Revert PageRank benchmark withdrawal (#2259, pyg-bench#337); freshness gate must label stale numbers, not remove them

Aperta
  • Qualcosa non va?
#2,264 2 commenti 0 reazioni 0 assegnatari Vedi su GitHub

I maintainer di solito rispondono entro 1 giorno

Nessuno ha ancora preso questa issue.

Valutazione

Difficoltà
4/5
Tempo stimato
3-5 giorni
Idoneità per principianti
35/100
Tipo di issue
Refactoring
Chiarezza
Specificata chiaramente
Stato di attività
Attiva
Stack tecnologico
json, python
Ambito
ci-cd, documentation, testing

Direzione di ricerca

First revert pyg-bench#337 then pygraphistry #2259 on paired branches, restoring docs/source/gfql/benchmark_filter_pagerank.rst and charts. Modify docs/test_bench_numbers.py and pyg-bench export_docs_numbers.py so stale cells keep measured-on date and label instead of being removed; add test that fails if a published cell is dropped without maintainer marker. Done when reverted files are byte-identical and docs CI stays green past 2026-11-04 with staleness labels.

Scritto dal modello di indicizzazione a partire dal testo della issue.

Descrizione

docs gfql p0

Summary

#2259 (merge 968066cf8) and its paired producer change graphistry/pyg-bench#337 (merge 9fdca5ce1) pulled the published filter→PageRank results from the docs because they had passed the 60-day freshness limit. That withdrawal was the wrong remedy, and both changes are being reverted. Old numbers should stay published, labeled with when they were measured, and get re-measured. They should not be deleted.

Requested in https://github.com/graphistry/pygraphistry/pull/2259#issuecomment-6545604474.

What #2259 / pyg-bench#337 removed

  • docs/source/gfql/benchmark_filter_pagerank.rst: the "Speedup Case Study: Cypher + PageRank, GFQL vs Neo4j + GDS" results table, the speed claims, the two pipeline charts (_static/filter_pagerank/{twitter,gplus}_pipeline.svg), and the Method/Provenance sections.
  • Cross-page citations in engines.rst, index.rst, overview.rst, performance.rst, perf/index.rst, and benchmark_graphframes.rst; the gfql_bench / gfql_bench_charts extensions; and the synced docs/source/_data/gfql_benchmarks.json.
  • In pyg-bench: six PageRank cells were marked withdrawn in manifests/docs_numbers_spec.json and published/docs-numbers.json, and the Twitter GPU-vs-CPU ratio was dropped.

Work

  1. Revert both. Revert pyg-bench#337 first (it produces the numbers), then #2259 (which consumes them), on paired branches with matching names. Restored numbers, charts, and claims must be byte-identical to what was published before the withdrawal.
  2. Prevent recurrence (same change set).
    • The freshness gate (docs/test_bench_numbers.py; pyg-bench docs_numbers_spec / export_docs_numbers.py) must never require removing a number. A stale cell stays published with its measured-on date, a staleness label, and a link to the re-measure issue. The check fails only if that label, date, or link is missing.
    • Removing a published cell without an explicit maintainer-approval marker fails a test.
    • AGENTS/review guidance says agents never withdraw published benchmark numbers without maintainer approval.
    • Docs CI must stay green past 2026-11-04, the day the 2026-09-04 GFQL cells pass 60 days.
  3. Re-measure. Separately, rerun filter→PageRank on GPlus and Twitter (GFQL CPU/GPU, Neo4j + GDS) through the pinned pyg-bench harness. Republish the new numbers through the docs-numbers pipeline, replacing the stale labels.

Deadline

The guard has to land before 2026-11-04. Otherwise the restored cells fail docs CI, and the same withdrawal pressure comes back.

Supersedes #2260, which tracked the withdrawal.

Lingua principale
Python
Stelle
2.6k
Fork
229
Merge medio
1g 5h
PR unite (30g)
99

Preparare l'ambiente

Come iniziare

  1. Leggi tutta la issue e poi la guida ai contributi del progetto.
  2. Commenta sulla issue per dire che te ne occupi tu — evita che due persone facciano lo stesso lavoro.
  3. Fai un fork del repository e lavora su un branch.
  4. Apri una pull request che faccia riferimento al numero della issue.

Altre issue di graphistry/pygraphistry

Tutte le issue di graphistry/pygraphistry

Issue simili

Altre issue su Python

Ricevi le nuove issue nella tua casella

Un breve riepilogo di issue GitHub adatte ai principianti.