[rfc] getting the dataflow library ready for submission

Hello all, I am getting close to having the Dataflow library ready to be submitted for review, and would like to respark some interest. The current version of the library offers a generic layer for dataflow programming that can be applied to various data transport mechanisms, and a Dataflow.Signals support layer for Boost.Signals that provides a number of general-purpose components and easy connectability. In the near future I will try to put together more real-world examples of how this library can be useful, but for now the docs have two possibly interesting examples: * an example of how to provide a support layer for another mechanism, with VTK as the victim. * an example of how to use the Dataflow.Signals layer with Boost.Asio and Boost.Serialization to make a distributed dataflow network. The docs are uploaded to: http://dancinghacker.com/code/dataflow/ Code is available from the boost sandbox (SOC/2007/signals), as well as the boost vault (Dataflow directory). All tests (with 1 exception) pass on GCC 4.0.1/darwin and MSVC8. The docs (and the library to some extent) still need work, but I hope what's out there is good enough to paint a picture of what this library is about and how it might be useful. All feedback is welcome! I will next work on an example that uses Boost.GIL to construct an image processing dataflow network. Best regards, Stjepan

How does dataflow lib address the (rather broad) topic of scheduling? I am reminded of the Ptolemy project from Berkeley.

On Nov 12, 2007 5:36 PM, Neal Becker
How does dataflow lib address the (rather broad) topic of scheduling? I am reminded of the Ptolemy project from Berkeley.
As it stands, not in a whole lot of ways :-) The generic dataflow layer doesn't address scheduling (except you can invoke a component). The Dataflow.Signals layer included with the library (which uses Boost.Signals to transport data) is pretty much self scheduling since data transport is coupled with component invocation. The only functionality that remotely relates to scheduling is in the threading-related components (mutex, condition, and timed_storage). Thanks for the Ptolemy reference! I was not aware of that project and it looks very interesting. Stjepan
participants (2)
-
Neal Becker
-
Stjepan Rajko