The ecCodes gate

The decoder is accepted only when its output agrees bit for bit with ecCodes, ECMWF’s reference implementation. The values must be identical, with no tolerance.

Why exact equality is possible

Decoded GRIB values are integers scaled by powers of two and ten, so every value assertion is exact equality, and an inexact double is a decoder bug. The gate forced two refinements.

• Scale factors use ecCodes’ iterated codes_power instead of Math.pow. The two differ by an ulp at decimal scale 6.
• The recorded mean is recomputed with numpy’s pairwise summation, because a naive left-to-right sum differs in the last ulp over millions of doubles.

The twenty-message corpus

The golden corpus in test/fixtures/ is twenty real messages harvested from every live feed the forecast engine reads: GFS, HRRR, NAM, HRDPS (continental and West 1 km), RDPS, GDPS, REPS, GEPS, RAQDPS. Each message is paired with an expectation sidecar derived from ecCodes (2.48.0). A sidecar holds:

• a sha256 of the decoded Float64Array (serialized as little-endian float64),
• exact statistics (count, missing count, min, max, mean),
• 200 evenly spread sampled values, plus known-missing indexes on bitmapped fields, and
• ecCodes’ own nearest-gridpoint answer for every catalogued site.

A decoder that reproduces a sidecar is bit-for-bit compatible with ecCodes on that message. The corpus was curated to cover the awkward shapes. It includes a two-submessage NCEP paired-wind message, a sparse bitmap (76 of 1,905,141 points masked), grid-relative ensemble wind for rotation validation, and the GEPS orography field whose values are decametres although its metadata says metres.

Why the corpus is frozen

The fixtures README, grib/test/fixtures/README.md, records the provenance URLs, the harvest process, and why the corpus cannot be regenerated. ECCC’s Datamart keeps roughly one day of files, so every ECCC source URL expired within ~24 h of harvest, and no harvester ships in this repository. The site coordinates in the sidecars are a frozen copy of the catalogue as it stood at harvest (2026-08-11). They are part of the golden data and do not track the live catalogue. The committed bytes are the ground truth, and the URLs record provenance only.

rdps-cape-sfc-jasper is kept beside the corpus and is not part of it. It is a JasPer-encoded RDPS field that a live smoke surfaced after the corpus froze. It answers to a different oracle (OpenJPEG.js, verified out-of-tree before that codec’s retirement). The golden suite’s enumeration therefore stays the twenty ecCodes messages, and the suites that gate this fixture name it directly. The j2k subset page explains its role on the JPEG 2000 side.

Where the gate runs

The golden suite lives in the package’s test/ directory alongside the module suites. The JPEG 2000 codec configurations sit behind the same gate. Every codec and strategy combination must reproduce the golden answers, in both full and sampled decodes (test/j2k-configs.test.ts).