Improve computations

[cmeyer]

Start here:

https://github.com/orgs/nion-software/projects/33

Summary 2022:

• Computations Project
• distinctions between fast and long computations; up-to-date, stale, incomplete stages (2023-11-02)
• separate data-like computation (with incremental functionality) from ui-like agent (commit phase) (2023-01-10)
• agent can have way of keep state, i.e. a set of graphics it controls (2023-01-10)
• could separate graphics like rectangle into the graphic component and a model data structure/item (2022-12-24)
• separate computation description (UI) from computation values from arguments passed to computations themselves (2022-12-23)
• may need a custom computation designer to replace old computation editor and allow custom computations (2022-12-23)
• data structures can be a fundamental items that gets registered (2022-12-12)
• separate data structure implementation from data structure idea (2022-12-12)
• data structures could be often implemented as a class which can read/write itself and resolve references (2022-12-12)
• no need to get more fancy (2022-12-12)
• separate the computation description/structure from the storage of values (2022-07-15)
• inputs more hierarchical (arrays, references, etc.) (2022-07-15)
• preferably make dependencies use seeds for simplified updates (2022-07-15)
• how to handle crops/masks? (2022-07-15)
• need a plan to eliminate 'source' relationship (2022-07-15)
• partial/overlapped computations for performance (2022-07-10)
• computation migrations for future stability (2022-07-10)
• api for maintenance (2022-07-10)
• implied computations, not explicitly saved, established when package loads (2022-07-10)
• combine data-structure and computation-values? (2022-07-10)
• need to see quickly what computations are attached to a data item (2022-07-02)
• thoughts about background model, data-structures (2022-07-02)
• there should be an API for finding computations associated with a data item (2022-07-02)
• distinguish between a particular piece of data which has a UUID and a stream of data which can also have a UUID (2022-05-20, #430)
• computations should only act on low level data to simplify them (2022-01-03)
• agents updating graphics should be a separate process, mostly standardized (2022-01-03)

[cmeyer]

A diagram of how elemental mapping structures get deleted (mostly accurate, but incomplete, trying out graphs).

flowchart LR
    EdgeRef -->|ref| SpectrumImage & Edge & EELSData & PickGraphic;
    PickGraphic -->|src| EELSData;
    EELSData -->|src| BackgroundComputation