Note that mymodule.cluster_analyzer in the comments to #205 is sort of a placeholder for extensibility that hasn't been well-specified yet. In #205, it implies the implementation of something like what is discussed in #208
But @peterkasson has pointed out that we would like to package chunks of work, like a sequence of command line operations, into a single object with inputs and outputs that can be used like a single work graph operation. This probably just means that the subgraph idea in #205 should be convertible to an Operation type.
I specifically say "Operation type" because we probably want to have the option of using the subgraph as a template to create multiple graph nodes. It gives us some freedom to continue to refine our concept of state and data flow across iterations.
follow up to #205
Note that
mymodule.cluster_analyzerin the comments to #205 is sort of a placeholder for extensibility that hasn't been well-specified yet. In #205, it implies the implementation of something like what is discussed in #208But @peterkasson has pointed out that we would like to package chunks of work, like a sequence of command line operations, into a single object with inputs and outputs that can be used like a single work graph operation. This probably just means that the subgraph idea in #205 should be convertible to an Operation type.
I specifically say "Operation type" because we probably want to have the option of using the subgraph as a template to create multiple graph nodes. It gives us some freedom to continue to refine our concept of state and data flow across iterations.