PEGO
A110: Expected-Set IDs
Repeated diagnostic strings and copied expectation lists make memo/error bookkeeping expensive. Compile-time IDs and windows of one stack reduce repeated storage.
Mechanism and Invariants
Description tables resolve integer IDs into messages. Each isolated region owns a window of the expectation stack; sets that must survive in memo or recovery state are copied into an arena, with reuse of equal consecutive kept sets where supported. Order, farthest position and message flags remain exact; IDs are internal representation, not user-visible diagnostics.
Scope and State
Default expectation-ID representation across closure, both VMs, generated Go Node/typed Go and TypeScript. Go arena allocation and JavaScript array ownership differ.
Reference Control
No universal same-source string-set comparator exists. The absence filter changes duplicate search, not stored diagnostic identity.
Evidence and Implementation
Historical record 004 reports Minilang allocation volume 59 to 42 MB and timing about 128 to 103 ms. Record 011 ports the representation among generated changes. Current per-backend isolated effects are unmeasured.
Evidence records: 004.
Source owners: record, isolate and keep,
Program.DescTable, generated expectation IDs, TypeScript description IDs.
The comparison policy distinguishes same-source controls, historical changes and full checkpoints.
Related mechanisms: M100.