UN/CEFACT splits its vocabulary in two — codes that measure (MTK a square metre, TNE a tonne) and codes that package (BX a box, PA a parcel). A unit read off an article belongs to one family or the other, and the caller rarely knows which.

unitCodeName( 'MTK' ) ; // 'Square Meter'
unitCodeName( 'BX'  ) ; // 'Box'
unitCodeName( 'ZZZ' ) ; // null

The lookup is case-sensitive : unitCodeName( 'mtk' ) returns null.

⚠️ The two families are flattened into a single name. The returned string carries no trace of the family it was resolved against, and that distinction is often the one that matters : a fee stated per square metre can be brought down to an article billed by the square metre ; it cannot be brought down to one billed by the box without knowing what the box holds. A caller that needs to tell a unit that measures from a unit that merely holds must ask MeasureCode::getName() directly and read its null as the signal — this helper is the wrong tool for that job.

The measures are searched first, so a code claimed by both families would resolve to its measure reading. No such code exists today : PT and DB used to collide, but only because MeasureCode declared them wrongly — in Rec 20 PT is a pint and DB a dry pound, never a Point nor a Decibel. Correcting the codes against the official dataset removed the ambiguity rather than arbitrating it.