fix/tariff-cov

Unbounded table count from CFE HTML response

services/utility/tariffs/cfe/src/handlers/tariff-collector.lambda.ts

extractRateCellTables iterates all matching tables with no upper bound before the zones.length guard fires. A malformed page could accumulate many tables in memory before the error is thrown. A simple early-exit (break-equivalent once tables.length > max_expected) would bound memory usage.

Unbounded cell count per scraped table row

services/utility/tariffs/cfe/src/handlers/tariff-collector.lambda.ts

No per-table row count cap on scraped cells. Unexpected page growth could produce unexpectedly long strings passed to cfeRateRowToComponents. Low risk since CFE controls the source, but a max-row guard is cheap defense-in-depth.

Scraped values stored in JSONB without explicit length cap

services/utility/tariffs/cfe/src/handlers/tariff-collector.lambda.ts

parseRateValue strips non-numeric characters but does not assert a length bound before persisting. In practice a valid CFE rate value is ≤10 chars after cleaning. Not exploitable, noted for defense-in-depth.

Error message zone names sourced from static config — confirmed safe

services/utility/tariffs/cfe/src/handlers/tariff-collector.lambda.ts

cfg.divisionGroup and zones.join() in the error template string originate from bundled divisions.json (not user input). No injection risk in current form; note this if zones ever sourced from external input.

Missing NORTE vs CENTRO price-distinctness assertion

services/utility/tariffs/cfe/__tests__/collector.test.ts:217

The 'gives each contained zone ITS OWN table's prices' test checks NORTE≠SUR and CENTRO≠SUR but never checks NORTE≠CENTRO. A bug that assigns table[0]'s prices to both NORTE and CENTRO (while giving SUR table[2]) would produce 3 distinct Set values and pass all current assertions. Add: expect(byZone.get('VALLE DE MEXICO NORTE')).not.toEqual(byZone.get('VALLE DE MEXICO CENTRO')).

Non-null assertion on data.rates[0]! may hide regression

services/utility/tariffs/cfe/__tests__/collector.test.ts:221

In 'gives each zone its own prices', data.rates[0]! throws a TypeError (not a test failure) if runCollection emits no rate entries. A guard expect(data.rates).toHaveLength(1) before the .map() would produce a clear, actionable failure instead.

Mismatch test division assertion is vacuously correct on undefined

services/utility/tariffs/cfe/__tests__/collector.test.ts:242

expect(data.rates[0]?.divisions ?? []).toHaveLength(0) passes both when rates is empty AND when rates[0] exists with an empty divisions array. Verify the production path — if a page-level error fires, rates may be completely empty, and adding expect(data.rates).toHaveLength(1) would pin the expected shape.

Table-to-zone index mapping correct but not explicitly pinned

services/utility/tariffs/cfe/__tests__/collector.test.ts

The test proves three zones have distinct prices but not which zone received which table's prices. A pairing inversion (NORTE ↔ CENTRO swap) passes all current assertions. Low risk given the index comes directly from divisions.json order, but a direct component equality check would fully close the regression surface.

Mismatch failure test does not assert errorCode

services/utility/tariffs/cfe/__tests__/collector.test.ts

The test validates stage and message but not errorCode (expected DIVISION_SCRAPE_FAILED). Adding this assertion would catch a future refactor that changes the code while leaving the message intact.

zones.forEach with index is less readable than a for-loop zip

services/utility/tariffs/cfe/src/handlers/tariff-collector.lambda.ts

zones.forEach((zone, i) => { rates.push({ ..., components: componentsByZone[i]! }) }) pairs two arrays by index with a non-null assertion. A classic for-loop or a pre-zip makes the pairing explicit and drops the !, since the length guard above already guarantees safety.

isTou derivation from .some() assumes all tables share TOU character

services/utility/tariffs/cfe/src/handlers/tariff-collector.lambda.ts

isTou = componentsByZone.some(c => ...) marks the whole page as TOU if any zone has TOU fields. This is correct for current CFE structure, but a short comment stating the assumption (all tables on one page are uniformly TOU or non-TOU) would save future readers from reconstructing the intent.