Skip to content

scratch.md

Latest commit

 

History

History
110 lines (101 loc) · 2.74 KB

scratch.md

File metadata and controls

110 lines (101 loc) · 2.74 KB
Raw

todo:

  • purge legacy converters:
    • ...
  • block generator
  • type hints for definitions
  • key warnings
  • bailouts for:
    • blockly json
    • docs markdown
  • dev server
  • clean gh:a, ensure tests run correctly

Action Specification

const actionSpecification = {
  // version of this JSON spec
  version: "1.0.0-beta.1",

  // job-level settings
  settings: {},

  // list of top-level expressions ("commands")
  expressions: [
    { conditional: {
      if: {
        compare: {
          left: { variable: "temperature" },
          comparator: ">=",
          right: 90
        }
      },
      then: { publish: {
        feed: { feed: "Feed:Thermostat" },
        value: 85
      }},
      else: {
        conditional: {
          if: {
            compare: {
              left: { variable: "temperature" },
              comparator: "<=",
              right: 70
            }
          },
          then: { publish: {
            feed: { feed: "Feed:Thermostat" },
            value: 75
          }}
        }
      }
    }}
  ]
}}

Validate/Parse/Save Routine

  • validate against json schema
    • catch broad issues of data well-formedness
  • validate permissions on all referenced resources
    • catch hacking and data leaks
  • validate IO account standing
    • no SMS without IO+!
  • find all "triggerables":
    • feed references
    • temporal references (dates, times, durations, schedules)
    • power-up features
  • hoist ids to a db column
    • model_references or something, ["Feed:12345", "PowerUp:54321"]
    • so those resources can query what actions reference them
  • clear/register observers
    • cron schedules
    • feed hooks
    • power-up apis

Execution

  • initialize:
    • set up observers:
      • watchdog timer
      • expression constrainer
      • system monitor
      • stats reporter
      • run logger (to support live-view of running job, or later replay)
    • set up a variable context
      • hash/openstruct/w/e with controlled access (namespacing, security)
      • pre-fill with global data, feed values, power-up info, etc
  • execute expressions[]
    • synchronously
      • later could be async via Threads, Fibers, or sidekiq jobs
    • only executes and notifies observers
      • use observers for all other desired behavior
  • cleanup:
    • ask all observers to finalize/report
      • timestamps, completion statuses, logs, etc
    • job status updates
      • disable, error, etc as appropriate
  • failure cleanup: expected errors the user might cause
    • show good error/log messages
    • help them fix their own problems
    • notify us as appropriate
  • error cleanup:
    • includes unexpected errors that:
      • need to come to our attention
      • need to be hidden from the user but give good messaging
      • maybe provide the user a way to reference this issue to us