Feat static tree

[ipa-jsk]

Use static tf tree (tf2) instead of publishing with fix rate.
Closes #14

[ipa-danb]

Suggested change

	print(f"updating element {element.name}")

[ipa-danb]

Suggested change

	parser.add_argument("-r", "--rate", type=int, help="Rate for broadcasting. Does not involve tf frames.")
	parser.add_argument("-r", "--rate", type=int, help="Rate for broadcasting. Does not involve tf frames. This argument is deprecated and will be removed in future versions.")

[ipa-danb]

Would also be nice to add a warning in line 319

[ipa-jsk]

Is it deprecated? One could still use it.

[ipa-danb]

Does this break the "delete" feature as it is implemented right now? Delete Frames will still be latched and wont vanish if you delete them?

[ipa-jsk]

Does this break the "delete" feature as it is implemented right now? Delete Frames will still be latched and wont vanish if you delete them?

This is true indeed.. well, with the current non-static publisher, the frames also are not deleted, but just not published anymore, such that a lookup causes an error after some time. This would not be possible with static trees.. see also this issue:

ros/geometry2#370

The delete feature still has the problem that you cannot create a frame with the same name again. So, its actually almost useless. Could also be an argument to get rid of the delete feature?

We could still implement both variants (static and non-static) as you suggested earlier, if there is any value in the deletion feature.

[ipa-danb]

Does this break the "delete" feature as it is implemented right now? Delete Frames will still be latched and wont vanish if you delete them?

This is true indeed.. well, with the current non-static publisher, the frames also are not deleted, but just not published anymore, such that a lookup causes an error after some time. This would not be possible with static trees.. see also this issue:

ros/geometry2#370

The delete feature still has the problem that you cannot create a frame with the same name again. So, its actually almost useless. Could also be an argument to get rid of the delete feature?

We could still implement both variants (static and non-static) as you suggested earlier, if there is any value in the deletion feature.

I mean the delete referes to deleting it from the frame editor, not from the TF tree. The only problem with the deleting is just that you cannot reassign it if it already exists in the TF tree. With the current implementation, i see it as a bug that you cannot reassign a frame. We can fix it by clearing the frame buffer ( as implemented in #39 ). For static Transformer we would need a new concept for saveguarding reassignment, e.g., by tracking which frame has ever been published by the frame editor (and allow assignment then).

[ipa-jsk]

The only problem with the deleting is just that you cannot reassign it if it already exists in the TF tree.

Are you sure? I thought you can just publish a new transform. I'll test.

[ipa-danb]

The only problem with the deleting is just that you cannot reassign it if it already exists in the TF tree.

Are you sure? I thought you can just publish a new transform. I'll test.

The issue lies with the implementation in frame editor. it checks for frames in the buffer and only allows for ones that are not set

    def btn_add_clicked(self, checked):

       existing_frames = set(self.editor.all_frame_ids())

        name, ok = QtWidgets.QInputDialog.getText(self.widget, "Add New Frame", "Name:", QtWidgets.QLineEdit.Normal, "my_frame");

        while ok and name in existing_frames:
            name, ok = QtWidgets.QInputDialog.getText(self.widget, "Add New Frame", "Name (must be unique):", QtWidgets.QLineEdit.Normal, "my_frame")
        if not ok:
            return

Which makes sense because it prevents from publishing transform that jump around

[ipa-jsk]

-> idea: Check which frames are static and allow only reassigning them

[ipa-jsk]

-> idea: Check which frames are static and allow only reassigning them

@ipa-danb check new commit. It was more difficult than expected. This was the smartest way I could find.