route match algorithm failed to find path" when map matching a trip, map matching another trip caused the "Segment fault" error.

For maximum performance & output information we could try protobuf in-/output optionally. Probably quite a bit of work, if we want to do it properly, so I’d first try to benchmark the HTTP service to see if it’s worth it. Also, would be much nicer if all endpoints are protobuf compatible, e.g. matrix/isochrone & expansion are not.

I haaate the location stuff for Valhalla😅 Coords ALWAYS come as lists in one way or another so let’s at least give some convenience function that accepts a ([lon, lat], **kwargs) and returns a location object.

Another year passed, python 3.11 is out, so we should build for that as well. Thankfully it's not updated yet on Arch, but usually is within a few weeks, so let's get it built together with #19.

Hi,

I am map matching a large number of trips to the roads (more than 1000). I currently use a for-loop strategy to match one trip for each request. The issue is when there is a failed match, pyvalhalla will return an error message and stop the loop. Is there a configuration to avoid pyvalhalla to return the error message?

Or is there a better way than for loop to map match a large number of trips?

Thank you so much.