Component API Refinement List

[aseth1]

PR #810 addresses a few major issues that makes the current API unusable (see #748) but several tasks remain for making it robust and ready for the 4.0 release (i.e after the Hackathon) . These include:

• [x] implement a Component name validater (e.g ComponentPathName) to centralize name checking and to address #596
• [x] Strict enforcement against nameless Components. Many models and components have subcomponents without names (e.g. the GeometryPath of a PathActuator, Muscle). I am currently assigning a name instead of firing the ComponentHasNoName exception.
• [ ] Strict enforcement of component path naming for connectors (i.e. Connectors connectee_name). Currently the ComponentNotFoundOnSpecifedPath exception is being trapped by Connector::findAndConnect() and findComponent() is invoked.
• [ ] In both cases above, we need a mechanism to allow a "lenient" mode so older models can be read in and made to comply. This could take the form of a static OpenSim::NamingStrictness enum
• [ ] Optimizing the subcomponent iterators (e.g. cache the tree or other data structure) such that the traversal tree is not rebuilt on ever getComponent call.
• [x] Getting access to connectors, inputs, outputs and state variables should follow the same naming convention and use the same traversal methods
• [x] providing macros to construct subcomponents and automatically provide accessors (e.g. by wrapping access by index, similar to properties) OR find a way to preserve subcomponents as data members who's owner is automatically reset to the new copied component and not stale or reset to nullptr.

[tkuchida]

• [x] verify that findComponent() is no longer misbehaving. The hopper example reports

Model'bouncing_leg'::findComponent() Found '/bouncing_leg/ground_block/yTranslation' as a match for: Component '/bouncing_leg/ground_block/yTranslation' of type Coordinate, but it is not on specified path.

which sounds like it can't find A but it found A instead.

[tkuchida]

Related to naming: see #390. There was also some discussion about assigning a unique unsigned int to each Component upon creation (to ensure connections are preserved even if a Component's full path changes) (see #1192); if so, a unique name (e.g., "component[uniqueComponentID#]") could be assigned by default.

[jenhicks]

@aseth1 Can you review the list and close issue if everything is fixed or tracked with a separate issue elsewhere?