stonier
commented
3 years ago For reference, connected to https://github.com/splintered-reality/py_trees/issues/305
Closed pitosalas closed 2 years ago
stonier
commented
3 years ago For reference, connected to https://github.com/splintered-reality/py_trees/issues/305
pitosalas
commented
3 years ago Yes:
import py_trees bb = py_trees.blackboard.Client()
For example…Thanks…
Pito Salas Faculty, Computer Science Brandeis University
On Nov 26, 2020, at 12:31 PM, Daniel Stonier notifications@github.com wrote:
I'm not actually using melodic, but did just enough to upgrade those branches for python3, pass tests and run demos so you might be running into some teething problems.
Got a script or example I could potentially reproduce your problems with?
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or unsubscribe.
stonier
commented
3 years ago Aye, just found your other issue :)
There's a large compatibility matrix (bottom of the readme) that associates py_trees/platform/ROS version. If you're using the ROS build farm's deb for melodic, then it's 0.6.x for py_trees and 0.5.x for py_trees_ros. Looks like you're trying to use the blackboard api for >1.x.y versions, which would generate attribute errors (class to use in 0.5.x is Blackboard, whereas class to use in >1.x.yisClient`).
Perhaps check you're trying to learn from the documentation for the correct version?
Or digging through the correct branch for the sources:
pitosalas
commented
3 years ago Ah I see! So the version that I get from ROS is the older version with the less fancy blackboard, yes?
Is there a way to get ROS to locate the newer version of py_trees?
Pito Salas Faculty, Computer Science Brandeis University
p.s. By the way, I forgot to compliment you on a really nice package, with good design, documentation, test cases, etc. Well done!
On Nov 26, 2020, at 12:45 PM, Daniel Stonier notifications@github.com wrote:
Aye, just found your other issue :)
There's a large compatibility matrix (bottom of the readme) that associates py_trees/platform/ROS version. If you're using the ROS build farm's deb for melodic, then it's 0.6.x for py_trees and 0.5.x for py_trees_ros. Looks like you're trying to use the blackboard api for >1.x.y versions, which would generate attribute errors.
Perhaps check you're trying to learn from the documentation for the correct version?
• http://docs.ros.org/en/melodic/api/py_trees/html/ • http://docs.ros.org/en/melodic/api/py_trees_ros/html/ — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or unsubscribe.
stonier
commented
3 years ago Aye, re-inventing the blackboard to be something better than a bunch of unobservable global variables was the last big push to 1.0.
Will get back to you later today re adopting a newer version (family calling!)
pitosalas
commented
3 years ago Hi Daniel
Thanks so much for your help!
Based on your feedback I tried to make a little progress by commenting out the blackboard stuff and using 0.6.8.
There I found that setup_with_descendants doesn't exist yet (fine) and I changed it on a guess to bt_root.setup(timeout=15). This produced (in Py 2.7) this error:
[CLOUD]~/catkin_ws$ rosrun prrexamples stroller_control.py
('init ', 'Turn2Target')
('init ', 'ApproachTarget')
('init ', 'TrackFixedDistance')
[DEBUG] Find Wall : Sequence.setup()
Traceback (most recent call last):
File "/my_ros_data/catkin_ws/src/prrexamples/scripts/stroller/stroller_control.py", line 74, in <module>
bt_root.setup(timeout=15)
File "/opt/ros/melodic/lib/python2.7/dist-packages/py_trees/composites.py", line 84, in setup
new_result = child.setup(timeout)
TypeError: setup() takes exactly 1 argument (2 given)
[CLOUD]~/catkin_ws$I mention this in case it is useful to you!
I look forward to your help in getting the 1.0 version to work in py2.7 (and if not, at least getting over this little hump :)
Best,
Pito Salas Faculty, Computer Science Brandeis University
On Nov 26, 2020, at 12:53 PM, Daniel Stonier notifications@github.com wrote:
Aye, re-inventing the blackboard to be something better than a bunch of unobservable global variables was the last big push to 1.0.
Will get back to you later today re adopting a newer version (family calling!)
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or unsubscribe.
stonier
commented
3 years ago ^Looks like you might have an ill-formed child. What type is that child that is failing? If it's your own class, then make sure it's inheriting and overwriting the setup method correctly (i.e. with timeout argument).
TLDR; It's likely there is no out-of-the box upgrade path for you on Melodic.
The 1.0+ libraries aren't compatible with the additional ROS1 tooling (messaging & services, blackboard & tree watchers, loggers, rqt_py_trees etc). They skipped straight to ROS2 and there is an entirely upgraded set of debug/viz tooling co-existing with the core libraries there.
That all could be backported, but there is a significant amount of effort required and ostensibly should be done by or in concert with someone making heavy use of them in ROS1 to iron out the wrinkles. I can't really do that without a local use case stress testing it, so opted not to do so to avoid making it unstable. There is also the problem of several companies relying on the pre-1.0 ROS1 api so a major version bump was out of the question. It could be released under a different namespace e.g. py_trees2 - but again, effort and instability.
On the other hand, if you're just planning to use py_trees, that's pretty portable, but would require a little bit of work. A few patches to make it python2 and catkin compatible.
pitosalas
commented
3 years ago Hi Daniel
Thanks a lot for taking the time to look at it.
To play back my understanding -- py_tree_ros 1.0 is moved to ROS2 and going back is not an option
But Py_tree 1.0 has a dependency on python 3 which might not be too hard to reverse. Can you give me guidance on where the key changes would be and whether it would be worth attempting (that is cost of back porting vs. benefits of additional features etc in 1.0?)
Thanks!
Pito Salas Faculty, Computer Science Brandeis University
On Nov 26, 2020, at 10:30 PM, Daniel Stonier notifications@github.com wrote:
^Looks like you might have an ill-formed child. What type is that child that is failing? If it's your own class, then make sure it's inheriting and overwriting the setup method correctly (i.e. with timeout argument).
TLDR; It's likely there is no out-of-the box upgrade path for you on Melodic.
The 1.0+ libraries aren't compatible with the additional ROS1 tooling (messaging & services, blackboard & tree watchers, loggers, rqt_py_trees etc). They skipped straight to ROS2 and there is an entirely upgraded set of debug/viz tooling co-existing with the core libraries there.
That all could be backported, but there is a significant amount of effort required and ostensibly should be done by or in concert with someone making heavy use of them in ROS1 to iron out the wrinkles. I can't really do that without a local use case stress testing it, so opted not to do so to avoid making it unstable. There is also the problem of several companies relying on the pre-1.0 ROS1 api so a major version bump was out of the question. It could be released under a different namespace e.g. py_trees2 - but again, effort and instability.
On the other hand, if you're just planning to use py_trees, that's pretty portable, but would require a little bit of work. A few patches to make it python2 and catkin compatible.
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or unsubscribe.
pitosalas
commented
3 years ago Hi Daniel
Is there a place where I can discuss use questions about py_trees. I don't think posting issues in the GitHub is the right way. Please advise? Thanks!
Pito Salas Faculty, Computer Science Brandeis University
On Nov 26, 2020, at 11:53 PM, Pito Salas pitosalas@gmail.com wrote:
Hi Daniel
Thanks a lot for taking the time to look at it.
To play back my understanding -- py_tree_ros 1.0 is moved to ROS2 and going back is not an option
But Py_tree 1.0 has a dependency on python 3 which might not be too hard to reverse. Can you give me guidance on where the key changes would be and whether it would be worth attempting (that is cost of back porting vs. benefits of additional features etc in 1.0?)
Thanks!
Pito Salas Faculty, Computer Science Brandeis University
On Nov 26, 2020, at 10:30 PM, Daniel Stonier notifications@github.com wrote:
^Looks like you might have an ill-formed child. What type is that child that is failing? If it's your own class, then make sure it's inheriting and overwriting the setup method correctly (i.e. with timeout argument).
TLDR; It's likely there is no out-of-the box upgrade path for you on Melodic.
The 1.0+ libraries aren't compatible with the additional ROS1 tooling (messaging & services, blackboard & tree watchers, loggers, rqt_py_trees etc). They skipped straight to ROS2 and there is an entirely upgraded set of debug/viz tooling co-existing with the core libraries there.
That all could be backported, but there is a significant amount of effort required and ostensibly should be done by or in concert with someone making heavy use of them in ROS1 to iron out the wrinkles. I can't really do that without a local use case stress testing it, so opted not to do so to avoid making it unstable. There is also the problem of several companies relying on the pre-1.0 ROS1 api so a major version bump was out of the question. It could be released under a different namespace e.g. py_trees2 - but again, effort and instability.
On the other hand, if you're just planning to use py_trees, that's pretty portable, but would require a little bit of work. A few patches to make it python2 and catkin compatible.
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or unsubscribe.
stonier
commented
3 years ago Python 2/3 Compatibilty for PyTrees 2.x
I don't think there would be too much involved in creating PR's for python2 compatibility on the devel branch. Tests and demos provide pretty good coverage, so you could start with them. The lads at CARLA are interested in this too. The critical piece is in making sure there is some CI job running tests to ensure it doesn't regress.
If you're just starting out though, I'd recommend sticking to the 0.6 version. Fundamentally it's the same api and workflow, with rough edges (especially around decorators & blackboard), so at least it'll get you familiar with api and development workflow before deciding you're seriously enough committed to working out what a backport looks like.
Use Case Questions
Hadn't really thought about a chat / forum, but it could certainly use something - questions do trickle in as issues and as you say, it's a rather sub-optimal means of fast-forwarding a conversation. I think gitter might serve as a lightweight experiment on that front. I'll see if I can set something up.
stonier
commented
3 years ago A gitter room now available at: https://gitter.im/splintered-reality/py_trees
If it wasn't for covid, we likely could have done :coffee: :)
pitosalas
commented
3 years ago Thanks, see you on the list!
Pito Salas Faculty, Computer Science Brandeis University
On Nov 29, 2020, at 11:22 PM, Daniel Stonier notifications@github.com wrote:
A gitter room now available at: https://gitter.im/splintered-reality/py_trees
If it wasn't for covid, we likely could have done ☕ :)
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or unsubscribe.
pitosalas
commented
3 years ago Hi Daniel
Python 2/3 Compatibilty for PyTrees 2.x
I don't think there would be too much involved in creating PR's for python2 compatibility on the devel branch. Tests and demos provide pretty good coverage, so you could start with them. The lads at CARLA are interested in this too. The critical piece is in making sure there is some CI job running tests to ensure it doesn't regress.
If you're just starting out though, I'd recommend sticking to the 0.6 version. Fundamentally it's the same api and workflow, with rough edges (especially around decorators & blackboard), so at least it'll get you familiar with api and development workflow before deciding you're seriously enough committed to working out what a backport looks like.
That's just what I did and I have something working. I will have detail questions about the exact behavior of control flow under different scenarios. Also I learned about catkin_virtualenv which would let me run py_trees. (Not py_trees_ros of course).
Use Case Questions
Hadn't really thought about making a chat room / forum available, but it could certainly use something - questions do trickle back here as issues and as you say, it's a rather sub-optimal way of fast-forwarding a conversation. I suspect there's even more that just don't get started. I think gitter might serve as a lightweight experiment on that front. I'll see if I can set something up. In the meantime, you can ping me on my email (as listed in the package.xml in the root directory).
Thanks!
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or unsubscribe.
This is just too trivial but how do I install py_trees_ros on Melodic? I am getting weird attribute errors which lead me to conclude that it is not installed (correctly.) But sudo apt install ors-melodic- ... doesn't work in any variation. I am overlooking something but I am can't pin it down!