mkjaergaard / darc Goto Github PK

Many non-templated classes are defined in header files. This code should be moved to a proper cpp file and investigate if some header dependencies can be removed.

Add support for a namespaces. Use same symbols as ros namespaces. Add ability to run a node and a component in a specific namespace.

Right now the node class has get???Manager for each available primitive type, e.g. pubsub, procedures, and timers.
This dependency should be removed. Add some generic interface to access a manager.

Create an executable (Node) what just prints out all nodes, topics and procedures.

A Signal is similar to publisher except it is:
Not using dataflow model, but is a variable that always has a value and can be read at any time.
Event handlers are triggered on changes, not on new data published.
Changes are guaranteed to be delivered to subscribers.

Can be used to provide information such as current state or values that are only required to be read in certain situations.

Go through the code and find places where errors are not handled properly.
Add a descriptive assert or warning printout.

Calls should go through Node class

To reflect a more network-stack-alike architecture.

To create a source checkout people can use to try out darc.

To reflect the current state of DARC.

Wrap Node class and other stuff that is not templated around message types to make the component executable and lib files smaller

Use control messages to exchange pub/sub information between nodes

Add support for a "Procedure" (ProcedureServer/ProcedureClient) or similar.
Works as a ros action/ros service. Should support guaranteed delivery.

For the final intern presentation.
Cool stuff with the python interface, some introspectrum, CPU usage stats, start/stopping, darc_info calls.

Right now the strings for the topics (tags) are handled a bit in the remote and the local subscribers. This should optimally be moved to a central place, where tags are mapped to topic ID's.