Repository navigation
enrich(search,#13410): prose interpretative VRP-Logistics App-17/17b (densite 766/658/1074 -> 1215/1220/1247) - #16024
Conversation
…(densite 766/658/1074 -> 1215/1220/1247) Tranche densite sur le trio VRP-Logistics de Search/Applications/Hybrid : 19 cellules markdown d'interpretation placees APRES la cellule de code lue, chacune ancre sur sa sortie executee (borne ceil(328/60)=6 ; NN 595.83 ; CI 585.89 ; 2-opt 583.03 / +12.80 ; recuit 569.24 / +13.79 ; OR-Tools 403.45 charges 60/60 x5 + 43/45 ; decomposition de l'ecart 3.00 vs 10.04 ; Tours C# 36,37..176,92 ; OR-Tools C# 510,59 vs Python 510,47 -- budgets 3 s vs 5 s, NuGet 9.11 vs pip). Diff en insertions pures ; cellules code byte-identiques a origin/main (verifie par comparaison JSON) ; enrich_quality_ci rc=0 x3 ; detect_repeated_prose : findings pre-existants (intro/summary), reduits sur App-17 (1->0) et le twin Python (5->3). See #13410 Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Golden-Set Execution (H.7 P3)✅ 8/8 notebooks passed (certified reproducible)
Pinned lockfile: |
|
Scope = notebooks CHANGED in this PR, not the whole corpus. Explicit |
Notebook outputs-required (H.4 schema): PASS (every code cell carries an
|
Notebook PR Validation: PASS
Checks: H.1 (no errors), H.3 (execution_count), C.1 (no banned patterns) |
|
G-VAR-2/3 GENRE signals (advisory, non bloquant, #10020).
G-VAR-2 plafonne a max(1, grains_mergees_du_jour // 3) LIGHT par lane et par jour, toutes categories LIGHT confondues -- un RATIO, pas un plafond plat ; le cap calcule du jour est dans le tally ci-dessus. G-VAR-3 interdit deux genres LIGHT consecutifs. Les signaux ci-dessus rendent le fait VISIBLE (labels |
…e densite #13410 Les deux cotes de la paire enregistree (App-17 python, App-17b Csharp) ont recu des cellules markdown interpretatives (code byte-identique) -- le gate --per-pair sortait DRIFT INTRODUCED. Attestation --update passee en DERNIER operation (#8957) : python_sha 507af7c, csharp_sha 417ae30 (verifies = blob SHA au HEAD). See #13410 See #8057 Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Path-collision (organ #13359/#13615)Cette PR #16024 (
|
…uctural fold (#16112) The existing organ (#15901) measures source VOLUME. A cell whose source folds entirely into a comment is invisible to it by construction: on the founding case (PR #16097, Lean-18 cell 40cb37d5) the head GROWS 1132 -> 1728 chars because the same write that stripped the newlines appended a 639-char recovery note, so the magnitude gate stops before any floor. The BLOCKING notebook-cell-source-parses guard is blind too -- a fully commented cell parses clean -- which is why it stayed green on a cell whose code had disappeared. Adds two structural signals to the SAME check-run, no new fast-lane wiring: * `emptied` matched cell whose statement count went from > 0 to 0. * `orphan-output` non-empty outputs with 0 statements and no IPython magic (base-free by construction -- the issue's 3rd criterion). The existing exemptions speak about the base->head RELATION, so they arbitrate the comparative signal only; `orphan-output` is intra-cell and stays true whatever happens to the code elsewhere. Both exemption fractions measure 0.00 on the founding case, so nothing was suppressed. `_scan_line` / `_strip_ipython_magics` / `_is_python_kernel` are imported from check_cell_source_parses rather than reimplemented, so the organs cannot disagree about the same cell. Criterion 1 of the issue (unterminated-item count) is REFUTED by measurement, not implemented: per-character serialization already exists on main (GenAI/Texte/21_LoRA_FineTuning.ipynb cell 69b296cb -> 802 unterminated items, 10 statements, real output, healthy). Histogram over the 11970 code cells of main's 953 Python notebooks: {0: 11970, 1: 1, 802: 1}. The signal is withdrawn rather than shipped with a threshold; the folds it targeted are covered twice without it (code-first fold -> the blocking syntax guard; comment-first fold -> `emptied`). Calibration: 0 structural findings on main's full Python corpus; 0 structural findings on the 18 notebooks changed by 12 merged PRs (#16080, #16071, #16069, #16067, #16041, #16039, #16027, #16024, #16021, #16020, #16013, #16012). A first sweep without the no-magic guard flagged 6 cells, all `# comment` + `!python`/`%pip` -- the magic IS the producer, so the guard is measured necessity, not caution. Evidence: 30 tests green in the file, 89 across the sibling organ files; `--self-test` OK with BOTH founding-case replays firing (volume #15901 and structure #16110); live end-to-end run at #16097's head reports the exact measured finding with RC=1. Co-authored-by: Claude Sonnet 5 <noreply@anthropic.com>
…17046) * fix(density,#17040): App-17 VRP-Logistics Python — 1 paraphrase supprimee, 1 parasite editorial retire Suppression de cellules 'Lecture' ajoutees par la campagne de densite (#16024) qui ne portent pas d'interpretation utile sur la sortie lue. * App-17 : suppression 'pourquoi aucune bibliotheque de solveur dans les imports' (f8fba7ef) — paraphrase sans interpretation de l'absence d'OR-Tools dans les imports ; le code de la cellule 6a0adef8 documente deja la strategie. * App-17b-Python : suppression 'le contrat numpy seul' (68a43a2f) — paraphrase du commentaire introductif 'from-scratch' deja explicite dans la section d'introduction du notebook. * App-17b-Python : retrait d'une phrase parasite ('La meme mesure, refaite depuis la solution de l'insertion au moindre cout, en dirait moins encore') inseree au milieu du paragraphe de la cellule 38855cac par la generation automatique — residue editorial sans lien logique. Cellules 'Lecture' conservees : interpretationnelles, chiffrees et positionnees apres la cellule code dont elles lisent la sortie (regle b respectee, regle a OK car aucune lecture ne se chevauche sur une meme sortie). Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> * fix(density): restore App-17b transition and attest twin parity Read both App-17 twins at the PR head; code, outputs and execution counts remain unchanged. Restore the dropped sentence ending and re-attest the markdown-only App-17 change. Co-Authored-By: Claude Code <noreply@anthropic.com> --------- Co-authored-by: Claude Sonnet 5 <noreply@anthropic.com>
2 participants
Grain: DEEP/notebook-python — lane myia-po-2026:CoursIA — prev: LIGHT/tooling #15948
Tranche densite #13410 — VRP-Logistics (Search/Applications/Hybrid)
Rotation de famille (R6) : apres ML (#16007), DecInfer .NET (#16012) et SMT/Z3 (#16021), cette tranche couvre la famille Search, jamais touchee aujourd'hui. Candidats choisis sur mesure LIVE (
pedagogy_density.pysurorigin/main80fc3a815) — le baseline #13410 etant un snapshot Phase-1 fige, jamais retro-fit.App-17-VRP-Logistics.ipynb(python3)App-17b-VRP-Logistics-Python.ipynb(python3)App-17b-VRP-Logistics-Csharp.ipynb(.net-csharp)19 cellules markdown d'interpretation placees APRES la cellule de code lue, chacune ancre sur sa sortie executee :
is_validcomme oracle independant ; criterium d'insertion glouton et sa myopie ; charges OR-Tools60/60x5 +43/60+45/60vs glouton48..60(tasser, pas equilibrer) ; mecanisme du 2-opt avant sa mesure ; decomposition de l'ecart417.49 -> 414.49 -> 403.45(le 2-opt ne recupere que3.00des13.04unités, moins d'un quart) ; lecture geographique de la carte du solveur.ceil(328/60) = 6— compte de vehicules structurellement optimal, tout l'ecart se joue sur la distance ; validateur independant ;595.83(myopie NN) ;-1.7%CI ;+12.802-opt et son plafond intra-tournee ;569.24recuit (relocate inter-tournees + Metropolis) ; « plus compactes » relu structurellement.176,92contre cinq secteurs courts36,37..107,08; OR-Tools510,59vs Python510,47— deux bindings du meme moteur C++ (NuGet9.11/ pip), budgets distincts (3 s/5 s), moins de0,12d'ecart : convergence robuste a l'hote.Validation
json.dumps ensure_ascii=False, indent=1)origin/mainenrich_quality_ci.py --base --headdetect_repeated_prose.pypedagogy_density.py(LIVE, post-commit)Preflight collisions :
gh pr listsur chemins Search -> #15999 (Part2-CSP dll), #15995 (MGS-29), #15955 (09c/09d markdown-only) — aucun chevauchement VRP/Hybrid. Notebooks deja executes (exec = code, 0 erreur) -> exception C.2 markdown-only, aucune re-execution requise.Rectification twin : le preflight initial avait cherche
twin_pairs.d/a la racine du depot (inexistant) et conclu « aucune entree Search » — le registre vit en reality dansscripts/notebook_tools/twin_pairs.d/, ou la paireApp-17 VRP Logistics(App-17 python <-> App-17b-Csharp,native-both) est enregistree. Les deux cotes ayant recu du markdown, le gate--per-pairsortait DRIFT INTRODUCED ; rebaseline prescribe passee en DERNIER operation (commitba28d3d10,python_sha 507af7c1/csharp_sha 417ae30a, verifies = blob SHA au HEAD). Re-verification locale :157 paire(s) | OK=154 INTRO=0(les 3 PRE restants sont pre-existants, hors perimetre #8264). Le twinApp-17b-Pythonn'est PAS dans le registre (gap documente dans l'entree meme).See #13410
🤖 Generated with Claude Code