mrgnc => triage => release
current modernity in system design is best left with the concise description, complexity
i am developing an application that runs on windows (at first) and integrates data flows from hetregenous sources while permitting random users to conduct analytics on pipelines consructed from these random data flows
the requirement for something like this will lead you towards providing a palette of data sources, a palette of operations, and a mechanism for property mapping and projection of arbitrary objects
additionally the ingress and egress operations need to be schedulable and filterable, re-authenticate semi-autonomously, expose a human readable log etc
long before the clown wars looked obsolete one would have approached the problem with a few asynchronous interfaces and a service oriented architecture
but autonomous services need autonomous management. that's where this effort comes in
this is not a rebel open source full featured workflow manager. it's a low footprint workflow manager for coders.
enjoy
Current state as of the timestamp on the document:
-
building towards supporting flowchart type process definitions, with bpm obviously in mind
-
nothing resembling running code with a 'built-in' gui currently in sight
-
major wins, specification of an 'active compound document' subsystem wherein ingressed data is 'effectively' typed and composable with attachable UI behaviour
that is, tuples expose their properties, can exist in mixed collections, and contain type information about their assignable consumers
that is, a taxonomy has been defined based around process definitions that contain pipelines that contain pipeline tools that can contain other pipeline tools with current archetypes of Pipeline, Workflow and Activity supported by Func<x,x> userSuppliedPredicates for execution tree scheduling decisions
For now, look at the code, maybe try compiling it and implementing the interfaces and extending the abstract classes and finding the pain points.
Use Visual Studio