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
By including pybind11/stl.h, we can have Pybind11 directly translate STL container elements (even those that are typedef'd) rather than go the roundabout way of defining a custom opaque type that may collide with another type.
For example, recently #1069 added a Rot3Vector to the wrapped code and this shouldn't be necessary. We should just be able to say std::vector<gtsam::Rot3> and use a regular list on the python side. This makes the wrapper more powerful and more intuitive.
The downside is that this will considerably change the Python API (using python based data types rather than our custom types) but it is the more beneficial approach in the long term.
Motivation
The hassle we had with #1069 is some pretty good motivation.
Pitch
I took the day off yesterday and worked a bit on this to see how much effort it was, and I made surprisingly great progress. I can make a PR as early as today.
The text was updated successfully, but these errors were encountered:
If it is very much API breaking I want to hold off until next major version (5.x). The book development can't tolerate this amount of disruption at this time.. Also, IROS....
Feature
By including
pybind11/stl.h
, we can have Pybind11 directly translate STL container elements (even those that are typedef'd) rather than go the roundabout way of defining a custom opaque type that may collide with another type.For example, recently #1069 added a
Rot3Vector
to the wrapped code and this shouldn't be necessary. We should just be able to saystd::vector<gtsam::Rot3>
and use a regular list on the python side. This makes the wrapper more powerful and more intuitive.The downside is that this will considerably change the Python API (using python based data types rather than our custom types) but it is the more beneficial approach in the long term.
Motivation
The hassle we had with #1069 is some pretty good motivation.
Pitch
I took the day off yesterday and worked a bit on this to see how much effort it was, and I made surprisingly great progress. I can make a PR as early as today.
The text was updated successfully, but these errors were encountered: