PEGO
R100: Document Repetition Resumption
A long repetition can spend reparse time looking up every unchanged element. Saved run records let it reuse a prefix, rematch around one edit and splice in an unchanged suffix.
Mechanism and Invariants
Repetitions of at least 16 elements record positions, examined ranges, expectations and values. After an eligible single edit, resumption restarts near the affected element and rejoins when new and old boundaries align. Reused Nodes move under the Document contract; capture/variable and error ownership retain conservative eligibility. Multiple unparsed edits, aborts and reset/rollover can invalidate the saved run. Bytecode run sites are discovered from the module.
Scope and State
Default eligible long repetitions in closure, recursive VM and iterative VM Documents. Batch parsing, streams and generated Go/TypeScript are outside this resumption mechanism.
Reference Control
No maintained universal resume-disable same-source control exists. A fresh parse is a different operation and cannot isolate all run bookkeeping costs.
Evidence and Implementation
Historical record 054 measures a 100,000-line closure insert/delete reparse from 11.0 to 5.1 ms and long CSV from 15.1 to 8.4 ms. Saved runs cost about 80 bytes per element; short-repetition Minilang Documents cost about 2% more in that sample. Record 059 adds VM resumption. Benefit depends on length and rejoin opportunities, not every Document operation.
Source owners: run records and resumeRepeat, VM repetition run sites, saved runs and eligibility.
The comparison policy distinguishes same-source controls, historical changes and full checkpoints.