Keys and sections

ecCodes keys

k = f.keys(0)
k['bitsPerValue'], k['typeOfLevel'], k['stepRange']

f.keys(pick) returns one field's keys under the names ecCodes, pygrib and cfgrib use, read straight from the message's bytes. That covers the identification keys (discipline, parameterCategory, shortName), the level keys (typeOfFirstFixedSurface, scaledValueOfFirstFixedSurface, typeOfLevel, level), grid keys (Ni, Nj, latitudeOfFirstGridPointInDegrees, LoVInDegrees), time keys (forecastTime, typeOfStatisticalProcessing, lengthOfTimeRange, stepRange, stepType), ensemble keys (perturbationNumber) and packing keys (packingType, referenceValue, bitsPerValue, numberOfMissing).

The values equal ecCodes' on every key of every test field, about 17,000 checks. The one exception is negative potential-vorticity levels, which ecCodes reads as unsigned.

Raw sections

For anything not covered by a named key, f.sections(pick) returns the field's sections as a dict of section number to bytes, each whole with its length and number first, so you can read any octet yourself. For a message holding several fields, these are the sections in force for that field.