Should allow "drilling down" into nested model frames #293

EricCousineau-TRI · 2020-05-28T19:15:13Z

Towards #278

Should happen after #283

Some options:

Just "hoist" duplicates of the frames (and frame graphs) with {model_name}::{frame_name}
Actually try to parse the name and actually navigate down

The text was updated successfully, but these errors were encountered:

Signed-off-by: Steve Peters <[email protected]>

scpeters · 2020-08-11T05:48:01Z

Some User Story ideas for this feature:

Loading a model that contains :: syntax for referencing frames inside a nested model in the following places:
- //joint/parent
- //joint/child
- //frame/@attached_to
- //pose/@relative_to
Resolving a SemanticPose to a nested frame (i.e. link->SemanticPose.Resolve(pose, "nestedModel::nestedLink");)
Resolving the attached-to body for a frame attached to a nested link (i.e. it returns "nestedModel::nestedLink")
Resolving the attached-to body for a frame attached to a nested model to the canonical link of that nested model
Resolving a joint parent/child to a nested link (i.e. it returns "nestedModel::nestedLink")
Support passing scoped names to the following functions:
- Model::ModelByName
- Model::LinkByName
- Model::JointByName
- Model::FrameByName
Getting a SemanticPose object corresponding to a scoped frame name (i.e. Errors Model::SemanticPoseByName(std::unique_ptr<SemanticPose> &_semanticPose, const std::string _scopedFrameName)). We could actually implement an unscoped version of this in libsdformat9 that just checks the Links, Joints, and Frames in the current model for matches with the passed name and returns the SemanticPose of the named object.

scpeters · 2020-08-12T19:17:00Z

Support passing scoped names to the following functions:
* Model::ModelByName

* Model::LinkByName

* Model::JointByName

* Model::FrameByName

I've prototyped a variation on this in scpeters@9eec062 by adding Model::*ByNestedName and Model::*NestedNameExists methods that accept a :: delimited name while leaving the existing APIs unchanged.

EricCousineau-TRI · 2020-08-13T19:23:19Z

For supergraph + subgraph assembly, scoping, and queries, options for recursion:

Copy + Graft
Scoped Alias + Propagate
Pros/cons are in maintaining invariants and ease of local checks (e.g. encapsulation at each level).

For element queries, e.g. LinkByName vs. LinkByNestedName, LinkByName being able to resolve with "::" may be the best for backwards compatibility (the legacy composition behavior for include) as well as being able to easily resolve elements (e.g. Joint::ResolveChildLinkX()).

Signed-off-by: Steve Peters <[email protected]>

scpeters · 2020-08-14T01:02:41Z

For element queries, e.g. LinkByName vs. LinkByNestedName, LinkByName being able to resolve with "::" may be the best for backwards compatibility (the legacy composition behavior for include) as well as being able to easily resolve elements (e.g. Joint::ResolveChildLinkX()).

I reverted the new APIs in my prototype branch in bd618e9 and added :: syntax to the existing APIs instead.

scpeters/sdformat@32b3c39...nested_model_dom

Nested model elements (`//model/model`) are currently supported in the SDFormat 1.7 spec, but are not supported by the DOM API or frame semantics operations in libsdformat 9.2. Per Amendment 1 of the SDFormat 1.7 proposal (sdformat.org/tutorials?tut=pose_frame_semantics_proposal#amendment-1-directly-nested-models), this PR adds support for nested models in the DOM API and frame semantics (fixing #283) through three steps: * adding `Model::Model*` methods for accessing nested models via the DOM API (047ec96) * loading nested models in `Model::Load` (b57fea2, step 3 of model parsing stages (sdformat.org/tutorials?tut=pose_frame_semantics_proposal#1-model)) * supporting nested models (`//model/model`) in frame semantics as opaque frames to match how models in the world scope (`//world/model`) are treated (85e0b4f, steps 6-9 of model parsing stages (sdformat.org/tutorials?tut=pose_frame_semantics_proposal#1-model)) The first two steps are straightforward, while the 3rd step adds new behavior. This behavior is added to `libsdformat9` because it is compatible with the SDFormat 1.7 spec (which supports nested models) and makes the treatment of models more consistent with regard to frame semantics. In libsdformat 9.2.0, a `//world/model` supports frame semantics: they have frames that can be referenced by name with `//pose/@relative_to` and `//world/frame/@attached_to` values can resolve to a `//world/model`. This extends that same behavior to nested `//model/model` elements; they now have their own frames in the frame and pose graphs, and a `//model/frame` is permitted to attach to a model. This does not add support for referencing elements within a model via the `::` syntax in the frame semantics or DOM APIs. That will be added in SDFormat 1.8 (see #293). Signed-off-by: Steve Peters <[email protected]>

Nested model elements (`//model/model`) are currently supported in the SDFormat 1.7 spec, but are not supported by the DOM API or frame semantics operations in libsdformat 9.2. Per Amendment 1 of the SDFormat 1.7 proposal (sdformat.org/tutorials?tut=pose_frame_semantics_proposal#amendment-1-directly-nested-models), this PR adds support for nested models in the DOM API and frame semantics (fixing gazebosim#283) through three steps: * adding `Model::Model*` methods for accessing nested models via the DOM API (047ec96) * loading nested models in `Model::Load` (b57fea2, step 3 of model parsing stages (sdformat.org/tutorials?tut=pose_frame_semantics_proposal#1-model)) * supporting nested models (`//model/model`) in frame semantics as opaque frames to match how models in the world scope (`//world/model`) are treated (85e0b4f, steps 6-9 of model parsing stages (sdformat.org/tutorials?tut=pose_frame_semantics_proposal#1-model)) The first two steps are straightforward, while the 3rd step adds new behavior. This behavior is added to `libsdformat9` because it is compatible with the SDFormat 1.7 spec (which supports nested models) and makes the treatment of models more consistent with regard to frame semantics. In libsdformat 9.2.0, a `//world/model` supports frame semantics: they have frames that can be referenced by name with `//pose/@relative_to` and `//world/frame/@attached_to` values can resolve to a `//world/model`. This extends that same behavior to nested `//model/model` elements; they now have their own frames in the frame and pose graphs, and a `//model/frame` is permitted to attach to a model. This does not add support for referencing elements within a model via the `::` syntax in the frame semantics or DOM APIs. That will be added in SDFormat 1.8 (see gazebosim#293). Signed-off-by: Steve Peters <[email protected]>

Nested model elements (`//model/model`) are currently supported in the SDFormat 1.7 spec, but are not supported by the DOM API or frame semantics operations in libsdformat 9.2. Per Amendment 1 of the SDFormat 1.7 proposal (sdformat.org/tutorials?tut=pose_frame_semantics_proposal#amendment-1-directly-nested-models), this PR adds support for nested models in the DOM API and frame semantics (fixing #283) through three steps: * adding `Model::Model*` methods for accessing nested models via the DOM API (047ec96) * loading nested models in `Model::Load` (b57fea2, step 3 of model parsing stages (sdformat.org/tutorials?tut=pose_frame_semantics_proposal#1-model)) * supporting nested models (`//model/model`) in frame semantics as opaque frames to match how models in the world scope (`//world/model`) are treated (85e0b4f, steps 6-9 of model parsing stages (sdformat.org/tutorials?tut=pose_frame_semantics_proposal#1-model)) The first two steps are straightforward, while the 3rd step adds new behavior. This behavior is added to `libsdformat9` because it is compatible with the SDFormat 1.7 spec (which supports nested models) and makes the treatment of models more consistent with regard to frame semantics. In libsdformat 9.2.0, a `//world/model` supports frame semantics: they have frames that can be referenced by name with `//pose/@relative_to` and `//world/frame/@attached_to` values can resolve to a `//world/model`. This extends that same behavior to nested `//model/model` elements; they now have their own frames in the frame and pose graphs, and a `//model/frame` is permitted to attach to a model. This does not add support for referencing elements within a model via the `::` syntax in the frame semantics or DOM APIs. That will be added in SDFormat 1.8 (see #293). Signed-off-by: Steve Peters <[email protected]>

Nested model elements (`//model/model`) are currently supported in the SDFormat 1.7 spec, but are not supported by the DOM API or frame semantics operations in libsdformat 9.2. Per Amendment 1 of the SDFormat 1.7 proposal (sdformat.org/tutorials?tut=pose_frame_semantics_proposal#amendment-1-directly-nested-models), this PR adds support for nested models in the DOM API and frame semantics (fixing gazebosim#283) through three steps: * adding `Model::Model*` methods for accessing nested models via the DOM API (047ec96) * loading nested models in `Model::Load` (b57fea2, step 3 of model parsing stages (sdformat.org/tutorials?tut=pose_frame_semantics_proposal#1-model)) * supporting nested models (`//model/model`) in frame semantics as opaque frames to match how models in the world scope (`//world/model`) are treated (85e0b4f, steps 6-9 of model parsing stages (sdformat.org/tutorials?tut=pose_frame_semantics_proposal#1-model)) The first two steps are straightforward, while the 3rd step adds new behavior. This behavior is added to `libsdformat9` because it is compatible with the SDFormat 1.7 spec (which supports nested models) and makes the treatment of models more consistent with regard to frame semantics. In libsdformat 9.2.0, a `//world/model` supports frame semantics: they have frames that can be referenced by name with `//pose/@relative_to` and `//world/frame/@attached_to` values can resolve to a `//world/model`. This extends that same behavior to nested `//model/model` elements; they now have their own frames in the frame and pose graphs, and a `//model/frame` is permitted to attach to a model. This does not add support for referencing elements within a model via the `::` syntax in the frame semantics or DOM APIs. That will be added in SDFormat 1.8 (see gazebosim#293). Signed-off-by: Steve Peters <[email protected]>

Nested model elements (`//model/model`) are currently supported in the SDFormat 1.7 spec, but are not supported by the DOM API or frame semantics operations in libsdformat 9.2. Per Amendment 1 of the SDFormat 1.7 proposal (sdformat.org/tutorials?tut=pose_frame_semantics_proposal#amendment-1-directly-nested-models), this PR adds support for nested models in the DOM API and frame semantics (fixing gazebosim#283) through three steps: * adding `Model::Model*` methods for accessing nested models via the DOM API (047ec96) * loading nested models in `Model::Load` (b57fea2, step 3 of model parsing stages (sdformat.org/tutorials?tut=pose_frame_semantics_proposal#1-model)) * supporting nested models (`//model/model`) in frame semantics as opaque frames to match how models in the world scope (`//world/model`) are treated (85e0b4f, steps 6-9 of model parsing stages (sdformat.org/tutorials?tut=pose_frame_semantics_proposal#1-model)) The first two steps are straightforward, while the 3rd step adds new behavior. This behavior is added to `libsdformat9` because it is compatible with the SDFormat 1.7 spec (which supports nested models) and makes the treatment of models more consistent with regard to frame semantics. In libsdformat 9.2.0, a `//world/model` supports frame semantics: they have frames that can be referenced by name with `//pose/@relative_to` and `//world/frame/@attached_to` values can resolve to a `//world/model`. This extends that same behavior to nested `//model/model` elements; they now have their own frames in the frame and pose graphs, and a `//model/frame` is permitted to attach to a model. This does not add support for referencing elements within a model via the `::` syntax in the frame semantics or DOM APIs. That will be added in SDFormat 1.8 (see gazebosim#293). An expected console message was updated in a test while merging forward to libsdformat11. Signed-off-by: Steve Peters <[email protected]>

Nested model elements (`//model/model`) are currently supported in the SDFormat 1.7 spec, but are not supported by the DOM API or frame semantics operations in libsdformat 9.2. Per Amendment 1 of the SDFormat 1.7 proposal (sdformat.org/tutorials?tut=pose_frame_semantics_proposal#amendment-1-directly-nested-models), this PR adds support for nested models in the DOM API and frame semantics (fixing #283) through three steps: * adding `Model::Model*` methods for accessing nested models via the DOM API (047ec96) * loading nested models in `Model::Load` (b57fea2, step 3 of model parsing stages (sdformat.org/tutorials?tut=pose_frame_semantics_proposal#1-model)) * supporting nested models (`//model/model`) in frame semantics as opaque frames to match how models in the world scope (`//world/model`) are treated (85e0b4f, steps 6-9 of model parsing stages (sdformat.org/tutorials?tut=pose_frame_semantics_proposal#1-model)) The first two steps are straightforward, while the 3rd step adds new behavior. This behavior is added to `libsdformat9` because it is compatible with the SDFormat 1.7 spec (which supports nested models) and makes the treatment of models more consistent with regard to frame semantics. In libsdformat 9.2.0, a `//world/model` supports frame semantics: they have frames that can be referenced by name with `//pose/@relative_to` and `//world/frame/@attached_to` values can resolve to a `//world/model`. This extends that same behavior to nested `//model/model` elements; they now have their own frames in the frame and pose graphs, and a `//model/frame` is permitted to attach to a model. This does not add support for referencing elements within a model via the `::` syntax in the frame semantics or DOM APIs. That will be added in SDFormat 1.8 (see #293). An expected console message was updated in a test while merging forward to libsdformat11. Signed-off-by: Steve Peters <[email protected]>

EricCousineau-TRI · 2020-08-20T18:41:40Z

@azeey is looking into the global graph.

EricCousineau-TRI · 2020-09-03T18:33:01Z

Per f2f, summary on global graph is to couple the scoping and encapsulation to call sites (e.g. at each query to large graph from elements) -- vs. having some special aliasing view.

Nested model elements (`//model/model`) are currently supported in the SDFormat 1.7 spec, but are not supported by the DOM API or frame semantics operations in libsdformat 9.2. Per Amendment 1 of the SDFormat 1.7 proposal (sdformat.org/tutorials?tut=pose_frame_semantics_proposal#amendment-1-directly-nested-models), this PR adds support for nested models in the DOM API and frame semantics (fixing gazebosim#283) through three steps: * adding `Model::Model*` methods for accessing nested models via the DOM API (047ec96) * loading nested models in `Model::Load` (b57fea2, step 3 of model parsing stages (sdformat.org/tutorials?tut=pose_frame_semantics_proposal#1-model)) * supporting nested models (`//model/model`) in frame semantics as opaque frames to match how models in the world scope (`//world/model`) are treated (85e0b4f, steps 6-9 of model parsing stages (sdformat.org/tutorials?tut=pose_frame_semantics_proposal#1-model)) The first two steps are straightforward, while the 3rd step adds new behavior. This behavior is added to `libsdformat9` because it is compatible with the SDFormat 1.7 spec (which supports nested models) and makes the treatment of models more consistent with regard to frame semantics. In libsdformat 9.2.0, a `//world/model` supports frame semantics: they have frames that can be referenced by name with `//pose/@relative_to` and `//world/frame/@attached_to` values can resolve to a `//world/model`. This extends that same behavior to nested `//model/model` elements; they now have their own frames in the frame and pose graphs, and a `//model/frame` is permitted to attach to a model. This does not add support for referencing elements within a model via the `::` syntax in the frame semantics or DOM APIs. That will be added in SDFormat 1.8 (see gazebosim#293). Signed-off-by: Steve Peters <[email protected]>

EricCousineau-TRI · 2020-09-24T18:27:41Z

Per f2f, global graph w/ tests working.

Currently working on how to validate graphs (e.g. handling the no-longer-opaque nested models)
Has scoped graph (via weak_ptr), adds some scoping context.

EricCousineau-TRI changed the title ~~Allow "drilling down" into nested model frames~~ Should allow "drilling down" into nested model frames May 28, 2020

EricCousineau-TRI added this to the SDFormat 1.8 / libsdformat10 milestone May 28, 2020

EricCousineau-TRI mentioned this issue May 28, 2020

[epic] proposal: Should provide better composition behavior #278

Closed

17 tasks

scpeters mentioned this issue Jul 17, 2020

[sdf9] Support nested models in DOM and frame semantics #316

Merged

EricCousineau-TRI assigned scpeters Jul 23, 2020

scpeters added a commit to scpeters/sdformat that referenced this issue Jul 24, 2020

Add comment referencing gazebosim#293.

4e99441

Signed-off-by: Steve Peters <[email protected]>

scpeters added a commit to scpeters/sdformat that referenced this issue Aug 13, 2020

Add comment referencing gazebosim#293.

9d4fd6d

Signed-off-by: Steve Peters <[email protected]>

EricCousineau-TRI mentioned this issue Aug 20, 2020

Should support nested canonical links for libsdformat11 (with explicit specification) #343

Closed

EricCousineau-TRI assigned azeey and unassigned scpeters Sep 10, 2020

scpeters mentioned this issue Sep 10, 2020

Make included models work as nested models via DOM in libsdformat10 #284

Closed

EricCousineau-TRI mentioned this issue Sep 17, 2020

Should support nested canonical links for libsdformat (implicit for 9/10, implicit+explicit for 11) #342

Closed

azeey mentioned this issue Oct 1, 2020

Allow "drilling down" into nested model frames #381

Merged

azeey closed this as completed in #381 Nov 18, 2020

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Should allow "drilling down" into nested model frames #293

Should allow "drilling down" into nested model frames #293

EricCousineau-TRI commented May 28, 2020

scpeters commented Aug 11, 2020

scpeters commented Aug 12, 2020

EricCousineau-TRI commented Aug 13, 2020 •

edited

Loading

scpeters commented Aug 14, 2020 •

edited

Loading

EricCousineau-TRI commented Aug 20, 2020

EricCousineau-TRI commented Sep 3, 2020

EricCousineau-TRI commented Sep 24, 2020

Should allow "drilling down" into nested model frames #293

Should allow "drilling down" into nested model frames #293

Comments

EricCousineau-TRI commented May 28, 2020

scpeters commented Aug 11, 2020

scpeters commented Aug 12, 2020

EricCousineau-TRI commented Aug 13, 2020 • edited Loading

scpeters commented Aug 14, 2020 • edited Loading

EricCousineau-TRI commented Aug 20, 2020

EricCousineau-TRI commented Sep 3, 2020

EricCousineau-TRI commented Sep 24, 2020

EricCousineau-TRI commented Aug 13, 2020 •

edited

Loading

scpeters commented Aug 14, 2020 •

edited

Loading