Hacktoberfest 2026: los issues que los mantenedores marcaron para octubre, abiertos y aptos para principiantes. Explorar issues de Hacktoberfest

HermitCrab analysis: keep equivalent-analysis merging on while tracing

Abierto
  • ¿Algo no está bien?
#565 1 comentario 0 reacciones 0 asignados Ver en GitHub

Los mantenedores suelen responder en 1 día

Nadie ha tomado este issue todavía.

Evaluación

Dificultad
5/5
Tiempo estimado
Más de una semana
Aptitud para principiantes
30/100
Tipo de issue
Nueva funcionalidad
Claridad
Bastante claro
Estado de actividad
Activo
Stack tecnológico
csharp
Área
backend, machine-learning

Línea de trabajo

Start by reading AnalysisStratumRule and the census harness, then compare censusmerge with census across all grammars. Reconcile the synthesis-step claim and verify conformance verdicts against PR #480; review how trace callbacks and counts are reported for PR #534. Done means measurements support the change, trace behavior is agreed with the Try-a-Word product owner, and conformance remains identical.

Escrito por el modelo de indexación a partir del texto del issue.

Descripción

enhancement Hermit Crab

HermitCrab turns off equivalent-analysis merging while tracing, so a traced parse repeats lookups that an untraced parse folds. The trigger is any traced parse (Try-a-Word style): analyses reaching lexical lookup rise by 1.67x on Mbugwe and 2.19x on Aweti. In steps the effect is small, so expected value is low.

Problem (census, steps)

AnalysisStratumRule skips merging when tracing ("Don't merge if tracing because it messes up the tracing"). Census, traced unmerged versus merged: analyses reaching lookup Mbugwe 9,334 to 15,617 (x1.67), Aweti x2.19. Analysis event totals move by only -0.2% to 2.4%, because most of the extra work is lookup-side and lookups are 0.7% of Mbugwe events. The synthesis-side effect is unverified (the census says synthesis steps are unchanged, which is hard to reconcile with lookup-driven synthesis; it needs rechecking before any claim).

Proposed change

Keep merging on while tracing and record folded duplicates as one node with a count. This is a product decision on how Try-a-Word shows folded duplicates, not only an optimization.

Soundness

Merging is already used untraced and is parse-identical there (ledger rows 4, 5, 16 family; PR #493). The open soundness question is the trace: callbacks for a merged duplicate are skipped, so LexicalLookupCompleted counts and trace-node counts change meaning, which affects PR #534's reported counts.

Memory bound

One count field per folded trace node. No cache.

Expected value (low, possibly negative)

Analysis events change by only 0.2 to 2.4% (the plan lists it as -0.2..2.4%, so the sign is uncertain); lookups divided by 1.67 (Mbugwe) to 2.19 (Aweti) under tracing. Memory gain probably modest (lookups are 0.7% of Mbugwe events), unmeasured.

Measurement plan

Census harness: censusmerge mode versus census mode on all grammars, report events, lookups and synthesis steps separately; first reconcile the synthesis-steps claim.

Kill criterion

Close if the synthesis-step reconciliation shows no reduction, or if the product owner of Try-a-Word does not want folded duplicates in the trace.

Dependencies

None. Interacts with PR #534 (trace nodes) and T1 (trace memory).

Related

  • PR #490: HermitCrab optimization round record (ledger rows cited above are in it).
  • PR #480: HermitCrab conformance suite. Conformance per-fixture verdicts must be identical with and without the change.
  • Units: deterministic analysis events and synthesis steps, never ms. "% less work" = (off - on) / off per grammar, never pooled. Census figures are ceilings unless stated. Ranges marked "est." are estimates that the shadow phase replaces with measurements.
  • Owner constraint: no memo or cache; any new structure must be grammar-sized and at most a few MB.

🤖 Generated with Claude Code

Lenguaje dominante
C#
Estrellas
29
Forks
16
Merge medio
6 d 13 h
PR fusionados (30 d)
12

Preparar el entorno

  • Sin Dockerfile ni archivo de Docker Compose
  • Tiene una plantilla de pull request
  • Sin guía de contribución

Primeros pasos

  1. Lee el issue completo y luego la guía de contribución del proyecto.
  2. Comenta en el issue que vas a ocuparte — evita que dos personas hagan lo mismo.
  3. Haz un fork del repositorio y trabaja en una rama.
  4. Abre un pull request que haga referencia al número del issue.

Más de sillsdev/machine

Todos los issues de sillsdev/machine

Issues similares

Más issues de C#

Recibe los nuevos issues en tu correo

Un resumen breve de issues de GitHub para principiantes.