Data in
Data records and notices
Rows
A CSV file supplies rows to a declared input shape. Its header names fields; rows supply values of their
declared sorts. For example, thing,kind supplies facts about which kind a thing belongs to. The header
counts as line 1 when an answer cites a row.
An unmapped header is refused as data/unmapped, an ambiguous one as data/ambiguous, and a value of the
wrong declared sort as data/sort. Similar column names do not supply a mapping by themselves.
A file's rows establish no completeness claim about missing rows. A finished file supplies its complete captured contents. A file still arriving may gain more rows, and answers depending on those missing rows can remain pending.
Covers and claims
The data owner's notices name the author, telling date and any completeness or omission claims. Backticks preserve file and header spelling. A cover is a dated claim about the file's extent, not a finding that the world has no other facts.
`crm.csv` was told by the lead desk on 2026-10-01.
`crm.csv` is complete for lead through 2026-10-01.
`crm.csv` is complete for recent through 2026-10-01.
The author notice is required for a cover. A cover after its own telling is refused as
cover/after-telling. For an undated class list, a cover that stops before the telling date closes no
absence at the telling date. A file still arriving cannot simultaneously carry a completeness claim for
that class: the combination is refused as data/complete-pending.
A declared input can receive rows; a class computed only by the document's rules cannot be declared
complete by an input file. That claim is refused as data/complete-not-a-class.
What changes an answer
A cover can settle the absence of an input row within its scope. It does not close an unstated reason in the document. A sufficient condition says when a conclusion holds, not that no other reason could hold. The lead examples separate these cases using the same document and different data notices.