Shared Look-Ahead Preparation

Purpose

This document defines the current shared-prefix identification metadata produced by the preparation layer.

The intent is to prepare future duplicated-work reduction research while preserving the current parser runtime contract.

Shared-prefix metadata does not change parser execution.

Scope and boundaries

Current shared-prefix preparation is:

  • deterministic;
  • conservative;
  • structural;
  • advisory metadata only.

Current shared-prefix preparation is not:

  • shared-prefix execution;
  • parser graph execution;
  • continuation replay;
  • scheduler control;
  • parse-tree control;
  • diagnostics authority.

Metadata semantics

Shared-prefix metadata means only:

  • participating alternatives begin with structurally similar token/reference prefixes;
  • prefix depth can be described in ordered structural token form.

Shared-prefix metadata does not mean:

  • alternatives are semantically equivalent;
  • alternatives are interchangeable;
  • one alternative can be executed on behalf of another.

Structural token preparation

Structural token sequences are computed by AlternativeStructuralPrefixExtractor during preparation, before scheduler orchestration begins.

The extractor inspects grammar model objects (RuleContent, Sequence, RuleRef, LiteralMatch) and produces lightweight AlternativeStructuralDescriptor records carrying only token name strings. These descriptors are forwarded to ParserSharedPrefixPlanFactory, which aggregates them without accessing grammar model objects.

AlternativeScheduler receives pre-computed descriptors and forwards them; it does not inspect rule content, traverse grammar structures, or extract tokens itself.

Conservative detection model

Detection is intentionally under-approximated.

It reports shared prefixes when they are structurally obvious from rule metadata (for example, identical leading RuleRef and LiteralMatch sequences).

It does not attempt semantic equivalence or deep inference over complex nested constructs.

Determinism and non-authority

For a given grammar snapshot, produced shared-prefix metadata is deterministic.

Produced metadata remains non-authoritative:

  • parser acceptance remains owned by ParserEngine;
  • scheduling remains sequential and deterministic;
  • diagnostics ownership remains unchanged.

Future intent

This metadata enables future, explicitly gated design exploration for duplicated-work reduction.

Any future execution activation requires dedicated design, tests, and roadmap updates, and is out of scope for this preparation step.

Edit this page