Error dissolution with layered recovery strategy

[lyzkov]

It is highly recommended to propagate faults through a dedicated error pipeline instead of stream output to take advantage of Combine consistency leveraging try operators in a throwable manner.

Faults should pass over at most three recovery layers before displaying the recovery message in the exception view.

1. System space - providing debug instructions to resolve a source of recurring error.
2. Application space - providing severity that promotes fluency over verbosity so triviality bloats are diminished.
3. User space - providing exceptional flow enabling users to decide about critical actions (system alert with buttons) or recovery view to excuse a cumbersome hit.

The recovery strategy should involve:

• Ignoring erroring occurrence:
  • Empty completion of a single value stream.
  • Continuation of multiple values stream.
• Retrying from an erroring occurrence which results in subscribing to upstream once again.
• Passing error downstream through transcription gate providing contextual description about nuisance.

For handler convenience, it should be possible to materialize erroring stream completion to Result type.

As a bonus point, recovery could involve debugging mode that collects logs with traces on every step.

The above strategy, metaphorically reduced to a battleship game, adopts Abort, Retry, Fail? anti-pattern and assesses the holistic approach of dealing with bugs. Primely reintroduced and documented in SinkEmAll asks for exhaustive exploration.

Please take care and good luck! ^^

[lyzkov]

Take look at https://github.com/danielt1263/CLE-Architecture-Tools/blob/main/Utilities/ErrorRouter.swift