Dataflow networks have application in various forms of stream processing, for example for parallel processing of multimedia data. A great deal of success has been found in the static scheduling of dataflow networks, providing predictable runtime performance and minimizing the overhead of synchronization. As such applications become larger and more complex, there is a need for mechanisms for composing such networks from components, without violating interface constraints required for safe and correct execution. Such composition will sometimes be done dynamically, including loading and replacing dataflow subgraphs at run-time. A metalanguage for expressing such composition mechanisms is considered, where metaprograms provide the glue for combining dataflow graphs.
Metaprograms are related to dataflow graphs by a type and effect system that captures the firing behavior of the dataflow programs. The type system ensures the correctness of the composition of dataflow graphs, ensuring the correct scheduling of actors in the dataflow graph, in a context where the scheduler is defined as part of the application.