LA Time Machine: persistent indexed history buffer for large captures

[vailuc]

Background

After merging the SWD/JTAG/Modbus protocol decoders, the LA tool can now decode high-sample-count captures (e.g., the 20M-sample JTAG sigrok-dumps fixture). Rendering and decoding the entire buffer on every UI update is becoming slow.

Goal

Build a persistent, indexed history buffer that makes large captures searchable and scrubbable without UI lag.

Open questions

• Storage format: in-memory chunked arrays, IndexedDB, or a custom binary index?
• Index granularity: per-sample, per-decoded-frame, or fixed-time windows?
• Decoding should be lazy / on-demand for large captures.

Acceptance criteria

• A 20M-sample JTAG capture can be opened and scrubbed without the UI freezing.
• Decoded frames are indexed so users can jump to the next/previous transaction.
• The buffer design works for the existing UART/I2C/SPI/CAN/1-Wire/SWD/JTAG/Modbus decoders.

Related

• Part of the post-feat/la-protocols roadmap.
• See .devin/plans/la-roadmap-time-machine-and-beyond.md in the local workspace.