Function derive_enum_dataframe 

pub(super) fn derive_enum_dataframe(
    row_name: &Ident,
    input: &DeriveInput,
    data: &DataEnum,
    df_name: &Ident,
    attrs: &DataFrameAttrs,
) -> Result<TokenStream>

Derive DataFrameRow for an enum with #[dataframe(align)].

Generates a companion struct where every column is Vec<Option<T>>, with None fill for fields absent in a given variant. This is the enum counterpart of super::derive_struct_dataframe.

Generated items

• Companion struct {Name}DataFrame with Vec<Option<T>> columns (field-name union)
• Optional _tag: Vec<String> column for variant discrimination
• impl IntoDataFrame (companion → R data.frame) and impl ColumnarFrame (rows ↔ the pure-Rust companion: from_rows / from_rows_par)
• impl From<Vec<Enum>> (sequential row->column transposition)
• impl DataFrameRowSplit on the enum — the split (one-data.frame-per-variant) representation, surfaced as rows.into_dataframe_split() via IntoDataFrameSplit’s blanket impl

The write/read verbs are the trait surface — rows.into_dataframe()? (IntoDataFrame / AsDataFrameExt), rows.into_dataframe_split() (IntoDataFrameSplit), Vec::<Row>::from_dataframe(&df)? (FromDataFrame) or the one-call Row::try_from_dataframe(sexp) reader, and the pure-Rust ColumnarFrame companion verbs. The row type keeps two #[doc(hidden)] helpers: to_dataframe(rows) -> companion for the nested-companion write path (Inner::to_dataframe(..) without naming the inner companion type) and to_dataframe_split(rows), the body holder behind the DataFrameRowSplit bridge.

Variant support

• Named variants ({ field: T }): fields contribute by name to the unified schema
• Tuple variants ((T, U)): fields are named _0, _1, etc.
• Unit variants: contribute no columns (only tag if present)

Auto-expand fields

Fields with #[dataframe(expand)] on Vec<T> types get dynamic column counts determined at runtime from the maximum row length across all rows. These are tracked separately from the static ColumnRegistry.

Returns Err if the enum has no variants or if type conflicts arise without #[dataframe(conflicts = "string")].