Functional stackable substrates in D
These are templates as in the "Range" paradigm, which operate on concurrent processing stacks. Substrates are arranged in a heirarchy and connected via channels.
Substrates are grouped via containers, and implemented using Stacks.
A container is a tree of substrate definitions, and is generated by the compiler. A client would create the stacking method overrides for each type that they want to use for a substrate binding. The stack generates match functions for the input bindings and defines a fiber function for each potential invokation. These are converted into a tuple of matching rules, and we can add our own rules and fibers.
Channel inputs are tuple patterns which evoke signals on concurrent processing fibres. The fibres are created by the Stack and when an input event becomes available, each matching fibre receives a message containing the input value or state from the channel.
The stack also maintains a concurrent "Environment" which is local to a fiber, and provides a means of accessing global state data. The enviroment is constant during the executuion of a particular fiber but may be modified during a yield state. And to a custom fiber function, the Environment is the only parameter supplied by a stack. The input and output vectors are bound to the environment context.
Each output is bound to a function, and the stack can implement a default output funtion that indexes a table of instance bindings and works as a pipe, sending an input message to a different substrate.
Examples
Generic CRUD bindings. This example creates a set of bindings that can attach to a webserver object of some type to perform the routing of crud requests.
Non-blocking server stack Configurable assets that implement a TCP / HTTP stack which is configurable to use various types of socket.