PEGO
G120: LSP Position Checkpoints
An editor repeatedly converting positions on a long line should not scan that line from the start each time. Checkpoints bound UTF-16/code-point conversion work.
Mechanism and Invariants
The language-server text index records cumulative UTF-16 and code-point counts every 64 bytes. Conversions start at a suitable checkpoint and scan the remaining local segment. UTF-8 boundaries, supplementary characters and line/column behavior remain in the text-index contract.
Scope and State
Default language-server position conversion only. Closure, VMs, generated Go and TypeScript parser matching are not applicable; this is editor index work.
Reference Control
No maintained same-source no-checkpoint comparator exists.
Evidence and Implementation
Record 065 reports a one-megabyte single-line grammar analyzed/tokenized in about 0.35 s versus over 10 s, while ordinary grammar analysis remains around 11 ms. The adversarial long-line result does not quantify parsing-runtime speed.
Evidence records: 065.
Source owners: text index checkpoints and conversions.
The comparison policy distinguishes same-source controls, historical changes and full checkpoints.
Related mechanisms: M090.