We should be able to make things more brief by using C++ namespaces to drive the pydrake module organization.
We pay the cost of non-granular dependencies (e.g. you need VTK, OSPRay, etc. just to import pydrake.math), but meh, whatevs, we haven't solved #6760 for a while, so we shouldn't hold back usability based on a 3-year-old issue.
This formulation may also resolve #12055, and possibly incorporate #7912
We should also consider a shortened import for pydrake, similar to how open3d, pandas, numpy, tensorflow, etc. suggest.
Slack conversations (a bit piecemeal, latest first):
Update (2020-10-26)
Per Drake Developer meeting (doc), it is a bit painful to use full spellings for symbol imports the flavors listed here: https://github.com/RobotLocomotion/drake/blob/v0.24.0/doc/python_bindings.rst#whats-available-from-python
We should be able to make things more brief by using C++ namespaces to drive the
pydrake
module organization.We pay the cost of non-granular dependencies (e.g. you need VTK, OSPRay, etc. just to import
pydrake.math
), but meh, whatevs, we haven't solved #6760 for a while, so we shouldn't hold back usability based on a 3-year-old issue.This formulation may also resolve #12055, and possibly incorporate #7912
We should also consider a shortened import for
pydrake
, similar to howopen3d
,pandas
,numpy
,tensorflow
, etc. suggest.Slack conversations (a bit piecemeal, latest first):
Old (2019-07-17)
From this comment: https://github.com/robotlocomotion/drake/pull/11783#pullrequestreview-263001864
At present, we have things like:
Ideally, we should move them to reflect the module organization, either directly in
pydrake.common
, or underneath it as:\cc @ggould-tri