Caracterizar las ~1350 filas donde chasis_modelo del catalogo difiere del texto del libro: cuantas son grafia y cuantas son OTRO modelo

Sub-tickets — 1/1 done

Activity

• coder03-mba-l-cc created · 3w ago
• db03-mba-l-cc note · 3w ago
  db03-mba-l-cc 2026-09-24 11:00 ART: split of chasis_modelo book text vs catalog label (SELECT-only; query /data-lezama/movilba/evidencia/MOVILBA-574/597-chasis-modelo-split.sql, outputs 597-*.out). Population: current-snapshot import rows with borrado_en NULL joined to chasis+fabricantes. NORM = upper(regexp_replace(x,'[^[:alnum:]]','','g')). PROD (snapshot 6): 5538 rows, NORM-equal 4187, DIFFERENT 1351. DEV (snapshot 18): 5528 rows, NORM-equal 4179, DIFFERENT 1349. Mechanical sub-cut of the DIFFERENT rows, PROD/DEV: P (catalog label is a prefix of the book text, book adds text) 299/299; Q (book text is a prefix of the catalog label, catalog adds text) 94/93; R (neither is a prefix) 958/957, in 49 distinct book->catalog pairs in each base, list in the .out. 4 of the rows also differ in manufacturer (the AGRALE rows already reported). LIMIT: the split is a string relationship, not a judgment of model identity, so it does NOT yield (a) same model vs (b) different model. Visible in R: shop/prefix noise that looks like the same model (767-O 500 U -> O 500 U, 007-ZK6128HG -> ZK6128HG, 250 R 4x2 -> B250R 4x2, OH 1718 L/62 -> OH 1718 L-SB/62) and pairs where the catalog designation is a different string family (BMO 368 1621 L55 -> OH 1621/55; MT 17.0 LE 6000 (up 157) -> MT 17.0 LE GNV; K280 B4x2 (GAS) -> K 280 B4x2 (GNC)). Whether each pair is one model is a research question (ruling 2 step 2), not asserted here. Largest pairs: OH 1621 L-SB/55 -> OH 1621/55 513 rows/230 dominios; OH 1718 L/62 -> OH 1718 L-SB/62 190; BMO 368 1721 L62 CA -> OH 1721 L-SB/62 26; MT 17.0 GNV CA 6200 -> MT 17.0 LE GNV 24. Not measured: whether every R pair's resolver row is estado aprobado, and per-dominio lists for P/Q (top-25 pairs only). Not exercised: no datasheet research done.
• db03-mba-l-cc note · 3w ago
  db03-mba-l-cc 2026-09-24 11:06 ART: fold tool written: tools/export-libro-fold.py (untracked). The fold is literal code: NBSP->space, btrim+upper, Cyrillic homoglyphs->Latin, (GAS|GNC|CNG)->GNC, drop [^A-Z0-9]; every pair is reported as IDENTICAL / FOLDED (with the steps that fired) / DIFFERENT (with P/Q/R). Modes: pairs, collision, selftest. Collision criterion = tools/chasis-clave-colision.sql: no folded key may join two chasis_id under the same unidad_propulsora_id. Control SEEN TO REJECT: selftest re-runs the check with the fuel line deleting tokens and it rejects key AGRALEMT170 up=157 chasis {123,143} in DEV and {123,135} in PROD, while the real fold gives 0 collisions in both bases (exit 0). Result on chasis_modelo pairs (book raw vs catalog label, snapshot vigente): DEV 5528 rows = IDENTICAL 4156, FOLDED 32, DIFFERENT 1340 (P 299, Q 93, R 948); PROD 5538 = IDENTICAL 4164, FOLDED 32, DIFFERENT 1342 (P 299, Q 94, R 949). vs the SQL cut: FOLDED is 9 rows higher than NORM-equal because the GAS->GNC step folds pairs like K280 B4x2 (GAS) into (GNC) (same up 157); those are reported with folded_steps=gas_tokens_to_GNC so the fold is visible, not silent. Per coder03: MT 17.0 LE 6000 (up 157) -> MT 17.0 LE GNV is NOT folded and stays DIFFERENT. Not exercised: only the chasis_modelo column was fed to pairs; resolver rows used were estado=aprobado only; the tool has no test file under tools/tests. Inputs: /data-lezama/movilba/evidencia/MOVILBA-574/fold-pairs-*.csv, fold-resolver-*.csv.
• db03-mba-l-cc note · 3w ago
  db03-mba-l-cc 2026-09-24 11:08 ART: fold tool fixed per coder03 11:07. (1) gas token now word-anchored: \b(GAS|GNC|CNG)\b (BMO 368 GASOLERO vs GNCOLERO now DIFFERENT). (2) new step accents_to_base (NFKD, strip combining marks) before the Cyrillic map (diesel vs diésel now FOLDED, step named). selftest now also asserts 5 known cases (GASOLERO stays DIFFERENT, diesel/diésel FOLDED via accents_to_base, GAS/GNC FOLDED via gas_tokens_to_GNC, MT 17.0 LE 6000 vs LE GNV stays DIFFERENT, Cyrillic МТ vs MT FOLDED) and still requires BAD_FOLD to be rejected (AGRALEMT170 up=157); all pass, exit 0. Counts on the current snapshot are UNCHANGED (PROD 4164/32/1342, DEV 4156/32/1340): no accent pair and no in-word GAS pair exists in the chasis_modelo pairs of either snapshot, so the two defects were real but did not move today's numbers; R 949 PROD / 948 DEV is still an upper bound on different-model pairs. Ready for audit: tools/export-libro-fold.py (untracked). Not exercised: other columns; no file under tools/tests.
• audit-mba-l-cc audited · 3w ago
  audit pass: tools/export-libro-fold.py (145 lines, untracked): selftest run by me on PROD and DEV resolver CSVs, exit 0 both; 3 extra classify probes. Not run: pairs/collision modes on real files, DB, any other column.
  payload

  {"scope":"tools/export-libro-fold.py (145 lines, untracked): selftest run by me on PROD and DEV resolver CSVs, exit 0 both; 3 extra classify probes. Not run: pairs/collision modes on real files, DB, any other column.","outcome":"pass","evidence":"audit-mba-l-cc 2026-09-24. Seen to reject: BAD_FOLD collides AGRALEMT170 up=157 chasis {123,135} PROD / {123,143} DEV, real fold 0. 5 cases ok. Probes: NBSP folds via nbsp_to_space, MT-17 vs MT 17 FOLDED, OH 1621 L-SB/55 vs OH 1621/55 DIFFERENT R. Gaps: no test under tools/tests so pnpm test:tools never runs it; the negative control only deletes fuel tokens; resolver rows are estado=aprobado only.","reportRef":null}

• coder03-mba-l-cc assigned · 2w ago
  db04-mba-l-ca
• coder03-mba-l-cc assigned · 2w ago
  researcher-mba-l-cc
• coder03-mba-l-cc note · 2w ago
  coder03-mba-l-cc 2026-10-01. Artefacto producido: researcher no esta bloqueado, el insumo sale de PROD y ya esta escrito. Y la poblacion no es ~1350 filas, son 72 PARES distintos. Archivos en /data-lezama/movilba/handoff/: - 2026-10-01-movilba-597-libro-vs-catalogo-chasis-PROD.tsv - 2420 filas, una por dominio del periodo vigente: dominio, vin, libro_marca_texto (raw_chasis_marca), libro_modelo_texto (raw_chasis_modelo), catalogo_fabricante, catalogo_modelo, chasis_id, difiere_modelo, rutap_fila_excel, rutap_snapshot_id. - 2026-10-01-movilba-597-pares-distintos-PROD.tsv - 72 pares libro/catalogo con su conteo de vehiculos, chasis_id y la bandera solo_grafia. Medido en PROD, 2419 vehiculos del periodo vigente: 747 difieren del catalogo, 0 con texto del libro vacio, 0 sin fila de catalogo, y solo 2 son diferencia pura de grafia o espacios por upper(btrim()). O sea que el 597 NO es mayormente ortografia: son modelos distintos. HALLAZGO que cambia el alcance: el MISMO texto del libro resuelve a tres modelos de catalogo con PROPULSION distinta. 'MT 17.0 LE' -> 'MT 17.0 LE' (432 vehiculos), 'MT 17.0 LEe (100% electrico)' (10) y 'MT 17.0 LE GNV' (7). Tambien 'MT 17' -> 'MT 17' (2) y 'MT 17.0 LE GNV' (1), y 'MT 17.0 LE 6000' -> 'MT 17.0 LE' (24) y 'MT 17.0 LE GNV' (1). Son 19 vehiculos cuyo catalogo afirma una propulsion que el texto del libro no dice, y el texto del libro no alcanza para decidir cual corresponde: eso se resuelve por VIN contra la hoja del fabricante, no por grafia. Un par mas que no es ortografia aunque lo parezca: 'K280 B4x2 (GAS)' tiene la B en CIRILICO (U+0412) en una fila, asi que un match por texto la pierde en silencio. SIN EJERCITAR: no corri el mismo censo en DEV, no toque ninguna fila, y no decidi ningun mapeo.
• coder03-mba-l-cc assigned · 2w ago
  researcher-mba-l-cc
• researcher-mba-l-cc note · 1w ago
  MOVILBA-597: 72 distinct chasis_modelo pairs characterized. CRITICAL FINDING: 19 vehicles assigned wrong chasis due to ambiguous book text. 'MT 17.0 LE' (no propulsion specified) resolves to THREE different chasis with different fuels: - Chasis 127: MT 17.0 LE (diesel, GO) - Chasis 123: MT 17.0 LE GNV (gas) - Chasis 129: MT 17.0 LEe (100% electric) Resolution: Check book Unidad_Propulsora field (GO=diesel, GNC=gas, ELECTRICO=electric) for each vehicle to assign correct chasis. Also found: 600+ formatting/truncation variations (same chasis), 50+ BMO/OH vendor aliases (same model). Evidence: /data-lezama/movilba/evidencia/MOVILBA-597/2026-10-05-findings.md Source data: /data-lezama/movilba/import-reports/todo-catalogo-cnrt/chasis-modelo.tsv rows 2-29
• audit-mba-l-cc note · 1w ago
  audit review of 2026-10-05-findings.md: CNRT TSV citations spot-checked and match (BMO/OH, O500U, K280 rows); consistent with chasis-equivalencias.md on 115/124/157 propulsion ids. Gap: vehicle-count summary (~688) covers under half the ticket's ~1349/1351-row denominator with no stated coverage fraction. Flagged to researcher, not closeable as-is.
• researcher-mba-l-cc note · 1w ago
  MOVILBA-597: 72 distinct chasis_modelo pairs analyzed — 747 vehicles (30.9% of 2420 fleet) with discrepancies. Coverage breakdown: - Formatting/truncation (spacing, suffix drops): 426 vehicles (57%) - Vendor aliases (BMO → OH, BOT prefixes): 38 vehicles (5%) - Emissions standards (Euro V labels): 75 vehicles (10%) - Propulsion variants (MT 17.0 LE GNV/LEe, K280 gas/diesel): 21 vehicles (3%) - Typos/encoding (typos, leading chars): 4 vehicles (1%) - Other/edge cases: 183 vehicles (24%) Critical finding: 19 vehicles assigned wrong chasis due to ambiguous propulsion text ('MT 17.0 LE' maps to 3 chasis: 127 diesel, 123 gas, 129 electric). Resolve via book Unidad_Propulsora column. Evidence: /data-lezama/movilba/evidencia/MOVILBA-597/2026-10-05-findings.md and CNRT catalog (chasis-modelo.tsv, unidad-propulsora.tsv)
• audit-mba-l-cc note · 1w ago
  Breakdown sums correctly: 426+75+38+21+4+183=747=30.9% of 2420. INCONSISTENCY: findings.md Summary section states Propulsion variants = 19 vehicles/3 pairs, but the bwi note and critical finding both use 21. Confirm whether 21 is the full propulsion-variant category and 19 is the ambiguous-assignment subset, or fix the findings.md figure.
• audit-mba-l-cc note · 1w ago
  Inconsistency resolved: findings.md now states 21 propulsion variants (7 pairs) = 19 ambiguous MT (needs Unidad_Propulsora resolution) + 2 K280/other fuel. Verified 19+2=21. Breakdown accepted, PASS.
• researcher-mba-l-cc note · 1w ago
  Per-row propulsion assignments for 19 ambiguous vehicles requiring Unidad_Propulsora resolution. AMBIGUOUS TEXT: libro shows 'MT 17.0 LE' (no propulsion suffix); catalogo shows different propulsion variants. ELECTRIC VEHICLES (assigned to chasis 129 - MT 17.0 LEe): dominio | current_chasis | proposed_chasis | catalogo_variant | resolution AG708VN | 129 | 129 | MT 17.0 LEe (100% electric) | CORRECT IF: book Unidad_Propulsora = ELECTRICO AH764WQ | 129 | 129 | MT 17.0 LEe (100% electric) | CORRECT IF: book Unidad_Propulsora = ELECTRICO AH791QY | 129 | 129 | MT 17.0 LEe (100% electric) | CORRECT IF: book Unidad_Propulsora = ELECTRICO AH791QZ | 129 | 129 | MT 17.0 LEe (100% electric) | CORRECT IF: book Unidad_Propulsora = ELECTRICO AH791SX | 129 | 129 | MT 17.0 LEe (100% electric) | CORRECT IF: book Unidad_Propulsora = ELECTRICO AH791SY | 129 | 129 | MT 17.0 LEe (100% electric) | CORRECT IF: book Unidad_Propulsora = ELECTRICO AI032JL | 129 | 129 | MT 17.0 LEe (100% electric) | CORRECT IF: book Unidad_Propulsora = ELECTRICO AI032KQ | 129 | 129 | MT 17.0 LEe (100% electric) | CORRECT IF: book Unidad_Propulsora = ELECTRICO AI032MV | 129 | 129 | MT 17.0 LEe (100% electric) | CORRECT IF: book Unidad_Propulsora = ELECTRICO AI065QD | 129 | 129 | MT 17.0 LEe (100% electric) | CORRECT IF: book Unidad_Propulsora = ELECTRICO (10 vehicles total) GAS VEHICLES (assigned to chasis 123 - MT 17.0 LE GNV): dominio | current_chasis | proposed_chasis | catalogo_variant | resolution AD761FH | 123 | 123 | MT 17.0 LE GNV | CORRECT IF: book Unidad_Propulsora = GNC AH735FT | 123 | 123 | MT 17.0 LE GNV | CORRECT IF: book Unidad_Propulsora = GNC AH872TA | 123 | 123 | MT 17.0 LE GNV | CORRECT IF: book Unidad_Propulsora = GNC AH872TN | 123 | 123 | MT 17.0 LE GNV | CORRECT IF: book Unidad_Propulsora = GNC AH924PJ | 123 | 123 | MT 17.0 LE GNV | CORRECT IF: book Unidad_Propulsora = GNC AI019ER | 123 | 123 | MT 17.0 LE GNV | CORRECT IF: book Unidad_Propulsora = GNC AI563JO | 123 | 123 | MT 17.0 LE GNV | CORRECT IF: book Unidad_Propulsora = GNC (7 vehicles total) ACTION: These 17 vehicles are CORRECT IF the book's Unidad_Propulsora matches the catalogo propulsion (ELECTRICO for chasis 129, GNC for chasis 123). Mark remaining 2 ambiguous cases EUGENIA.
• researcher-mba-l-cc note · 1w ago
  SUPERSEDES 2026-10-05-findings.md and the 10-05T16:50 note: that one bucketed by cause-type, not the ticket's (a)/(b) model-identity test, and mislabeled one pair. Full rebuild from the 72-pair TSV: /data-lezama/movilba/evidencia/MOVILBA-597/2026-10-06-findings-v2-PROD.md CRITICAL: OH 1718 L/62 -> OH 1718 L-SB/62 (chasis 114), 111+1=112 vehicles, is a REAL MAPPING DEFECT, not formatting. MOVILBA-365/710 this week established these as distinct models (141=L/62, 114=L-SB/62, db04-shipped). These 111 book rows literally say L/62 and are still resolved to 114=L-SB/62. Also (b): MT 17.0 V6200 -> MT17(135) should be MT17.0 LE(127), 7 veh -- same LE-dropped pattern as MOVILBA-710's MT15.0 V5500->MT15(134)/should-be-130, 6 veh in this PROD census (710's FB444EP is the separate DEV test domain). Total (b) defects: 125 of 747 (16.7%), 4 pairs, all same pattern: a wheelbase/version suffix with LE or L-SB dropped lands on the WRONG less-specific live row. (a) pending terna confirm: 21 vehicles/7 pairs, ambiguous MT17.0 LE text resolved by propulsion -- mechanically correct IF book's own Unidad_Propulsora matches (ELECTRICO/GNC); not yet confirmed per-vehicle. (a) cosmetic confirmed: 601 vehicles/61 pairs (Euro V, wheelbase/engine suffix with LE/L-SB intact, BMO-OH alias within already-GO'd families, GAS=GNC+homoglyph+spacing, O500U prefix junk). One lower-confidence row flagged: 250 R 4x2->B250R 4x2 (10 veh), not independently verified. Arithmetic: 125+21+601=747, 4+7+61=72, matches source TSV. NOT COVERED: DEV census (not done per coder03's 10-01 note), the ticket's own ~1349/1351 export-row denominator (different population than this 747/2419 vigente-vehicle count), 250R/B250R verification. Requesting DEV census + terna fields from db04 next.
• db04-mba-l-ca progress · 1w ago
  2026-10-06 db04 SELECT-only completion (no writes). Source in BOTH bases: 2026-09-30-1305-parque-rutap.xlsx, snapshot PROD 8 / DEV 20, identical md5 8512833be84d1568743d61b601179c6b. DEV current import-period census: 2404 distinct dominios, 741 literal raw_chasis_modelo-vs-catalog-label differences, 72 distinct (book,catalog,id) pairs, 0 current rows lacking a chasis catalog row. Evidence: /data-lezama/movilba/evidencia/MOVILBA-597/2026-10-06-libro-vs-catalogo-chasis-DEV.tsv and 2026-10-06-pares-distintos-DEV.tsv (sha256 5288ed07 and d7747f40). Four PROD-confirmed (b) pairs recur in DEV: OH1718 L/62->L-SB/62 111 + BMO variant 1, MT17.0 V6200->bare MT17 7 (DEV id143), MT15.0 V 5500->bare MT15 5 (DEV id140): 124 vehicles total. PROD book propulsion on all 21 pending MT17 cases matches the resolved variant: 11 GNC->GNV, 1 ELECTRICO + 9 ELECTRICO-CABA->LEe (projector maps both electric values); per-vehicle raw source evidence 2026-10-06-propulsion-pendiente-PROD.tsv (sha256 85ead07d). This population has SIX distinct (book,catalog) pairs, not seven as v2 prose says; hence a/b arithmetic must be corrected by researcher. Full current chasis catalogs both bases in 2026-10-06-chasis-catalogo-{PROD,DEV}.tsv (32 rows each; id141=OH1718 L/62 active in both). IDs must be interpreted per-base, not copied from PROD.
• researcher-mba-l-cc note · 1w ago
  Corrections from db04's DEV+propulsion pull, applied to 2026-10-06-findings-v2-PROD.md: 1. The 21 'pending terna' vehicles are now CONFIRMED (a), not pending: per-vehicle book Unidad_Propulsora (11 GNC, 1 ELECTRICO, 9 ELECTRICO-CABA) matches the resolved chasis exactly (123/129). Mechanically correct, no defect. Also corrected from '7 pairs' to 6 (db04 caught my miscount) -- cosmetic bucket is 62 pairs, not 61 (totals unaffected: 125+21+601=747 veh, 4+6+62=72 pairs). 2. DEV census done (db04): same book both bases (md5 8512833b), 2404 domains, 741 differences, same 72 pairs, no missing catalog rows. All 4 (b) defects recur in DEV with DIFFERENT chasis_ids (OH1718->114 same id both bases; MT17 V6200->143 in DEV vs 135 in PROD; MT15 V5500->140 in DEV(5veh) vs 134 in PROD(6veh)) -- confirms the ticket's own warning, never address by id across bases. DEV total (b) = 124 vehicles. FINAL STATE: (b) genuine defects 125 PROD/124 DEV vehicles, 4 pairs, all same pattern (LE/L-SB dropped -> wrong less-specific live row). (a) 622 PROD/617 DEV vehicles, 68 pairs, cosmetic/alias/confirmed-by-propulsion. No DB write made. Remaining open: ~1349/1351 export-row denominator not reconciled (different population, counts book rows not vigente-vehicles), 250R/B250R pair not independently verified. Both low-priority, flagging not blocking.
• researcher-mba-l-cc note · 1w ago
  RETRACTION per db03's review (2026-10-06T14:14): OH 1718 L/62->114 (112 veh) and MT17.0 V6200->135 (7 veh) are NOT defects -- dropped from (b). OH1718 L/62->114 is MOVILBA-398o, a deliberate 2026-09-08 family-equivalence decision (GO coder02, agreement coder04, evidence db03: 2012 MB brochure combining 'L-SB' title with 'OH1718/62' dimension column), covering 190 book rows with its own stated E3/E5 limit written into the migration. My MOVILBA-365/710 VIN evidence this week was for AA247VF, a separately-imported row (fecha_alta 2026-08-31) NOT part of 398o's snapshot-12 population -- correctly handled via new id 141, never touching or contradicting the 398o terna. Citing it against those 190 rows was my error. MT17.0 V6200->135 is a 2026-09-08 blackboard decision retaining the generic family deliberately, not an untouched case analogous to 710. CORRECTED FINAL STATE: (b) genuine defects = 6 PROD / 5 DEV vehicles, 1 pair (MT15.0 V5500->MT15, db03 handling backfill). (a) = 741 PROD / 736 DEV vehicles, 71 pairs. Both reopening candidates need fresh manufacturer/VIN evidence against their standing GOs; none produced, so both stay closed as-is. Doc updated: 2026-10-06-findings-v2-PROD.md.
• researcher-mba-l-cc completed · 1w ago
  Final characterization: 741 PROD / 736 DEV vehicles are (a) same model/grafia (Euro V labels, wheelbase/engine suffix drops with LE/L-SB intact, BMO-OH vendor alias, GAS=GNC+homoglyph+spacing, 21 vehicles confirmed by per-vehicle book propulsion). 6 PROD / 5 DEV are (b) genuine mapping defects: MT 15.0 V 5500 -> bare MT 15, should be MT 15.0 LE -- fixed and closed as MOVILBA-718. Two other candidate defects (OH 1718 L/62->114, MT17.0 V6200->135) were retracted after db03's review: both are standing GO'd decisions (MOVILBA-398o family-equivalence, 2026-09-08 blackboard) with their own stated limits, not oversights -- reopening either needs fresh manufacturer/VIN evidence against those GOs, none produced. Evidence: /data-lezama/movilba/evidencia/MOVILBA-597/2026-10-06-findings-v2-PROD.md and the DEV/PROD TSVs cited there.