pub const CATEGORICAL_CELL_SENTINEL: char = '\u{0}';Expand description
Prefix a typed Python frame stamps onto a cell that originates from a
genuinely-categorical source column (string / object / categorical dtype).
The column-major inference (infer_and_encode_column_major) and the
schema-guided predict ingest (gam-pyffi::string_records_from_rows) both
strip this prefix before recording or matching a level; its presence forces
the column to Categorical even when every label parses as a number, so a
string column labeled “0”,“1”,“2” is one centred factor level per label
rather than a numeric ramp (#1317 / #1318). A leading NUL never appears in a
numeric literal, so an untyped CSV/array frame (no prefix) is unaffected.